OpenProjects.html

<!--#include virtual="header.incl" -->

<div class="www_sectiontitle">Open LLVM Projects</div>

<ul>
  <li>Google Summer of Code Ideas & Projects
    <ul>
      <li>
        <a href="#gsoc25">Google Summer of Code 2025</a>
        <ul>
          <li><b>LLVM Core</b>
            <ul>
              <li><a href="#llvm-abi-lowering">Introduce an ABI lowering library</a></li>
              <li><a href="#byte-type">Byte type</a></li>
           </ul>
          <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
            <ul>
              <li><a href="#clang-driver-modules">Simple C++20 modules without a build system</a></li>
              <li><a href="#clang-improve-trapping-ubsan-2025">Usability Improvements for trapping Undefined Behavior Sanitizer (UBSan)</a></li>
              <li><a href="#improve-documentation-parsing-in-clang">Improve documentation parsing in Clang</a></li>
              <li><a href="#improve-symbol-discovery-clang-repl">Advanced symbol resolution and reoptimization for clang-repl</a></li>
            </ul>
          </li>
          <li><a href="http://lldb.llvm.org/"><b>LLDB</b></a>
            <ul>
              <li><a href="#rich-disassembler-for-lldb-2025">Rich disassembler for LLDB</a></li>
            </ul>
          </li>
          <li><a href="http://libc.llvm.org/"><b>LLVM libc</b></a>
            <ul>
              <li><a href="#bfloat16-libc">Bfloat16 in LLVM libc</a></li>
              <li><a href="#io_uring-libc">Direct I/O from the GPU with io_uring</a></li>
              <li><a href="#testing-math-libc">Profiling and testing the LLVM libc GPU math</a></li>
            </ul>
          </li>
          <li><a href="https://clangir.org"><b>ClangIR</b></a>
            <ul>
              <li><a href="#clangir-codegen-tests">Validate existing Clang CodeGen test coverage with ClangIR</a></li>
              <li><a href="#clangir-codegen-upstreaming">Participate in ClangIR upstreaming</a></li>
            </ul>
          </li>
          <li><a href="https://clang.llvm.org/docs/ClangStaticAnalyzer.html"><b>Clang Static Analyzer</b></a>
            <ul>
              <li><a href="#csa-lifetime-annotations">Teach the Clang Static Analyzer to understand lifetime annotations</a></li>
            </ul>
          </li>
          <li><a href="https://enzyme.mit.edu"><b>Enzyme</b></a>
            <ul>
              <li><a href="#rust-enzyme-improvements">Improve Enzyme reliability and compile times for Rust</a></li>
            </ul>
          </li>
          <li><a href="https://discourse.llvm.org/t/offload-new-subproject-pending-move-of-libomptarget/78185"><b>Offload</b></a> (Former <a href="https://openmp.llvm.org/"><b>OpenMP Offloading</b></a>)
            <ul>
              <li><a href="#offload-remarks-visualization">LLVM Compiler Remarks Visualization Tool for Offload Proposal</a></li>
            </ul>
          </li>
        </ul>
      </li>

      <li>
        <a href="#gsoc24">Google Summer of Code 2024</a>
        <ul>
          <li><b>LLVM Core</b>
            <ul>
              <li><a href="#remove_ub_tests">Remove undefined behavior from tests</a></li>
              <li><a href="#spirv_tablegen">Automatically generate TableGen file for SPIR-V instruction set</a></li>
              <li><a href="#bitstream_cas">LLVM bitstream integration with CAS (content-addressable storage)</a></li>
              <li><a href="#three_way_comparison">Add 3-way comparison intrinsics</a></li>
              <li><a href="#llvm_www">Improve the LLVM.org Website Look and Feel</a></li>
              <li><a href="#parameter-tuning">The 1001 thresholds in LLVM</a></li>
           </ul>
          <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
            <ul>
              <li><a href="#clang-repl-out-of-process">Out-of-process execution for clang-repl</a></li>
              <li><a href="#clang-plugins-windows">Support clang plugins on Windows</a></li>
              <li><a href="#clang-on-demand-parsing">On Demand Parsing in Clang</a></li>
              <li><a href="#clang-doc-improve-usability">Improve Clang-Doc Usability</a></li>
            </ul>
          <li><a href="http://lldb.llvm.org/"><b>LLDB</b></a>
            <ul>
              <li><a href="#rich-disassembler-for-lldb">Rich disassembler for LLDB</a></li>
            </ul>
          <li><a href="http://openmp.llvm.org/"><b>(OpenMP) Offload</b></a>
            <ul>
              <li><a href="#gpu-delta-debugging">GPU Delta Debugging</a></li>
              <li><a href="#offload-libcxx">Offloading libcxx</a></li>
              <li><a href="#gpu-libc">Performance tuning the GPU libc</a></li>
              <li><a href="#gpu-first">Improve GPU First Framework</a></li>
            </ul>
          <li><a href="https://clangir.org"><b>ClangIR</b></a>
            <ul>
              <li><a href="#clangir-gpu">Compile GPU kernels using ClangIR</a></li>
            </ul>
          <li><a href="http://libc.llvm.org/"><b>LLVM libc</b></a>
            <ul>
              <li><a href="#half-precision-libc">Half precision in LLVM libc</a>
            </ul>
          </ul>

      </li>

      <li>
        <a href="#gsoc23">Google Summer of Code 2023</a>
        <ul>
          <li>
            <b>LLVM Core</b>
            <ul>
               <li><a href="#llvm_new_jitlink_reopt">Re-optimization using JITLink</a></li>
               <li><a href="#llvm_new_jitlink_backends">JITLink new backends</a></li>
               <li><a href="#llvm_improving_compile_times">Improving compile times</a></li>
               <li><a href="#llvm_addressing_rust_optimization_failures">Addressing Rust optimization failures</a></li>
               <li><a href="#llvm_mlgo_latency_model">Better performance models for MLGO training</a></li>
               <li><a href="#llvm_mlgo_passes_2023">Machine Learning Guided Ordering of Compiler Optimization Passes</a></li>
               <li><a href="#llvm_map_value_to_src_expr">Map LLVM values to corresponding source-level expressions</a></li>
            </ul>
          </li>
          <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
            <ul>
              <li><a href="#clang-repl-out-of-process">Out-of-process execution for clang-repl</a>
              <li><a href="#clang_analyzer_taint_analysis">Improve and Stabilize the Clang Static Analyzer's "Taint Analysis" Checks</a></li>
              <li><a href="#clang-repl-autocompletion">Implement autocompletion in clang-repl</a>
              <li><a href="#clang-modules-build-daemon">Modules build daemon: build system agnostic support for explicitly built modules</a></li>
              <li><a href="#clang-extract-api-categories">ExtractAPI Objective-C categories</a></li>
              <li><a href="#clang-extract-api-cpp-support">ExtractAPI C++ Support</a></li>
              <li><a href="#clang-extract-api-while-building">ExtractAPI while building</a></li>
              <li><a href="#clang-improve-diagnostics2">Improve Clang diagnostics</a></li>
              <li><a href="#clang-tutorials-clang-repl">Tutorial development with clang-repl</a></li>
              <li><a href="#clang-repl-wasm">Add WebAssembly Support in clang-repl</a></li>
              </li>
            </ul>
          </li>

          <li>
            <b>LLD</b>
            <ul>
              <li><a href="#llvm_lld_embedded">LLD Linker Improvements for Embedded Targets</a></li>
            </ul>
          </li>

          <li>
            <b>MLIR</b>
            <ul>
              <li><a href="#llvm_mlir_presburger_opt">Optimizing MLIR’s Presburger library</a></li>
              <li><a href="#llvm_mlir_query">Interactively query MLIR IR</a></li>
            </ul>
          </li>

          <li>
            <b>Code Coverage</b>
            <ul>
              <li><a href="#llvm_code_coverage">Support a hierarchical directory structure in generated coverage html reports</a></li>
              <li><a href="#llvm_patch_coverage">Patch based test coverage for quick test feedback</a></li>
            </ul>
          </li>

          <li>
            <b>ClangIR</b>
            <ul>
              <li><a href="#clangir">Build and run SingleSource benchmarks using ClangIR</a></li>
            </ul>
          </li>

          <li>
            <b><a href="https://enzyme.mit.edu">Enzyme</a></b>
            <ul>
              <li><a href="#enzyme_tblgen_extension">Move additional Enzyme Rules to Tablegen</a></li>
            </ul>
          </li>

        </ul>
      </li>

      <li>
        <a href="#gsoc22">Google Summer of Code 2022</a>
        <ul>
          <li>
            <b>LLVM Core</b>
            <ul>
              <li><a href="#llvm_shared_jitlink">Implement a shared-memory based JITLinkMemoryManager for out-of-process JITting</a></li>
              <li><a href="#llvm_build_jit_tutorial">Modernize the LLVM "Building A JIT" tutorial series</a></li>
              <li><a href="#llvm_jit_new_format">Write JITLink support for a new format/architecture</a></li>
              <li><a href="#llvm_instrumentaion_for_compile_time">Instrumentation of Clang/LLVM for Compile Time</a></li>
              <li><a href="#llvm_lto_dependency_info">Richer symbol dependency information for LTO</a></li>
              <li><a href="#llvm_mlgo_passes">Machine Learning Guided Ordering of Compiler Optimization Passes</a></li>
              <li><a href="#llvm_mlgo_loop">Learning Loop Transformation Heuristics</a></li>
              <li><a href="#llvm_module_inliner">Evaluate and Expand the Module-Level Inliner</a></li>
              <li><a href="#llvm_undef_load">Remove undef: move uninitialized memory to poison</a></li>
              <li><a href="#llvm_abi_export">Add ABI/API export annotations to the LLVM build</a></li>
            </ul>
          </li>

          <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
            <ul>
              <li><a href="#clang-template-instantiation-sugar">Extend clang AST to
                provide information for the type as written in template
                instantiations</a>
              </li>
              <li><a href="#clang-sa-structured-bindings">Implement support for
                C++17 structured bindings in the Clang Static Analyzer</a>
              </li>
              <li><a href="#clang-improve-diagnostics">Improve Clang Diagnostics</a>
              </li>
            </ul>
          </li>

          <li>
            <a href="https://polly.llvm.org"><b>Polly</b></a>
            <ul>
              <li><a href="#polly_npm">Completely switch to new pass manager</a></li>
            </ul>
          </li>

          <li>
            <b><a href="https://enzyme.mit.edu">Enzyme</a></b>
            <ul>
              <li><a href="#enzyme_tblgen">Move Enzyme Instruction Transformation Rules to Tablegen</a></li>
              <li><a href="#enzyme_vector">Vector Reverse-Mode Automatic Differentiation</a></li>
              <li><a href="#enzyme_pm">Enable The New Pass Manager</a></li>
            </ul>
          </li>
        </ul>
      </li>

      <li>
        <a href="#gsoc21">Google Summer of Code 2021</a>
        <ul>
          <li>
            <b>LLVM Core</b>
            <ul>
              <li><a href="#llvm_distributing_lit">Distributed lit testing</a></li>
              <li><a href="#llvm_loop_heuristics">Learning Loop Transformation Heuristics</a></li>
              <li><a href="#llvm_ir_fuzzing">Fuzzing LLVM-IR Passes</a></li>
              <li><a href="#llvm_ir_assume"><tt>llvm.assume</tt> the missing pieces</a></li>
              <li><a href="#llvm_shared_jitlink">Implement a shared-memory based JITLinkMemoryManager for out-of-process JITting</a></li>
              <li><a href="#llvm_build_jit_tutorial">Modernize the LLVM "Building A JIT" tutorial series</a></li>
              <li><a href="#llvm_jit_new_format">Write JITLink support for a new format/architecture</a></li>
              <li><a href="#llvm_ir_issues">Fix fundamental issues in LLVM's IR</a></li>
              <li><a href="#llvm_utilize_loopnest">Utilize LoopNest Pass</a></li>
            </ul>
          </li>

          <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
            <ul>
              <li><a href="#clang-template-instantiation-sugar">Extend clang AST to
                provide information for the type as written in template
                instantiations</a>
              </li>
            </ul>
          </li>

          <li>
            <b>OpenMP</b>
            <ul>
              <li><a href="#openmp_gpu_jit">JIT-ing OpenMP GPU kernels transparently</a></li>
            </ul>
          </li>

          <li>
            <b>OpenACC</b>
            <ul>
              <li><a href="#openacc_rt_diagnostics">OpenACC Diagnostics from the OpenMP Runtime</a></li>
            </ul>
          </li>
          <li>
            <b><a href="https://polly.llvm.org">Polly</a></b>
            <ul>
              <li><a href="#polly_isl_bindings">Use official isl C++ bindings</a></li>
            </ul>
          </li>

          <li>
            <b><a href="https://enzyme.mit.edu">Enzyme</a></b>
            <ul>
              <li><a href="#enzyme_blas">Integrate custom derivatives of BLAS, Eigen, and similar routines into Enzyme</a></li>
              <li><a href="#enzyme_swift">Integrate Enzyme into Swift to provide high-performance differentiation in Swift</a></li>
              <li><a href="#enzyme_fixed">Differentiation of Fixed-Point Arithmetic</a></li>
              <li><a href="#enzyme_rust">Integrate Enzyme into Rust to provide high-performance differentiation in Rust</a></li>
            </ul>
          </li>

          <li>
            <b>Clang Static Analyzer</b>
            <ul>
              <li><a href="#static_analyzer_profling">Clang Static Analyzer performance profiling</a></li>
              <li><a href="#static_analyzer_constraint_solver">Clang Static Analyzer constraint solver improvements</a></li>
            </ul>
          </li>

          <li>
            <b>LLDB</b>
            <ul>
              <li><a href="#lldb_diagnostics">A structured approach to diagnostics in LLDB</a></li>
            </ul>
          </li>
        </ul>
      </li>

      <li>
        <a href="#gsoc20">Google Summer of Code 2020</a>
          <ul>
            <li>
              <b>LLVM Core</b>
              <ul>
                <li><a href="#llvm_optimized_debugging">Improve debugging of optimized code</a></li>
                <li><a href="#llvm_ipo">Improve inter-procedural analyses and optimizations</a></li>
                <li><a href="#llvm_par">Improve parallelism-aware analyses and optimizations</a></li>
                <li><a href="#llvm_dbg_invariant">Make LLVM passes debug info invariant</a></li>
                <li><a href="#llvm_mergesim">Improve MergeFunctions to incorporate MergeSimilarFunction patches and ThinLTO Support</a></li>
                <li><a href="#llvm_dwarf_yaml2obj">Add DWARF support to yaml2obj</a></li>
                <li><a href="#llvm_hotcold">Improve hot cold splitting to aggressively outline small blocks</a></li>
                <li><a href="#llvm_pass_order">Advanced Heuristics for Ordering Compiler Optimization Passes</a></li>
                <li><a href="#llvm_ml_scc">Machine learning and compiler optimizations: using inter-procedural analysis to select optimizations</a></li>
                <li><a href="#llvm_postdominators">Add PostDominatorTree in LoopStandardAnalysisResults</a></li>
                <li><a href="#llvm_loopnest">Create loop nest pass</a></li>
                <li><a href="#llvm_instdump">Instruction properties dumper and checker</a></li>
                <li><a href="#llvm_movecode">Unify ways to move code or check if code is safe to be moved</a></li>
              </ul>
            <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
              <ul>
                <li><a href="#clang-template-instantiation-sugar">Extend clang AST to
                    provide information for the type as written in template
                    instantiations</a>
                </li>
                <li><a href="#clang-sa-cplusplus-checkers">Find null smart pointer dereferences
                    with the Static Analyzer</a>
                </li>
              </ul>
            </li>
            <li><a href="http://lldb.llvm.org/"><b>LLDB</b></a></li>
            <ul>
              <li><a href="#lldb-autosuggestions">Support autosuggestions in LLDB's command line</a></li>
              <li><a href="#lldb-more-completions">Implement the missing tab completions for LLDB's command line</a></li>
              <li><a href="#lldb-reimplement-lldb-cmdline">Reimplement LLDB's command-line commands using the public SB API.</a></li>
              <li><a href="#lldb-batch-testing">Add support for batch-testing to the LLDB testsuite.</a></li>
            </ul>
            <li>
              <b>MLIR</b>
              <ul>
                <li>See the <a href="https://mlir.llvm.org/getting_started/openprojects/">MLIR open project list</a></li>
              </ul>
            </li>
          </ul>
      </li>

      <li>
        <a href="#gsoc19">Google Summer of Code 2019</a>
        <ul>
          <li>
            <b>LLVM Core</b>
            <ul>
              <li><a href="#debuginfo_codegen_mismatch">Debug Info should have no
                  effect on codegen</a></li>
              <li><a href="#llvm_function_attributes">Improve (function) attribute
                  inference</a></li>
              <li><a href="#improve_binary_utilities">Improve LLVM binary utilities
              </a></li>
            </ul>
          </li>
          <li><a href="http://clang.llvm.org/"><b>Clang</b></a>
            <ul>
              <li><a href="#clang-astimporter-fuzzer">Implement an ASTImporter
                  fuzzer</a>
              </li>
              <li><a href="#improve-autocompletion">Improve shell autocompletion
                  for Clang</a>
              </li>
              <li><a href="#analyze-llvm">Apply the Clang Static Analyzer to LLVM-based
                  Projects</a>
              </li>
              <li><a href="#header-generation">Generate annotated sources based on
                  LLVM-IR analyses</a>
              </li>
            </ul>
          </li>
        </ul>
      </li>

      <li><a href="#gsoc18">Google Summer of Code 2018</a></li>
      <li><a href="#gsoc17">Google Summer of Code 2017</a></li>
    </ul>
  </li>

  <li><a href="#what">What is this?</a></li>
  <li><a href="#subprojects">LLVM Subprojects: Clang and more</a></li>
  <li><a href="#improving">Improving the current system</a>
  <ol>
    <li><a href="#target-desc">Factor out target descriptions</a></li>
    <li><a href="#code-cleanups">Implementing Code Cleanup bugs</a></li>
    <li><a href="#programs">Compile programs with the LLVM Compiler</a></li>
    <li><a href="#llvmtest">Add programs to the llvm-test suite</a></li>
    <li><a href="#benchmark">Benchmark the LLVM compiler</a></li>
    <li><a href="#statistics">Benchmark Statistics and Warning System</a></li>
    <li><a href="#coverage">Improving Coverage Reports</a></li>
    <li><a href="#misc_imp">Miscellaneous Improvements</a></li>
  </ol></li>

  <li><a href="#new">Adding new capabilities to LLVM</a>
  <ol>
    <li><a href="#llvm_ir">Extend the LLVM intermediate representation</a></li>
    <li><a href="#pointeranalysis">Pointer and Alias Analysis</a></li>
    <li><a href="#profileguided">Profile-Guided Optimization</a></li>
    <li><a href="#compaction">Code Compaction</a></li>
    <li><a href="#xforms">New Transformations and Analyses</a></li>
    <li><a href="#codegen">Code Generator Improvements</a></li>
    <li><a href="#misc_new">Miscellaneous Additions</a></li>
  </ol></li>

  <li><a href="#using">Project using LLVM</a>
  <ol>
    <li><a href="#machinemodulepass">Add a MachineModulePass</a></li>
    <li><a href="#encodeanalysis">Encode Analysis Results in MachineInstr IR</a></li>
    <li><a href="#codelayoutjit">Code Layout in the LLVM JIT</a></li>
    <li><a href="#fieldlayout">Improved Structure Splitting and Field Reordering</a></li>
    <li><a href="#slimmer">Finish the Slimmer Project</a></li>
  </ol></li>
</ul>

<div class="doc_author">
  <p>Written by the <a href="/">LLVM Team</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc25">Google Summer of Code 2025</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>
    Welcome prospective Google Summer of Code 2025 Students! This document is
    your starting point to finding interesting and important projects for LLVM,
    Clang, and other related sub-projects. This list of projects is not only
    developed for Google Summer of Code, but open projects that really need
    developers to work on and are very beneficial for the LLVM community.
  </p>

  <p>We encourage you to look through this list and see which projects excite
    you and match well with your skill set. We also invite proposals not on this
    list. More information and discussion about GSoC can be found in
    <a href="https://discourse.llvm.org/c/community/gsoc" target="_blank">
      discourse
    </a>. If you have questions about a particular project please find the
    relevant entry in discourse, check previous discussion and ask. If there is
    no such entry or you would like to propose an idea please create a new
    entry. Feedback from the community is a requirement for your proposal to be
    considered and hopefully accepted.
  </p>

  <p>The LLVM project has participated in Google Summer of Code for many years
    and has had some very successful projects. We hope that this year is no
    different and look forward to hearing your proposals. For information on how
    to submit a proposal, please visit the Google Summer of Code main
    <a href="https://summerofcode.withgoogle.com/">website.</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="rich-disassembler-for-lldb-2025">Rich Disassembler for LLDB</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>Use the variable location information from the debug info to annotate LLDB’s disassembler (and `register read`) output with the location and lifetime of source variables. The rich disassembler output should be exposed as structured data and made available through LLDB’s scripting API so more tooling could be built on top of this. In a terminal, LLDB should render the annotations as text.</p>

<p><b>Expected outcomes</b></p>

For example, we could augment the disassembly for the following function

<pre>
frame #0: 0x0000000100000f80 a.out`main(argc=1, argv=0x00007ff7bfeff1d8) at demo.c:4:10 [opt]
  1   void puts(const char*);
  2   int main(int argc, char **argv) {
  3    for (int i = 0; i < argc; ++i)
→ 4      puts(argv[i]);
  5    return 0;
  6   }
(lldb) disassemble
a.out`main:
...
  0x100000f71 <+17>: movl  %edi, %r14d
  0x100000f74 <+20>: xorl  %r15d, %r15d
  0x100000f77 <+23>: nopw  (%rax,%rax)
→  0x100000f80 <+32>: movq  (%rbx,%r15,8), %rdi
  0x100000f84 <+36>: callq 0x100000f9e ; symbol stub for: puts
  0x100000f89 <+41>: incq  %r15
  0x100000f8c <+44>: cmpq  %r15, %r14
  0x100000f8f <+47>: jne 0x100000f80 ; <+32> at demo.c:4:10
  0x100000f91 <+49>: addq  $0x8, %rsp
  0x100000f95 <+53>: popq  %rbx
...
</pre>

<p>using the debug information that LLDB also has access to (observe how the source variable i is in r15 from [0x100000f77+slide))</p>


<pre>
$ dwarfdump demo.dSYM --name  i
demo.dSYM/Contents/Resources/DWARF/demo: file format Mach-O 64-bit x86-64
0x00000076: DW_TAG_variable
 DW_AT_location (0x00000098:
 [0x0000000100000f60, 0x0000000100000f77): DW_OP_consts +0, DW_OP_stack_value
 [0x0000000100000f77, 0x0000000100000f91): DW_OP_reg15 R15)
 DW_AT_name ("i")
 DW_AT_decl_file ("/tmp/t.c")
 DW_AT_decl_line (3)
 DW_AT_type (0x000000b2 "int")
</pre>

to produce output like this, where we annotate when a variable is live and what its location is:

<pre>
(lldb) disassemble
a.out`main:
...                                                               ; i=0
  0x100000f74 <+20>: xorl  %r15d, %r15d                           ; i=r15
  0x100000f77 <+23>: nopw  (%rax,%rax)                            ; |
→  0x100000f80 <+32>: movq  (%rbx,%r15,8), %rdi                   ; |
  0x100000f84 <+36>: callq 0x100000f9e ; symbol stub for: puts    ; |
  0x100000f89 <+41>: incq  %r15                                   ; |
  0x100000f8c <+44>: cmpq  %r15, %r14                             ; |
  0x100000f8f <+47>: jne 0x100000f80 ; <+32> at t.c:4:10          ; |
  0x100000f91 <+49>: addq  $0x8, %rsp                             ; i=undef
  0x100000f95 <+53>: popq  %rbx
</pre>

<p>The goal would be to produce output like this for a subset of unambiguous cases, for example, variables that are constant or fully in registers.</p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Adrian Prantl aprantl@apple.com (primary contact)
  <li>Jonas Devlieghere jdevlieghere@apple.com
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Good understanding of C++
  <li>Familiarity with using a debugger on the terminal
  <li>Need to be familiar with all the concepts mentioned in the example above
  <li>Need to have a good understanding of at least one assembler dialect for machine code (x86_64 or AArch64).
</ul>

<p>Desired:</p>

<ul>
  <li>Compiler knowledge including data flow and control flow analysis is a plus.
  <li>Being able to navigate debug information (DWARF) is a plus.
</ul>

<p><b>Size of the project:</b></p>

<p>medium (~175h)</p>

<p><b>Project difficulty:</b></p>

<p>hard</p>

<p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/rich-disassembler-for-lldb/76952">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="bfloat16-libc">Bfloat16 in LLVM libc</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b></p>

  <p>
    <a href="https://en.wikipedia.org/wiki/Bfloat16_floating-point_format">Bfloat16</a>
    is a recently developed floating point format tailored to machine
    learning and AI, and in the latest <a href="https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/n4950.pdf">C++23 standard</a>,
    it is officially standardized as std::bfloat16_t. Its support could be found
    in many modern hardware, ranging from CPUs of all the major vendors
    including Intel, AMD, Apple, and Amazon, to GPUs (nVidia and AMD GPUs) and
    Google TPUs. On the software side, it is supported in all major accelerator
    libraries, such as CUDA, ROCm, oneAPI, PyTorch, and Tensorflow. The goal for
    this project is to implement bfloat16 math functions in the LLVM libc
    library.
  </p>

  <p><b>Expected result:</b></p>

  <ul>
    <li> Setup the generated headers properly so that the type and the functions
         can be used with various compilers (+versions) and architectures. </li>
    <li> Implement generic basic math operations supporting bfloat16 data types
         that work on supported architectures: x86_64, arm (32 + 64), risc-v
         (32 + 64), and GPUs. </li>
    <li> Implement specializations using compiler builtins or special hardware
         instructions to improve their performance whenever possible. </li>
    <li> If time permits, we can start investigating higher math functions for
         bfloat16. </li>
  </ul>

  <p><b>Skills:</b></p>

  <p>
    Basic C & C++ skills + Interest in knowing / learning more about the
    delicacy of floating point formats.
  </p>

  <p><b>Project size:</b> Large </p>

  <p><b>Difficulty:</b> Easy/Medium </p>

  <p><b>Confirmed Mentors:</b>
    <a href="mailto:lntue.h@gmail.com">Tue Ly</a>,
    <a href="mailto:its.overmighty@gmail.com">Nicolas Celik</a>,
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/libc-gsoc-2025-bfloat16-in-llvm-libc/84469">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="io_uring-libc">Direct I/O from the GPU with io_uring</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b></p>

  <p>
    Modern GPUs are capable of unified addressing with the host. We currently use
    this to provide I/O support using the <a href="https://libc.llvm.org/gpu/rpc.html">RPC interface</a>. However, this
    requires a dedicated user thread on the CPU to handle the server code. We want
    to explore alternatives to providing I/O from the GPU using the Linux 
    <a href="https://en.wikipedia.org/wiki/Io_uring">io_uring</a> interface.
  </p>

  <p>
    This interface is a ring buffer designed to accelerate syscalls. However, it
    provides a <a href="https://unixism.net/loti/tutorial/sq_poll.html#sq-poll">polling mode</a>
    that allows the kernel to flush the ring buffer without the user initiating a
    system call. We should be able to register <code>mmap()</code> memory with the
    GPU using <a href="https://github.com/ROCm/ROCR-Runtime/blob/amd-staging/runtime/hsa-runtime/inc/hsa_ext_amd.h#L3022">AMD</a>
    and <a href="https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__UNIFIED.html">NVIDIA</a>
    API calls. This interface should allow us to implement a rudimentary
    read/write interface which can be thought of as the same as the syscall on the
    CPU. That can then be used to implement a whole file interface.
  </p>

  <p><b>Expected result:</b></p>

  <ul>
    <li>An implementation of <code>pwrite</code> and <code>pread</code> that runs on the GPU.</li>
    <li>Support for <code>printf</code> by forwarding <code>snprintf</code> into <code>pwrite</code>.</li>
    <li>If time permits, exploring GPU file APIs.</li>
  </ul>

  <p><b>Skills:</b></p>

  <p>
    Basic C & C++ skills + access to a GPU, Linux kernel knowledge, GPU knowledge.
  </p>

  <p><b>Project size:</b> Small </p>

  <p><b>Difficulty:</b> Hard </p>

  <p><b>Confirmed Mentors:</b>
    <a href="mailto:joseph.huber@amd.com">Joseph Huber</a>, 
    <a href="mailto:i@tianshilei.me">Tian Shilei</a>
  </p>

  <p><b>Discourse:</b>
    <a href=https://discourse.llvm.org/t/libc-gsoc-2025-direct-i-o-from-the-gpu-with-io-uring/84569>URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="testing-math-libc">Profiling and testing the LLVM libc GPU math</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b></p>

  <p>
    The LLVM C library provides implementations of math functions. We want to profile these against existing implementations, 
    such as CUDA's <code>libdevice</code>, ROCm's <code>device libs</code>, and OpenCL's <code>libclc</code>. 
    Last year we worked on some interfaces to support these tests, now they need to be refined and filled out for the interesting functions.
  </p>

  <p>
    Additionally, we want to verify the accuracy of these functions when run on the GPU via brute force testing. 
    The goal is to verify that the implementations are correct and at least conformant to the error ranges in the 
    <a href="https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_C.html#relative-error-as-ulps">OpenCL standard</a>. 
    This will require a set of unit tests written in the <code>offload/</code> project, ideally using the new API that @callumfare is working on.
  </p>

  <p><b>Expected result:</b></p>

  <ul>
    <li>Final performance results similar to <a href="https://dl.acm.org/doi/fullHtml/10.1145/3624062.3624166">Old results</a> 
        but with the more optimized functions and higher accuracy.</li>
    <li>A test suite that can do brute force testing to confirm that the implementations are conformant.</li>
  </ul>

  <p><b>Skills:</b></p>

  <p>
    Basic C & C++ skills + access to a GPU, some math knowledge
  </p>

  <p><b>Project size:</b> Small </p>

  <p><b>Difficulty:</b> Easy / Medium </p>

  <p><b>Confirmed Mentors:</b>
    <a href="mailto:joseph.huber@amd.com">Joseph Huber</a>, 
    <a href="mailto:lntue.h@gmail.com">Tue Ly</a>
  </p>

  <p><b>Discourse:</b>
    <a href=https://discourse.llvm.org/t/libc-gsoc-2025-profiling-and-testing-of-the-llvm-libc-gpu-math/84570>URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clangir-codegen-tests">Validate existing Clang CodeGen test coverage with ClangIR</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    The <a href="https://clangir.org">ClangIR</a> (CIR) project aims to
    establish a new intermediate representation (IR) for Clang. Built on top of
    MLIR, it provides a dialect for C/C++ based languages in Clang, and the
    necessary infrastructure to emit it from the Clang AST, as well as a
    lowering path to the LLVM-IR dialect. ClangIR <a href="https://discourse.llvm.org/t/rfc-upstreaming-clangir/76587">upstreaming</a>
    is currently in progress.
  </p><p>
    In order to give community more frequent updates it'd be great if we can
    report ClangIR progress by measuring the coverage of existing Clang's
    CodeGen tests in face of a ClangIR enabled pipeline. By collecting
    information on crashing, passing or failing tests we can come up with a
    metric that is easier to report and understand, provide entry points for
    newcomers looking for tasks and help the project by classifying existing issues.
    
