Skip to content

[LLVM] Bump to 75aa01b89 #257

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
makslevental merged 3 commits into from
Dec 2, 2025
Merged

[LLVM] Bump to 75aa01b89 #257

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
08a049f
[LLVM] Bump to 75aa01b89
github-actions[bot] Dec 1, 2025
2ec3231
fix linalg.fill
makslevental Dec 1, 2025
ca1aa4c
fix example
makslevental Dec 2, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 10 additions & 10 deletions projects/eudsl-python-extras/examples/cuda_matmul_opt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -415,7 +415,7 @@ def sgemm_shared_mem_1d_block_tiling[
inner_row_B = tid / BN

thread_results = memref.alloca((TM,), dtype)
linalg.fill(0, thread_results)
linalg.fill(0.0, thread_results)

for bk_idx in range_(0, K, BK):
# Move blocktile to beginning of A's row and B's column
Expand Down Expand Up @@ -483,13 +483,13 @@ def sgemm_shared_mem_2d_block_tiling[
stride_B = num_threads_blocktile // BN

thread_results = memref.alloca((TM, TN), dtype)
linalg.fill(0, thread_results)
linalg.fill(0.0, thread_results)

reg_M = memref.alloca((TM,), dtype)
linalg.fill(0, reg_M)
linalg.fill(0.0, reg_M)

reg_N = memref.alloca((TN,), dtype)
linalg.fill(0, reg_N)
linalg.fill(0.0, reg_N)

for bk_idx in range_(0, K, BK):
A_ = A[c_row : c_row + BM, bk_idx : bk_idx + BK]
Expand Down Expand Up @@ -579,13 +579,13 @@ def sgemm_shared_mem_2d_block_tiling_vectorize[
inner_row_B = tid / (BN // VECTOR_WIDTH)

thread_results = memref.alloca((TM, TN), dtype)
linalg.fill(0, thread_results)
linalg.fill(0.0, thread_results)

reg_M = memref.alloca((TM,), dtype)
linalg.fill(0, reg_M)
linalg.fill(0.0, reg_M)

reg_N = memref.alloca((TN,), dtype)
linalg.fill(0, reg_N)
linalg.fill(0.0, reg_N)

for bk_idx in range_(0, K, BK):
A_ = A[c_row : c_row + BM, bk_idx : bk_idx + BK]
Expand Down Expand Up @@ -708,13 +708,13 @@ def sgemm_warp_tiling[

# allocate thread-local cache for results in registerfile
thread_results = memref.alloca((WMITER * TM, WNITER * TN), dtype)
linalg.fill(0, thread_results)
linalg.fill(0.0, thread_results)

reg_M = memref.alloca((WMITER, TM), dtype)
linalg.fill(0, reg_M)
linalg.fill(0.0, reg_M)

reg_N = memref.alloca((WNITER, TN), dtype)
linalg.fill(0, reg_N)
linalg.fill(0.0, reg_N)

for bk_idx in range_(0, K, BK):
A_ = A[c_row : c_row + BM, bk_idx : bk_idx + BK]
Expand Down
12 changes: 6 additions & 6 deletions projects/eudsl-python-extras/tests/dialect/test_transform.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -702,7 +702,7 @@ def matmul_i8_i8(
B: T.tensor(K, N, T.i8()),
):
empty = tensor.empty(M, N, T.i8())
filled = linalg_dialect.fill(arith.constant(0), outs=[empty])
filled = linalg_dialect.fill(arith.constant(0, type=T.i8()), outs=[empty])
return linalg.matmul(A, B, filled)

@module(attrs={"transform.target_tag": StringAttr.get("payload")})
Expand Down Expand Up @@ -856,8 +856,8 @@ def main(variant_op: any_op_t()):
module attributes {transform.target_tag = "payload"} {
func.func @matmul_i8_i8(%arg0: tensor<16x256xi8>, %arg1: tensor<256x256xi8>) -> tensor<16x256xi8> {
%0 = tensor.empty() : tensor<16x256xi8>
%c0_i32 = arith.constant 0 : i32
%1 = linalg.fill ins(%c0_i32 : i32) outs(%0 : tensor<16x256xi8>) -> tensor<16x256xi8>
%c0_i8 = arith.constant 0 : i8
%1 = linalg.fill ins(%c0_i8 : i8) outs(%0 : tensor<16x256xi8>) -> tensor<16x256xi8>
%2 = linalg.matmul {cast = #linalg.type_fn<cast_signed>} ins(%arg0, %arg1 : tensor<16x256xi8>, tensor<256x256xi8>) outs(%1 : tensor<16x256xi8>) -> tensor<16x256xi8>
return %2 : tensor<16x256xi8>
}
Expand Down Expand Up @@ -924,7 +924,7 @@ def matmul_i8_i8(
B: T.tensor(K, N, T.i8()),
):
empty = tensor.empty(M, N, T.i8())
filled = linalg_dialect.fill(arith.constant(0), outs=