    Existing Clang CodeGen tests live in clang/test/CodeGen* and can be found
    in different states regarding ClangIR support:
    <ul>
      <li><b>FileCheck fails</b>. LLVM IR builds but FileCheck fails to match output
      <ul>
        <li> LLVM IR differs because ClangIR pipeline is emitting different IR
        (e.g. different instructions are used, missing attributes). Issues
        need to be created and ClangIR needs to be fixed. </li>
        <li> LLVM IR differs because CHECK lines need be made more flexible (LLVM-IR
        dialect output is different, SSA value names, order of attributes, etc).
        It's possible a tool like llvm-canon might be of good use here. </li>
      </ul></li>
      <li><b>Test crash / error</b>. ClangIR doesn't support some C/C++ construct or LLVM lowering
      hasn't been implemented.</li>
      <li><b>Test pass</b>. Yay!</li>
    </ul>

  <p>
    In order to retrieve the information above, the student needs to make
    changes to Clang's testing infra (LIT configs, scripts, tests, ???)
    such that it's easier to replay the same invocations with ClangIR enabled,
    compare against traditional pipeline result or retrieve special directives
    from tests.
  </p>

  <p>
    It's not clear what is the best methodology just yet, but it's
    expected that submitted proposals that want to be taken seriously should
    present few possible ideas on how to achieve this, prior discussion with
    other members of the community is encouraged. The student is also expected
    to interact with the ClangIR community, file github issues, investigate
    and/or make changes to failing codegen tests.
  </p>

  <p><b>Expected result:</b>
    <ol>
    <li>Build the infrastructure to run tests and collect results.</li>
    <li>Present the results in a way that can be placed on a webpage.</li>
    <li>File issues or change check lines for 50% of the "FileCheck fails" category
    above. The only subdirectories that need consideration for the moment are:
    <pre>
    clang/test/CodeGen
    clang/test/CodeGenCXX
    clang/test/CodeGenOpenCL
    clang/test/CodeGenCUDA
    </pre>
    </li>
    </ol>
    <ul>
    <li> Bonus point: find ways to automate/facilitate changes to tests, put PRs to
    fix problems in ClangIR.</li>
    </ul>
  </p>

  <p><b>Skills:</b>
    Python, intermediate C++ programming skills and familiarity with basic compiler
    design concepts are required. Prior experience with LLVM IR, MLIR, Clang or
    ClangIR programming is a big plus, but willingness to learn is also a
    possibility.
  </p>

  <p><b>Project size:</b> Large</p>

  <p><b>Difficulty:</b>Medium</p>

  <p><b>Potential Mentors:</b>
    <a href="https://github.com/bcardosolopes">Bruno Cardoso Lopes</a>
    <a href="https://github.com/andykaylor">Andy Kaylor</a>
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/clangir-gsoc2025-validate-existing-clang-codegen-test-coverage-with-clangir/84481">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clangir-upstreaming">Participate in ClangIR Upstreaming</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    <a href="https://clangir.org">ClangIR</a> is a new, MLIR_based intermediate
    representation of C and C++ code. It has been developed in an LLVM incubator
    project, but work is now underway to migrate the code from the incubator to
    the main LLVM repository. As the code is moved, it must be updated to align
    with LLVM coding standards and quality expectations. The goal for this
    project is to participate in the ClangIR upstreaming process and help
    improve both the code and the upstreaming process.
  </p><p>
    The ClangIR project intends to unlock the possibility of better
    optimization, analysis, and diagnostics for C and C++ code by adding new
    abstractions that more closely model the source constructs, preserving
    more details than are available in standard LLVM IR. The ClangIR dialect
    is already being used to solve real-world problems using the implementation
    available in the ClangIR incubator, but we need to move this into the
    main LLVM repository in order to make this functionality available to a
    larger audience.
  </p><p>
    This project will be an opportunity to gain hands-on experience with MLIR
    development with a focus on day-to-day software engineering discipline.
    Participants will work side-by-side with other LLVM contributors to achieve
    a common goal, and in the process will gain a deep understanding of the
    ClangIR dialect.
  </p>

  <p><b>Expected result:</b>
    <ol>
    <li>Migrate ClangIR support for C and C++ language features into the main
        LLVM repository</li>
    <li>Improve the quality of code as it is being migrated</li>
    <li>Suggest ways to improve the migration process</li>
    </ol>
  </p>

  <p><b>Skills:</b>
    Proficiency with modern C++ programming and familiarity with basic compiler
    design concepts are required. Prior experience with LLVM IR, MLIR, Clang or
    ClangIR programming is a big plus, but since the goal of this project is to
    gain such experience, it is not a prerequisite.
  </p>

  <p><b>Project size:</b> Medium to Large</p>

  <p><b>Difficulty:</b>Medium</p>

  <p><b>Potential Mentors:</b>
    <a href="https://github.com/andykaylor">Andy Kaylor</a>
    <a href="https://github.com/bcardosolopes">Bruno Cardoso Lopes</a>
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/clangir-gsoc-2025-clangir-upstreaming/84766">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="csa-lifetime-annotations">Teach the Clang Static Analyzer to understand lifetime annotations</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    The <a href="https://clang.llvm.org/docs/ClangStaticAnalyzer.html">Clang Static Analyzer</a> (CSA) 
    can already find a wide range of temporal memory errors. These checks often have hardcoded
    knowledge about the behavior of some APIs. For example, the <code>cplusplus.InnerPointer</code> checker knows
    the semantics of <code>std::string::data</code>. The Clang community introduced some lifetime
    annotations including <code>[[clang::lifetimebound]]</code> and <code>[[clang::lifetime_capture_by(X)]]</code>
    and made <a href="https://discourse.llvm.org/t/lifetime-analysis-improvements-in-clang/81374">many improvements</a>
    to Clang's default warnings. Unfortunately, the compiler's warnings only do statement local analysis.
    The CSA is capable of advanced inter-procedural analysis. Generalizing the existing checks like 
    <code>cplusplus.InnerPointer</code> could enable the analyzer to find even more errors in annotated code.
    This can become even more impactful once the standard library
    <a href="https://github.com/llvm/llvm-project/pull/112751">gets annotated</a>.
  </p>

  <p><b>Expected result:</b>
    <ol>
    <li>Identify the checks that can benefit from the <code>[[clang::lifetimebound]]</code> and <code>[[clang::lifetime_capture_by(X)]]</code> annotations.</li>
    <li>Extend those checks to support these annotations.</li>
    <li>Make sure the generated bug reports are high quality, the diagnostics properly explain how the analyzer took these annotations into account.</li>
    <li>Validate the results on real world projects.</li>
    <li>Potentially warn about faulty annotations (stretch goal).</li>
    </ol>
  </p>

  <p><b>Skills:</b>
    Intermediate C++ programming skills and familiarity with basic compiler
    design concepts are required. Prior experience with Clang or
    CSA programming is a big plus, but willingness to learn is also a
    possibility.
  </p>

  <p><b>Project size:</b> Large</p>

  <p><b>Difficulty:</b>Hard</p>

  <p><b>Potential Mentors:</b>
    <a href="https://github.com/Xazax-hun">Gabor Horvath</a>
    <a href="https://github.com/steakhal">Balazs Benics</a>
    <a href="https://github.com/isuckatcs">Daniel Domjan</a>
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/clang-static-analyzer-gsoc-2025-teach-the-clang-static-analyzer-to-understand-lifetime-annotations/84487">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-driver-modules">Simple C++20 modules without a build system</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description</b></p>
  
  <p>Currently there is no easy way to take a collection of source files using C++20 modules and build an executable from them. This makes it hard to create simple tests or tiny programs using C++20 modules without first setting up a build system. This project's goal is to extend the extremely simple build system in Clang's driver to handle these cases.</p>
  <p>This can be done by using Clang's existing support for scanning for C++20 modules to discover the dependencies between the source files that have been passed in, and then build them in that order, passing in the right PCM files where needed. This may also be extended to support explicitly building Clang modules discovered via module map files too.</p>

  <p><b>Expected outcomes</b></p>

  <p>Invoking clang similarly to <pre>clang -o program -std=c++20 main.cpp A.cppm B.cppm</pre> should compile successfully where each translation-unit only imports modules defined in other source files on the command line, or the standard library. This should add no overhead to cases where modules are not used.</p>

  <p><b>Confirmed mentors and their contacts</p></b>

  <ul>
    <li><a href="https://github.com/Bigcheese">Michael Spencer</a></li>
  </ul>
  
  <p><b>Required skills</b></p>

  <p>Intermediate knowledge of C++; familiarity with how C++ code is built. Familiarity with C++20 modules is an asset, but not required.</p>
    
  <p><b>Size of the project:</b></p>
  
  <p>medium (~175h)</p>
  
  <p><b>Project difficulty:</b></p>
  
  <p>medium</p>
  
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-gsoc-2025-support-simple-c-20-modules-use-from-the-clang-driver-without-a-build-system/84511">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-improve-trapping-ubsan-2025">Usability Improvements for trapping Undefined Behavior Sanitizer (UBSan)</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description</b></p>
  <p>Undefined Behavior Sanitizer (UBSan) is a useful compilation mode in Clang for finding uses of undefined behavior (e.g. signed integer overflow) and problematic C/C++ code (e.g. unsigned integer overflow). The default version of UBSan uses a compiler runtime that only works in userspace (e.g. it won’t work in the kernel or for embedded applications) and is not considered secure enough for use in production environments.  To handle these other environments UBSan provides a trapping mode that emits trap instructions that immediately halts the application rather than calling into the UBSan runtime which normally diagnoses the problem and then carries on execution.</p>

  <p>Unfortunately trapping UBSan has some deficiencies which make it hard to use. In particular:</p>
  <ul>
    <li>Clang silently ignores the <tt>-fsanitize-trap=undefined</tt> flag when it's passed without <tt>-fsanitize=undefined</tt>. This project would fix this as a “warm up task” to get familiar with the Clang codebase.</li>
    <li>When a UBSan trap is hit with the debugger attached it is not convenient to figure out the reason UBSan trapped. For x86_64 and arm64 some information is encoded in the instruction but decoding this is very inconvenient. While LLDB could be taught to look at the instruction and decode the meaning this is brittle because it depends on undocumented compiler ABI. Instead we can build upon the <tt>__builtin_verbose_trap</tt> work to encode the reason for trapping inside the debug information.</li>
  </ul>

  <p><b>Expected outcomes</b></p>
  <ul>
    <li>When the <tt>-fsanitize-trap=undefined</tt> flag is passed on its own the compiler silently ignores it. Currently Clang requires that the <tt>-fsanitize-trap=</tt> flag is also passed. Clang should be taught to warn about this.</li>
    <li>Teach Clang to emit the UBSan trap reasons in debug information on UBSan trap instructions similar to how <tt>__builtin_verbose_trap</tt> works.</li>
    <li>Confirm LLDB is able to recognize the UBSan trap reasons and add tests for this.</li>
  </ul>


  <p><b>Confirmed mentors and their contacts</p></b>

  <ul>
    <li><a href="https://github.com/delcypher">Dan Liew</a></li>
    <li><a href="https://github.com/Michael137">Michael Buch</a></li>
  </ul>

  <p><b>Required skills</b></p>

  <p>Good understanding of C++</p>

  <p><b>Desirable skills</b></p>

  <ul>
    <li>Familiarity with UBSan</li>
    <li>Familiarity with LLDB</li>
  </ul>

  <p><b>Size of the project:</b></p>

  <p>small (~10h). but can be extended if time allows</p>

  <p><b>Project difficulty:</b></p>

  <p>Easy. This project would be good to a beginner to LLVM. If we end up having lots of extra time we can investigate improving the UBSan reasons by encoding other useful information that the compiler knows into the trap reason.</p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-gsoc-2025-usability-improvements-for-trapping-undefined-behavior-sanitizer/84568">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="improve-documentation-parsing-in-clang">Improve documentation parsing in Clang</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b>
    <a href=https://clang.llvm.org/extra/clang-doc.html>Clang-Doc</a> is a
    C/C++ documentation generation tool created as an alternative for Doxygen
    and built on top of LibTooling. This effort started in 2018 and critical
    mass has landed in 2019, but the development has been largely stagnant
    mostly due to a lack of resources until last year when the development
    restarted as a successful Google Summer of Code project.
    <br /> <br />
    The tool is built on top of LibTooling and leverages Clang parsers which
    supports parsing of Doxygen commands in documentation comments (this
    support is also used in the implementation of Clang’s
    <pre>-Wdocumentation</pre> which can be used to validate the content of
    documentation comments during
    compilation).
    <br /> <br />
    Unfortunately, Clang’s documentation parser is incomplete and has several issues:
    <ul>
      <li>Not all Doxygen commands are supported, limiting the Clang-Doc’s usability.
      <li>Not all C/C++ constructs are currently handled, most notably C++20 features such as concepts.
      <li>Markdown support in documentation comments introduced in Doxygen version 1.8.0 is missing.
    </ul>
  </p>

  <p><b>Expected result:</b>
    The goal of this project is to implement the missing features in Clang’s
    documentation parser as well as their handling in Clang-Doc to improve the
    quality of the generated documentation. The eventual goal is for the LLVM
    project to start using Clang-Doc for generating its reference
    documentation, but before we can do that we need to ensure that all
    required features are implemented.
    <br /><br />
    Successful proposals should focus not only on addressing the existing
    limitations, but also draw inspiration for other potential improvements
    from other documentation tools such as <a href=https://hdoc.io/>hdoc</a>, <a
    href=https://github.com/standardese/standardese>standardese</a>, <a
    href=https://github.com/chromium/subspace/tree/main/subdoc>subdoc</a> or
    <a href=https://cs.opensource.google/fuchsia/fuchsia/+/main:tools/cppdocgen/>cppdocgen</a>.
    <br /><br />
    Over the course of the project, the candidate will have an opportunity to
    gain significant experience with LLVM and Clang internals (including lexer
    and parser) and C/C++ language.
  </p>

  <p><b>Skills:</b>
    Intermediate knowledge of C++; interest in compilers and parsers. Previous
    experience with Clang/LibTooling is a bonus but not required.
  </p>

  <p><b>Project size:</b> Either medium or large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/petrhosek>Petr Hosek</a>,
    <a href=https://github.com/ilovepi>Paul Kirth</a>
  </p>

  <p><b>Discourse:</b> <a href=https://discourse.llvm.org/t/improve-documentation-parsing-in-clang/84513>URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="improve-symbol-discovery-clang-repl">Advanced symbol resolution and reoptimization for clang-repl</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b>
    The Clang compiler is part of the LLVM compiler infrastructure and supports
    various languages such as C, C++, ObjC and ObjC++. The design of LLVM and
    Clang enables them to be used as libraries, and has led to the creation of
    an entire compiler-assisted ecosystem of tools. The relatively friendly
    codebase of Clang and advancements in the JIT infrastructure in LLVM further
    enable research into different methods for processing C++ by blurring the
    boundary between compile time and runtime. Challenges include incremental
    compilation and fitting compile/link time optimizations into a more dynamic
    environment. Incremental compilation pipelines process code chunk-by-chunk
    by building an ever-growing translation unit. Code is then lowered into the
    LLVM IR and subsequently run by the LLVM JIT. Such a pipeline allows
    creation of efficient interpreters. The interpreter enables interactive
    exploration and makes the C++ language more user friendly. Clang-Repl is one
    example.
  </p>

  <p><b>Expected result:</b>
    The project aims to develop a robust mechanism for resolving missing symbols
    by dynamically identifying and loading the appropriate shared objects or
    static archives. Additionally, it will explore use cases where adapting
    symbols based on execution profiles leads to measurable performance
    improvements, optimizing the efficiency of just-in-time compilation and
    dynamic execution environments.
  </p>

  <p><b>Skills:</b>
    Intermediate knowledge of C++, Understanding of LLVM and the LLVM JIT in particular
  </p>

  <p><b>Project size:</b>Either medium or large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>
  </p>

  <p><b>Discourse:</b> <a https://discourse.llvm.org/t/gsoc2025-advanced-symbol-resolution-and-reoptimization-for-clang-repl>URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="rust-enzyme-improvements">Improve Enzyme reliability and compile times for Rust</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description</b></p>
 
  <p>Enzyme requires good information about the memory layout of types. LLVM-IR is intentionally opaque, e.g. `&f32` and `&f64` both have the LLVM-IR type `ptr`.
  Enzyme is generally able to infer the underlying type (e.g. f32 vs f64) through usage analysis, but that process is slow and can in some cases fail. 
  To make autodiff more robust, we should lower either MIR or THIR type information into LLVM-IR metadata. This analysis is recursive, for example `&[T]` is a fat pointer
  and therefore will be represented as a (ptr, int) pair in LLVM-IR. In this case the algorithm should recursively also analyze `T` and generate metadata for it.</p>

  <p>The function <a href="https://github.com/rust-lang/rust/blob/d4bdd1ed551fed0c951eb47b4be2c79d7a02d181/compiler/rustc_monomorphize/src/partitioning/autodiff.rs#L30">here</a> can be extended to generate metadata from rusts Mid-level IR (MIR). A prototype of the parser was implemented <a href="https://github.com/EnzymeAD/rust/blob/58fee1abf3f2cd0e73ee8b98e53869d6fc3ba604/compiler/rustc_middle/src/ty/mod.rs#L2826">here</a> and can be used for inspiration. Various LLVM-IR examples for the metadata which we want to generate can be found in <a href="https://github.com/EnzymeAD/Enzyme/blob/main/enzyme/test/TypeAnalysis">this</a> test folder. Look for annotations in the style of ` {[-1]:Pointer, [-1,0]:Float@float}`.</p>

  <p>The online compiler <a href="https://enzyme.mit.edu/explorer/">Explorer</a> fork can be used to trigger related bugs, starting with "can not deduce type of X".</p>

  <p><b>Expected outcomes</b></p>

  <p>The participant should find and select some interesting testcases, in which Enzyme either fails to differentiate an example due to inssuficient Type Information, or takes unreasonable long times (e.g. > 20x slower than compiling the code without autodiff). In the second case, a profiler should be used to verify that Enzyme causes a long compile time due to type analysis. The participant should then write (or later extend) the Type parser to generate the correct metadata, such that Enzyme can handle the new testcases. The LIT testcases should be added to the rust compiler, to avoid further regressions.</p>

  <p>Examples for code that currently is not handled correctly can be discussed in the project proposal phase.</p>

  <p><b>Confirmed mentors and their contacts</p></b>

  <ul>
    <li><a href="https://github.com/ZuseZ4">Manuel Drehwald</a></li>
    <li><a href="https://github.com/oli-obk">Oli Scherer</a></li>
    <li><a href="https://github.com/jdoerfert">Johannes Doerfert</a></li>
  </ul>
  
  <p><b>Required skills</b></p>

  <p>Intermediate knowledge of Rust and C++; Familiarity with profilers or LLVM metadata is an asset, but not required.</p>
    
  <p><b>Size of the project:</b></p>
  
  <p>medium (~175h)</p>
  
  <p><b>Project difficulty:</b></p>
  
  <p>medium</p>
  
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/gsoc2025-improve-rust-enzyme-reliability-and-compile-times/84523">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm-abi-lowering">Introduce an ABI lowering library</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    Currently, every LLVM-based frontend that wants to support calling into C
    code (FFI) needs to re-implement a substantial amount of complex call ABI
    handling. The goal of this project is to introduce an LLVM ABI library,
    which can be reused across different frontends, including Clang. More
    details on the motivation and a broad outline of the design are available
    in the <a href="https://discourse.llvm.org/t/rfc-an-abi-lowering-library-for-llvm/84495/1">corresponding RFC</a>.
  </p>

  <p>
    The initial phase of the project will be to implement a prototype that can
    handle at least the x86_64 System V ABI. This will involve implementing the
    ABI type system, mapping of Clang types to ABI types and moving at least
    part of the X86 ABIInfo implementation from Clang to the new ABI library.
    This is to demonstrate general feasibility, figure out design questions and
    analyze compilation-time impact.
  </p>

  <p>
    Assuming the results from the prototype are positive, the next step would
    be to upstream the implementation by splitting it into smaller PRs. Finally,
    the implementation can be expanded to cover additional targets, ultimately
    removing Clang's ABI handling code entirely.
  </p>

  <p><b>Expected result:</b>
    The minimum result is a prototype for the x86_64 ABI. The maximum result is
    fully upstreamed support for all targets. The expected result is somewhere
    in the middle between those two.
  </p>

  <p><b>Skills:</b>
    Intermediate C++. Some familiarity with LLVM is a plus, but not required.
  </p>

  <p><b>Project size:</b> Large</p>

  <p><b>Difficulty:</b> Hard</p>

  <p><b>Confirmed mentors:</b>
    <a href="https://github.com/nikic">Nikita Popov</a>
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/llvm-introduce-an-abi-lowering-library/84554">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="byte-type">Byte type</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    LLVM IR can't represent implementations of memcpy, memcmp, etc correctly
    due to the lack of a way to represent raw memory. This project aims to
    add a new 'byte' type to the LLVM IR to represent raw memory.
  </p>

  <p>
    In addition to adding the new type, the project involves changing clang
    to lower chars to the new b8 type instead of i8, fixing incorrect lowerings
    of memory intrinsics, and tracking down the performance regressions.
  </p>

  <p>
    There is already a
    <a href="https://github.com/georgemitenkov/GSoC-2021">prototype</a>
    implementation of the byte type for an older version of LLVM.
    More information
    <a href="https://gist.github.com/georgemitenkov/3def898b8845c2cc161bd216cbbdb81f">here</a>.
  </p>

  <p><b>Expected result:</b>
    The minimum result is a port of the existing prototype to the current LLVM,
    fixing all known incorrect optimizations, add support for the byte type to
    Alive2, and a performance analysis.
  </p>

  <p><b>Skills:</b>
    Intermediate C++, familiarity with LLVM, profiling.
  </p>

  <p><b>Project size:</b> Large</p>

  <p><b>Difficulty:</b> Hard</p>

  <p><b>Confirmed mentors:</b>
    <a href="https://web.ist.utl.pt/nuno.lopes/">Nuno Lopes</a>
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/gsoc-2025-byte-type/84636">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="offload-remarks-visualization">LLVM Compiler Remarks Visualization Tool for Offload Proposal</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    LLVM offers different information via remarks, profiling, or runtime debug annotations 
    (e.g., LIBOMPTARGET_INFO=-1). However, as projects increase, dissecting this information 
    becomes difficult for users. For example, it is difficult to collect all this data, 
    visualize it, and learn from it when building large projects.
  </p>
  
  <p>
    Currently, some of this information—for example, compilation remarks—can be exported 
    in JSON format. We want to create a tool to visualize, aggregate, and summarize the 
    information. To aid accelerator development, we will start with the offload project 
    as the primary candidate.
  </p>

  <p>
    Similar tools, such as opt-viewer, can be used as references and starting points.
  </p>

  <p><b>Expected outcomes:</b>
    The expected outcome is a tool (e.g., in the form of a compiler wrapper such as 
    ccache) that will allow the dump of all the compiler-generated information in 
    JSON format and organize it in a project structure.
  </p>

  <p>
    The tool should generate an HTML-based report to help visualize the remarks. We
    envision a small client-server application using Python to spawn a local server 
    as the visualization's front end. The server will expose the different reports 
    and perform early analysis and aggregation.
  </p>

  <p>
    Additionally, the tool should be designed so that, in the future, the analysis of 
    the remarks can provide generalized guidelines for the developer (e.g., show the 
    most common remark, use LLM models to explain actions, etc.). The client (HTML viewer) 
    will display the aggregated data, in-line remarks, profile information, etc. We do 
    not expect the project to have all the features at the end of the GSoC but to 
    serve as a placeholder for growth in the future.
  </p>

  <p>
    In particular, the outcomes of this project should be:
    <ul>
      <li>
        Together with the mentors, help the design of the compiler wrapper, 
        data storage layer, and client/server infrastructure. This includes the 
        server API. The outcome of this task is a design document (similar to an RFC).
      </li>
      <li>
        Create a compiler wrapper that will dump different information in JSON 
        format into the data storage layer (e.g., folders).
      </li>
      <li>
        Create a simple server layer that exposes the backend API to the front 
        end. Python is the right way to do this, but we welcome other suggestions 
        that align with the LLVM project. We would like to avoid relying on 
        external projects (e.g., Flask) to avoid adding more dependencies to 
        the LLVM project.
      </li>
      <li>
        Create a simple client-side visualization tool that can be extended in 
        the future to show more reports.
      </li>
    </ul>
  </p>

  <p><b>Mentors:</b>
    <a href="https://github.com/shiltian">@shiltian</a>, 
    <a href="https://github.com/jdoerfert">@jdoerfert</a>, 
    <a href="https://github.com/josemonsalve2">@josemonsalve2</a>
  </p>

  <p><b>Required/desired skills:</b>
    <ul>
      <li>
        Basic understanding of the LLVM Compiler to be able to generate 
        compiler remarks, profiling data, and other information from the 
        compiler.
      </li>
      <li>
        Proficiency in Python and C++.
      </li>
      <li>
        Full-stack web development.
      </li>
    </ul>
  </p>

  <p><b>Project size:</b> Large</p>

  <p><b>Difficulty:</b> Easy</p>

  <p><b>Discourse Link:</b> 
    <a href="https://discourse.llvm.org/t/gsoc-offload-llvm-compiler-remarks-visualization-tool-for-offload-proposal/84596">[GSoC][Offload]LLVM Compiler Remarks Visualization Tool for Offload Proposal</a>
  </p>
</div>

    
<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc24">Google Summer of Code 2024</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>
    Google Summer of Code 2024 was yet another successful one for LLVM project. For the
    list of accepted and completed projects, please take a look into Google
    Summer of
    Code <a href="https://summerofcode.withgoogle.com/archive/2024/organizations/llvm-compiler-infrastructure">website</a>.
  </p>
</div>

<div class="www_text">
  <p>
    Welcome prospective Google Summer of Code 2024 Students! This document is
    your starting point to finding interesting and important projects for LLVM,
    Clang, and other related sub-projects. This list of projects is not only
    developed for Google Summer of Code, but open projects that really need
    developers to work on and are very beneficial for the LLVM community.
  </p>

  <p>We encourage you to look through this list and see which projects excite
    you and match well with your skill set. We also invite proposals not on this
    list. More information and discussion about GSoC can be found in
    <a href="https://discourse.llvm.org/c/community/gsoc" target="_blank">
      discourse
    </a>. If you have questions about a particular project please find the
    relevant entry in discourse, check previous discussion and ask. If there is
    no such entry or you would like to propose an idea please create a new
    entry. Feedback from the community is a requirement for your proposal to be
    considered and hopefully accepted.
  </p>

  <p>The LLVM project has participated in Google Summer of Code for several years
    and has had some very successful projects. We hope that this year is no
    different and look forward to hearing your proposals. For information on how
    to submit a proposal, please visit the Google Summer of Code main
    <a href="https://summerofcode.withgoogle.com/">website.</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="remove_ub_tests">Remove undefined behavior from tests</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Many of LLVM's unit tests have been reduced automatically from larger tests.
    Previous-generation reduction tools used undef and poison as placeholders
    everywhere, as well as introduced undefined behavior (UB).
    Tests with UB are not desirable because 1) they are fragile since in the
    future the compiler may start optimizing more aggressively and break the
    test, and 2) it breaks translation validation tools such as
    <a href="https://github.com/AliveToolkit/alive2/">Alive2</a> (since it's
    correct to translate a fuction that is always UB into anything).

    <br />
    The major steps include:
    <ol>
      <li>Replace known patterns such as branch on undef/poison, memory accesses
          with invalid pointers, etc with non-UB patterns.</li>
      <li>Use Alive2 to detect further patterns (by searching for tests that are
          always UB).</li>
      <li>Report any LLVM bug found by Alive2 that is exposed when removing
          UB.</li>
    </ol>
  </p>

  <p><b>Expected result:</b>
    The majority of LLVM's unit tests will be free of UB.</p>

  <p><b>Skills:</b>
    Experience with scripting (Python or PHP) is required.
    Experience with regular expressions is encouraged.
  </p>

  <p><b>Project size:</b> Either medium or large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b> <a href="https://web.ist.utl.pt/nuno.lopes/">Nuno Lopes</a></p>
  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/gsoc-2004-remove-undefined-behavior-from-tests/77236">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="spirv_tablegen">Automatically generate TableGen file for SPIR-V instruction set</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The existing file that describes the SPIR-V instruction set in LLVM was
    manually created and is not always complete or up to date. Whenever new
    instructions need to be added to the SPIR-V backend, the file must be
    amended. In addition, since it is not created in a systematic way, there are
    often slight discrepancies between how an instruction is described in the
    SPIR-V spec and how it is declared in the TableGen file. Since SPIR-V
    backend developers often use the spec as a reference when developing new
    features, having a consistent mapping between the specification and TableGen
    records will ease development. This project proposes creating a script
    capable of generating a complete TableGen file that describes the SPIR-V
    instruction set given the JSON grammar available in the
    KhronosGroup/SPIRV-Headers repository, and updating SPIR-V backend code to
    use the new definitions. The specific method used for translating the JSON
    grammar to TableGen is left up to the discretion of the applicant, however,
    it should be checked into the LLVM repository with well-documented
    instructions to replicate the translation process so that future maintainers
    will be able to regenerate the file when the grammar changes. Note that the
    grammar itself should remain out-of-tree in its existing separate
    repository.
  </p>

  <p><b>Expected result:</b>
    <ul>
      <li>The SPIR-V instruction set's definition in TableGen is replaced with
      one that is autogenerated.</li>
      <li>A script and documentation are written that support regenerating the
      definitions as needed given the JSON grammar of the SPIR-V instruction
      set.</li>
      <li>Usage of the SPIR-V instruction set in the SPIR-V backend updated to
      use the new autogenerated definitions.</li>
    </ul>
  </p>

  <p><b>Skills:</b>
    Experience with scripting and an intermediate knowledge of C++. Previous
    experience with LLVM/TableGen is a bonus but not required.
  </p>

  <p><b>Project size:</b> Medium (175 hour)</p>

  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/sudonatalie/">Natalie Chouinard</a>,
    <a href="https://github.com/keenuts/">Nathan Gauër</a></p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/clang-automatically-generate-tablegen-file-for-spir-v-instruction-set/76369">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="bitstream_cas">LLVM bitstream integration with CAS (content-addressable storage)</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The LLVM bitstream file format is used for serialization of intermediate
    compiler artifacts, such as LLVM IR or Clang modules. There are situations
    where multiple bitstream files store identical information, and this
    duplication leads to increased storage requirements.
    <br><br>
    This project aims to integrate the LLVM CAS library into the LLVM bitstream
    file format. If we factor out the frequently duplicated part of a bitstream
    file into a separate CAS object, we can replace all copies with a small
    reference to the canonical CAS object, saving storage.
    <br><br>
    The primary motivating use-case for this project is the dependency scanner
    that's powering "implicitly-discovered, explicitly-built" Clang modules.
    There are real-world situations where even coarse de-duplication on the
    block level could halve the size of the scanning module cache.
  </p>

  <p><b>Expected result:</b>
    There's a way to configure the LLVM bitstream writer/reader to use CAS as
    the backing storage.
  </p>

  <p><b>Skills:</b>
    Intermediate knowledge of C++, some familiarity with data serialization, self-motivation.
  </p>

  <p><b>Project size:</b> Medium or large</p>

  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/jansvoboda11/">Jan Svoboda</a>,
    <a href="https://github.com/cachemeifyoucan/">Steven Wu</a></p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/llvm-bitstream-integration-with-cas-content-addressable-storage/76757">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="three_way_comparison">Add 3-way comparison intrinsics</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <a href="https://en.wikipedia.org/wiki/Three-way_comparison">3-way comparisons</a>
    return the values -1, 0 or 1 depending on whether the values compare lower,
    equal or greater. They are exposed in C++ via the spaceship operator
    (operator&lt;=&gt;) and in Rust via the PartialOrd and Ord traits.
    Currently, such comparisons produce sub-optimal codegen and optimization
    results in some cases.
    <br/><br/>
    The goal of this project is to resolve these optimization issues by
    implementing new 3-way comparison intrinsics, as described in
    <a href="https://discourse.llvm.org/t/rfc-add-3-way-comparison-intrinsics/76685">[RFC] Add 3-way comparison intrinsics</a>.
    The implementation steps are broadly:
    <ol>
      <li>Add the intrinsics to LLVM IR.</li>
      <li>Implement legalization/expansion support in SelectionDAG and
          GlobalISel.</li>
      <li>Implement optimization support in ConstantFolding, InstSimplify,
          InstCombine, CorrelatedValuePropagation, IndVarSimplify,
          ConstraintElimination, IPSCCP, and other relevant transforms.</li>
      <li> Make use of the intrinsics via InstCombine canonicalization or
          direct emission in clang/rustc.</li>
    </ol>
    Adding new target-independent intrinsics is a good way of becoming familiar with a broad slice of LLVM!
  </p>

  <p><b>Expected result:</b>
    Support for the intrinsics in the backend and the most important
    optimization passes. Ideally full integration starting at the frontend.
  </p>

  <p><b>Skills:</b> Intermediate knowledge of C++ </p>

  <p><b>Project size:</b> Medium or large</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/nikic">Nikita Popov</a>,
    <a href="https://github.com/dc03">Dhruv Chawla</a></p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/llvm-add-3-way-comparison-intrinsics/76807">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_www">Improve the LLVM.org Website Look and Feel</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The llvm.org website serves as the central hub for information about the
    LLVM project, encompassing project details, current events, and relevant
    resources. Over time, the website has evolved organically, prompting the
    need for a redesign to enhance its modernity, structure, and ease of
    maintenance.
    <br/><br/>
    The goal of this project is to create a contemporary and coherent static
    website that reflects the essence of LLVM.org. This redesign aims to improve
    navigation, taxonomy, content discoverability, mobile device support,
    accessibility, and overall usability. Given
    the critical role of the website in the community, efforts will be made to
    engage with community members, seeking consensus on the proposed changes.
  </p>

  <p><b>Expected result:</b>
    A modern, coherent-looking website that attracts new prospect users and
    empowers the existing community with better navigation, taxonomy, content
    discoverability, and overall usability. Since the website is a critical
    infrastructure and most of the community will have an opinion this project
    should try to engage with the community building community consensus on the
    steps being taken. Suggested approach:
    <ul>
      <li>Conduct a comprehensive content audit of the existing website.</li>
      <li>Select appropriate technologies, preferably static site generators
        like Hugo or Jekyll.</li>
      <li>Advocate for a separation of data and visualization, utilizing formats
        such as YAML and Markdown to facilitate content management without
        direct HTML coding.</li>
      <li>Present three design mockups for the new website, fostering open
        discussions and allowing time for alternative proposals from interested
        parties.</li>
      <li>Implement the chosen design, incorporating valuable feedback from the
        community.</li>
      <li>Collaborate with content creators to integrate or update content as
        needed.</li>
    </ul>
    The successful candidate should commit to regular participation in weekly
    meetings, deliver presentations, and contribute blog posts as requested.
    Additionally, they should demonstrate the ability to navigate the community
    process with patience and understanding.
  </p>

  <p><b>Skills:</b>
    Knowledge in the area of web development with static site generators.
    Knowledge in html, css, bootstrap, and markdown. Patience and self-motivation.
  </p>

  <p><b>Difficulty:</b> Hard</p>

  <p><b>Project size:</b> Large</p>

  <p><b>Confirmed Mentors:</b>
    <a href=https://github.com/tlattner>Tanya Lattner</a>,
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>
  </p>
  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/improve-the-llvm-org-website-look-and-feel/76864">URL</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-repl-out-of-process">Out-of-process execution for clang-repl</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The Clang compiler is part of the LLVM compiler infrastructure and supports
    various languages such as C, C++, ObjC and ObjC++. The design of LLVM and
    Clang enables them to be used as libraries, and has led to the creation of
    an entire compiler-assisted ecosystem of tools. The relatively friendly
    codebase of Clang and advancements in the JIT infrastructure in LLVM further
    enable research into different methods for processing C++ by blurring the
    boundary between compile time and runtime. Challenges include incremental
    compilation and fitting compile/link time optimizations into a more dynamic
    environment.
    <br /> <br />
    Incremental compilation pipelines process code chunk-by-chunk by building an
    ever-growing translation unit. Code is then lowered into the LLVM IR and
    subsequently run by the LLVM JIT. Such a pipeline allows creation of
    efficient interpreters. The interpreter enables interactive exploration and
    makes the C++ language more user friendly. Clang-Repl is one example.
    <br /> <br />
    Clang-Repl uses the Orcv2 JIT infrastructure within the same process. That
    design is efficient and easy to implement however it suffers from two
    significant drawbacks. First, it cannot be used in devices which do not have
    sufficient resources to host the entire infrastructure, such as the arduino
    due (see this
    <a href="https://compiler-research.org/meetings/#caas_10Mar2022">talk</a>
    for more details). Second, crashes in user codes mean that the entire
    process crashes, hindering overall reliability and ease of use.
    <br /> <br />
    This project aims to move Clang-Repl to an out-of-process execution model
    in order to address both of these issues.
  </p>

  <p><b>Expected result:</b>
    Implement an out-of-process execution of statements with Clang-Repl;
    Demonstrate that Clang-Repl can support some of the ez-clang use-cases;
    Research into approaches to restart/continue the session upon crash;
    As a stretch goal design a versatile reliability approach for crash recovery;
  </p>

  <p><b>Skills:</b>
    Intermediate knowledge of C++, Understanding of LLVM and the LLVM JIT in particular
  </p>

  <p><b>Project size:</b>Either medium or large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
  </p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-out-of-process-execution-for-clang-repl/68225">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-plugins-windows">Support clang plugins on Windows</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b>
    The Clang compiler is part of the LLVM compiler infrastructure and supports
    various languages such as C, C++, ObjC and ObjC++. The design of LLVM and

    Clang allows the compiler to be extended with plugins[1]. A plugin makes it
    possible to run extra user defined actions during a compilation. Plugins
    are supported on unix and darwin but not on windows due to some specifics of
    the windows platform.
    <br /> <br />
    This project would expose the participant to a broad cross section of the LLVM codebase. It involves exploring the API surface, classifying the interfaces as being public or private, and annotating that information to the API declarations. It would also expose the participant to details and differences of different platforms as this work is cross-platform (Windows, Linux, Darwin, BSD, etc). The resulting changes would improve LLVM on Linux and Windows while enabling new functionality on Windows.
  </p>

  <p><b>Expected result:</b>
    This project aims to allow make clang -fplugin=windows/plugin.dll work. The
    implementation approach should extend the working prototype [3] and extend
    the annotation tool [4]. The successful candidate should be prepared to
    attend a weekly meeting, make presentations and prepare blog posts upon
    request.
  </p>

  <p><i>Further reading</i><br />
    [1] https://clang.llvm.org/docs/ClangPlugins.html
    <br />
    [2] https://discourse.llvm.org/t/clang-plugins-on-windows
    <br />
    [3] https://github.com/llvm/llvm-project/pull/67502
    <br />
    [4] https://github.com/compnerd/ids
  </p>

  <p><b>Skills:</b>
    Intermediate knowledge of C++, Experience with Windows and its compilation
    and linking model.
  </p>

  <p><b>Project size:</b>Either medium or large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/compnerd>Saleem Abdulrasool</a>
  </p>


  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/support-clang-plugins-on-windows/76408">URL</a>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-on-demand-parsing">On Demand Parsing in Clang</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b> Clang, like any C++ compiler, parses a
    sequence of characters as they appear, linearly. The linear character
    sequence is then turned into tokens and AST before lowering to machine
    code. In many cases the end-user code uses a small portion of the C++
    entities from the entire translation unit but the user still pays the price
    for compiling all of the redundancies.
    <br /> <br />
    This project proposes to process the heavy compiling C++ entities upon using
    them rather than eagerly. This approach is already adopted in Clang’s
    CodeGen where it allows Clang to produce code only for what is being
    used. On demand compilation is expected to significantly reduce the
    compilation peak memory and improve the compile time for translation units
    which sparsely use their contents. In addition, that would have a
    significant impact on interactive C++ where header inclusion essentially
    becomes a no-op and entities will be only parsed on demand.
    <br /> <br />
    The Cling interpreter implements a very naive but efficient
    cross-translation unit lazy compilation optimization which scales across
    hundreds of libraries in the field of high-energy physics.
    <br /> <br />
    <pre>
// A.h
#include &lt;string&gt;
#include &lt;vector&gt;
template &lt;class T, class U = int&gt; struct AStruct {
  void doIt() { /*...*/ }
  const char* data;
  // ...
};

template&lt;class T, class U = AStruct&lt;T&gt;&gt;
inline void freeFunction() { /* ... */ }
inline void doit(unsigned N = 1) { /* ... */ }

// Main.cpp
#include &quot;A.h&quot;
int main() {
  doit();
  return 0;
}
    </pre>
    <br /> <br />
    This pathological example expands to 37253 lines of code to process. Cling
    builds an index (it calls it an autoloading map) where it contains only
    forward declarations of these C++ entities. Their size is 3000 lines of
    code. The index looks like:
    <pre>
// A.h.index
namespace std{inline namespace __1{template &lt;class _Tp, class _Allocator&gt; class __attribute__((annotate(&quot;$clingAutoload$vector&quot;)))  __attribute__((annotate(&quot;$clingAutoload$A.h&quot;)))  __vector_base;
  }}
...
template &lt;class T, class U = int&gt; struct __attribute__((annotate(&quot;$clingAutoload$A.h&quot;))) AStruct;
    </pre>
    <br /> <br />
    Upon requiring the complete type of an entity, Cling includes the relevant
    header file to get it. There are several trivial workarounds to deal with
    default arguments and default template arguments as they now appear on the
    forward declaration and then the definition. You can read more in [1].
    <br /> <br />
    Although the implementation could not be called a reference implementation,
    it shows that the Parser and the Preprocessor of Clang are relatively
    stateless and can be used to process character sequences which are not
    linear in their nature. In particular namespace-scope definitions are
    relatively easy to handle and it is not very difficult to return to
    namespace-scope when we lazily parse something. For other contexts such as
    local classes we will have lost some essential information such as name
    lookup tables for local entities. However, these cases are probably not very
    interesting as the lazy parsing granularity is probably worth doing only for
    top-level entities.
    <br /> <br />
    Such implementation can help with already existing issues in the standard
    such as CWG2335, under which the delayed portions of classes get parsed
    immediately when they're first needed, if that first usage precedes the end
    of the class. That should give good motivation to upstream all the
    operations needed to return to an enclosing scope and parse something.
    <br /> <br />
    <b>Implementation approach</b>: Upon seeing a tag definition during parsing
    we could create a forward declaration, record the token sequence and mark it
    as a lazy definition. Later upon complete type request, we could re-position
    the parser to parse the definition body. We already skip some of the
    template specializations in a similar way [2, 3].
    <br /> <br />
    Another approach is every lazy parsed entity to record its token stream and
    change the Toks stored on LateParsedDeclarations to optionally refer to a
    subsequence of the externally-stored token sequence instead of storing its
    own sequence (or maybe change CachedTokens so it can do that
    transparently). One of the challenges would be that we currently modify the
    cached tokens list to append an "eof" token, but it should be possible to
    handle that in a different way.
    <br /> <br />
    In some cases, a class definition can affect its surrounding context in a
    few ways you'll need to be careful about here:
    <br /> <br />
    1) `struct X` appearing inside the class can introduce the name `X` into the
        enclosing context.
    <br /> <br />
    2) `static inline` declarations can introduce global variables with
        non-constant initializers that may have arbitrary side-effects.
    <br /> <br />
    For point (2), there's a more general problem: parsing any expression can
    trigger a template instantiation of a class template that has a static data
    member with an initializer that has side-effects. Unlike the above two
    cases, I don't think there's any way we can correctly detect and handle such
    cases by some simple analysis of the token stream; actual semantic analysis
    is required to detect such cases. But perhaps if they happen only in code
    that is itself unused, it wouldn't be terrible for Clang to have a language
    mode that doesn't guarantee that such instantiations actually happen.
    <br /> <br />
    Alternative and more efficient implementation could be to make the lookup
    tables range based but we do not have even a prototype proving this could be
    a feasible approach.
  </p>

  <p><b>Expected result:</b>
    <ul>
      <li>Design and implementation of on-demand compilation for non-templated functions</li>
      <li>Support non-templated structs and classes</li>
      <li>Run performance benchmarks on relevant codebases and prepare report</li>
      <li>Prepare a community RFC document</li>
      <li>[Stretch goal] Support templates</li>
    </ul>

    The successful candidate should commit to regular participation in weekly
    meetings, deliver presentations, and contribute blog posts as
    requested. Additionally, they should demonstrate the ability to navigate the
    community process with patience and understanding.
  </p>

  <p><i>Further reading</i><br/>
    [1] https://github.com/root-project/root/blob/master/README/README.CXXMODULES.md#header-parsing-in-root
    <br />
    [2] https://github.com/llvm/llvm-project/commit/b9fa99649bc99
    <br />
    [3] https://github.com/llvm/llvm-project/commit/0f192e89405ce
  </p>

  <p><b>Skills:</b>
    Knowledge of C++, Deeper understanding of how Clang works,
    knowledge of Clang AST and Preprocessor.
  </p>

  <p><b>Project size:</b>Large</p>
  <p><b>Difficulty:</b> Hard</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/mizvekov>Matheus Izvekov</a>
  </p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/on-demand-parsing-in-clang/76912">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-doc-improve-usability">Improve Clang-Doc Usability</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b>
    <a href=https://clang.llvm.org/extra/clang-doc.html>Clang-Doc</a> is a
    C/C++ documentation generation tool created as an alternative for Doxygen
    and built on top of LibTooling. This effort started in 2018 and critical
    mass has landed in 2019, but the development has been largely dormant since
    then, mostly due to a lack of resources.
    <br /> <br />
    The tool can currently generate documentation in Markdown and HTML formats,
    but the tool has some structural issues, is difficult to use, the generated
    documentation has usability issues and is missing several key features:
    <ul>
      <li>Not all C/C++ constructs are currently handled by the Markdown and
        HTML emitter limiting the tool’s usability.</li>
      <li>The generated HTML output does not scale with the size of the
        codebase making it unusable for larger C/C++ projects.</li>
      <li>The implementation does not always use the most efficient or
        appropriate data structures which leads to correctness and performance
        issues.</li>
      <li>There is a lot of duplicated boiler plate code which could be
        improved with templates and helpers.</li>
    </ul>
  </p>

  <p><b>Expected result:</b>
    The goal of this project is to address the existing shortcomings and
    improve the usability of Clang-Doc to the point where it can be used to
    generate documentation for large scale projects such as LLVM. The ideal
    outcome is that the LLVM project will use Clang-Doc for generating its <a
    href=https://llvm.org/doxygen/>reference documentation</a>.
    <br /><br />
    Successful proposals should focus not only on addressing the existing
    limitations, but also draw inspiration for other potential improvements
    from other similar tools such as <a href=https://hdoc.io/>hdoc</a>, <a
    href=https://github.com/standardese/standardese>standardese</a>, <a
    href=https://github.com/chromium/subspace/tree/main/subdoc>subdoc</a> or
    <a href=https://cs.opensource.google/fuchsia/fuchsia/+/main:tools/cppdocgen/>cppdocgen</a>.
  </p>

  <p><b>Skills:</b>
    Experience with web technologies (HTML, CSS, JS) and an intermediate
    knowledge of C++. Previous experience with Clang/LibTooling is a bonus but
    not required.
  </p>

  <p><b>Project size:</b> Either medium or large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/petrhosek>Petr Hosek</a>,
    <a href=https://github.com/ilovepi>Paul Kirth</a>
  </p>

  <p><b>Discourse:</b> <a href=https://discourse.llvm.org/t/improve-clang-doc-usability/76996>URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="rich-disassembler-for-lldb">Rich Disassembler for LLDB</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>Use the variable location information from the debug info to annotate LLDB’s disassembler (and `register read`) output with the location and lifetime of source variables. The rich disassembler output should be exposed as structured data and made available through LLDB’s scripting API so more tooling could be built on top of this. In a terminal, LLDB should render the annotations as text.</p>

<p><b>Expected outcomes</b></p>

For example, we could augment the disassembly for the following function

<pre>
frame #0: 0x0000000100000f80 a.out`main(argc=1, argv=0x00007ff7bfeff1d8) at demo.c:4:10 [opt]
  1   void puts(const char*);
  2   int main(int argc, char **argv) {
  3    for (int i = 0; i < argc; ++i)
→ 4      puts(argv[i]);
  5    return 0;
  6   }
(lldb) disassemble
a.out`main:
...
  0x100000f71 <+17>: movl  %edi, %r14d
  0x100000f74 <+20>: xorl  %r15d, %r15d
  0x100000f77 <+23>: nopw  (%rax,%rax)
→  0x100000f80 <+32>: movq  (%rbx,%r15,8), %rdi
  0x100000f84 <+36>: callq 0x100000f9e ; symbol stub for: puts
  0x100000f89 <+41>: incq  %r15
  0x100000f8c <+44>: cmpq  %r15, %r14
  0x100000f8f <+47>: jne 0x100000f80 ; <+32> at demo.c:4:10
  0x100000f91 <+49>: addq  $0x8, %rsp
  0x100000f95 <+53>: popq  %rbx
...
</pre>

<p>using the debug information that LLDB also has access to (observe how the source variable i is in r15 from [0x100000f77+slide))</p>


<pre>
$ dwarfdump demo.dSYM --name  i
demo.dSYM/Contents/Resources/DWARF/demo: file format Mach-O 64-bit x86-64
0x00000076: DW_TAG_variable
 DW_AT_location (0x00000098:
 [0x0000000100000f60, 0x0000000100000f77): DW_OP_consts +0, DW_OP_stack_value
 [0x0000000100000f77, 0x0000000100000f91): DW_OP_reg15 R15)
 DW_AT_name ("i")
 DW_AT_decl_file ("/tmp/t.c")
 DW_AT_decl_line (3)
 DW_AT_type (0x000000b2 "int")
</pre>

to produce output like this, where we annotate when a variable is live and what its location is:

<pre>
(lldb) disassemble
a.out`main:
...                                                               ; i=0
  0x100000f74 <+20>: xorl  %r15d, %r15d                           ; i=r15
  0x100000f77 <+23>: nopw  (%rax,%rax)                            ; |
→  0x100000f80 <+32>: movq  (%rbx,%r15,8), %rdi                   ; |
  0x100000f84 <+36>: callq 0x100000f9e ; symbol stub for: puts    ; |
  0x100000f89 <+41>: incq  %r15                                   ; |
  0x100000f8c <+44>: cmpq  %r15, %r14                             ; |
  0x100000f8f <+47>: jne 0x100000f80 ; <+32> at t.c:4:10          ; |
  0x100000f91 <+49>: addq  $0x8, %rsp                             ; i=undef
  0x100000f95 <+53>: popq  %rbx
</pre>

<p>The goal would be to produce output like this for a subset of unambiguous cases, for example, variables that are constant or fully in registers.</p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Adrian Prantl aprantl@apple.com (primary contact)
  <li>Jonas Devlieghere jdevlieghere@apple.com
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Good understanding of C++
  <li>Familiarity with using a debugger on the terminal
  <li>Need to be familiar with all the concepts mentioned in the example above
  <li>Need to have a good understanding of at least one assembler dialect for machine code (x86_64 or AArch64).
</ul>

<p>Desired:</p>

<ul>
  <li>Compiler knowledge including data flow and control flow analysis is a plus.
  <li>Being able to navigate debug information (DWARF) is a plus.
</ul>

<p><b>Size of the project.</b></p>

<p>medium (~175h)</p>

<p><b>An easy, medium or hard rating if possible</b></p>

<p>hard</p>
<p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/rich-disassembler-for-lldb/76952">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="gpu-delta-debugging">GPU Delta Debugging</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>
LLVM-reduce, and similar tools perform delta debugging but are less useful if many implicit constraints exist and violation could easily lead to errors similar to the cause that is to be isolated. This project is about developing a GPU-aware version, especially for execution time bugs, that can be used in conjunction with LLVM/OpenMP GPU-record-and-replay, or simply a GPU loader script, to minimize GPU test cases more efficiently and effectively.
</p>

<p><b>Expected outcomes</b></p>

  <p>A tool to reduce GPU errors without loosing the original error. Optionally, other properties could be the focus of the reduction, not only errors. </p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Parasyris, Konstantinos parasyris1@llnl.gov
  <li>Johannes Doerfert jdoerfert@llnl.gov
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Good understanding of C++
  <li>Familiarity with GPUs and LLVM-IR
</ul>

<p>Desired:</p>

<ul>
  <li>Compiler knowledge including data flow and control flow analysis is a plus.
  <li>Experience with debugging and bug reduction techniques (llvm-reduce) is helpful
</ul>

<p><b>Size of the project.</b></p>

<p>medium</p>

<p><b>An easy, medium or hard rating if possible</b></p>

<p>medium</p>
  
<p><b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/gsoc-2024-gpu-delta-debugging/77237">URL</a>
</p>
  
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="offload-libcxx">Offloading libcxx</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>
Modern C++ defines parallel algorithms as part of the standard library, like `std::transform_reduce(std::execution::par_unseq, vec.begin(), vec.end(), 0, std::plus<int>, …)`. In this project we want to extend an implementation of those that is using OpenMP, including GPU offload, where reasonable. While some algorithms might be amenable to GPU offload via a pure (wrapper) runtime solution, we know others, especially those featuring user provided functors, will also require static program analysis and potentially transformation for additional data management. The goal of the project is to explore different algorithms and the options we have to execute them on the host as well as on accelerator devices, esp. GPUs, automatically via OpenMP.
</p>

<p><b>Expected outcomes</b></p>

  <p> Improvements to the prototype support of offloading in libcxx. Evaluations against other offloading approaches and documentation on the missing parts and shortcommings. </p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Johannes Doerfert jdoerfert@llnl.gov
  <li>Tom Scogland scogland1@llnl.gov
  <li>Tom Deakin tom.deakin@bristol.ac.uk
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Good understanding of C++ and C++ standard algorithms
  <li>Familiarity with GPUs and (OpenMP) offloading
</ul>

<p>Desired:</p>

<ul>
  <li>Experience with libcxx (development).
  <li>Experience debugging and profiling GPU code.
</ul>

<p><b>Size of the project.</b></p>

<p>large</p>

<p><b>An easy, medium or hard rating if possible</b></p>

<p>medium</p>

<p><b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/gsoc-2024-offloading-libcxx/77238">URL</a>
</p>
  
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="parameter-tuning">The 1001 thresholds in LLVM</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>
LLVM has lots of thresholds and flags to avoid "costly cases". However, it is unclear if these thresholds are useful, their value is reasonable, and what impact they really have. Since there are a lot, we cannot do a simple exhaustive search. In some prototype work we introduced a C++ class that can replace hardcoded values and offers control over the threshold, e.g., you can increase the recursion limit via a command line flag from the hardcoded "6" to a different number. In this project we want to explore the thresholds, when they are hit, what it means if they are hit, how we should select their values, and if we need different "profiles".
</p>

<p><b>Expected outcomes</b></p>

  <p> Statistical evidence on the impact of various thresholds inside of LLVM's code base, including compile time changes, impact on transformations, and performance measurements. </p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Jan Hueckelheim jhueckelheim@anl.gov
  <li>Johannes Doerfert jdoerfert@llnl.gov
  <li>William Moses wmoses@mit.edu
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Profiling skills and knowledge of statistical reasoning
</ul>

<p>Desired:</p>

<ul>
  <li>Good understanding of the LLVM code base and optimization flow
</ul>

<p><b>Size of the project.</b></p>

<p>medium</p>

<p><b>An easy, medium or hard rating if possible</b></p>

<p>easy</p>

<p><b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/gsoc-2024-the-1001-thresholds-in-llvm/77235">URL</a>
</p>
  
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="gpu-libc">Performance tuning the GPU libc</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>
We have begun work on a libc library targeting GPUs. This will allow users to call functions such as malloc or memcpy while executing on the GPU. However, it is important that these implementations be functional and performant. The goal of this project is to benchmark the implementations of certain libc functions on the GPU. Work would include writing benchmarks to test the current implementations as well as writing more optimal implementations.
</p>

<p><b>Expected outcomes</b></p>

  <p> In-depth performance for libc functions. Overhead of GPU-to-CPU remote procedure calls. More optimal implementations of 'libc' functions. </p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Joseph Huber joseph.huber@amd.com
  <li>Johannes Doerfert jdoerfert@llnl.gov
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Profiling skills and understanding of GPU architecture
</ul>

<p>Desired:</p>

<ul>
  <li>Experience with libc utilities
</ul>

<p><b>Size of the project.</b></p>

<p>small</p>

<p><b>An easy, medium or hard rating if possible</b></p>

<p>easy</p>

<p><b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/libc-gsoc-2024-performance-and-testing-in-the-gpu-libc/77042">URL</a>
</p>

</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="gpu-first">Improve GPU First Framework</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><b>Description</b></p>

<p>
  <a href="https://arxiv.org/abs/2306.11686">GPU First</a> is a methodology and framework that can enable any existing host code to execute the entire program on a GPU without any modification from users.
  The goal of this project is two folded:
  1) Port <a href="https://github.com/shiltian/llvm-project/tree/direct_gpu_compilation">host code</a> to handle RPC to the new plugin and rewrite it with the host RPC framework introduced in the GPU LibC project.
  2) Explore the support for MPI among multiple thread blocks on a single GPU, or even multiple GPUs.
</p>

<p><b>Expected outcomes</b></p>

  <p> More efficient GPU First framework that can support both NVIDIA and AMD GPUs. Optionally, upstream the framework. </p>

<p><b>Confirmed mentors and their contacts</p></b>

<ul>
  <li>Shilei Tian i@tianshilei.me
  <li>Johannes Doerfert jdoerfert@llnl.gov
  <li>Joseph Huber joseph.huber@amd.com
</ul>

<p><b>Required / desired skills</b></p>

<p>Required:</p>

<ul>
  <li>Good understanding of C++ and GPU architecture
  <li>Familiarity with GPUs and LLVM IR
</ul>

<p>Desired:</p>

<ul>
  <li>Good understanding of the LLVM code base and OpenMP target offloading
</ul>

<p><b>Size of the project.</b></p>

<p>medium</p>

<p><b>An easy, medium or hard rating if possible</b></p>

<p>medium</p>

<p><b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/openmp-gsoc-2024-improve-gpu-first-framework/77048">URL</a>
</p>

</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clangir-gpu">Compile GPU kernels using ClangIR</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
    Heterogeneous programming models such as
    <a href="https://sycl.tech">SYCL</a>,
    <a href="https://www.openmp.org">OpenMP</a> and
    <a href="https://www.openacc.org">OpenACC</a> help developers to offload
    computationally intensive kernels to GPUs and other accelerators.
    <a href="https://mlir.llvm.org">MLIR</a> is expected to unlock new
    high-level optimisations and better code generation for the next generation
    of compilers for heterogeneous programming models. However, the availability
    of a robust MLIR-emitting C/C++ frontend is a prerequisite for these
    efforts.
  </p><p>
    The <a href="https://clangir.org">ClangIR</a> (CIR) project aims to
    establish a new intermediate representation (IR) for Clang. Built on top of
    MLIR, it provides a dialect for C/C++ based languages in Clang, and the
    necessary infrastructure to emit it from the Clang AST, as well as a
    lowering path to the LLVM-IR dialect. Over the last year, ClangIR has
    evolved into a mature incubator project, and a recent
    <a href="https://discourse.llvm.org/t/rfc-upstreaming-clangir/76587">RFC</a>
    on upstreaming it into the LLVM monorepo has seen positive comments and
    community support.
  </p><p>
    The overall goal of this GSoC project is to identify and implement missing
    features in ClangIR to make it possible to compile GPU kernels in the
    <a href="https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_C.html">OpenCL C language</a>
    to LLVM-IR for the
    <a href="https://registry.khronos.org/SPIR-V">SPIR-V</a> target. The OpenCL
    to SPIR-V flow is a great environment for this project because a) it is
    <a href="https://clang.llvm.org/docs/OpenCLSupport.html">already supported</a>
    in Clang and b) OpenCL's work-item- and work-group-based programming model
    still captures modern GPU architectures well. The contributor will extend
    the AST visitors, the dialect and the LLVM-IR lowering, to add support e.g.
    for multiple address spaces, vector and custom floating point types, and the
    <code>spir_kernel</code> and <code>spir_func</code> calling conventions.
  </p><p>
    A good starting point for this work is the
    <a href="https://github.com/sgrauerg/polybenchGpu/tree/master/OpenCL">Polybench-GPU</a>
    benchmark suite. It contains self-contained small- to medium sized OpenCL
    implementations of common algorithms. We expect only the device code (*.cl
    files) to be compiled via ClangIR. The existing OpenCL support in Clang can
    be used to create lit tests with reference LLVM-IR output to guide the
    development. Optionally, the built-in result verification and time
    measurements in Polybench could also be used to assess the correctness and
    quality of the generated code.
  </p>

  <p><b>Expected result:</b>
    Polybench-GPU's
    <a href="https://github.com/sgrauerg/polybenchGpu/blob/master/OpenCL/2DCONV/2DConvolution.cl"><code>2DCONV</code></a>,
    <a href="https://github.com/sgrauerg/polybenchGpu/blob/master/OpenCL/GEMM/gemm.cl"><code>GEMM</code></a> and
    <a href="https://github.com/sgrauerg/polybenchGpu/blob/master/OpenCL/CORR/correlation.cl"><code>CORR</code></a>
    OpenCL kernels can be compiled with ClangIR to LLVM-IR for SPIR-V.
  </p>

  <p><b>Skills:</b>
    Intermediate C++ programming skills and familiarity with basic compiler
    design concepts are required. Prior experience with LLVM IR, MLIR, Clang or
    GPU programming is a big plus, but willingness to learn is also a
    possibility.
  </p>

  <p><b>Project size:</b> Large</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/jopperm">Julian Oppermann</a>,
    <a href="https://github.com/Naghasan">Victor Lom&uuml;ller</a>,
    <a href="https://github.com/bcardosolopes">Bruno Cardoso Lopes</a>
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/clangir-compile-gpu-kernels-using-clangir/76984">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="half-precision-libc">Half precision in LLVM libc</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b></p>

  <p>
    Half precision is an IEEE 754 floating point format that has been widely
    used recently, especially in machine learning and AI. It has been
    standardized as _Float16 in the latest C23 standard, bringing its support to
    the same level as float or double data types. The goal for this project is
    to implement C23 half precision math functions in the LLVM libc library.
  </p>

  <p><b>Expected result:</b></p>

  <ul>
    <li> Setup the generated headers properly so that the type and the functions
         can be used with various compilers (+versions) and architectures. </li>
    <li> Implement generic basic math operations supporting half precision data
         types that work on supported architectures: x86_64, arm (32 + 64),
         risc-v (32 + 64), and GPUs. </li>
    <li> Implement specializations using compiler builtins or special hardware
         instructions to improve their performance whenever possible. </li>
    <li> If time permits, we can start investigating higher math functions for
         half precision. </li>
  </ul>

  <p><b>Skills:</b></p>

  <p>
    Intermediate C++ programming skills and familiarity with basic compiler
    design concepts are required. Prior experience with LLVM IR, MLIR, Clang or
    GPU programming is a big plus, but willingness to learn is also a
    possibility.
  </p>

  <p><b>Project size:</b> Large </p>

  <p><b>Difficulty:</b> Easy/Medium</p>

  <p><b>Confirmed Mentors:</b>
    <a href="mailto:lntue@google.com">Tue Ly</a>,
    <a href="mailto:joseph.huber@amd.com">Joseph Huber</a>,
  </p>

  <p><b>Discourse:</b>
    <a href="https://discourse.llvm.org/t/libc-gsoc-2024-half-precision-in-llvm-libc/77027">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc23">Google Summer of Code 2023</a>
</div>
<!-- *********************************************************************** -->


<div class="www_text">
  <p>
    Google Summer of Code 2023 was very successful for LLVM project. For the
    list of accepted and completed projects, please take a look into Google
    Summer of
    Code <a href="https://summerofcode.withgoogle.com/archive/2023/organizations/llvm-compiler-infrastructure">website</a>.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>LLVM</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_new_jitlink_reopt">Re-optimization using JITLink</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    In Just-In-Time compilers we often choose a low optimization level to
    minimize compile time and improve launch times and latencies, however some
    functions (which we call hot functions) are used very frequently and for
    these functions it is worth optimizing more heavily. In general hot
    functions can only be identified at runtime (different inputs will cause
    different functions to become hot), so the aim of the reoptimization project
    is to build infrastructure to (1) detect hot functions at runtime and (2)
    compile them a second time at a higher optimization level, hence the name
    "re-optimization".
    <br /><br />
    There are many possible approaches to both parts of this problem. E.g. hot
    functions could be identified by sampling, or using existing profiling
    infrastructure, or by implementing custom instrumentation. Reoptimization
    could be applied to whole functions, or outlining could be used to enable
    optimization of portions of functions. Re-entry into the JIT infrastructure
    from JIT’d code might be implemented on top of existing lazy compilation, or
    via a custom path.
    <br /><br />
    Whatever design is adopted, the goal is that the infrastructure should be
    generic so that it can be used by other LLVM API clients, and should support
    out-of-process JIT-compilation (so some of the solution will be implemented
    in the ORC runtime).

  <p><b>Expected result:</b>
    <ul>
      <li>Improve ergonomics of indirection – ideally all forms of indirection
        (for re-optimization, lazy compilation, and procedure-linkage-tables)
        should be able to share a single stub (and/or binary rewriting metadata)
        at runtime.</li>
      <li>Implement basic re-optimization on top of the tidied up
        indirection.</li>
      <li>(Stretch goal) Garbage-collect unoptimized code that is no longer
        needed once the optimized version is available.</li>
    </ul>

  <p><b>Desirable skills:</b>
    Intermediate C++; Understanding of LLVM and the LLVM JIT in particular.

  <p><b>Project size:</b> Large.</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/lhames>Lang Hames</a></p>
  </p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/re-optimization-using-jitlink/68260">URL</a>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_new_jitlink_backends">JITLink new backends</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    JITLink is LLVM's new JIT linker API -- the low-level API that transforms
    compiler output (relocatable object files) into ready-to-execute bytes in
    memory. To do this JITLink’s generic linker algorithm needs to be
    specialized to support the target object format (COFF, ELF, MachO), and
    architecture (arm, arm64, i386, x86-64). LLVM already has mature
    implementations of JITLink for MachO/arm64, MachO/x86-64, ELF/x86-64,
    ELF/aarch64 and COFF/x86-64, while the implementations for ELF/riscv,
    ELF/aarch32 and COFF/i386 are still relatively new.
    <br />
    You can either work on an entirely new architecture like PowerPC or eBPF,
    or complete one of the recently added JITLink implementations. In both cases
    you will likely reuse the existing generic code for one of the target object
    formats. You will also work on relocation resolution, populate PLTs and GOTs
    and wire up the ORC runtime for your chosen target.
    <br />

  <p><b>Expected result:</b>
    Write a JITLink specialization for a not-yet-supported or incomplete
    format/architecture such as PowerPC, AArch32 or eBPF.

  <p><b>Desirable skills:</b>
    Intermediate C++; Understanding of LLVM and the LLVM JIT in particular;
    familiarity with your chosen format/architecture, and basic linker concepts
    (e.g. sections, symbols, and relocations).

  <p><b>Project size:</b> Large.</p>

  <p><b>Difficulty:</b>Medium</p>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/lhames>Lang Hames</a></p>
    <a href=https://github.com/weliveindetail>Stefan Gränitz</a></p>
  </p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/jitlink-new-backends/68223">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_improving_compile_times">Improving compile times</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    While the primary job of a compiler is to produce fast code (good run-time
    performance), it is also important that optimization doesn’t take too much
    time (good compile-time performance). The goal of this project is to improve
    compile-time without hurting optimization quality.
    <br />
    The general approach to this project is:
    <ol>
      <li>Pick a workload to optimize. For example, this could be a file from
        <a href="https://github.com/llvm/llvm-test-suite/tree/main/CTMark">CTMark</a>
        compiled in a certain build configuration (e.g. <code>-O0 -g</code> or
        <code>-O3 -flto=thin</code>).</li>
      <li>Collect profiling information. This could involve compiler options like
        <code>-ftime-report</code> or <code>-ftime-trace</code> for a high-level
        overview, as well as <code>perf record</code> or
        <code>valgrind --tool=callgrind</code> for a detailed profile.</li>
      <li>Identify places that are unexpectedly slow. This is heavily workload
        dependent.</li>
      <li>Try to optimize an identified hotspot, ideally without impacting generated
        code. The <a href="https://llvm-compile-time-tracker.com/">compile-time tracker</a>
        can be used to quickly evaluate impact on CTMark.</li>
    </ol>
    As a disclaimer, it should be noted that outside of pathological cases,
    compilation doesn’t tend to have a convenient hotspot where 90% of the time
    is spent, instead it is spread out across many passes. As such, individual
    improvements also tend to have only small impact on overall compile-time.
    Expect to do 10 improvements of 0.2% each, rather than one improvement of 2%.
  </p>

  <p><b>Expected result:</b>
    Substantial improvements on some individual files (multiple percent), and a
    small improvement on overall geomean compile-time.</p>

  <p><b>Desirable skills:</b>
    Intermediate C++. Familiarity with profiling tools (especially if you are
    not on Linux, in which case I won’t be able to help).</p>

  <p><b>Project size:</b> Either medium or large.</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Confirmed Mentor:</b> <a href="https://github.com/nikic">Nikita Popov</a></p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/llvm-improving-compile-times/68094">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_addressing_rust_optimization_failures">Addressing Rust optimization failures</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The <a href="https://www.rust-lang.org/">Rust programming language</a> uses
    LLVM for code generation, and heavily relies on LLVM’s optimization
    capabilities. However, there are many cases where LLVM fails to optimize
    typical code patterns that are emitted by rustc. Such issues are reported
    using the <a href="https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3AI-slow">I-slow</a>
    and/or <a href="https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3AA-LLVM">A-LLVM</a> labels.
    <br />
    The usual approach to fixing these issues is:
    <ol>
      <li>Inspect the <code>--emit=llvm-ir</code> output on
        <a href="https://rust.godbolt.org/">Godbolt</a>.</li></li>
      <li>Create an LLVM IR test case that is not optimized when run through
        <code>opt -O3</code>.</li>
      <li>Identify a minimal missing transform and prove its correctness
        using <a href="https://alive2.llvm.org/ce/">alive2</a>.</li>
      <li>Identify which LLVM pass or passes could perform the transform.</li>
      <li>Add necessary test coverage and implement the transform.</li>
      <li>(Much later: Check that the issue is really resolved after the next
        major LLVM version upgrade in Rust.)</li>
    </ol>
    The goal of this project is to address some of the less hard optimization
    failures. This means that in some cases, the process would stop after step 3
    or 4 without proceeding to implementation, because it’s unclear how the issue
    could be addressed, or it would take a large amount of effort. Having an
    analysis of the problem is still valuable in that case.
  </p>

  <p><b>Expected result:</b>
    Fixes for a number of easy to medium Rust optimization failures. Preliminary
    analysis for some failures even if no fix was implemented.</p>

  <p><b>Desirable skills:</b>
    Intermediate C++ for implementation. Some familiarity with LLVM (at least
    ability to understand LLVM IR) for analysis. Basic Rust knowledge (enough
    to read, but not write Rust).</p>

  <p><b>Project size:</b> Either medium or large.</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Confirmed Mentor:</b> <a href="https://github.com/nikic">Nikita Popov</a></p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/llvm-addressing-rust-optimization-failures-in-llvm/68096">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-repl-autocompletion">Implement autocompletion in clang-repl</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The Clang compiler is part of the LLVM compiler infrastructure and supports
    various languages such as C, C++, ObjC and ObjC++. The design of LLVM and
    Clang enables them to be used as libraries, and has led to the creation of
    an entire compiler-assisted ecosystem of tools. The relatively friendly
    codebase of Clang and advancements in the JIT infrastructure in LLVM further
    enable research into different methods for processing C++ by blurring the
    boundary between compile time and runtime. Challenges include incremental
    compilation and fitting compile/link time optimizations into a more dynamic
    environment.
    <br /> <br />
    Incremental compilation pipelines process code chunk-by-chunk by building an
    ever-growing translation unit. Code is then lowered into the LLVM IR and
    subsequently run by the LLVM JIT. Such a pipeline allows creation of
    efficient interpreters. The interpreter enables interactive exploration and
    makes the C++ language more user friendly. The incremental compilation mode
    is used by the interactive C++ interpreter, Cling, initially developed to
    enable interactive high-energy physics analysis in a C++ environment.
    <br /> <br />
    <a href="https://compiler-research.org/">Our group</a> puts efforts to
    incorporate and possibly redesign parts of Cling in Clang mainline through a
    new tool, clang-repl. The project aims at the design and implementation of
    robust autocompletion when users type C++ at the prompt of clang-repl.
    For example:
    <pre>
      [clang-repl] class MyLongClassName {};
      [clang-repl] My&lt;tab&gt;
      // list of suggestions.
    </pre>
  </p>

  <p><b>Expected result:</b>
    There are several foreseen tasks:
    <ul>
      <li>Research the current approaches for autocompletion in clang such as
        clang -code-completion-at=file:col1:col2.</li>
      <li>Implement a version of the autocompletion support using the partial
        translation unit infrastructure in clang’s libInterpreter.</li>
      <li>Investigate the requirements for semantic autocompletion which takes
        into account the exact grammar position and semantics of the code. Eg:
        <pre>
          [clang-repl] struct S {S* operator+(S&) { return nullptr;}};
          [clang-repl] S a, b;
          [clang-repl] v = a + &lt;tab&gt; // shows b as the only acceptable choice here.
        </pre>
      </li>
      <li>Present the work at the relevant meetings and conferences.</li>
  </p>

  <p><b>Project size:</b>Large.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>
  </p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-repl-implement-autocompletion-in-clang-repl/60364">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-modules-build-daemon">Modules build daemon: build system agnostic support for explicitly built modules</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b> Clang currently handles modules independently in each
    <code>clang</code> instance using the filesystem for synchronization of which instance builds
    a given module. This has many issues with soundness and performance due to tradeoffs made for
    module reuse and filesystem contention.</p>
  <p>Clang has another way of building modules, explicitly built modules, that currently requires
    build system changes to adopt. Here the build system determines which modules are needed, for
    example by using <a href="https://github.com/llvm/llvm-project/tree/main/clang/tools/clang-scan-deps">clang-scan-deps</a>,
    and ensures those modules are built before running the <code>clang</code> compile task that
    needs them.</p>
  <p>In order to allow adoption of this new way of building modules without major build system work
    we need a module build daemon. With a small change to the command line, clang will connect to
    this daemon and ask for the modules it needs. The module build daemon then either returns an
    existing valid module, or builds and then returns it.</p>
  <p>There is an existing open source dependency scanning daemon that is in a llvm-project fork.
    This only handles file dependencies, but has an IPC mechanism. This IPC system could be used as
    a base for the modules build daemon, but does need to be extended to work on Windows.</p>
  <p><b>Expected result:</b> A normal project using Clang modules with an existing build system
    (like Make or CMake) can be built using only explicitly built modules via a modules build
    daemon.</p>
  <p><b>Desirable skills:</b> Intermediate C++ programming skills; familiarity with compilers;
    familiarity with Clang is an asset, but not required.</p>
  <p><b>Project size:</b> 175h or 350h depending on reuse of IPC</p>
  <p><b>Difficulty:</b> medium</p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/Bigcheese">Michael Spencer</a>,
    <a href="https://github.com/jansvoboda11">Jan Svoboda</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-modules-build-daemon-build-system-agnostic-support-for-explicitly-built-modules/68224">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-extract-api-categories">ExtractAPI Objective-C categories</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b> <a href="https://github.com/apple/swift-docc">Swift-DocC</a> is
  the canonical documentation compiler for the Swift OSS project. However
  Swift-DocC is not Swift specific and
  uses <a href="https://github.com/apple/swift-docc-symbolkit/blob/main/openapi.yaml">SymbolKit</a>'s
  languaguage agnostic JSON-based symbol graph format to understand which
  symbols are available in the code, this way any language can be supported by
  Swift-DocC as long as there is a symbol graph generator.</p>
  <p>Clang supports symbol graph generation for C and Objective-C as described
  in <a href="https://discourse.llvm.org/t/rfc-clang-support-for-api-information-generation-in-json/58845">[RFC]
  clang support for API information generation in JSON</a>. Today, support for
  Objective-C categories is not complete, on one hand if the category extends a
  type in the current module, the category members are assumed to belong to the
  extended type itself. On the other hand, if the extended type belongs to
  another module the category is ignored. Nonetheless, it is common to extend
  types belonging to other modules in Objective-C as part of the public API of
  the module. The goal of this project is to extend the symbol graph format to
  accommodate Objective-C categories and to implement support for generating
  this information both through clang and through libclang.</p>
  <p><b>Expected result:</b> Adding the necessary support to clang's symbol graph
  generator and in libclang for describing categories of symbols defined in
  other modules. This might involve additions to SymbolKit that would need to be
  discussed with that community.</p>
  <p><b>Desirable skills:</b> Intermediate C++ programming skills; familiarity
  with clang and Objective-C are assets but not required.</p>
  <p><b>Project size:</b> Medium</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/daniel-grumberg">Daniel Grumberg</a>,
    <a href="https://github.com/zixu-w">Zixu Wang</a>,
    <a href="https://github.com/ributzka">Juergen Ributzka</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-extractapi-objective-c-categories/68370">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-extract-api-cpp-support">ExtractAPI C++ Support</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b> <a href="https://github.com/apple/swift-docc">Swift-DocC</a> is
  the canonical documentation compiler for the Swift OSS project. However
  Swift-DocC is not Swift specific and
  uses <a href="https://github.com/apple/swift-docc-symbolkit/blob/main/openapi.yaml">SymbolKit</a>'s
  languaguage agnostic JSON-based symbol graph format to understand which
  symbols are available in the code, this way any language can be supported by
  Swift-DocC as long as there is a symbol graph generator.</p>
  <p>Clang supports symbol graph generation for C and Objective-C as described
  in <a href="https://discourse.llvm.org/t/rfc-clang-support-for-api-information-generation-in-json/58845">[RFC]
  clang support for API information generation in JSON</a>.</p>
  <p>Currently the emitted symbol graph format does not support various C++
  constructs such as templates and exceptions and the symbol graph generator
  does not fully understand C++. This project aims to introduce support for
  various C++ constructs in the symbol graph format and to implement support for
  generating this data in clang.</p>
  <p><b>Expected result:</b> Adding the necessary support to clang's symbol graph
  generator and in libclang for describing categories of symbols defined in
  other modules. This will involve additions to SymbolKit that would need to be
  discussed with that community.</p>
  <p><b>Desirable skills:</b> Intermediate C++ programming skills; familiarity
  with clang and Objective-C are assets but not required.</p>
  <p><b>Project size:</b> Large</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/daniel-grumberg">Daniel Grumberg</a>,
    <a href="https://github.com/zixu-w">Zixu Wang</a>,
    <a href="https://github.com/ributzka">Juergen Ributzka</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/extractapi-c-support/68371">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-extract-api-while-building">ExtractAPI while building</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b> <a href="https://github.com/apple/swift-docc">Swift-DocC</a> is
  the canonical documentation compiler for the Swift OSS project. However
  Swift-DocC is not Swift specific and
  uses <a href="https://github.com/apple/swift-docc-symbolkit/blob/main/openapi.yaml">SymbolKit</a>'s
  languaguage agnostic JSON-based symbol graph format to understand which
  symbols are available in the code, this way any language can be supported by
  Swift-DocC as long as there is a symbol graph generator.</p>
  <p>Clang supports symbol graph generation for C and Objective-C as described
  in <a href="https://discourse.llvm.org/t/rfc-clang-support-for-api-information-generation-in-json/58845">[RFC]
  clang support for API information generation in JSON</a>.</p>
  <p>Currently users can use clang to generate symbol graph files using
  the <code>clang -extract-api</code> command line interface or generating
  symbol graphs for a specific symbol using the libclang interface. This project
  would entail adding a third mode that would generate the symbol graph output
  as a side-effect of a regular compilation job. This can enable using the
  symbol graph format as a light weight alternative to clang Index or clangd
  for code intelligence services.</p>
  <p><b>Expected result:</b> Enable generating symbol graph files during a
  regular compilation (or module build); provide a tool to merge symbol graph
  files in the same way a static linker links individual object files; Extend
  clang Index to support all the information contained by symbol graph
  files.</p>
  <p><b>Desirable skills:</b> Intermediate C++ programming skills; familiarity
  with clang and Objective-C are assets but not required.</p>
  <p><b>Project size:</b> Medium</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/daniel-grumberg">Daniel Grumberg</a>,
    <a href="https://github.com/zixu-w">Zixu Wang</a>,
    <a href="https://github.com/ributzka">Juergen Ributzka</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-extractapi-while-building/68372">URL</a></p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-improve-diagnostics2">Improve Clang diagnostics</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description:</b>
  The diagnostics clang emits are ultimately its interface to the developer. While the diagnostics are generally good, there are some rough edges that need to be ironed out. Some cases can be improved by special-casing them in the compiler as well.
  </p>

  <p>
  As one can see from Clang’s issue tracker, there are <a href="https://github.com/llvm/llvm-project/issues?page=2&q=is%3Aopen+is%3Aissue+label%3Aclang%3Adiagnostics">lots of issues</a> open against clang’s diagnostics.
  </p>

  <p>
  This project does not aim to implement one big feature but instead focuses on smaller, incremental improvements to Clang’s diagnostics.
  </p>

  <p>
  Possible example issues to resolve:
  <ul>
    <li><a href="https://github.com/llvm/llvm-project/issues/59872">Calling nullptr function pointer in a constexpr function results in poor diagnostic</a></li>
    <li><a href="https://github.com/llvm/llvm-project/issues/58601">Print name of uninitialized subobject (instead of type)</a></li>
    <li><a href="https://github.com/llvm/llvm-project/issues/57906">https://github.com/llvm/llvm-project/issues/57906</a></li>
    <li><a href="https://github.com/llvm/llvm-project/issues/57337">clang(++) unhelpful frame-larger-than warning, very small stack frame exceeding very large limit</a></li>
    <li>Any other diagnostics issue you find interesting or ran into personally.</li>
  </ul>
  </p>


  <p><b>Expected outcomes</b>:
  At least three fixed smaller diagnostics issues, or one larger implemented diagnostics improvement.
  </p>

  <p><b>Confirmed Mentor:</b><a href=https://github.com/tbaederr>Timm Bäder</a>

  <p><b>Desirable skills:</b>
    <ul>
      <li>Intermediate C++ knowledge.</li>
      <li>Preferably experience in the Clang code base, since the issues mentioned can have their root cause in various parts of it.</li>
      <li>Preferably an already working local LLVM build</li>
    </ul>
  </p>

  <p><b>Project type:</b> Medium/200 hr</p>

  <p><b>Discourse</b>
  <a href="https://discourse.llvm.org/t/improve-clang-diagnostics-2/68900/3">URL</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-tutorials-clang-repl">Tutorial development with clang-repl</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description:</b>
    The Clang compiler is part of the LLVM compiler infrastructure and supports
    various languages such as C, C++, ObjC and ObjC++. The design of LLVM and
    Clang enables them to be used as libraries, and has led to the creation of
    an entire compiler-assisted ecosystem of tools. The relatively friendly
    codebase of Clang and advancements in the JIT infrastructure in LLVM further
    enable research into different methods for processing C++ by blurring the
    boundary between compile time and runtime. Challenges include incremental
    compilation and fitting compile/link time optimizations into a more dynamic
    environment.
  </p>

  <p>
    Incremental compilation pipelines process code chunk-by-chunk by building an
    ever-growing translation unit. Code is then lowered into the LLVM IR and
    subsequently run by the LLVM JIT. Such a pipeline allows creation of
    efficient interpreters. The interpreter enables interactive exploration and
    makes the C++ language more user friendly. The incremental compilation mode
    is used by the interactive C++ interpreter, Cling, initially developed to
    enable interactive high-energy physics analysis in a C++ environment.
  </p>

  <p>
    We invest efforts to incorporate and possibly redesign parts of Cling in
    Clang mainline through a new tool, clang-repl. The project aims implementing
    tutorials demonstrating the capabilities of the project and investigating
    adoption of clang-repl in xeus-clang-repl prototype allowing to write C++
    in Jupyter.
  </p>

  <p><b>Expected result:</b>
    There are several foreseen tasks:
    <ul>
      <li>Write several tutorials demostrating the current capabilities of
        clang-repl.</li>
      <li>Investigate the requirements for adding clang-repl as a backend to
        xeus-cling.</li>
      <li>Improve the xeus kernel protocol for clang-repl.</li>
      <li>Prepare a blog post about clang-repl and possibly Jupyter.
        Present the work at the relevant meetings and conferences.</li>
  </p>


  <p><b>Confirmed Mentor:</b>
    <a href="https://github.com/vgvassilev">Vassil Vassilev</a>
    <a href="https://github.com/davidlange6">David Lange</a>

  <p><b>Desirable skills:</b>
    Intermediate C++; Understanding of Clang and the Clang API in particular
  </p>

  <p><b>Project type:</b> Medium</p>

  <p><b>Discourse</b>
  <a href="https://discourse.llvm.org/t/clang-repl-tutorial-development-with-clang-repl/60365">URL</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-repl-wasm">Add WebAssembly Support in clang-repl</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description:</b>
    The Clang compiler is part of the LLVM compiler infrastructure and supports
    various languages such as C, C++, ObjC and ObjC++. The design of LLVM and
    Clang enables them to be used as libraries, and has led to the creation of
    an entire compiler-assisted ecosystem of tools. The relatively friendly
    codebase of Clang and advancements in the JIT infrastructure in LLVM further
    enable research into different methods for processing C++ by blurring the
    boundary between compile time and runtime. Challenges include incremental
    compilation and fitting compile/link time optimizations into a more dynamic
    environment.
  </p>

  <p>
    Incremental compilation pipelines process code chunk-by-chunk by building an
    ever-growing translation unit. Code is then lowered into the LLVM IR and
    subsequently run by the LLVM JIT. Such a pipeline allows creation of
    efficient interpreters. The interpreter enables interactive exploration and
    makes the C++ language more user friendly. The incremental compilation mode
    is used by the interactive C++ in Jupyter via the xeus kernel protocol.
    Newer versions of the protocol allow possible in-browser execution allowing
    further possibilities for clang-repl and Jupyter.
  </p>

  <p>
    We invest efforts to incorporate and possibly redesign parts of Cling in
    Clang mainline through a new tool, clang-repl. The project aims to add
    WebAssembly support in clang-repl and adopt it in xeus-clang-repl to aid
    Jupyter-based C++.
  </p>

  <p><b>Expected result:</b>
    There are several foreseen tasks:
    <ul>
      <li>Investigate feasibility of generating WebAssembly in a similar way to
        the new <a href="https://reviews.llvm.org/D146389">interactive CUDA support</a>.</li>
      <li>Enable generating WebAssembly in clang-repl.</li>
      <li>Adopt the feature in xeus-clang-repl.</li>
      <li>Prepare a blog post about clang-repl and possibly Jupyter.
        Present the work at the relevant meetings and conferences.</li>
  </p>


  <p><b>Confirmed Mentor:</b>
    <a href="https://github.com/vgvassilev">Vassil Vassilev</a>
    <a href="https://github.com/alexander-penev">Alexander Penev</a>

  <p><b>Desirable skills:</b>
    Good C++; Understanding of Clang and the Clang API and the LLVM JIT in particular
  </p>

  <p><b>Project type:</b> Large</p>

  <p><b>Discourse</b>
  <a href="https://discourse.llvm.org/t/clang-repl-add-webassembly-support-in-clang-repl/69419">URL</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_lld_embedded">LLD Linker Improvements for Embedded Targets</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
    GNU toolchain is used widely for building embedded targets. There's a certain momentum in the
    Clang/LLVM community towards improving the Clang toolchain to support embedded targets. Using
    the Clang toolchain as an alternative can help us improve code quality, find and fix security
    bugs, improve developer experience and take advantage of the new ideas and the momentum
    surrounding the Clang/LLVM community in supporting embedded devices.
  </p>
  <p><b>A non-comprehensive list of improvements that can be made to LLD</b>:
  <ul>
    <li>
      <p><b>--print-memory-usage support</b></p>
      <p>"--print-memory-usage" in GCC provides a breakdown of the memory used in each memory region
        defined in the linker file. Embedded developers use this flag to understand the impact on
        memory. Often embedded systems define multiple memory regions with different space
        constraints. Supporting this in Clang toolchain will help projects that wish to use Clang
        toolchain for their projects.</p>
    </li>
    <li>
      <p><b>Linkmap</b></p>
      <p>Currently, the LLD linker's linkmap output is not as rich as the BFD linker output.
        Achieving feature parity on linkmap output will be highly
        useful in analyzing the binaries created by the LLD linker. Further, outputting linkmap in
        different formats (current LLD output, BFD, and JSON) can help build automation tools for
        investigating the artifacts produced by the linker.</p>
    </li>
    <li>
      <p><b>--print-gc-sections improvement</b></p>
      <p>When the "--print-gc-sections" flag is enabled, LLD prints the sections that were
        discarded during the linking process. This information currently does not include the
        mapping between the symbol and the section groups, which is useful for debugging.
        Preserving this information during the linking process will require modifications to
        internal linker data structures.</p>
    </li>
  </ul>
  <p><b>Project size:</b> Medium or Large</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Skills:</b> C++</p>
  <p><b>Expected result</b>:
    <ul>
      <li>Implementation of "--print-memory-usage" flag.</li>
      <li>Support for new linkmap output formats 1. BFD and 2. JSON. </li>
      <li>Improved "--print-gc-sections" output to include information about the surviving symbols.</li>
    </ul>
  </p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/Prabhuk">Prabhu Rajasekaran</a>
    <a href="https://github.com/petrhosek">Petr Hosek</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/lld-linker-improvements-for-embedded/68129">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_mlir_presburger_opt">Optimizing MLIR’s Presburger library </a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><strong>Description</strong>: MLIR’s Presburger Library, FPL (<a href="https://grosser.science/FPL">https://grosser.science/FPL</a>), provides mathematical abstractions for polyhedral compilation and analysis. The main abstraction that the library provides is a set of integer tuples defined by a system of affine inequality constraints. The library supports standard set operations over such sets. The result will be a set defined by another constraint system, possibly having more constraints. When many set operations are performed in sequence, the constraint system may become very large, negatively impacting performance. There are several potential ways to simplify the constraint system; however, this involves performing additional computations. Thus, spending more time on more aggressive simplifications may make each individual operation slower, but at the same time, insufficient simplifications can make sequences of operations slow due to an explosion in constraint system size. The aim of this project is to find the right balance between the two.</p>
<p><strong>The goals of this project:</strong></p>
<ul>
<li>Understand the library&#39;s performance in terms of runtime and output size.</li>
<li>Optimize the library by finding the best output size and performance tradeoff.</li>
</ul>
<p><strong>Expected outcomes</strong>:</p>
<ul>
<li>Benchmarking the performance and output constraint complexity of the primary operations of the library.</li>
<li>Implementing simplification heuristics.</li>
<li>A better understanding of which simplification heuristics improve overall performance enough to be worth the additional computational cost.</li>
</ul>
<p><strong>Desirable skills</strong>: Intermediate C++, Experience in benchmarking</p>
<p><strong>Project size</strong>: Large</p>
<p><strong>Difficulty</strong>: Medium</p>
<p><strong>Confirmed mentors</strong>: <a href="https://github.com/Groverkss">Kunwar Grover</a></p>
<p><strong>Discourse</strong>: <a href="https://discourse.llvm.org/t/mlir-optimizing-mlir-s-presburger-library/68213/1">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_mlir_query">Interactively query MLIR IR</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
<p><strong>Description</strong>:
  The project aims to develop an interactive query language for MLIR that enables developers to query the MLIR IR dynamically.
  The tool will provide a REPL (or command-line) interface to enable users to query various properties of MLIR code, such as
  "isConstant" and "resultOf". The proposed tool is intended to be similar to clang-query, which allows developers to match
  AST expressions in C++ code using a TUI with autocomplete and other features.
</p>
<p><strong>The goals of this project:</strong></p>
<ul>
<li>Understand the MLIR IR representation and common explorations user do.</li>
<li>Implement a REPL to execute queries over MLIR IR.</li>
</ul>
<p><strong>Expected outcomes</strong>:</p>
<ul>
<li>Standalone that can be used to interactively explore IR.</li>
<li>Implement common matchers that are usable by the tool.</li>
<li>(stretch) Enable extracting parts of the IR matched by query into self-contained IR snippets.</li>
</ul>
<p><strong>Desirable skills</strong>: Intermediate C++, Experience in writing/debugging peephole optimizations</p>
<p><strong>Project size</strong>: Either medium or large.</p>
<p><strong>Difficulty</strong>: Medium</p>
<p><strong>Confirmed mentors</strong>: <a href="https://github.com/jpienaar">Jacques Pienaar</a></p>
<p><strong>Discourse</strong>: <a href="https://discourse.llvm.org/t/gsoc-proposal-interactive-mlir-query-tool-to-make-exploring-the-ir-easier/69601">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_mlgo_latency_model">Better performance models for MLGO training</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
    We are using machine-guided compiler optimizations ("MLGO") for register allocation eviction and inlining for size, in
    real-life deployments. The ML models have been trained with reinforcement learning algorithms. Expanding to more
    performance areas is currently impeded by the poor prediction quality of our performance estimation models. Improving
    those is critical to the effectiveness of reinforcement learning training algorithms, and therefore to enabling applying
    MLGO systematically to more optimizations.
  </p>
  <p><b>Project size:</b> either 175 or 350 hr.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Skills:</b> C/C++, some compiler experience, some Python. ML experience is a bonus.</p>
  <p><b>Expected outcomes</b>: Better modeling of the execution environment by including additional runtime/profiling
    information, such as additional PMU data, LLC miss probabilities or branch mispredictions. This involves (1) building
    a data collection pipeline that covers additional runtime information, (2) modifying the ML models to allow processing
    this data, and (3) modifying the training and inference process for the models to make use this data.
  <p>Today, the models are almost pure static analysis; they see the instructions, but they make one-size-fits-all
    assumptions about the execution environment and the runtime behavior of the code. The goal of this project is to move
    from static analysis towards more dynamic models that better represent code the way it actually executes.</p>
  <p><b>Mentors</b>
    Ondrej Sykora, Mircea Trofin, Aiden Grossman
  </p>
  <p>
  <b>Discourse</b>
  <a href="https://discourse.llvm.org/t/better-performance-models-for-mlgo-training/68219">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang_analyzer_taint_analysis">Improve and Stabilize the Clang Static Analyzer's "Taint Analysis" Checks</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The Clang static analyzer comes with an experimental implementation of
    taint analysis, a security-oriented analysis technique built to warn
    the user about flow of attacker-controlled ("tainted") data into
    sensitive functions that may behave in unexpected and dangerous ways
    if the attacker is able to forge the right input. The programmer can address
    such warnings by properly "sanitizing" the tainted data in order to
    eliminate these dangerous inputs. A common example of a problem that can be
    caught this way is <a href="https://xkcd.com/327/">SQL injections</a>.
    A much simpler example, which is arguably much more relevant to users
    of Clang, is buffer overflow vulnerabilities caused by attacker-controlled
    numbers used as loop bounds while iterating over stack or heap arrays, or
    passed as arguments to low-level buffer manipulating functions such as
    <tt>memcpy()</tt>.
  </p>
  <p>
    Being a static symbolic execution engine, the static analyzer implements
    taint analysis by simply maintaining a list of "symbols" (named unknown
    numeric values) that were obtained from known taint sources during the
    symbolic simulation. Such symbols are then treated as potentially taking
    arbitrary concrete values, as opposed to the general case of taking an
    unknown subset of possible values. For example, division by a unchecked
    unknown value doesn't necessarily warrant a division by zero warning,
    because it's typically not known whether the value can be zero or not.
    However, division by an unchecked <i>tainted</i> value does immediately
    warrant a division by zero warning, because the attacker is free
    to pass zero as an input. Therefore the static analyzer's taint
    infrastructure consists of several parts: there is a mechanism for keeping
    track of tainted symbols in the symbolic program state, there is a way to
    define new sources of taint, and a few path-sensitive checks were taught to
    consume taint information to emit additional warnings (like the division
    by zero checker), acting as taint "sinks" and defining checker-specific
    "sanitization" conditions.
  </p>
  <p>
    The entire facility is flagged as experimental: it's basically a
    proof-of-concept implementation. It's likely that it can be made to work
    really well, but it needs to go through some quality control by running it
    on real-world source code, and a number of bugs need to be addressed,
    especially in individual checks, before we can declare it stable.
    Additionally, the tastiest check of them all – buffer overflow detection
    based on tainted loop bounds or size parameters – was never implemented.
    There is also a related check for array access with tainted index – which
    is, again, experimental; let's see if we can declare this one stable
    as well!
  </p>

  <p><b>Expected result:</b>
    A number of taint-related checks either enabled by default for all users
    of the static analyzer, or available as opt-in for users who care about
    security. They're confirmed to have low false positive rate on real-world
    code. Hopefully, the buffer overflow check is one of them.</p>

  <p><b>Desirable skills:</b>
    Intermediate C++ to be able to understand LLVM code. We'll run our analysis
    on some plain C code as well. Some background in compilers or security is
    welcome but not strictly necessary.
  </p>

  <p><b>Project size:</b> Either medium or large.</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Confirmed Mentors:</b>
  <a href="https://github.com/haoNoQ">Artem Dergachev</a>,
  <a href="https://github.com/xazax-hun">Gábor Horváth</a>,
  <a href="https://github.com/ziqingluo-90">Ziqing Luo</a>
  </p>

  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clang-improve-and-stabilize-the-static-analyzers-taint-analysis-checks/68235">URL</a>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_mlgo_passes_2023">Machine Learning Guided Ordering of Compiler Optimization Passes</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
  This continues the work of GSoC 2020 and <a href="https://summerofcode.withgoogle.com/archive/2021/projects/6411038932598784">2021</a>.

  Developers generally use standard optimization pipelines like -O2 and -O3 to
  optimize their code. Manually crafted heuristics are used to determine which
  optimization passes to select and how to order the execution of those passes.
  However, this process is not tailored for a particular program, or kind
  of program, as it is designed to perform “reasonably well” for any input.

  We want to improve the existing heuristics or replace the heuristics with
  machine learning-based models so that the LLVM compiler can provide a superior
  order of the passes customized per program.

  The last milestone enabled feature extraction, and started investigating training
  a policy for selecting a more appropriate pass pipeline.
  </p>
  <p><b>Project size:</b> either 175 or 350 hr.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Skills:</b> C/C++, some compiler experience. ML experience is a bonus.</p>
  <p><b>Expected outcomes</b>: Pre-trained model selecting the most economical
    optimization pipeline, with no loss in performance; hook-up of model in LLVM;
    (re-)training tool; come up with new optimization sequences through search or learning.</p>
  <p><b>Mentors</b>
    Tarindu Jayatilaka, Mircea Trofin, Johannes Doerfert
  </p>
  <p>
  <b>Discourse</b>
  <a href="https://discourse.llvm.org/t/machine-learning-guided-ordering-of-compiler-optimization-passes/60415">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_code_coverage">Support a hierarchical directory structure in generated coverage html reports</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b><br>
    Clang supports source-based coverage that shows which lines of code are covered by the executed tests
    <a href="https://clang.llvm.org/docs/SourceBasedCodeCoverage.html">[1]</a>.
    It uses llvm-profdata <a href="https://llvm.org/docs/CommandGuide/llvm-profdata.html">[2]</a> and
    llvm-cov <a href="https://llvm.org/docs/CommandGuide/llvm-cov.html">[3]</a> tools to generate coverage reports.
    llvm-cov currently generates a single top-level index HTML file.
    For example, a single top-level directory code coverage report
    <a href="https://lab.llvm.org/coverage/coverage-reports/index.html">[4]</a>
    for LLVM repo is published on a coverage bot.
    Top-level indexing causes rendering scalability issues in large projects,
    such as Fuchsia <a href="https://fuchsia.dev">[5]</a>.
    The goal of this project is to generate a hierarchical directory structure in generated coverage html reports
    to match the directory structure and solve scalability issues.
    Chromium uses its own post-processing tools to show a per-directory hierarchical structure for coverage results
     <a href="https://analysis.chromium.org/coverage/p/chromium">[6]</a>.
    Similarly, Lcov, which is a graphical front-end Gcov<a href="https://gcc.gnu.org/onlinedocs/gcc/Gcov.html">[7]</a>,
    provides a one-level directory structure to display coverage results <a href="https://llvm.org/reports/coverage/index.html">[8]</a>. <br>
    [1] <a href="https://clang.llvm.org/docs/SourceBasedCodeCoverage.html">Source-based code coverage</a><br>
    [2] <a href="https://llvm.org/docs/CommandGuide/llvm-profdata.html">llvm-profdata</a><br>
    [3] <a href="https://llvm.org/docs/CommandGuide/llvm-cov.html">llvm-cov</a><br>
    [4] <a href="https://lab.llvm.org/coverage/coverage-reports/index.html">LLVM coverage reports</a><br>
    [5] <a href="https://fuchsia.dev">Fuchsia</a><br>
    [6] <a href="https://analysis.chromium.org/coverage/p/chromium">Coverage summary for Chromium</a><br>
    [7] <a href="https://gcc.gnu.org/onlinedocs/gcc/Gcov.html">Gcov</a><br>
    [8] <a href="https://llvm.org/reports/coverage/index.html">Lcov coverage reports</a><br>
    [9] <a href="https://github.com/llvm/llvm-project/issues/54711">Issue #54711: Support per-directory index files for HTML coverage report</a></p>
  </p>
  <p><b>Expected result:</b> Implement a support in hierarchical directory structure in generated coverage html reports and show the usage of this feature in LLVM repo code coverage reports.</p>
  <p><b>Project size:</b> Medium or Large</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/gulfemsavrun">Gulfem Savrun Yeniceri</a>
    <a href="https://github.com/petrhosek">Petr Hosek</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/coverage-support-a-hierarchical-directory-structure-in-generated-coverage-html-reports/68239">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_map_value_to_src_expr">Map LLVM values to corresponding source-level expressions</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>

  Developers often use compiler generated remarks and analysis reports to optimize their code. While
  compilers in general are good at including source code positions (i.e line and column numbers) in the
  generated messages, it is useful if these generated messages also include the corresponding source-level
  expressions. The approach used by the LLVM implementation is to use a small set of intrinsic functions
  to define a mapping between LLVM program objects and the source-level expressions. The goal of this
  project is to use the information included within these intrinsic functions to either generate the
  source expression corresponding to LLVM values or to propose and implement solutions to get the same if
  the existing information is insufficient. Optimizing memory accesses in a program is important for
  application performance. We specifically intend to use compiler analysis messages that report
  source-level memory accesses corresponding to the LLVM load/store instructions that inhibit compiler
  optimizations. As an example, we can use this information to report memory access dependences that
  inhibit vectorization.
  </p>

  <p><b>Project size:</b> Medium</p>

  <p><b>Difficulty:</b> Medium</p>

  <p><b>Skills:</b> Intermediate C++, familiarity with LLVM core or willingness to learn the same.</p>

  <p><b>Expected result:</b> Provide an interface which takes an LLVM value and returns a string corresponding
    to the equivalent source-level expression. We are especially interested in using this interface to map
    addresses used in load/store instructions to equivalent source-level memory references.</p>

  <p><b>Confirmed Mentors:</b>
    Satish Guggilla (satish.guggilla@intel.com)
    Karthik Senthil (karthik.senthil@intel.com)
  </p>

  <p>
  <b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/map-llvm-values-to-corresponding-source-level-expressions/68450">URL</a>
  </p>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="clangir">Build and run SingleSource benchmarks using ClangIR</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project:</b><br>
    Clang codegen works by emitting LLVM IR using AST visitors. In the
    <a href="https://llvm.github.io/clangir/">ClangIR</a> project, we emit ClangIR (CIR)
    from AST visitors too (CIRGen), and then lower to (a) LLVM IR directly or, alternatively,
    (b) MLIR in-tree dialects. Lowering to LLVM is still quite immature and lacks many
    instructions, attributes and metadata support.

    ClangIR would greatly benefit from some level of parity with Clang AST →
    LLVM IR codegen quality, in both performance and build time. This is key
    for incrementally bridging correctness and performance testing, providing a
    baseline for future higher level optimizations on top of C/C++.

    A good starting point is to build and run simple benchmarks,
    measuring both generated code and build time performance. LLVM's llvm-test-suite contains scripts and
    machinery that easily allows checking correctness and collecting perf related data and its
    <a href="https://github.com/llvm/llvm-test-suite/tree/main/SingleSource">SingleSource</a>
    collection provide a set of simpler programs to build.

    In a nutshell, while working on this project the student will brigde the
    gap of CIR → LLVM lowering, and at times fix any lacking Clang AST → CIR
    support. The work is going to be done incrementally on top of SingleSource
    benchmarks, while measuring compiler build time and the performance of compiled
    programs.
  </p>

  <p><b>Skills:</b>
    Intermediate C++ programming skills; familiarity with compilers, LLVM IR,
    MLIR or Clang are a big plus, but willingness to learn is also a
    possibility.
  </p>
  <p><b>Expected result:</b>Build and run programs from the SingleSource subdirectory from the
  lvm-test-suite, collect and present results (perf and build time) against regular (upstream) clang codegen.</p>
  <p><b>Project size:</b> Large</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/bcardosolopes">Bruno Cardoso Lopes</a>
    <a href="https://github.com/lanza">Nathan Lanza</a>
  </p>
  <p><b>Discourse:</b> <a href="https://discourse.llvm.org/t/clangir-build-and-run-singlesource-benchmarks-using-clangir/68473">URL</a>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_tblgen_extension">Move additional Enzyme Rules to Tablegen</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Enzyme performs automatic differentiation (in the calculus sense) of LLVM programs. This enables users to use Enzyme to perform various algorithms such as back-propagation in ML or scientific simulation on existing code for any language that lowers to LLVM. The support for an increasing number of LLVM Versions (7-main), AD modes (Reverse, Forward, Forward-Vector, Reverse-Vector, Jacobian), and libraries (BLAS, OpenMP, MPI, CUDA, ROCm, ...) leads to a steadily increasing code base. In order to limit complexity and help new contributors we would like to express more parts of our core logic using LLVM Tablegen. The applicant is free to decide how to best map the program transformation abstractions within Enzyme to Tablegen.
  </p>
  <p><b>Expected results:</b>
     1. Extend the tablegen rule generation system within Enzyme to cover a new component beside of the AdjointGenerator
     <br/>
     2. Moving several existing rules to the new autogenerated system (e.g. LLVM instructions, LLVM intrinsics, MPI calls, ...
     <br/>
   </p>

  <p><b>Confirmed mentor:</b>
    <a href="https://github.com/zuseZ4">Manuel Drehwald</a>
    <a href="mailto:wmoses@mit.edu">William Moses</a>
  </p>
  <p><b>Desirable skills:</b>
    Good knowledge of C++, calculus, and LLVM and/or Clang, and/or MLIR internals. Experience with Tablegen, Enzyme or automatic differentiation would be nice, but can also be learned in the project.
  </p>
  <p><b>Project size:</b> Large</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Discourse</b> <a href="https://discourse.llvm.org/t/enzyme-move-additional-enzyme-rules-to-tablegen/69738">URL</a></p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_patch_coverage">Patch based test coverage for quick test feedback</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
    Most of the day to day tests in LLVM are regression tests executed by <a href="https://llvm.org/docs/CommandGuide/lit.html">Lit</a>, structured as source code or IR to be passed to some binary, rather than test code directly calling the code to be tested.
    This has many advantages but can make it difficult to predict which code path is executed when the compiler is invoked with a certain test input, especially for edge cases where error handling is involved.
    The goal of this project is to help developers create good test coverage for their patch and enable reviewers to verify that they have done so.
    To accomplish this we would like to introduce a tool that can be fed a patch as input, add coverage instrumentation for the affected source files, runs Lit tests, and records which test cases cause each counter to be executed.
    For each counter we can then report the number of test cases executing the counter, but perhaps more importantly we can also report the number of test cases executing the counter that are also changed in some way by the patch, since a modified line that results in the same test results isn’t properly tested, unless it’s intended to be a non-functional change.
    This can be implemented in three separate parts:

    <ol>
      <li>Adding an option to llvm-lit to emit the necessary test coverage data, divided per test case (involves setting a unique value to <a href="https://clang.llvm.org/docs/SourceBasedCodeCoverage.html#running-the-instrumented-program"><code>LLVM_PROFILE_FILE</code></a> for each RUN)
      <li>New tool to process the generated coverage data and the relevant git patch, and present the results in a user friendly manner
      <li>Adding a way to non-intrusively (without changing build configurations) enable coverage instrumentation to a build. By building the project normally, touching the files changed by the patch, and rebuilding with <a href="https://github.com/llvm/llvm-project/blob/93a1fc2e18b452216be70f534da42f7702adbe1d/clang/tools/driver/driver.cpp#L79-L105"><code>CCC_OVERRIDE_OPTIONS</code></a> set to <a href="https://clang.llvm.org/docs/SourceBasedCodeCoverage.html#compiling-with-coverage-enabled">add coverage</a> we can lower the overhead of generating and processing coverage of lines not relevant to the patch.
    </ol>

    The tooling in step 2 and 3 can be made completely agnostic of the actual test-runner, lowering the threshold for other test harnesses than Lit to implement the same functionality.
    If time permits adding this as a step in CI would also be helpful for reviewers.
  </p>

  <p><b>Project size:</b> Small or medium</p>

  <p><b>Difficulty:</b> Simple </p>

  <p><b>Skills:</b> Python for Lit, data processing and <a href="https://www.gnu.org/software/diffutils/manual/html_node/Unified-Format.html">diff</a> processing. No compiler experience necessary. </p>

  <p><b>Expected result:</b> Implement a new tool for use by the community. Developers get help finding uncovered edge cases during development, while also avoiding paranoid sprinkling of asserts or logs just to check that the code is actually executed. Reviewers can more easily check which parts of the patch are tested by each test. </p>

  <p><b>Confirmed Mentors:</b>
    <a href="https://github.com/hnrklssn">Henrik Olsson</a>
  </p>

  <p>
  <b>Discourse:</b>
  <a href="https://discourse.llvm.org/t/coverage-patch-based-test-coverage-for-quick-test-feedback/68628">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc22">Google Summer of Code 2022</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>
    Google Summer of Code 2022 was very successful for LLVM project. For the
    list of accepted and completed projects, please take a look into Google
    Summer of
    Code <a href="https://summerofcode.withgoogle.com/archive/2022/organizations/llvm-compiler-infrastructure">website</a>.
  </p>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>LLVM</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_shared_jitlink">Implement a shared-memory based JITLinkMemoryManager for out-of-process JITting</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Write a shared-memory based JITLinkMemoryManager.
    <br />
    LLVM’s JIT uses the JITLinkMemoryManager interface to allocate both working
    memory (where the JIT fixes up the relocatable objects produced by the
    compiler) and target memory (where the JIT’d code will reside in the target).
    JITLinkMemoryManager instances are also responsible for transporting
    fixed-up code from working memory to target memory. LLVM has an existing
    cross-process allocator that uses remote procedure calls (RPC) to allocate
    and copy bytes to the target process, however a more attractive solution
    (when the JIT and target process share the same physical memory) would be to
    use shared memory pages to avoid copies between processes.

  </p>

  <p><b>Expected results:</b>
    <ul>Implement a shared-memory based JITLinkMemoryManager:
      <li>Write generic LLVM APIs for shared memory allocation.</li>
      <li>
        Write a JITLinkMemoryManager that uses these generic APIs to allocate
        shared working-and-target memory.
      </li>
      <li>Make an extensive performance study of the approach.</li>
    </ul>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/lhames>Lang Hames</a></p>

  <p><b>Desirable skills:</b> Intermediate C++; Understanding of LLVM and the
    LLVM JIT in particular; Understanding of virtual memory management APIs.
  </p>

  <p><b>Project type:</b> Large</p>

  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/implement-a-shared-memory-based-jitlinkmemorymanager-for-out-of-process-jitting">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_build_jit_tutorial">Modernize the LLVM "Building A JIT" tutorial series</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The LLVM BuildingAJIT tutorial series teaches readers to build their own JIT
    class from scratch using LLVM’s ORC APIs, however the tutorial chapters have
    not kept pace with recent API improvements. Bring the existing tutorial
    chapters up to speed, write up a new chapter on lazy compilation (chapter
    code already available) or write a new chapter from scratch.
  </p>

  <p><b>Expected results:</b>
    <ul>
      <li>
        Update chapter text for Chapters 1-3 -- Easy, but offers a chance to get
        up-to-speed on the APIs.
      </li>
      <li>
        Write chapter text for Chapter 4 -- Chapter code is already available,
        but no chapter text exists yet.
      </li>
      <li>
        Write a new chapter from scratch -- E.g. How to write an out-of-process
        JIT, or how to directly manipulate the JIT'd instruction stream using
        the ObjectLinkingLayer::Plugin API.
      </li>
    </ul>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/lhames>Lang Hames</a></p>
  <p><b>Desirable skills:</b> Intermediate C++; Understanding of LLVM and the
    LLVM JIT in particular; Familiarity with RST (reStructed Text); Technical
    writing skills.
  </p>
  <p><b>Project type:</b> Medium</p>

  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/modernize-the-llvm-building-a-jit-tutorial-series">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_jit_new_format">Write JITLink support for a new format/architecture</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    JITLink is LLVM’s new JIT linker API -- the low-level API that transforms
    compiler output (relocatable object files) into ready-to-execute bytes in
    memory. To do this JITLink’s generic linker algorithm needs to be
    specialized to support the target object format (COFF, ELF, MachO), and
    architecture (arm, arm64, i386, x86-64). LLVM already has mature
    implementations of JITLink for MachO/arm64 and MachO/x86-64, and a
    relatively new implementation for ELF/x86-64. Write a JITLink implementation
    for a missing target that interests you. If you choose to implement support
    for a new architecture using the ELF or MachO formats then you will be able
    to re-use the existing generic code for these formats. If you want to
    implement support for a new target using the COFF format then you will need
    to write both the generic COFF support code and the architecture support
    code for your chosen architecture.
  </p>

  <p><b>Expected results:</b>
    Write a JITLink specialization for a not-yet-supported format/architecture.
  </p>
  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/weliveindetail>Stefan Gränitz</a>,
    <a href=https://github.com/lhames>Lang Hames</a>
  </p>
  <p><b>Desirable skills:</b> Intermediate C++; Understanding of LLVM and the
    LLVM JIT in particular; familiarity with your chosen format/architecture,
    and basic linker concepts (e.g. sections, symbols, and relocations).
  </p>
  <p><b>Project type:</b> Large</p>

  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/write-jitlink-support-for-a-new-format-architecture">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_instrumentaion_for_compile_time">Instrumentation of Clang/LLVM for Compile Time</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Every developer, at some point (usually while waiting for their program to
    compile), has asked "Why is it taking so long?"  This project is to seek an
    answer to this question.  There exists within LLVM, and by extension CLANG,
    a timing infrastructure that records events within the compiler.  However,
    its utilization is inconsistent and insufficient.  This can be improved by
    adding more instrumentation throughout LLVM and CLANG but one must be careful.
    Too much instrumentation, or instrumenting the wrong things, can be confusing
    and overwhelming, thus making it no more useful than not enough information.
    The trick is to find the right places to instrument and controlling the
    instrumentation.  Seeking out these key spots will take you through the
    entire compilation process, from preprocessing through to final code
    generation, and all phases between.  As you instrument the code, you will
    look at the data as you evolve it, which will further direct your search.
    You will develop new ways to control and filter the information to allow a
    better understanding of where the compiler is spending its time.  You will
    seek out and develop example test inputs that illustrate where the compiler
    can be improved, which will in turn, help direct your instrumenting and search.
    You will consider and develop ways of controlling the instrumentation to
    allow better understanding and detailed examination of phases of compilation.
    Through all of this, you will gain an understanding of how a compiler works,
    from front end processing, through the LLVM optimization pipeline, through
    to code generation.  You will see, and understand, the big picture of what
    is required to compile and optimize a C/C++ program, and in particular, how
    CLANG, LLVM and LLC accomplish these tasks.  Your mentors have a combined
    experience of approximately 25 years of compiler development and around 8
    years of experience with LLVM itself to help you on your quest.
  </p>

  <p><b>Expected results:</b>
    <ul>
      <li>Targetted expansion of the use of the existing timing infrastructure</li>
      <li>Identification of appropriate test inputs for improving compile time</li>
      <li>Identification of compile time hotspots</li>
      <li>New and improved methods of controlling the timing infrastructure</li>
    </ul>
  </p>
  <p><b>Confirmed Mentor:</b> Jamie Schmeiser, Whitney Tsang</p>
  <p><b>Desirable skills:</b> C++ programming skills; CLANG/LLVM knowledge an asset but not necessary; self motivated; curiosity; desire to learn</p>
  <p><b>Project type:</b>175 or 350 hour</p>
  <p><b>Difficulty Rating:</b>Easy - Medium</p>
  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/instrumentation-of-clang-llvm-for-compile-time">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_mlgo_passes">Machine Learning Guided Ordering of Compiler Optimization Passes</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
  This continues the work of GSoC 2020 and <a href="https://summerofcode.withgoogle.com/archive/2021/projects/6411038932598784">2021</a>.

  Developers generally use standard optimization pipelines like -O2 and -O3 to
  optimize their code. Manually crafted heuristics are used to determine which
  optimization passes to select and how to order the execution of those passes.
  However, this process is not tailored for a particular application, or kind
  of application, as it is designed to perform “reasonably well” for any input.

  We want to improve the existing heuristics or replace the heuristics with
  machine learning-based models so that the LLVM compiler can provide a superior
  order of the passes customized per application.

  The last milestone enabled feature extraction, and started investigating training
  a policy for selecting a more appropriate pass pipeline.
  </p>
  <p><b>Project size:</b> either 175 or 350 hr.</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Skills:</b> C/C++, some compiler experience. ML experience is a bonus.</p>
  <p><b>Expected outcomes</b>: Pre-trained model selecting the most economical
    optimization pipeline, with no loss in performance; hook-up of model in LLVM;
    (re-)training tool.</p>
  <p><b>Mentors</b>
    Tarindu Jayatilaka, Mircea Trofin, Johannes Doerfert
  </p>
  <p>
  <b>Discourse</b>
  <a href="https://discourse.llvm.org/t/machine-learning-guided-ordering-of-compiler-optimization-passes/60415">URL</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_mlgo_loop">Learning Loop Transformation Policies</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
  This project is a continuation of last <a href="https://summerofcode.withgoogle.com/archive/2021/projects/5732097817313280">year’s</a>.
  In 2021, the project achieved its first milestone - separating correctness
  decisions from policy decisions. This opens up the possibility of replacing
  the latter with machine-learned ones.

  Rough milestones: 1) select an initial set of features and use the existing ML
  Guided Optimizations (MLGO) infra to generate training logs; 2) define a reward
  signal, computable at compile time, to guide a reinforcement learning training loop;
  3) iterate through training and refine reward/feature set
  </p>
  <p><b>Project size:</b> either 175 or 350 hr, ideally 350 hr</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Skills:</b> C/C++, some compiler experience. ML experience is a bonus.</p>
  <p><b>Expected outcomes</b>: policy ('advisor') interface for loop unrolling,
    with current heuristic as default implementation; set up feature extraction
    for reinforcement learning training; set up a reward metric; set up training
    algorithm, and iterate over policy training</p>

  <p><b>Mentors</b>
    Johannes Doerfert, Mircea Trofin
  </p>
  <p>
  <b>Discourse</b>
  <a href="https://discourse.llvm.org/t/learning-loop-transformation-policies/60413">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_module_inliner">Evaluate and Expand the Module-Level Inliner</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
    LLVM's inliner is a bottom-up, strongly-connected component-level pass. This
    places limits on the order in which call sites are evaluated, which impacts
    the effectiveness of inlining.

    We now have a functional Module Inliner, as result of <a href="https://summerofcode.withgoogle.com/archive/2021/projects/5195658885070848">GSoC2021 work</a>.
    We want to call site priority schemes, effectiveness/frequency of running
    function passes after successful inlinings, interplay with the ML inline
    advisor, to name a few areas of exploration.
  </p>
  <p><b>Project size:</b> either 175 or 350 hr, ideally 350 hr, milestones allow
    for 175hr scoping</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Skills:</b> C/C++, some compiler experience.</p>
  <p><b>Expected outcomes</b>: Proposal and Evaluation of alternative traversal
    orders; evaluation of 'clustering' inlining decisions (inline more than one
    call site at a time); evaluation of effectiveness/frequency of function
    optimization passes after inlining
  </p>
  <p><b>Mentors</b>
    Kazu Hirata, Liqiang Tao, Mircea Trofin
  </p>
  <p>
  <b>Discourse</b>
  <a href="https://discourse.llvm.org/t/evaluate-and-expand-the-module-level-inliner/60525">URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_lto_dependency_info">Richer symbol dependency information for LTO</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    C and C++ programs are often composed of various object files produced from
    separately-compiled source files that are then linked together.
    When compiling one source file, knowledge that can be derived from the logic
    contained within the other source files would normally not be available.
    Link-time optimization, also known as LTO, is a way for optimization to be
    done using information from more than one source file.
  </p>
  <p>In LLVM, LTO is achieved by using LLVM bitcode objects as the output from
    the "compile" step and feeding those objects into the link step.
    LLVM's LTO operates in conjunction with the linker.
    The linker is invoked by the user and the linker in turn drives LLVM's LTO
    when it encounters LLVM bitcode files, getting information from LTO about
    what symbols a bitcode object defines or references.
    Information about what symbols are defined in or referenced from an object
    is necessary for the linker to perform symbol resolution, and a linker is
    normally able to extract such information from regular (non-bitcode) object
    files.
  </p>
  <p>The implied consequences of LLVM's LTO implementation
    with respect to linker GC
    (linker garbage collection) can be improved, especially for aggressive forms
    of linker GC with lazy inclusion of objects and sections.
    In particular, the symbols referenced but undefined by an LTO module are,
    to the linker, monolithic at the module level.
    At the same time, the symbols referenced but undefined by regular
    (non-LTO) objects are monolithic to LTO.
    Together, this means that the inclusion of an LTO module
    into the overall process potentially leads, in the linker's initial symbol
    resolution, to all the undefined symbols in that module being considered as
    referenced; in turn, additional artifacts (e.g., archive members) may be
    added into the resolution, which further leads to references that may
    resolve to symbols defined in LTO modules and a premature conclusion that
    the definition of these symbols are needed.
    This at least means potentially unnecessary codegen is being done for
    functions that will be garbage-collected in the end (waste of electricity
    and time).
  </p>
  <p>We acknowledge that an ideal implementation probably involves a "coroutine"
    like interaction between the linker and LTO codegen where information flows
    back and forth; however, such an endeavour is invasive to both linkers and
    to LLVM.
  </p>
  <p>We believe that by</p>
  <ul>
    <li>having the linker register, via an API to LTO, symbol reference "nodes"
    modelling the relationship between a symbol and the symbols that are
    referenced in turn from (the object file section containing) its
    linker-selected definition, and
    </li>
    <li>using that information in LTO processing,</li>
  </ul>
  <p>the LTO processing will be able to effectively identify a more accurate set
    of LTO symbols that are visible outside of the LTO unit.
    The linker merely needs to identify only exported symbols and entry points
    (such as the entry point for an executable and functions involved in
    initialization and finalization).
  </p>
  <p>Having the LLVM opt/codegen understand the dependency implications from the
    "outside world" is strictly better than the other direction: the symbols
    referred to by relocations in non-LTO code are pretty much fixed as compiled
    (whereas some references in LTO code may disappear with optimization).
  </p>

  <p><b>Expected results:</b></p>
  <ol>
    <li>Modification of the C++ LTO interface used by LLD to implement an
      interface to record the symbol reference dependency data (incorporating
      awareness of sections and comdats). This may additionally include a
      method to add LTO objects provisionally, simulating behaviours where
      linkers only add objects as needed.
    </li>
    <li>
      Modification of LTO to use new symbol reference information
      for definitions in regular objects when visiting definitions
      in the IR prior to the
      internalization pass to discover (transitive) symbol references and
      record the so-referenced symbols as being visible to regular objects.
      This may additionally include the "late" incorporation of LTO objects
      added provisionally into the merged LTO module.
    </li>
    <li>Modification of LLD (for ELF) to modify initial resolution to use the
      new interface as a replacement for setting
      <code>VisibleToRegularObj</code>
      except for entry point functions (including C++ dynamic initialization and
      finalization).
    </li>
  </ol>

  <p><b>Confirmed Mentors:</b>
    Sean Fertile,
    Hubert Tong,
    Wael Yehia
  </p>

  <p><b>Desirable skills:</b>
    Intermediate C++;
    basic linker concepts (e.g., symbols, sections, and relocations)
  </p>

  <p><b>Project size:</b> 350 hours</p>

  <p><b>Difficultly:</b> Medium/Hard</p>

  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/richer-symbol-dependency-information-for-lto/60335">
      URL</a>
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_undef_load">Remove undef: move uninitialized memory to poison</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
    The existence of the undef value in LLVM prevents several optimizations,
    even in programs where it is not used. Therefore, we have been trying to
    move all uses of undef to poison so we can eventually remove undef from
    LLVM.<br/>
    This project focuses on uninitialized memory: right now the semantics of
    LLVM is that loading a value from uninitilized memory yields an undef value.
    This prevents, for example, SROA/mem2reg from optimizing conditional loads
    as phi(undef, %x) cannot be replaced with x, as %x might be poison.<br/>
    This project consists in devising a consistent semantics for uninitialized
    (based on existing proposals), an upgrade plan for LLVM, and implementing
    the changes in LLVM and clang.
    In clang the changes should be specific to bit-fields.<br/>
    For more information see the following
    <a href="https://github.com/llvm/llvm-project/issues/52930">discussion</a>
    and/or contact the mentor.<br/>
    Further reading:
    <a href="https://web.ist.utl.pt/nuno.lopes/pubs/llvmmem-oopsla18.pdf">introduction to LLVM's memory model</a>.
  </p>
  <p><b>Project size:</b> 350 hr</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Skills:</b> Intermediate C++</p>
  <p><b>Expected outcomes</b>:
    <ul>
      <li>Semantics for memory operations that removes the need for undef
        values</li>
      <li>Upgrade plan for LLVM and frontends</li>
      <li>Implementation of the proposed semantics in LLVM</li>
      <li>Implementation of auto-upgrade path for old LLVM IR files</li>
      <li>Implementation of fixes in clang to use the new IR features</li>
      <li>Benchmarking to check for regressions and/or perf improvements</li>
    </ul>
  </p>
  <p><b>Mentors:</b>
    <a href="https://web.ist.utl.pt/nuno.lopes/">Nuno Lopes</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a id="llvm_abi_export">Add API/ABI export annotations to the LLVM build</a>
</div>
<!-- *********************************************************************** -->
<div class="www_text">
  <p><b>Description of the project</b>
<p>Currently, all libraries inside LLVM export all their symbols publicly. When
linking statically against them, the linker will remove unused symbols and this
is not a problem.</p>

<p>When the libraries are built as shared libraries however, the number of exported
symbols is very large and symbols that are meant to be internal spill into the
public ABI of the shared libLLVM.so.</p>

<p>In this project, we’d like to change the default visibility of library symbols
to “hidden”, add an annotation macro to LLVM and use the macro to gradually move
the entire library in this direction. This will eventually enable building the
shared libLLVM.so on Windows as well.<p>

<p>In practice, this means adding -fvisibility=hidden to individual libraries and
annotating exported symbols with the LLVM export annotation.</p>

<p>We would like this work to be as unintrusive into other developer’s workflow as
possible, so starting with a small internal library would be beneficial,
e.g. one of the LLVM targets or IR passes.</p>

<p>For further reading, there is a Discourse thread avaiable that discusses the
idea behind this proposal:
<a href="https://discourse.llvm.org/t/supporting-llvm-build-llvm-dylib-on-windows/58891">
  Supporting LLVM_BUILD_LLVM_DYLIB on Windows</a>
as well as the linked Phabricator review with a patch implementing the functionality:
<a href="https://reviews.llvm.org/D109192">⚙ D109192 [WIP/DNM] Support:
  introduce public API annotation support</a>
None of this work has been committed yet but can be used as a starting point
for this proposal.</p>
  </p>
  <p><b>Project size:</b> Medium</p>
  <p><b>Difficulty:</b> Easy</p>
  <p><b>Skills:</b> Build systems, CMake, LLVM</p>
  <p><b>Expected outcomes</b>:
    <ul>
      <li>Export macro implemented and commited to LLVM</li>
      <li>At least one internal target ported to the new export scheme</li>
    </ul>
  </p>
  <p><b>Mentors:</b>
    Timm Bäder, Tom Stellard
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>Clang</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-template-instantiation-sugar">Extend clang AST to provide
    information for the type as written in template instantiations.</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b>
    When instantiating a template, the template arguments are canonicalized
    before being substituted into the template pattern. Clang does not preserve
    type sugar when subsequently accessing members of the instantiation.

    <pre>
    std::vector&lt;std::string&gt; vs;
    int n = vs.front(); // bad diagnostic: [...] aka 'std::basic_string&lt;char&gt;' [...]

    template&lt;typename T&gt; struct Id { typedef T type; };
    Id&lt;size_t&gt;::type // just 'unsigned long', 'size_t' sugar has been lost
    </pre>

    Clang should "re-sugar" the type when performing member access on a class
    template specialization, based on the type sugar of the accessed
    specialization. The type of vs.front() should be std::string, not
    std::basic_string&lt;char, [...]&gt;.
    <br /> <br />
    Suggested design approach: add a new type node to represent template
    argument sugar, and implicitly create an instance of this node whenever a
    member of a class template specialization is accessed. When performing a
    single-step desugar of this node, lazily create the desugared representation
    by propagating the sugared template arguments onto inner type nodes (and in
    particular, replacing Subst*Parm nodes with the corresponding sugar). When
    printing the type for diagnostic purposes, use the annotated type sugar to
    print the type as originally written.
    <br /> <br />
    For good results, template argument deduction will also need to be able to
    deduce type sugar (and reconcile cases where the same type is deduced twice
    with different sugar).
  </p>

  <p><b>Expected results: </b>
    Diagnostics preserve type sugar even when accessing members of a template
    specialization. T&lt;unsigned long&gt; and T&lt;size_t&gt; are still the
    same type and the same template instantiation, but
    T&lt;unsigned long&gt;::type single-step desugars to 'unsigned long' and
    T&lt;size_t&gt;::type single-step desugars to 'size_t'.</p>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/vgvassilev>Vassil Vassilev</a>,
    <a href=https://github.com/zygoloid>Richard Smith</a></p>

  <p><b>Desirable skills:</b>
    Good knowledge of clang API, clang's AST, intermediate knowledge of C++.
  </p>
  <p><b>Project type:</b> Large</p>

  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/clang-extend-clang-ast-to-provide-information-for-the-type-as-written-in-template-instantiations">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-sa-structured-bindings">Implement support for
    C++17 structured bindings in the Clang Static Analyzer</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b>
    Even though a lot of new C++ features are supported by the static analyzer
    automatically by the virtue of clang AST doing all the work under the hood,
    the C++17 "structured binding" syntax
    <pre>    auto [x, y] = ...;</pre>
    requires some extra work on the Static Analyzer side. The analyzer's transfer functions
    need to be taught about the new AST nodes, <a href="https://clang.llvm.org/doxygen/classclang_1_1BindingDecl.html">BindingDecl</a>
    and <a href="https://clang.llvm.org/doxygen/classclang_1_1DecompositionDecl.html">DecompositionDecl</a>,
    to work correctly in all <a href="https://en.cppreference.com/w/cpp/language/structured_binding">three interpretations</a>
    described by the Standard.
    <br /><br />
    Incomplete support for structured bindings is a common source of
    false positives in the uninitialized variable checker on modern C++ code,
    such as <a href="https://github.com/llvm/llvm-project/issues/42387">#42387</a>.
    <br /><br />
    It is likely that the Clang CFG also needs to be updated. Such changes in
    the CFG may improve quality of clang warnings outside of
    the Static Analyzer.
  </p>

  <p><b>Expected results: </b>
    The Static Analyzer correctly models structured binding and decomposition
    declarations. In particular, binding variables no longer appear
    uninitialized to the Static Analyzer's uninitialized variable checker.</p>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/haoNoQ>Artem Dergachev</a>,
    <a href=https://github.com/t-rasmud>Rashmi Mudduluru</a>,
    <a href=https://github.com/xazax-hun>Gábor Horváth</a>,
    <a href=https://github.com/Szelethus>Kristóf Umann</a>
  </p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++. Some familiarity with Clang AST and/or
    some static analysis background.
  </p>
  <p><b>Project size:</b> 350 hr</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/implement-support-for-c-17-structured-bindings-in-the-clang-static-analyzer/60588">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-improve-diagnostics">Improve Clang Diagnostics.</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description:</b>
  Clang Diagnostics, which issues Warnings and Errors to the programmer, are a critical
  feature of the compiler. Great diagnostics can have a significant impact on the
  user experience of the compiler and increase their productivity.
  </p>

  <p>Recent improvements in GCC
  <a href="https://developers.redhat.com/blog/2018/03/15/gcc-8-usability-improvements"> [1] </a>
  <a href="https://developers.redhat.com/blog/2019/03/08/usability-improvements-in-gcc-9/"> [2] </a>
  shows that there is significant headroom to improve diagnostics
  (and user interactions in general). It would be a very impactful project
  to survey and identify all the possible improvements to clang on this
  topic and start redesigning the next generation of our diagnostics.
  </p>

  <p>
  In addition, we will also make conclusions on issues reported on the LLVM Github Issue page labeled
  with <a href="https://github.com/llvm/llvm-project/labels/clang%3Adiagnostics"> clang-diagnostics</a>
  and if they need fixing, we will prepare patches otherwise simply close them.
  </p>

  <p><b>Expected outcomes</b>:
  Diagnostics will be improved:
    <ul>
      <li>Improve diagnostic aesthetics</li>
      <li>Cover missing diagnostics</li>
      <li>Reduce false positive rate</li>
      <li>Reword diagnostics</li>
    </ul>
  </p>

  <p><b>Confirmed Mentor:</b>
    <a href=https://github.com/AaronBallman>Aaron Ballman</a>,
    <a href=https://github.com/erichkeane>Erich Keane</a>,
    <a href=https://github.com/xgupta>Shivam Gupta</a></p></p>

  <p><b>Desirable skills:</b> C++ coding experience</p>
  <p><b>Project type:</b> Large/350 hr</p>

  <p><b>Discourse</b>
    <a href="https://discourse.llvm.org/t/improve-clang-diagnostics/61521">
      URL</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>Polly</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="polly_npm">Complete switch to new pass manager</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the Project:</b>
    While the standard Polly-enabled -O1/-O2/-O3 optimization pass pipelines work fine with the <a href="https://blog.llvm.org/posts/2021-03-26-the-new-pass-manager/">New Pass Manager</a> (NPM), some parts of Polly still only works with the legacy pass manager.
    This includes some passes such as -polly-export-jscop/-polly-export-jscop, regression testing, Polly-ACC, command line options such as -polly-show, the PassInstrumentation mechanism used by e.g. -print-after-all.
    LLVM (and Clang) have moved to NPM being the default and support for the legacy pass manager is deprecated, slowly degenerates and features getting removed.
    That is, all of Polly's functionality should eventually work with the NPM as well, and be prepared for the complete removal of the legacy pass manager.
    More details about the two pass managers found <a href="https://github.com/banach-space/llvm-tutor#about-pass-managers-in-llvm">here</a>.
  </p>
  <p><b>Expected results:</b>
    The goal is to make Polly more usable with using only the NPM. Milestones, not necessarily all to be reached in this GSoC, are:
    <br/>
    1. Make all of Polly's functionality available in the NPM (or decide to deprecate/remove it)
    <br/>
    2. Better integration into the NPM (such as supporting PassInstrumentation); If the NPM turns out to be inadequate, use only a monolothic function pass.
    <br/>
    3. Replace the legacy pass manager in regression tests.
    <br/>
    4. Be ready for complete removal of the legacy pass manager in LLVM.
  </p>
  <p><b>Confirmed mentor:</b>
    <a href="https://github.com/Meinersbur">Michael Kruse</a>
  </p>
  <p><b>Desirable skills:</b>
    Understanding of the C++ template pattern used by the new pass manager (<a href="https://en.wikipedia.org/wiki/Curiously_recurring_template_pattern">CRTP</a>, <a href="https://en.wikipedia.org/wiki/Mixin">Mixins</a>, etc).
    Familarity with how LLVM can be <a href="https://www.lurklurk.org/linkers/linkers.html">linked</a> (static, BUILD_SHARED_LIBS, and SHLIB/DYLIB) and its <a href="https://www.llvm.org/docs/WritingAnLLVMPass.html#building-pass-plugins">plugin loading machanisms</a> (static, -load and -load-pass-plugin).
    Ideally, already worked with LLVM's new pass manager.
  </p>
  <p><b>Project size:</b> Medium</p>
  <p><b>Difficulty:</b> Medium/Hard</p>
  <p><b>Discourse</b> <a href="https://discourse.llvm.org/t/61174">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>Enzyme</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_tblgen">Move Enzyme Instruction Transformation Rules to Tablegen</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Enzyme performs automatic differentiation (in the calculus sense) of LLVM programs. This enables users to use Enzyme to perform various algorithms such as back-propagation in ML or scientific simulation on existing code for any language that lowers to LLVM. The support for an increasing number of LLVM Versions (7-main), AD modes (Reverse, Forward, Forward-Vector, Reverse-Vector, Jacobian), and libraries (BLAS, OpenMP, MPI, CUDA, ROCm, ...) leads to a steadily increasing code base. In order to limit complexity and help new contributors we would like to express our core logic using LLVM Tablegen. The applicant is free to decide how to best map the program transformation abstractions within Enzyme to Tablegen.
  </p>
  <p><b>Expected results:</b>
     1. A working tablegen rule generation system within Enzyme
     <br/>
     2. Moving several existing rules to the new autogenerated system (e.g. LLVM instructions, LLVM intrinsics, BLAS calls, MPI calls, ...
     <br/>
   </p>

  <p><b>Confirmed mentor:</b>
    <a href="mailto:wmoses@mit.edu">William Moses</a>, Valentin Churavy
  </p>
  <p><b>Desirable skills:</b>
    Good knowledge of C++, calculus, and LLVM and/or Clang, and/or MLIR internals. Experience with Tablegen, Enzyme or automatic differentiation would be nice, but can also be learned in the project.
  </p>
  <p><b>Project size:</b> Large</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Discourse</b> <a href="https://discourse.llvm.org/t/enzyme-moving-instruction-rules-to-tablegen/61176">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_vector">Vector Reverse-Mode Automatic Differentiation</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Enzyme performs automatic differentiation (in the calculus sense) of LLVM programs. This enables users to use Enzyme to perform various algorithms such as back-propagation in ML or scientific simulation on existing code for any language that lowers to LLVM. Enzyme already implements forward and reverse mode automatic differentiation. Enzyme also implements vector forward mode automatic differentiation, which allows Enzyme to batch the derivative computation of several objects in a single call. The goal of this project is too extend this capability in order to perform vector reverse mode. In doing so, multiple sweeps of reverse mode automatic differentiation can be performed at the same time, reducing memory, time, and otherwise generally enabling further optimization.
  </p>
  <p><b>Expected results:</b>
     Vectorized version of reverse mode automatic differentiation
   </p>

  <p><b>Confirmed mentor:</b>
    <a href="mailto:wmoses@mit.edu">William Moses</a>, Tim Gymnich
  </p>
  <p><b>Desirable skills:</b>
     Good knowledge of C++ and some experience with LLVM API's. Experience with Enzyme or automatic differentiation would be nice, but can also be learned in the project.
  </p>
  <p><b>Project size:</b> Medium</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Discourse</b> <a href="https://discourse.llvm.org/t/enzyme-vector-reverse-mode-automatic-differentiation/61177">URL</a></p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_pm">Enable The New Pass Manager</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Enzyme is a compiler plugin for LLVM that performs automatic differentiation (in the calculus sense) of LLVM programs. This enables users to use Enzyme to perform various algorithms such as back-propagation in ML or scientific simulation on existing code for any language that lowers to LLVM.
    <br/>
    Enzyme integrates into frontends through the use of an LLVM plugin that can be loaded into Clang, LLVM (opt), the linker (lld), libraries (HIPRtc), directly loaded (Julia), among others (Flang, Rust, etc).
    <br/>
    While using various pieces of machinery from the new pass manager internally, Enzyme does not currently automatically register its transformation passes when using the new pass manager. This creates problems for users on LLVM 13 or above, where the new pass manager is run by default and may not understand why they get linker errors from their code not being differentiated (currently they must add a flag to specify the old pass manager).
    <br/>
    The goal of this project is to enable Enzyme to be called by the new pass manager in LLVM and generally create a coherent user experience.
    <br/>
  </p>

  </p>
  <p><b>Expected results:</b>
      1. Enzyme can be called by the new pass manager
      <br/>
      2. [Optional] Additional syntactic sugar that makes it easier to use Enzyme.
      <br/>
   </p>

  <p><b>Confirmed mentor:</b>
    <a href="mailto:wmoses@mit.edu">William Moses</a>, Valentin Churavy
  </p>
  <p><b>Desirable skills:</b>
     Good knowledge of C++, and LLVM. Experience with Enzyme would be nice, but can also be learned in the project.
  </p>
  <p><b>Project size:</b> Small</p>
  <p><b>Difficulty:</b> Medium</p>
  <p><b>Discourse</b> <a href="https://discourse.llvm.org/t/enzyme-enable-the-new-pass-manager/61178">URL</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc21">Google Summer of Code 2021</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>
    Welcome prospective Google Summer of Code 2021 Students! This document is your
    starting point to finding interesting and important projects for LLVM, Clang,
    and other related sub-projects. This list of projects is not only developed for
    Google Summer of Code, but open projects that really need developers to work on
    and are very beneficial for the LLVM community. </p>

  <p>We encourage you to look through this list and see which projects excite you
    and match well with your skill set. We also invite proposals not on this
    list. You must propose your idea to the LLVM community through our
    developers' mailing list (llvm-dev@lists.llvm.org or specific subproject mailing
    list). Feedback from the community is a requirement for your proposal to be
    considered and hopefully accepted.
  </p>

  <p>The LLVM project has participated in Google Summer of Code for several years
    and has had some very successful projects. We hope that this year is no
    different and look forward to hearing your proposals. For information on how to
    submit a proposal, please visit the Google Summer of Code
    main <a href="https://developers.google.com/open-source/gsoc/">website.</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>LLVM</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_distributing_lit">Distributed lit testing</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The LLVM lit test suites consist of thousands of small independent tests.
    Due to the number of tests, it can take a long time to run the full suite,
    even on a high-spec computer. Builds are already distributable across
    multiple computers available on the same network, using software such as
    distcc or icecream, so running tests on a single machine becomes a potential
    bottleneck. One way to speed up running of the tests could be to distribute
    test execution across many computers too. Lit provides a test sharding
    mechanism, which allows multiple computers to run parts of the same
    testsuite in tandem, but this currently assumes access to a single common
    filesystem, which may not be possible in all cases and a knowledge of which
    machines the suite can currently be run on.

    This project’s goal is to update the existing lit harness (or write a
    wrapper around it) to allow distribution of the tests in this way, with the
    idea that developers can write their own interface between the harness and
    the distribution system of their choice. This harness may need to be able to
    identify test dependencies such as input files and executables, send the
    tests to the distribution system (possibly in batches), and receive, collate
    and report the results to the user, in a similar manner to how lit already
    does.
  </p>

  <p><b>Expected results:</b> An easy to use harness as described above. Some
    evidence that given a distributed system, a user can expect to see test
    suite execution to speed up if they are using that harness.</p>

  <p><b>Confirmed mentor:</b> James Henderson</p>
  <p><b>Desirable skills:</b> Good knowledge of Python. Familiarity with LLVM
    lit testing. Some knowledge of distribution systems would also be
    beneficial.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_loop_heuristics">Learning Loop Transformation Heuristics</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    This is a short description, please reach out to Johannes (jdoerfert on
    IRC) and Mircea Trofin if it sounds interesting.

    We successfully introduced an ML framework for inliner decisions, now we want
    to expand the scope. In this project we will look at loop transformation
    heuristics, such as the unroll factor. As a motivational example we can look
    at a small trip count <a href="https://godbolt.org/z/Eeqcvs">dgemm</a> which
    we optimize pretty poorly. With the nounroll pragmas we do a better job but
    still not close to gcc.

    The project is open-ended and we could look at various passes/heuristics
    concurrently.
  </p>

  <p><b>Preparation resources:</b> The ML inliner framework in the LLVM code
  base as well as the <a href="https://arxiv.org/abs/2101.04808">paper</a>. LLVM
  transform passes (that are based on heuristics), e.g., loop unroll.</p>

  <p><b>Expected results:</b> Measurable better performance with a learned
  predictor, potentially a set of "classical" heuristics derived from the ML
  model.</p>

  <p><b>Confirmed Mentor:</b> Johannes Doerfert, Mircea Trofin</p>
  <p><b>Desirable skills:</b> Intermediate knowledge of ML, C++, self motivation.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_ir_fuzzing">Fuzzing LLVM-IR Passes</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    This is a short description, please reach out to Johannes (jdoerfert on
    IRC) if it sounds interesting.

    Fuzzing often reveals a myriad of bugs. CSmith (and others) showed how to do
    this with C-like languages and we have used <a
    href="https://www.youtube.com/watch?v=UBbQ_s6hNgg">LLVM-IR fuzzing</a> in
    the past successfully. In this project we will apply fuzzing to new passes
    that are in development, e.g., the Attributor pass. We want to find and fix
    crashes but also other bugs, including compile time performance problems.
  </p>

  <p><b>Preparation resources:</b> The <a
    href="https://llvm.org/docs/FuzzingLLVM.html#llvm-opt-fuzzer">LLVM fuzzer
    infrastructure</a>. LLVM passes that we might want to fuzz, e.g. the
  Attributor pass. Prior IR-Fuzzing work
  (https://www.youtube.com/watch?v=UBbQ_s6hNgg)</p>


  <p><b>Expected results:</b> Crashes, maybe also a way to catch non-crash
  bugs, including performance problems.</p>

  <p><b>Confirmed Mentor:</b> Johannes Doerfert</p>
  <p><b>Desirable skills:</b> Intermediate knowledge C++, self motivation.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_ir_assume"><tt>llvm.assume</tt> the missing pieces</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    This is a short description, please reach out to Johannes (jdoerfert on
    IRC) if it sounds interesting.

    <tt>llvm.assume</tt> is a powerful mechanism to retain knowledge. Since it
    inception it was improved already multiple times but there are major
    extensions still outstanding which we want to tackled in this project.
    An incomplete list of topics includes:
    <ul>
      <li> range-based assumptions, design idea 3) in the <a href="https://lists.llvm.org/pipermail/llvm-dev/2019-December/137632.html">RFC</a>. </li>
      <li> outline arbitrary assumption/assertion code, design idea 2) in the <a href="https://lists.llvm.org/pipermail/llvm-dev/2019-December/137632.html">RFC</a>. </li>
      <li> side-effect free assumptions, see <a href="https://reviews.llvm.org/D89054">this review</a>. </li>
      <li> more knowledge retention usages </li>
      <li> less interference with optimizations </li>
    </ul>

  </p>

  <p><b>Preparation resources:</b> The llvm.assumption usage, the assumption
  cache, the "enable-knowledge-retention" option, the <a
  href="https://lists.llvm.org/pipermail/llvm-dev/2019-December/137632.html">RFC</a>
  and <a href="https://reviews.llvm.org/D89054">this review</a>.
  </p>


  <p><b>Expected results:</b> New llvm.assume use cases, improved performance through knowledge retention, optimization based on assertions.</p>

  <p><b>Confirmed Mentor:</b> Johannes Doerfert</p>
  <p><b>Desirable skills:</b> Intermediate knowledge C++, self motivation.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_ir_issues">Fix fundamental issues in LLVM's IR</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    LLVM's IR has fundamental, long-standing issues. Many are related with
    undefined behaviors. Others are simply a fallout from underspecification
    and different interpretations by diffferent people.
    <a href="https://github.com/AliveToolkit/alive2">Alive2</a> is a tool that
    detects bugs in LLVM's optimizations automatically. Using Alive2, we track
    bugs exposed by the unit tests on a
    <a href="https://web.ist.utl.pt/nuno.lopes/alive2/">dashboard</a>.
  </p>

  <p><b>Expected results:</b>
    1) Report and fix bugs detected by Alive2.
    2) Pick one fundamental IR issue and
    make progress towards fixing it, including proposing fixes for the
    <a href="https://llvm.org/docs/LangRef.html">semantics</a>, testing
    fixes to the semantics by running Alive2 over the LLVM unit tests and
    medium-sized programs, test performance of semantic fixes and fix
    performance regressions.
  </p>
  <p><b>Confirmed Mentor:</b> Nuno Lopes, Juneyoung Lee</p>
  <p><b>Desirable skills:</b> Intermediate C++; willingness to learn about LLVM
    IR semantics; experience reading papers (preferred).
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_utilize_loopnest">Utilize LoopNest Pass</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    The idea of LoopNest pass is recently added, and there are no existing
    passes utilizing it. Before having LoopNest pass, if you want to write a
    pass that works on a loop nest, you have to pick from either a function
    pass or a loop pass. If you chose to write it as a function pass, then you
    lose the ability to add loops dynamically back to the pipeline. If you
    decide to write it as a loop pass, then you are wasting compile time to
    traverse to your pass and return right away when the given loop is not the
    outermost loop. In this project, we want to utilize the recently introduced
    LoopNest pass for passes intended for loop nest and have the same ability
    as the LoopPass to dynamically add loops to the pipeline. In addition,
    improve the current implementation of LoopNestPass when necessary.
  </p>
  <p><b>Expected results (possibilities):</b>
    Utilize LoopNest Pass for some existing transformations/analyses.
  </p>
  <p><b>Confirmed Mentors:</b>
    Whitney Tsang, Ettore Tiotto
  </p>
  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++, self-motivation.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="openmp_gpu_jit">JIT-ing OpenMP GPU kernels transparently</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    This is a short description, please reach out to Johannes (jdoerfert on
    IRC) if it sounds interesting.

    OpenMP GPU kernels are usually lowered to native binaries, e.g., cubin, and
    embedded into the host object. At runtime, OpenMP "plugins" will connect with
    the device driver, e.g., CUDA, to load and run such embedded binary images.
    In this project we want to develop a new plugin that takes LLVM-IR code, optimizes
    the IR with kernel parameters known only at runtime, and then generates the GPU
    binary for consumption by other plugins. Similar to the <a
      href=https://openmp.llvm.org/docs/design/Runtimes.html#remote-offloading-plugin>remote
      offload plugin</a> we can do this transparently to the user. In addition to the JIT
    infrastructure setup in the plugin we will need to embed the IR into the host object.

  </p>

  <p><b>Preparation resources:</b>OpenMP target offloading infrastructure, LLVM JIT infrastructure</a>.
  </p>


  <p><b>Expected results:</b> A JIT-capable offload plugin which can achieve superior performance when kernel specialization is enabling optimizations.</p>

  <p><b>Confirmed Mentor:</b> Johannes Doerfert</p>
  <p><b>Desirable skills:</b> Intermediate knowledge C++, JIT compilation, self motivation.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>OpenACC</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="openacc_rt_diagnostics">OpenACC Diagnostics from the OpenMP Runtime</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Clacc and Flacc are projects to introduce OpenACC support to Clang and
    Flang.  For that purpose, OpenACC runtime support is being developed on top
    of LLVM's OpenMP runtime.  However, diagnostics emitted by LLVM's OpenMP
    runtime are expressed in terms of OpenMP concepts, and so those diagnostics
    are not always meaningful to OpenACC users.  This project should address
    this issue in two steps:
    <ol>
      <li>
        Develop a mechanism that selects OpenACC versions of diagnostics that
        are emitted as a result of OpenACC-related calls into the runtime.  This
        mechanism should be general enough that it could be used for programming
        languages besides OpenMP and OpenACC.  One possible approach is to
        extend internationalization mechanisms already present in some
        components of the OpenMP runtime.
      </li>
      <li>
        Provide OpenACC translations for existing OpenMP diagnostics.  This step
        requires an understanding of the relationship between OpenACC and OpenMP
        as implemented in Clacc and Flacc.
      </li>
    </ol>
    Many components of OpenACC support that will depend upon this project have
    not yet been upstreamed and are under development.  A high-level
    understanding of those efforts is helpful for this project and can be
    provided by the mentors.  Nevertheless, this project can be completed in
    upstream LLVM's OpenMP runtime now independently of those efforts.
  </p>

  <p>
    <b>Expected results:</b> A version of upstream LLVM's OpenMP runtime that
    can emit OpenACC diagnostics as needed.
  </p>

  <p><b>Confirmed Mentors:</b> Valentin Clement, Joel E. Denny</p>

  <p>
    <b>Desirable skills:</b> Intermediate C++; Experience with OpenACC or
    OpenMP
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>Polly</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="polly_isl_bindings">Use official isl C++ bindings</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Polly use algorithms from the
    <a href="http://isl.gforge.inria.fr/">Integer Set Library (isl)</a>, which is a
    library written in C. It uses reference-counting for memory management.
    Getting reference counting correct is much easier in C++ using RAII,
    therefore we created a C++ binding for isl:
    <a href="https://github.com/llvm/llvm-project/blob/main/polly/lib/External/isl/include/isl/isl-noexceptions.h">isl-noexceptions.h</a>.
    Since then, isl also gained two official C++ bindings,
    <a href="https://github.com/llvm/llvm-project/blob/main/polly/lib/External/isl/include/isl/cpp.h">cpp.h</a>
    and
    <a href="https://github.com/llvm/llvm-project/blob/main/polly/lib/External/isl/include/isl/cpp-checked.h">cpp-checked.h</a>.

    We would like to replace the Polly-maintained C++ bindings with the upstream
    bindings. Unfortunately, this is not an in-place replacement. Differences
    include how errors are checked, method names, which functions are
    considered as operator/constructor overloads and the set of exported functions.
    This will require changing Polly's uses of the C++ bindings and submitting
    patches to isl to export additional functionality needed by Polly.
  </p>

  <p><b>Expected results:</b>
    Reduce the differences between the Polly-maintained isl-noexceptions.h
    bindings and one of the two C++ bindings that isl supports. Due to
    isl-noexceptions.h exporting more functions and classes than the upstream
    bindings do, a complete replacement will probably be out of reach, but
    even reducing the differences will reduce the maintenance cost of Polly's
    isl-noexceptions.h.
  </p>

  <p><b>Confirmed mentor:</b> Michael Kruse</p>
  <p><b>Desirable skills:</b>
    Deep knowledge of C++, in particular RAII and move-semantics. Interest in API design. Ideally, you already wrote some library's header file.
    Experience with the isl library would be nice, but can also be learned in the project.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>Enzyme</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_blas">Integrate custom derivatives of BLAS, Eigen, and similar routines into Enzyme</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <a href="https://enzyme.mit.edu/">Enzyme</a> performs automatic differentiation
    (in the calculus sense) of LLVM programs. This enables users to use Enzyme to
    perform various algorithms such as back-propagation in ML
    or scientific simulation on existing code for any language that lowers to LLVM.

    Enzyme does so by applying the chain rule to every instruction in every
    function called by the original function to be differentiated. While functional,
    this is not necessarily optimal for high-level matrix operations which may
    have algebraic properties for faster derivative computation.

    Enzyme also has a mechanism for specifying a custom gradient for a given function.
    If a custom derivative is available, Enzyme will use that rather than fallback
    to implementing its own.

    Many programs use BLAS libraries to efficiently compute matrix and tensor
    operations. This project would enable high-performance automatic differentiation
    of BLAS and similar libraries (such as Eigen) by specifying custom derivative
    rules for their operations.
  </p>

  <p><b>Expected results:</b>
    Efficient differentiation of BLAS and Eigen codes by writing custom
    derivative rules for matrix and tensor operations.
  </p>

  <p><b>Confirmed mentor:</b> <a href="mailto:wmoses@mit.edu">William Moses</a>, Johannes Doerfert</p>
  <p><b>Desirable skills:</b>
    Good knowledge of C++, calculus, and linear algebra. Experience with BLAS, Eigen,
    or Enzyme would be nice, but can also be learned in the project.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_swift">Integrate Enzyme into Swift to provide high-performance differentiation in Swift</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <a href="https://enzyme.mit.edu/">Enzyme</a> performs automatic differentiation
    (in the calculus sense) of LLVM programs. This enables users to use Enzyme to
    perform various algorithms such as back-propagation in ML
    or scientific simulation on existing code for any language that lowers to LLVM.

    While this functions for any frontend that emits LLVM IR, it may be desirable
    to have closer integration between Enzyme and the frontend for the sake of
    passing additional information and creating a better user experience.

    Swift provides automatic differentiation through the use of specifying custom
    derivative rules in the front-end. Enzyme could be integrated directly with
    Swift, differentiating the eventual LLVM, but it would lose out on all this
    additional information about custom derivatives. Moreover, calling into
    Enzyme naiively would be without Type checking, fine AD-specific debug information,
    or various other nice tools that Swift provides users of AD.

    This project would seek to integrate Enzyme and the Swift front end to provide
    both a nice user-experience for swift programmers who want to use Enzyme
    to enable high-performance automatic differentiation, and also to allow Enzyme
    to take advantage of derivative-specific metadata already available in swift.
  </p>

  <p><b>Expected results:</b>
    Creation of a custom type-checked linguistic construct in Swift for calling Enzyme.
    Mechanisms for passing Swift's differentiation-specific metadata for use by Enzyme.
  </p>

  <p><b>Confirmed mentor:</b> <a href="mailto:wmoses@mit.edu">William Moses</a>, Vassil Vassilev</p>
  <p><b>Desirable skills:</b>
    Good knowledge of C++ and Swift. Experience with Enzyme or automatic differentiation
    would be nice, but can also be learned in the project.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_fixed">Differentiation of Fixed-Point Arithmetic</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <a href="https://enzyme.mit.edu/">Enzyme</a> performs automatic differentiation
    (in the calculus sense) of LLVM programs. This enables users to use Enzyme to
    perform various algorithms such as back-propagation in ML
    or scientific simulation on existing code for any language that lowers to LLVM.

    In a variety of fields, it is desirable to compute on fixed-point values
    (e.g. integers) rather than floating point values. This avoid certain truncation
    errors that may be critical to a given application. Moreover, particular pieces
    of hardware may simply be more efficient on fixed point rather than floating
    point values.

    This project would seek to extend Enzyme to support differentiation of not only
    floating point base values, but also fixed point base values..
  </p>

  <p><b>Expected results:</b>
    Implementation of adjoints for LLVM fixed point intrinsics, requisite type analysis
    rules, and integration into a front-end for an end-to-end test.
  </p>

  <p><b>Confirmed mentor:</b> <a href="mailto:wmoses@mit.edu">William Moses</a></p>
  <p><b>Desirable skills:</b>
    Good knowledge of C++, caclulus, and LLVM internals. Experience with Enzyme or
    automatic differentiation would be nice, but can also be learned in the project.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="enzyme_rust">Integrate Enzyme into Rust to provide high-performance differentiation in Rust</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <a href="https://enzyme.mit.edu/">Enzyme</a> performs automatic differentiation
    (in the calculus sense) of LLVM programs. This enables users to use Enzyme to
    perform various algorithms such as back-propagation in ML
    or scientific simulation on existing code for any language that lowers to LLVM.

    While this functions for any frontend that emits LLVM IR, it may be desirable
    to have closer integration between Enzyme and the frontend for the sake of
    passing additional information and creating a better user experience.

    This project would seek to integrate Enzyme and the Rust front end to provide
    a nice user-experience for Rust programmers who want to use Enzyme
    to enable high-performance automatic differentiation. This also potentially
    involves integration of LLVM plugin support/custom codegen into rustc.
  </p>

  <p><b>Expected results:</b>
    Creation of a custom type-checked linguistic construct in Rust for calling Enzyme.
    Mechanisms for parsing Rust's Type information (represented as debug LLVM debug
    info) directly into type analysis.
  </p>

  <p><b>Confirmed mentor:</b> <a href="mailto:wmoses@mit.edu">William Moses</a></p>
  <p><b>Desirable skills:</b>
    Good knowledge of C++ and Rust. Experience with Enzyme or automatic differentiation
    would be nice, but can also be learned in the project.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="static_analyzer_profling">Clang Static Analyzer performance profiling</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <ul>
     <li>
       Chart how much time is spent in transfer functions – including (but not
       limited to!) checker callbacks.
     </li>
     <li>
       Add llvm Statistics and Timers for quickly obtaining precise and concise
       dumps without external profilers. Statistics on state splits might be
       particularly interesting!
     </li>
     <li>
       Measure time spent analyzing specific stack frames. Say, how much time
       do we spend inlining <tt>std::string</tt> methods? This time could be
       saved if we add custom  models for these methods instead.
     </li>
    </ul>
  </p>

  <p><b>Confirmed mentor:</b>  Artem Dergachev</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="static_analyzer_constraint_solver">Clang Static Analyzer constraint solver improvements</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    CSA has a small in-house constraint solver, it is pretty trivial, but super
    fast.  The goal is to support range-based logic for some of the symbolic
    operators, while keeping it linear.  Additionally, a unit-test framework
    can be designed specifically for testing constraint solvers (right now it’s
    tested rather awkwardly). This project has a couple of interesting
    properties.  It can be segmented into small chunks, and each of these
    chunks has a non-trivial solution.  It might introduce you to a world of
    solvers (it is a good idea to check your ideas with some heavy-weight
    solver such as z3). And because the existing solver is simple, there is a
    myriad of possible extensions to try.
  </p>

  <p><b>Confirmed mentor:</b>  Valeriy Savchenko</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="lldb_diagnostics">A structured approach to diagnostics in LLDB</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    <ul>
     <li>
       Design and integrate a new diagnostic abstraction (similar to
       clang::Diagnostic) to report errors, warnings and notes in a structured
       way.
     </li>
     <li>
       Allow us to differentiate between bugs (unexpected errors) and things
       the debugger simply doesn’t know (expected errors). Be smart about
       printing global errors only once. Have the ability of being verbose and
       have additional metadata (source location, DWARF unit, object file, etc,
       depending on the type of error and where it originated). </li>
     <li>
       Should be compatible and tightly integrated with the existing classes,
       such as the Status and CommandReturnObject.
     </li>
    </ul>
  </p>

  <p><b>Confirmed mentor:</b>
    <a href="mailto:teemperor@gmail.com,jonas@devlieghere.com?subject=[GSoC]%20LLDB%20Diagnostics">Jonas Devlieghere and Raphael Isemann</a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc20">Google Summer of Code 2020</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>
    LLVM participation in Google Summer of Code 2020 was very successful and resulted
    in many interesting projects contributed to LLVM. For the list of accepted and
    completed projects, please take a look into Google Summer of Code
    <a href="https://summerofcode.withgoogle.com/archive/2020/organizations/5902726635978752/">
      website.</a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>LLVM</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_ipo">Improve inter-procedural analyses and optimizations</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    This is a short description, please reach out to Johannes (jdoerfert on IRC)
    if it sounds interesting.

    During the GSoC'19 we build the Attributor framework to improve the
    inter-procedural capabilities of LLVM. This is useful on its own but
    especially in situations where inlining is impossible or undesirable.

    In this GSoC project we will look at capabilities not yet available in the
    Attributor and for the potential to connect the Attributor with existing
    intra- and inter-procedural optimizations.

    In this project there is a lot of freedom to determine the actual tasks but
    we will provide a pool of smaller and medium sized tasks that can be chosen
    from as well.
  </p>

  <p><b>Preparation resources:</b> The Attributor YouTube videos from the
  LLVM Developers Meeting 2019 and the recording of the IPO panel from the same
  meeting. The Attributor framework as well as other existing inter-procedural
  analyses and optimizations in LLVM.</p>

  <p><b>Expected results:</b> Measurable better IPO, especially visible in cases
                              where inlining is not an option or undesirable.</p>

  <p><b>Confirmed Mentor:</b> Johannes Doerfert</p>
  <p><b>Desirable skills:</b> Intermediate knowledge of C++, self motivation.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_par">Improve parallelism-aware analyses and optimizations</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    This is a short description, please reach out to Johannes (jdoerfert on IRC)
    if it sounds interesting.

    With the OpenMPOpt pass (<a href='https://reviews.llvm.org/D69930'>under
      review</a>) we started to teach the LLVM optimization pipeline about
    OpenMP parallelism encoded as OpenMP runtime calls.

    In this GSoC project we will look at capabilities not yet available in the
    OpenMPOpt pass and for the potential to connect existing intra- and
    inter-procedural optimizations, e.g. the Attributor.

    In this project there is a lot of freedom to determine the actual tasks but
    we will provide a pool of smaller and medium sized tasks that can be chosen
    from as well.
  </p>

  <p><b>Preparation resources:</b> The "Optimizing Indirections, using
  abstractions without remorse" video on YouTube from the LLVM Developers
  Meeting 2018. The paper "Compiler Optimizations for OpenMP" and "Compiler
  Optimizations For Parallel Programs" both by J. Doerfert and H. Finkel (the
  slides for these are potentially even more useful).</p>

  <p><b>Expected results:</b> Measurable better performance or program analysis
  results for parallel programs with a focus on OpenMP.</p>

  <p><b>Confirmed Mentor:</b> Johannes Doerfert</p>
  <p><b>Desirable skills:</b> Intermediate knowledge of C++, self motivation.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_dbg_invariant">Make LLVM passes debug info invariant</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Generating debug information is one of the fundamental tasks a compiler
    typically fulfills. It is clear that executable generated code should not
    depend on the presence of debug information.
    <br><br>
    Unfortunately there are known cases in LLVM were code generation differs
    depending on whether debug information is enabled (`-g`) or not. These kind
    of bugs can lead to bad debug experience ranging from unexpected execution
    behaviour to the point of programs running fine in debug mode while crashing
    without debug information.
    <br><br>
    The issue has likely not a single cause but is triggered during different
    passes on different architectures. One such reason is the insertion of Call
    Frame Information (CFI) in the compiler backend during frame lowering and
    other later passes. The presence of CFI instructions seems to change
    instruction scheduling which therefore leads to different generated code.
  </p>

  <p><b>Preparation resources:</b>
    <ul>
      <li>
        <a href="https://llvm.org/PR37728">PR37728</a> is a
        meta-bug that collects several related issues of differing codegen.
      </li>
      <li>
        <a href="https://llvm.org/PR37240">PR37240</a> is a
        bug discussing the CFI issue mentioned above.
      </li>
      <li>
        The following
        <a href="http://lists.llvm.org/pipermail/llvm-dev/2019-September/135433.html">
        RFC</a> discusses some possible mitigation strategies and gives some
        background information on the CFI issue.
      </li>
    </ul>
  </p>
  <p><b>Expected results:</b>
    <ul>
    <li>
      Write some tooling based on existing scripts to automatically generate
      examples of differing codegen. This is intended as a starting task to get
      to know the existing LLVM tools, learn to read LLVM's internal outputs etc.
    </li>
    <li>
      Choose one or more (depending on the difficulty) bugs that cause codegen
      differences and try to provide patches to fix them. We would be
      particularly interested in the mentioned CFI issue but working on some of
      the other related bugs is also absolutely fine.
    </li>
    </ul>
  </p>

  <p><b>Confirmed Mentors:</b> Paul Robinson and David Tellenbach</p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++, some familarity with general computer
    architecture, some familarity with the x86 or Arm/AArch64 instruction set.
  </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_mergesim">Improve MergeFunctions to incorporate MergeSimilarFunction patches and ThinLTO Support</a>
</div>
<!-- *********************************************************************** -->

 <div class="www_text">
  <p><b>Description of the project:</b> MergeSimilarFunctions pass is able to
    merge not just identical functions, but also functions with small differences in
    their instructions to reduce code size. It does this by inserting control flow
    and an additional argument in the merged function to account for the
    differences.

    This work was presented at
    the <a href="http://llvm.org/devmtg/2013-11/#talk3">LLVM Dev Meeting in
    2013</a> A more detailed description was published in a paper at
    <a href="http://dl.acm.org/citation.cfm?id=2597811">LCTES 2014</a>. The code
    was released to the community at the time. Meanwhile, the pass has been in
    production use at QuIC for the past few years and has been actively
    maintained internally. In order to magnify the impact of
    MergeSimilarFunctions, it has been ported to ThinLTO and the patches have
    been upstreamed (see stack of 5 patches mentioned below). But instead of
    replacing the existing MergeFunctions pass in LLVM-upstream the community
    suggested we improve the existing one with the ideas from
    MergeSimilarFunctions.  And then leverage the ThinLTO on top of that. The
    MergeSimilarFunction used in ThinLTO gives impressive code size reduction
    across a wide range of workloads and the work was presented at
    <a href="https://llvm.org/devmtg/2018-10/talk-abstracts.html#talk2">LLVM-dev
    2018</a>. The LLVM project would greatly benefit from this code size
    optimization as most embedded systems (think SmartPhones) applications are
    constrained on code-size.
  </p>
  <p><b>Preparation resources:</b>
  <ul>
    <li>
      Stack of patches:
      <ul>
        <li>
          <a href="https://reviews.llvm.org/D52896">MergeSimilarFunctions 1/n: a code size pass to merge functions with small differences</a>
        </li>
        <li>
          <a href="https://reviews.llvm.org/D52898">[Porting MergeSimilarFunctions 2/n] Changes to DataLayout</a>
        </li>
        <li>
          <a href="https://reviews.llvm.org/D52966">[Merge SImilar Function ThinLTO 3/n] Add hash code to function summary</a>
        </li>
        <li>
          <a href="https://reviews.llvm.org/D53253">[Merge SImilar Function ThinLTO 4/n] Make merge function decisions before the thin-lto stage</a>
        </li>
        <li>
          <a href="https://reviews.llvm.org/D53254">[Merge SImilar Function ThinLTO 5/n] Set up similar function to be imported</a>
        </li>
      </ul>
      The paches can be easily applied to LLVM-trunk and would give a developer a decent head start ;).
    </li>
    <li>List of llvm-dev mailing list posts on previous discussions around Merge Functions
      <ul>
        <li><a href="http://lists.llvm.org/pipermail/llvm-dev/2019-January/129835.html">Link1</li>
        <li><a href="http://lists.llvm.org/pipermail/llvm-dev/2019-March/131066.html">Link2</li>
        <li><a href="http://lists.llvm.org/pipermail/llvm-dev/2019-February/129863.html">Link3</li>
        <li><a href="http://lists.llvm.org/pipermail/llvm-dev/2019-January/129832.html">Link4</li>
      </ul>
    </li>
    <li>
      <a href="http://dl.acm.org/citation.cfm?id=2597811">The original paper: LCTES 2014</a>
    </li>
    <li>
      <a href="https://llvm.org/devmtg/2018-10/talk-abstracts.html#talk2">Video and slides of the presentation</a>
    </li>
  </ul>
  </p>
  <p><b>Expected results:</b>
    <ul>
      <li>
    Improve MergeFunctions to have feature parity with MergeSimilarFunctions.
      </li>
      <li>
    Enable MergeFunctions to ThinLTO.
      </li>
    </ul>
  </p>

  <p><b>Confirmed Mentors:</b>Aditya Kumar (hiraditya on IRC and phabricator), JF Bastien (jfb on phabricator)</p>

  <p><b>Desirable skills:</b>
    Course on compiler design, SSA Representation,
    Intermediate knowledge of C++, Familiarity with LLVM Core.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="llvm_dwarf_yaml2obj">Add DWARF support to yaml2obj</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    LLVM provides a tool called yaml2obj which coverts a YAML document into an
    object file, for various different file formats such as ELF, COFF and
    Mach-O, along with obj2yaml which does the inverse. The tool is commonly
    used to test parts of LLVM, as YAML is often easier to use to describe an
    object file than raw assembly and more maintainable than a pre-built binary.
    DWARF is a debugging file format commonly used by LLVM. Many of the tests
    for LLVM’s DWARF emission are written in assembly, but it would be nicer to
    write them in YAML. However, yaml2obj does not properly support emission of
    DWARF sections. This project is to add functionality to yaml2obj to make
    writing test inputs for DWARF tests simpler, particularly for ELF objects.
  </p>

  <p><b>Preparation resources:</b>
    Reading up on the DWARF file format will be useful, in particular the
    standards available at http://dwarfstd.org/Download.php. Also, familiarising
    yourself with the basics of the ELF file format, as described here
    https://www.sco.com/developers/gabi/latest/contents.html, may be beneficial.
  </p>
  <p><b>Expected results:</b>
    The ability to use yaml2obj to generate DWARF sections for object files.
    Particularly important is ensuring the input YAML can be more easily
    understood than the equivalent assembly.
  </p>

  <p><b>Confirmed Mentors:</b> James Henderson</p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++.
  </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_hotcold">Improve hot cold splitting to aggressively outline small blocks</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
 <p><b>Description of the project:</b> Hot Cold Splitting in LLVM is an IR level
   function splitting transformation. The goal of hot/cold splitting is to improve
   the memory locality of code and helps reduce startup working set. The splitting pass
   does this by identifying cold blocks and moving them into separate functions. Because it
   is implemented at the IR level all the back end target benefit from it.

   It is a relatively new optimization and it was recently presented at
   the <a href="https://llvm.org/devmtg/2019-10/talk-abstracts.html#tech8">LLVM Dev Meeting in
   2019</a> and the slides are <a href="https://llvm.org/devmtg/2019-10/slides/Kumar-HotColdSplitting.pdf">here</a>
   Because most of the benefit comes from outlining small blocks e.g., __assert_rtn. The goal of this project
   is to identify potential blocks via static analysis e.g., exception handling code, optimizing personality functions.

   Use cost-model to ensure outlining reduces the code size of the caller, use tail call whenever appropriate to save
   instructions.

 </p>
 <p><b>Preparation resources:</b>
 <ul>
   <li>
     <a href="http://lists.llvm.org/pipermail/llvm-dev/2019-January/129606.html">Update on hot cold splitting</a>
   </li>
   <li>
     The following two papers provide earlier work on hot cold splitting. While these papers are a good start, LLVM's
     HCS has completely different implementation in two aspects a) It is implemented at IR level and outlines basic
     blocks as function rather than naked branches. b) It is based on regions and outlines a set of basic blocks.
     <ul>
       <li>
         <a href="http://pages.cs.wisc.edu/~fischer/cs701.f05/code.positioning.pdf">Original paper on hot cold splitting by
           Pettis and Hansen.</a>Section 5 on procedure splitting is interesting one. It has nice examples ;) to help
         understand why HCS works.
       </li>
       <li>
         <a href="https://www.cs.cmu.edu/afs/cs/academic/class/15745-s07/www/papers/p80-cohn.pdf">Paper on hot cold
           splitting</a> The paper provides some details on one approach to split functions. This is helpful to get a
         different perspective and may help get new ideas.
       </li>
     </ul>
   </li>
   <li>
     <a href="https://llvm.org/devmtg/2019-10/talk-abstracts.html#tech8">Video and slides of the presentation</a>
   </li>
 </ul>
 </p>
 <p><b>Expected results:</b>
   <ul>
     <li>
       Improve Hot Cold Splitting to detect and outline cold blocks from program via static analysis or profile
       information. Use appropriate cost model to weigh benefit of HCS.
       In case compile time overhead becomes quadratic, come up with a cost model to detect when quadratic behavior
       gets triggered and bail out based on a compiler flag.
     </li>
   </ul>
 </p>

 <p><b>Confirmed Mentors:</b>Aditya Kumar (hiraditya on IRC and phabricator)</p>

 <p><b>Desirable skills:</b>
   Course on compiler design, SSA Representation,
   Intermediate knowledge of C++, Familiarity with LLVM Core.
 </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_pass_order">Advanced Heuristics for Ordering Compiler Optimization Passes</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
 <p><b>Description of the project:</b>
Selecting optimization passes for given application is very important but
non-trivial problem because of the huge size of the compiler transformation
space (incl. pass ordering). While the existing heuristics can provide high
performance code for certain applications, they cannot easily benefit a wide
range of application codes. The goal of the project is to learn the interplay
between LLVM transformation passes and code structures, then improve the
existing heuristics (or replace the heuristics with machine learning-based
models) so that the LLVM compiler can provide a superior order of the passes
customized per application.
 </p>
 <p><b>Expected results (possibilities):</b>
 <ul>
   <li>
Insights about (implicit) dependences between existing passes.
   </li>
   <li>
New pass pipelines (think -O3a, -O3b, ...) selectable by the user that tend to perform substantially better on certain kinds of programs.
   </li>
   <li>
An improved LLVM pass heuristic or new machine learning-based models that can select
the best order for LLVM transformation passes based on code structures.
   </li>
  </ul>
 </p>

 <p><b>Preparation resources:</b>
 <ul>
   <li>
HERCULES: Strong Patterns towards More Intelligent Predictive Modeling, Eunjung Park; Christos Kartsaklis; John Cavazos, IEEE ICPP’14
https://ieeexplore.ieee.org/abstract/document/6957226
   </li>

   <li>
Predictive Modeling in a Polyhedral Optimization Space, Eunjung Park, John Cavazos, Louis-Noël Pouchet, Cédric Bastoul, Albert Cohen & P. Sadayappan, IJPP’13
https://link.springer.com/article/10.1007/s10766-013-0241-1
   </li>

   <li>
Machine Learning in Compiler Optimization, Zheng Wang and Michael O’Boyle, IEEE Magazine 2018.
https://ieeexplore.ieee.org/document/8357388
   </li>
 </ul>
 </p>

 <p><b>Confirmed Mentors:</b>EJ Park, Giorgis Georgakoudis, Johannes Doerfert</p>

 <p><b>Desirable skills:</b>
    C++, Python, experience with LLVM and learning-based prediction preferable.
 </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_ml_scc">Machine learning and compiler optimizations: using inter-procedural analysis to select optimizations</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
 <p><b>Description of the project:</b>
Current machine learning models for compiler optimization select the best
optimization strategies for functions based on isolated per function analysis.
In this approach, the constructed models are not aware of any relationships
with other functions around it (callers or callees) which can be helpful to
decide the best optimization strategies for each function. In this project, we
want to explore the SCC (Strongly Connected Components) call graph to add
inter-procedural features in constructing machine learning-based models to find
the best optimization strategies per function. Moreover, we want to explore the
case that it is helpful to group strongly related functions together and
optimize them as a group, instead of per function.
 </p>
 <p><b>Expected results (possibilities):</b>
 <ul>
   <li>
Improved heuristics for existing (inter-procedural) passes, e.g. to weight inlining versus function cloning based on code features.
   </li>
   <li>
Machine learning models to select the best optimizations using code features
and inter-procedural analysis. This model can be used for functions in
isolation or groups of functions, e.g., CGSCCs.
   </li>
 </ul>
 </p>

 <p><b>Preparation resources:</b>
 <ul>
   <li>
HERCULES: Strong Patterns towards More Intelligent Predictive Modeling, Eunjung Park; Christos Kartsaklis; John Cavazos, IEEE ICPP’14
https://ieeexplore.ieee.org/abstract/document/6957226
   </li>

   <li>
Predictive Modeling in a Polyhedral Optimization Space, Eunjung Park, John Cavazos, Louis-Noël Pouchet, Cédric Bastoul, Albert Cohen & P. Sadayappan, IJPP’13
https://link.springer.com/article/10.1007/s10766-013-0241-1
   </li>

   <li>
Machine Learning in Compiler Optimization, Zheng Wang and Michael O’Boyle, IEEE Magazine 2018.
https://ieeexplore.ieee.org/document/8357388
   </li>
 </ul>
 </p>

 <p><b>Confirmed Mentors:</b>EJ Park, Giorgis Georgakoudis, Johannes Doerfert</p>

 <p><b>Desirable skills:</b>
    C++, Python, experience with LLVM and learning-based prediction preferable.
 </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_postdominators"></a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    There is currently no easy way to use the result of
    PostDominatorTreeAnalysis in a loop pass, as PostDominatorTreeAnalysis is a
    function analysis, and it is not included in LoopStandardAnalysisResults. If one adds
    PostDominatorTreeAnalysis in LoopStandardAnalysisResults, then all loop passes
    need to preserve it, meaning that all loop passes need to make sure the result is up to
    date. In this project, we want to modify some commonly used utilities to generate a
    list of updates, which can be consume by different updaters, e.g. DomTreeUpdater to
    update DominatorTree and PostDominatorTree, and MSSAU to update MemorySSA,
    etc, instead of only updating the DominatorTree. In additional, we want to change
    existing loop passes to preserve the PostDominatorTree. Finally, adding
    PostDominatorTree in LoopStandardAnalysisResults.
  </p>
  <p><b>Expected results (possibilities):</b>
    PostDominatorTree added in LoopStandardAnalysisResults, and
    can be used by loop passes. More common utilities change to generate list of updates
    to be easily obtained by different updaters.
  </p>
  <p><b>Confirmed Mentors:</b>
    Whitney Tsang, Ettore Tiotto, Bardia Mahjour
  </p>
  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++, self-motivation.
  </p>
  <p><b>Preparation resources:</b>
    <a href="https://reviews.llvm.org/rL336163"></a>
    <a href="http://llvm.org/doxygen/classllvm_1_1DomTreeUpdater.html"></a>
    <a href="https://llvm.org/doxygen/classllvm_1_1PostDominatorTreeAnalysis.html"></a>
    <a href="http://llvm.org/doxygen/structllvm_1_1LoopStandardAnalysisResults.html"></a>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_loopnest">Create LoopNest Pass</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Currently if you want to write a pass that works on a loop
    nest, you have to pick from either a function pass or a loop pass. If you chose to write
    it as a function pass, then you lose the ability to add loops dynamically back to the
    pipeline. If you decide to write it as a loop pass, then you are wasting compile time to
    traverse to your pass and return right away when the given loop is not the outermost
    loop. In this project, we want to create a LoopNestPass, where transformations
    intended for loop nest can inherit from it, and have the same ability as the LoopPass to
    dynamically add loops to the pipeline. In addition, create all the adaptors requires to
    add loop nest passes at different points of the pass builder.
  </p>
  <p><b>Expected results (possibilities):</b>
    Transformations/Analyses can be written as LoopNestPass,
    without compromising compile time or usability.
  </p>
  <p><b>Confirmed Mentors:</b>
    Whitney Tsang, Ettore Tiotto
  </p>
  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++, self-motivation.
  </p>
  <p><b>Preparation resources:</b>
    <a href="https://reviews.llvm.org/D68789"</a>
    <a href="https://llvm.org/doxygen/classllvm_1_1PassBuilder.html"</a>
  </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_instdump">Instruction properties dumper and checker</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    TableGen is flexible and allow the end-user to define and set common properties of
    records (instructions). Every target has dozens or hundreds of such instruction
    properties. As target code evolve, the td files become more and more complicated,
    it become harder to see whether the setting of some properties is necessary, even
    correct or not. eg: whether hasSideEffects property is correctly set on all
    instructions?

    One can manually search through the TableGen-generated files; or write some
    script to run TableGen and matching the output for some specific properties, but a
    standalone utility that can dump and check instruction properties
    systematically (eg: also allow target to define some verification rules) might be
    better from a build-process-management standpoint. This can help to find quite
    some hidden bugs and hence improve the overall codegen code quality. In
    addition, the utility can be used to write regression tests for instruction
    properties, which will increase the quality and precision of LLVM's
    regression tests.
  </p>
  <p><b>Expected results (possibilities):</b>
    A standalone llvm tool or utility that can dump and check instruction properties systematically
  </p>
  <p><b>Confirmed Mentors:</b>
    Hal Finkel, Jinsong Ji , Qingshan Zhang
  </p>
  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++, self-motivation.
  </p>
</div>

<!-- *********************************************************************** -->

<div class="www_subsubsection">
  <a name="llvm_movecode">Unify ways to move code or check if code is safe to be moved</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
    Determining whether it is safe to move code around is
    implemented in several transformations in LLVM (e.g. canSinkOrHoistInst in LICM,
    or makeLoopInvariant in Loop). Each of these implementations may return different
    results for a given query, making code motion safety checks inconsistent and
    duplicated. On the other hand, the mechanism for doing the actual code motion is also
    different in each transformation. Code duplication causes maintenance problems and
    increases the time taken to write new transformation. In this project, we want to first
    identify all the existing ways in loop transformations (could be function or loop pass)
    to check if code is safe to move, and to move code, and create a standardize way to do
    so.
  </p>
  <p><b>Expected results (possibilities):</b>
    A standardize/superset of all the existing ways in loop
    transformations of checking if code is safe to be moved and to move <code class=""></code>
  </p>

  <p><b>Confirmed Mentors:</b>
    Whitney Tsang, Ettore Tiotto, Bardia Mahjour
  </p>
  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++, self-motivation.
  </p>
  <p><b>Preparation resources:</b>
    <a href="https://github.com/llvm/llvm-project/blob/master/llvm/include/llvm/Transforms/Utils/CodeMoverUtils.h"></a>
    <a href="https://llvm.org/doxygen/LICM_8cpp_source.html"></a>
    <a href="https://llvm.org/doxygen/classllvm_1_1Loop.html"></a>
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>MLIR</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
<p>All the items in the list of
<a href="https://mlir.llvm.org/getting_started/openprojects/">open projects</a>
are opened to GSOC. Feel free to propose your own ideas as well on
<a href="https://llvm.discourse.group/c/llvm-project/mlir">Discourse</a>.
</p></div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-sa-cplusplus-checkers">Find null smart pointer dereferences
                                        with the Static Analyzer</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b>
    The Clang Static Analyzer already knows how to prevent crashes caused by
    null pointer dereference in arbitrary code, however it often "gives up"
    when the code is too complicated. In particular, implementation details
    of C++ standard classes, even simple ones such as smart pointers
    or optionals, may be too convoluted for the Analyzer to fully understand.
    Moreover, the exact behavior depends on which implementation of
    the Standard Library is used (e.g., GNU libstdc++ or LLVM's own libc++).
  </p>
  <p>
    We can enable the Analyzer to find more bugs in modern C++ code
    by teaching it explicitly about the behavior of C++ standard classes,
    and therefore skipping the whole process in which the Analyzer
    tries to understand all the implementation details on its own.
    For example, we could teach it that a default-constructed smart pointer
    is null, and any attempt to dereference it would result in a crash.
    The project would therefore consist in manually providing implementations
    for various methods of standard classes.
  </p>

  <p><b>Expected results: </b>
    We want the Static Analyzer to emit warnings when a null smart pointer
    dereference would occur in the code. For example:
    <pre>
    #include &lt;memory&gt;

    int foo(bool flag) {
      std::unique_ptr&lt;int&gt; x;  <i>// note: Default constructor produces a null unique pointer;</i>

      if (flag)                <i>// note: Assuming 'flag' is false;</i>
        return 0;              <i>// note: Taking false branch</i>

      return *x;               <i>// warning: Dereferenced smart pointer 'x' is null.</i>
    }
    </pre>
    We should be able to cover at least one class fully, for example, <tt>std::unique_ptr</tt>,
    and then see if we can generalize our results to other classes, such as <tt>std::shared_ptr</tt>
    or the C++17 <tt>std::optional</tt>.
  </p>


  <p><b>Confirmed Mentor:</b> Artem Dergachev, Gábor Horváth</p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++.
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>LLDB</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="lldb-autosuggestions">Support autosuggestions in LLDB's command line</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b> LLDB's command line offers several convenience
    features that are inspired by features of UNIX shells such as tab completions or a command history.
    One feature that is not implemented yet are 'autosuggestions'. These are suggestions
    for possible commands that the user might want to type, but unlike tab completions they
    are displayed directly behind the cursor while the user is typing a command. A good demonstration
    how this could look like are the autosuggestions implemented in <a href="https://fishshell.com">fish shell</a>.
  </p>
  <p>
    This project is about implementing autosuggestions in LLDB's editline-based command shell.
  </p>
  <p><b>Confirmed Mentor:</b>
    <a href="mailto:teemperor@gmail.com,jonas@devlieghere.com?subject=[GSoC]%20Autosuggestions">Jonas Devlieghere and Raphael Isemann</a></p>
  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="lldb-more-completions">Implement the missing tab completions for LLDB's command line</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b> LLDB's command line offers several convenience
    features that are inspired by features of UNIX shells such as tab completions for commands.
    These tab completions are implemented by a completion engine that is not only used by the
    command line interface of LLDB, but also by graphical interfaces for LLDB such as IDEs.

    While the tab completions in LLDB are really useful, they are currently not implemented for
    all commands and their respective arguments. This project is about implementing the remaining
    completions for the commands in LLDB which will greatly improve the user experience of LLDB.
    Improving existing completions is also part of the project.

    Note that the completions are not static list of strings but often require inspecting and
    understanding the internal state of LLDB. As LLDB commands and their tab completions cover
    all aspects of LLDB, this project offers a great way to get an overview of all the functionality
    in LLDB.
  </p>
  <p><b>Confirmed Mentor:</b><a href="mailto:teemperor@gmail.com?subject=[GSoC]%20Completions">Raphael Isemann</a></p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++.
  </p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="lldb-reimplement-lldb-cmdline">Reimplement LLDB's command-line commands
  using the public SB API.</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b> Just as LLVM is a library to
    build compilers, LLDB is a library to build debuggers. LLDB vends
    a stable, public SB API. Due to historic reasons the LLDB command
    line interface is currently implemented on top of LLDB's private
    API and it duplicates a lot of functionality that is already
    implemented in the public API. Rewriting LLDB's command line
    interface on top of the public API would simplify the
    implementation, eliminate duplicate code, and most importantly
    reduce the testing surface.
  </p>
  <p>
    This work will also provide an opportunity to clean up the SB API
    of commands that have accrued too many overloads over time and
    convert them to make use of option classes to both gather up all
    the variants and also future-proof the APIs.
  </p>
  <p><b>Confirmed Mentor:</b>Adrian Prantl and Jim Ingham</p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of C++.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="lldb-batch-testing">Add support for batch-testing to the LLDB
    testsuite.</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b>One of the tensions in the
    testsuite is that spinning up a process and getting it to some
    point is not a cheap operation, so you'd like to do a bunch of
    tests when you get there.  But the current testsuite bails at the
    first failure, so you don't want to do many tests since the
    failure of one fails all the others. On the other hand, there are
    some individual test assertions where the failure of the assertion
    <em>should</em> cause the whole test to fail.  For example, if you
    fail to stop at a breakpoint where you want to check some variable
    values, then the whole test should fail.  But if your test then
    wants to check the value of five independent locals, it should be
    able to do all five, and then report how many of the five variable
    assertions failed. We could do this by adding <em>Start</em>
    and <em>End</em> markers for a batch of tests, do all the tests in
    the batch without failing the whole test, and then report the
    error and fail the whole test if appropriate. There might also be
    a nice way to do this in Python using scoped objects for the test
    sections.
  </p>
  <p><b>Confirmed Mentor:</b> Jim Ingham</p>

  <p><b>Desirable skills:</b>
    Intermediate knowledge of Python.
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc19">Google Summer of Code 2019</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>Google Summer of Code 2019 contributed a lot to the LLVM project. For the list of
    accepted and completed projects, please take a look into Google Summer of Code
    <a href="https://summerofcode.withgoogle.com/archive/2019/organizations/5682474363912192/">website.
    </a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>LLVM</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="debuginfo_codegen_mismatch">Debug Info should have no
          effect on codegen</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project:</b>
      Adding Debug Info (compiling with `clang -g`) shouldn't change the
      generated code at all. Unfortunately we have bugs. These are usually not
      too hard to fix and a good way to discover new part of the codebase!
      We suggest building object files both ways and disassembling the
          text sections, which will give cleaner diffs than comparing .s files.
  </p>

  <p><b>Expected results:</b> Reduced test cases, bug reports with analysis
          (e.g., which pass is responsible), possibly patches.</p>

  <p><b>Confirmed Mentor:</b> Paul Robinson</p>
  <p><b>Desirable skills:</b> Intermediate knowledge of C++, some familiarity
        with x86 or ARM instruction set.</p>
</div>


<!-- *********************************************************************** -->
<div class="www_subsection">
  <a>Clang</a>
</div>
<!-- *********************************************************************** -->

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="clang-astimporter-fuzzer">Implement an ASTImporter fuzzer</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b>
    Clang contains an ASTImporter which allows moving declarations and
    statements from one Clang AST to another. This is for example used for
    static analysis across translation units and in LLDB's expression
    evaluator.
  </p>
  <p>
    The current ASTImporter works as intended when moving simple C code from
    one AST to another. However, more complicated declarations such as C++'s
    OOP features and templates are not fully implemented and can cause crashes
    or invalid AST nodes. The bug reports related to these crashes are often
    filed against LLDB's expression evaluator and are rarely submited with a
    minimal reproducer. This makes improving ASTImporter a time-consuming and
    tedious task.
  </p>
  <p>
    This project is about writing a fuzzer to proactively discover these
    ASTImporter bugs and provide minimal reproducers which make understanding
    and fixing the underlying bug easier.
  </p>
  <p>
    A possible implementation of such a fuzzer and driver could look like this:

  <ul>
    <li>Generate some source code that can be imported (either fully randomly
        or based on existing source code from a user-given code corpus).</li>
    <li>Import randomly a few declarations from this AST. The AST in which
        they are imported to can already be populated with declarations.</li>
    <li>Run Clang's code generator over our imported AST.</li>
    <li>If we hit an assert during the import or CodeGen steps we probably
        found an ASTImporter bug.</li>
    <li>The fuzzer driver should now reduce the size of the source code
        until it is as small as possible and still reproduces the crash (e.g.
        by running Creduce with an automatically generated test script).</li>
    <li>The reproducer should now be stored in a format so that it can just be
        copied into Clang's regression test suite for the ASTImporter (see
        the <a href="https://github.com/llvm/llvm-project/tree/master/clang/test/Import">clang/test/Import/</a> directory).
        The reproducer must still reproduce the found bug when run as part
        of the test suite.
        </li>
  </ul>
  This is just one possible approach and students are welcome to submit their
  own ideas on how the fuzzer should operate. Approaches that allow to
  automatically verify more aspects of the imported AST (e.g. the source
  locations of AST nodes, size of RecordDecls) are encouraged. The fuzzer and
  driver should be implemented in C++ and/or Python.
  </p>
  <p><b>Confirmed Mentor:</b> Raphael Isemann, Shafik Yaghmour</p>
  <p><b>Desirable skills:</b> Intermediate knowledge of C++.</p>
</div>

<!-- *********************************************************************** -->
<div class="www_subsubsection">
  <a name="improve-autocompletion">Improve shell autocompletion for Clang</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p><b>Description of the project: </b> Clang has a newly implemented autocompletion feature which details can be found at <a href="http://blog.llvm.org/2017/09/clang-bash-better-auto-completion-is.html">LLVM blog</a>. We would like to improve this by adding more flags to autocompletion, supporting more shells (currently it supports only bash) and exporting this feature to other projects such as llvm-opt. Accepted student will be working on Clang Driver, LLVM Options and shell scripts.
  </p>

  <p><b>Expected Results:</b> Autocompletion working on bash and zsh, support llvm-opt options.</p>

  <p><b>Confirmed Mentor:</b> Yuka Takahashi and Vassil Vassilev</p>

  <p><b>Desirable skills:</b>
  Intermediate knowledge of C++ and shell scripting
  </p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc18">Google Summer of Code 2018</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>Google Summer of Code 2018 contributed a lot to the LLVM project. For the list of
  accepted and completed projects, please take a look into Google Summer of Code
  <a href="https://summerofcode.withgoogle.com/archive/2018/organizations/5263452624912384/">website.
  </a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="gsoc17">Google Summer of Code 2017</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">
  <p>Google Summer of Code 2017 contributed a lot to the LLVM project. For the list of
  accepted and completed projects, please take a look into Google Summer of Code
  <a href="https://summerofcode.withgoogle.com/archive/2017/organizations/6215410651234304/">website.
  </a></p>
</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="what">What is this?</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">

<p>This document is meant to be a sort of "big TODO list" for LLVM.  Each
project in this document is something that would be useful for LLVM to have, and
would also be a great way to get familiar with the system.  Some of these
projects are small and self-contained, which may be implemented in a couple of
days, others are larger.  Several of these projects may lead to interesting
research projects in their own right.  In any case, we welcome all
contributions.</p>

<p>If you are thinking about tackling one of these projects, please send a mail
to the <a href="http://lists.llvm.org/mailman/listinfo/llvm-dev">LLVM
Developer's</a> mailing list, so that we know the project is being worked on.
Additionally this is a good way to get more information about a specific project
or to suggest other projects to add to this page.
</p>

<p>The projects in this page are open-ended. More specific projects are
filed as unassigned enhancements in the <a href="https://github.com/llvm/llvm-project/issues/">
LLVM bug tracker</a>. See the
<a href="https://github.com/llvm/llvm-project/issues?q=is%3Aopen+is%3Aissue+no%3Aassignee">list of currently outstanding issues</a>
if you wish to help improve LLVM.</p>

</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="subprojects">LLVM Subprojects: Clang and More</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">

<p>In addition to hacking on the main LLVM project, LLVM has several subprojects,
   including Clang and others.  If you are interested in working on these, please
   see their "Open projects" page:</p>

<ul>
<li>The <a href="http://clang.llvm.org/OpenProjects.html">Clang Open
    Projects</a> list.</li>
<li>The <a href="http://polly.llvm.org/projects.html">Polly Open
    Projects</a> list.</li>
<li>The <a href="http://sva.cs.illinois.edu/projects.html">SAFECode Open
    Projects</a> list.</li>
</ul>

</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="improving">Improving the current system</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">

<p>Improvements to the current infrastructure are always very welcome and tend
to be fairly straight-forward to implement.  Here are some of the key areas that
can use improvement...</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="target-desc">Factor out target descriptions</a>
</div>

<div class="www_text">

<p>Currently, both Clang and LLVM have a separate target description infrastructure,
with some features duplicated, others "shared" (in the sense that Clang has to create
a full LLVM target description to query specific information).</p>

<p>This separation has grown in parallel, since in the beginning they were quite
different and served disparate purposes. But as the compiler evolved, more and
more features had to be shared between the two so that the compiler would behave
properly. An example is when targets have default features on speficic configurations
that don't have flags for. If the back-end has a different "default" behaviour
than the front-end and the latter has no way of enforcing behaviour, it
won't work.</p>

<p>An alternative would be to create flags for all little quirks, but first, Clang
is not the only front-end or tool that uses LLVM's middle/back ends, and second,
that's what "default behaviour" is there for, so we'd be missing the point.</p>

<p>Several ideas have been floating around to fix the Clang driver WRT recognizing
architectures, features and so on (table-gen it, user-specific configuration files,
etc) but none of them touch the critical issue: sharing that information with the
back-end.</p>

<p>Recently, the idea to factor out the target description infrastructure from
both Clang and LLVM into its own library that both use, has been floating around.
This would make sure that all defaults, flags and behaviour are shared, but would
also reduce the complexity (and thus the cost of maintenance) a lot. That would
also allow all tools (lli, llc, lld, lldb, etc) to have the same behaviour
across the board.</p>

<p>The main challenges are:</p>

<ul>
  <li>To make sure the transition doesn't destroy the delicate balance on any
  target, as some defaults are implicit and, some times, unknown.</li>
  <li>To be able to migrate one target at a time, one tool at a time and still
  keep the old infrastructure intact.</li>
  <li>To make it easy for detecting target's features for both front-end and
  back-end features, and to merge both into a coherent set of properties.</li>
  <li>To provide a bridge to the new system for tools that haven't migrated,
  especially the off-the-tree ones, that will need some time (one release,
  at least) to migrate..</li>
</ul>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="code-cleanups">Implementing Code Cleanup bugs</a>
</div>

<div class="www_text">

<p>
The <a href="https://github.com/llvm/llvm-project/issues/">LLVM bug tracker</a> occasionally
has <a href="https://github.com/llvm/llvm-project/labels/code-cleanup">"code-cleanup" bugs</a> filed in it.
Taking one of these and fixing it is a good way to get your feet wet in the
LLVM code and discover how some of its components work.  Some of these include
some major IR redesign work, which is high-impact because it can simplify a lot
of things in the optimizer.
</p>

<p>
Some specific ones that would be great to have:

<ul>
<li><a href="/PR10367">Fix the design of GlobalAlias to not require dest type to match source type</a></li>
<li><a href="/PR10368">Redesign ConstantExpr's</a></li>
<li><a href="/PR11944">Static constructors should be purged from LLVM</a></li>
</ul>
</p>

<p>Additionally, there are performance improvements in LLVM that need to get
fixed. These are marked with the <tt>slow-compile</tt> keyword. Use
<a href="https://github.com/llvm/llvm-project/issues?q=is%3Aopen+is%3Aissue+label%3Aslow-compile">
this LLVM bug tracker query</a>
to find them.</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="llvmtest">Add programs to the llvm-test testsuite</a>
</div>

<div class="www_text">

<p>
The <a href="docs/TestingGuide.html#wholeprograms">llvm-test</a> testsuite is
a large collection of programs we use for nightly testing of generated code
performance, compile times, correctness, etc.  Having a large testsuite gives
us a lot of coverage of programs and enables us to spot and improve any
problem areas in the compiler.</p>

<p>
One extremely useful task, which does not require in-depth knowledge of
compilers, would be to extend our testsuite to include <a href=
"http://nondot.org/sabre/LLVMNotes/#benchmarks">new programs and benchmarks</a>.
In particular, we are interested in cpu-intensive programs that have few
library dependencies, produce some output that can be used for correctness
testing, and that are redistributable in source form.  Many different programs
are suitable, for example, see <a
href="http://nondot.org/sabre/LLVMNotes/#benchmarks">this list</a> for some
potential candidates.
</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="programs">Compile programs with the LLVM Compiler</a>
</div>

<div class="www_text">

<p>We are always looking for new testcases and benchmarks for use with LLVM.  In
particular, it is useful to try compiling your favorite C source code with LLVM.
If it doesn't compile, try to figure out why or report it to the <a
href="http://lists.llvm.org/pipermail/llvm-bugs/">llvm-bugs</a> list.  If you
get the program to compile, it would be extremely useful to convert the build
system to be compatible with the LLVM Programs testsuite so that we can check it
into SVN and the automated tester can use it to track progress of the
compiler.</p>

<p>When testing a code, try running it with a variety of optimizations, and with
all the back-ends: CBE, llc, and lli.</p>

</div>


<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="benchmark">Benchmark the LLVM compiler</a>
</div>

<div class="www_text">

<p>Find benchmarks either using our <a
href="/nightlytest/">test results</a> or on your own,
where LLVM code generators do not produce optimal code or where another
compiler produces better code.  Try to minimize the test case that demonstrates
the issue.  Then, either <a href="https://github.com/llvm/llvm-project/issues/">submit a
bug</a> with your testcase and the code that LLVM produces vs. the code that it
<em>should</em> produce, or even better, see if you can improve the code
generator and submit a patch.  The basic idea is that it's generally quite easy
for us to fix performance problems if we know about them, but we generally don't
have the resources to go finding out why performance is bad.</p>

</div>


<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="statistics">Benchmark Statistics and Warning System</a>
</div>

<div class="www_text">

<p>The <a href='http://llvm.org/perf/db_default/v4/nts/recent_activity'>
LNT perf database</a> has some nice features like detect moving average,
standard deviations, variations, etc. But the report page give too much emphasis
on the individual variation (where noise can be higher than signal), eg.
<a href='http://llvm.org/perf/db_default/v4/nts/graph?plot.0=10.341.3&highlight_run=8943'>
this case</a>.</p>

<p>The first part of the project would be to create an analysis tool that would
track moving averages and report:
<ul>
 <li>If the current result is higher/lower than the previous moving average by
     more than (configurable) S standard deviations</li>
 <li>If the current moving average is more than S standard deviations of the
     Base run</li>
 <li>If the last A moving averages are in constant increase/decrease of more
     than P percent</li>
</ul>

<p>The second part would be to create a web page which would show all related
benchmarks (possibly configurable, like a dashboard) and show the basic statistics
with red/yellow/green colour codes to show status and links to more detailed
analysis of each benchmark.</p>

<p>A possible third part would be to be able to automatically cross reference
different builds, so that if you group them by architecture/compiler/number
of CPUs, this automated tool would understand that the changes are more common
to one particular group.</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="coverage">Improving Coverage Reports</a>
</div>

<div class="www_text">

<p>The <a href='http://llvm.org/reports/coverage/'>
LLVM Coverage Report</a> has a nice interface to show what source lines are
covered by the tests, but it doesn't mentions which tests, which revision and
what architecture is covered.</p>

<p>A project to renovate LCOV would involve:
<ul>
 <li>Making it run on a buildbot, so that we know what commits / architectures
     are covered</li>
 <li>Update the web page to show that information</li>
 <li>Develop a system that would report every buildbot build into the web page
     in a searchable database, like LNT</li>
</ul>

<p>Another idea is to enable the test suite to run all built backends, not only
   the host architecture, so that coverage report can be built in a fast machine
   and have one report per commit without needing to update the buildbots.</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="misc_imp">Miscellaneous Improvements</a>
</div>

<div class="www_text">

<ol>

<li>Completely rewrite bugpoint.  In addition to being a mess, bugpoint suffers
from a number of problems where it will "lose" a bug when reducing.  It should
be rewritten from scratch to solve these and other problems.</li>
<li><a href="https://llvm.org/PR2116">Add support for
transactions to the PassManager</a> for improved bugpoint.</li>
<li><a href="https://llvm.org/PR539">Improve bugpoint to
support running tests in parallel on MP machines</a>.</li>
<li>Add MC assembler/disassembler and JIT support to the SPARC port.</li>
<li>Move more optimizations out of the <tt>-instcombine</tt> pass and into
InstructionSimplify.  The optimizations that should be moved are those that
do not create new instructions, for example turning <tt>sub i32 %x, 0</tt>
into <tt>%x</tt>.  Many passes use InstructionSimplify to clean up code as
they go, so making it smarter can result in improvements all over the place.</li>
</ol>

</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="new">Adding new capabilities to LLVM</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">

<p>Sometimes creating new things is more fun than improving existing things.
These projects tend to be more involved and perhaps require more work, but can
also be very rewarding.</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="llvm_ir">Extend the LLVM intermediate representation</a>
</div>

<div class="www_text">

<p>Many proposed <a href="http://nondot.org/sabre/LLVMNotes/">extensions and
improvements to LLVM core</a> are awaiting design and implementation.</p>

<ol>
<li><a href="http://nondot.org/sabre/LLVMNotes/DebugInfoImprovements.txt">Improvements
for Debug Information Generation</a></li>
<li><a href="/PR1269">EH support for non-call exceptions</a></li>
<li>Many ideas for feature requests are stored in LLVM bugzilla.  Search
<a href="https://github.com/llvm/llvm-project/issues?q=is%3Aissue+is%3Aopen+label%3Anew-feature">for bugs with a "new-feature" keyword</a>.</li>
</ol>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="pointeranalysis">Pointer and Alias Analysis</a>
</div>

<div class="www_text">

<p>We have a <a href="docs/AliasAnalysis.html">strong base for development</a> of
both pointer analysis based optimizations as well as pointer analyses
themselves.  We want to take advantage of this:</p>

<ol>
<li>The globals mod/ref pass does an inexpensive bottom-up context sensitive
  alias analysis.  There are some inexpensive things that we could do to better
  capture the effects of functions that access pointer arguments.  This can be
  really important for C++ methods, which spend lots of time accessing pointers
  off 'this'.</li>

<li>The alias analysis API supports the getModRefBehavior method, which allows
  the implementation to give details analysis of the functions. For example, we
  could implement <a href="/PR1604">full knowledge of
    printf/scanf</a> side effects, which would be useful.  This feature is in
  place but not being used for anything right now.</li>

<li>We need some way to reason about errno.  Consider a loop like this:

<pre>
    for ()
      x += sqrt(loopinvariant);
</pre>

<p>We'd like to transform this into:</p>

<pre>
    t = sqrt(loopinvariant);
    for ()
      x += t;
</pre>

<p>This transformation is safe, because the value of errno isn't
otherwise changed in the loop and the exit value of errno from the
loop is the same.  We currently can't do this, because sqrt clobbers
errno, so it isn't "readonly" or "readnone" and we don't have a good
way to model this.</p>

<p>The important part of this project is figuring out how to describe
errno in the optimizer: each libc #defines errno to something different
it seems.  Maybe the solution is to have a __builtin_errno_addr() or
something and change sys headers to use it.</p>

<li>There are lots of ways to optimize out and <a
href="/PR452">improve handling of
memcpy/memset</a>.</li>

</ol>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="profileguided">Profile-Guided Optimization</a>
</div>

<div class="www_text">

<p>We now have a unified infrastructure for writing profile-guided
transformations, which will work either at offline-compile-time or in the JIT,
but we don't have many transformations.  We would welcome new profile-guided
transformations as well as improvements to the current profiling system.
</p>

<p>Ideas for profile-guided transformations:</p>

<ol>
<li>Superblock formation (with many optimizations)</li>
<li>Loop unrolling/peeling</li>
<li>Profile directed inlining</li>
<li>Code layout</li>
<li>...</li>
</ol>

<p>Improvements to the existing support:</p>

<ol>
<li>The current block and edge profiling code that gets inserted is very simple
and inefficient.  Through the use of control-dependence information, many fewer
counters could be inserted into the code.  Also, if the execution count of a
loop is known to be a compile-time or runtime constant, all of the counters in
the loop could be avoided.</li>

<li>You could implement one of the "static profiling" algorithms which analyze a
piece of code an make educated guesses about the relative execution frequencies
of various parts of the code.</li>

<li>You could add path profiling support, or adapt the existing LLVM path
profiling code to work with the generic profiling interfaces.</li>
</ol>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="compaction">Code Compaction</a>
</div>

<div class="www_text">
<p>LLVM aggressively optimizes for performance, but does not yet optimize for code size.
With a new ARM backend, there is increasing interest in using LLVM for embedded systems
where code size is more of an issue.
</p>

<p>Someone interested in working on implementing code compaction in LLVM might want to read
<a href="http://citeseer.ist.psu.edu/425696.html">this</a> article, describing using
link-time optimizations for code size optimization.
</p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="xforms">New Transformations and Analyses</a>
</div>

<div class="www_text">

<ol>
  <li>Implement a Loop Dependence Analysis Infrastructure<br>
    - Design some way to represent and query dep analysis</li>
  <li>Value range propagation pass</li>
  <li>More fun with loops:
    <a href="http://www.cs.ualberta.ca/~amaral/cascon/CDP04/tal.html">
      Predictive Commoning
    </a>
  </li>
  <li>Type inference (aka. devirtualization)</li>
  <li><a href="http://nondot.org/sabre/LLVMNotes/BuiltinUnreachable.txt">Value
      assertions</a> (also <a href="/PR810">PR810</a>).</li>
</ol>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="codegen">Code Generator Improvements</a>
</div>

<div class="www_text">

<ol>
<li>Generalize target-specific backend passes that could be target-independent,
    by adding necessary target hooks and making sure all IR/MI features (such as
    register masks and predicated instructions) are properly handled. Enable these
    for other targets where doing so is demonstrably beneficial.
    For example:
      <ol><li>lib/Target/Hexagon/RDF*</li>
          <li>lib/Target/AArch64/AArch64AddressTypePromotion.cpp</li>
     </ol>
    </li>
<li>Merge the delay slot filling logic that is duplicated into (at least)
    the Sparc and Mips backends into a single target independent pass.
     Likewise, the branch shortening logic in several targets should be merged
     together into one pass.</li>
<li>Implement 'stack slot coloring' to allocate two frame indexes to the same
    stack offset if their live ranges don't overlap.  This can reuse a bunch of
    analysis machinery from LiveIntervals.  Making the stack smaller is good
    for cache use and very important on targets where loads have limited
    displacement like ppc, thumb, mips, sparc, etc.  This should be done as
    a pass before prolog epilog insertion.  This is now done for register
    allocator temporaries, but not for allocas.</li>
<li>Implement 'shrink wrapping', which is the intelligent placement of callee
    saved register save/restores.  Right now PrologEpilogInsertion always saves
    every (modified) callee save reg in the prolog and restores it in the
    epilog, however, some paths through a function (e.g. an early exit) may
    not use all regs.  Sinking the save down the CFG avoids useless work on
    these paths. Work has started on this, please inquire on llvm-dev.</li>
<li>Implement interprocedural register allocation. The CallGraphSCCPass can be
    used to implement a bottom-up analysis that will determine the *actual*
    registers clobbered by a function. Use the pass to fine tune register usage
    in callers based on *actual* registers used by the callee.</li>
<li>Add support for 16-bit x86 assembly and real mode to the assembler and
    disassembler, for use by BIOS code. This includes both 16-bit instruction
    encodings as well as privileged instructions (lgdt, lldt, ltr, lmsw, clts,
    invd, invlpg, wbinvd, hlt, rdmsr, wrmsr, rdpmc, rdtsc) and the control and
    debug registers.
</ol>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="misc_new">Miscellaneous Additions</a>
</div>

<div class="www_text">

<ol>
<li>Port the <a href="http://www-sop.inria.fr/mimosa/fp/Bigloo/">Bigloo</A>
Scheme compiler, from Manuel Serrano at INRIA Sophia-Antipolis, to
output LLVM bytecode. It seems that it can already output .NET
bytecode, JVM bytecode, and C, so LLVM would ostensibly be another good
candidate.</li>
<li>Write a new frontend for some other language (Java? OCaml? Forth?)</li>
<li>Random test vector generator: Use a C grammar to generate random C code,
e.g., <a href="http://code.google.com/p/quest-tester/">quest</a>;
run it through llvm-gcc, then run a random set of passes on it using opt.
Try to crash <tt><a href="/docs/CommandGuide/html/opt.html">opt</a></tt>. When
<tt>opt</tt> crashes, use <tt><a
href="/docs/CommandGuide/html/bugpoint.html">bugpoint</a></tt> to reduce the
test case and post it to a website or mailing list.  Repeat ad infinitum.</li>
<li>Add sandbox features to the Interpreter: catch invalid memory accesses,
  potentially unsafe operations (access via arbitrary memory pointer) etc.
</li>
<li>Port <a href="http://valgrind.org">Valgrind</a> to use LLVM code generation
  and optimization passes instead of its own.</li>
<li>Write LLVM IR level debugger (extend Interpreter?)</li>
<li>Write an LLVM Superoptimizer.  It would be interesting to take ideas from
    this superoptimizer for x86:
<a href="http://theory.stanford.edu/~aiken/publications/papers/asplos06.pdf">paper #1</a> and <a href="http://theory.stanford.edu/~sbansal/superoptimizer.html">paper #2</a> and adapt them to run on LLVM code.<p>

It would seem that operating on LLVM code would save a lot of time
because its semantics are much simpler than x86.  The cost of operating
on LLVM is that target-specific tricks would be missed.<p>

The outcome would be a new LLVM pass that subsumes at least the
instruction combiner, and probably a few other passes as well.  Benefits
would include not missing cases missed by the current combiner and also
more easily adapting to changes in the LLVM IR.<p>

All previous superoptimizers have worked on linear sequences of code.
It would seem much better to operate on small subgraphs of the program
dependency graph.</li>
</ol>

</div>

<!-- *********************************************************************** -->
<div class="www_sectiontitle">
  <a name="using">Projects using LLVM</a>
</div>
<!-- *********************************************************************** -->

<div class="www_text">

  <p>
  In addition to projects that enhance the existing LLVM infrastructure, there
  are projects that improve software that uses, but is not included with, the
  LLVM compiler infrastructure.  These projects include open-source software
  projects and research projects that use LLVM.  Like projects that enhance the
  core LLVM infrastructure, these projects are often challenging and rewarding.
  </p>

</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="encodeanalysis">Encode Analysis Results in MachineInstr IR</a>
</div>

<div class="www_text">
  <p>
  At least one project (and probably more) needs to use analysis information
  (such as call graph analysis) from within a MachineFunctionPass, however,
  most analysis passes operate at the LLVM IR level.  In some cases, a value
  (e.g., a function pointer) cannot be mapped from the MachineInstr level back
  to the LLVM IR level reliably, making the use of existing LLVM analysis
  passes from within a MachineFunctionPass impossible (or at least brittle).
  </p>

  <p>
  This project is to encode analysis information from the LLVM IR level into
  the MachineInstr IR when it is generated so that it is available to a
  MachineFunctionPass.  The exemplar is call graph analysis (useful for
  control-flow integrity instrumentation, analysis of code reuse defenses, and
  gadget compilers); however, other LLVM analyses may be useful.
  </p>
</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="codelayoutjit">Code Layout in the LLVM JIT</a>
</div>

<div class="www_text">
  <p>
  Implement an on-demand function relocator in the LLVM JIT. This can help
  improve code locality using runtime profiling information. The idea is to use
  a relocation table for every function.  The relocation entries need to be
  updated upon every function relocation (take a look at
  <a href="https://people.cs.umass.edu/~emery/pubs/stabilizer-asplos13.pdf">
  this article</a>).
  A (per-function) basic block reordering would be a useful extension.
  </p>
</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="fieldlayout">Improved Structure Splitting and Field Reordering</a>
</div>

<div class="www_text">
  <p>
  The goal of this project is to implement better data layout optimizations
  using the model of reference affinity.  This
  <a href="http://www.cs.rochester.edu/~cding/Documents/Publications/pldi04.pdf">
  paper</a>
  provides some background information.
  </p>
</div>

<!-- ======================================================================= -->
<div class="www_subsubsection">
  <a name="slimmer">Finish the Slimmer Project</a>
</div>

<div class="www_text">
  <p>
  Slimmer is a prototype tool, built using LLVM, that uses dynamic analysis to
  find potential performance bugs in programs.  Development on Slimmer started
  during Google Summer of Code in 2015 and resulted in an initial prototype,
  but evaluation of the prototype and improvements to make it portable and
  robust are still needed.  This project would have a student pick up and
  finish the Slimmer work.  The source code of Slimmer and
  its current documentation can be found at its
  <a href="https://github.com/james0zan/Slimmer">Github</a> web page.
  </p>
</div>

<!-- *********************************************************************** -->

<hr>

<!--#include virtual="footer.incl" -->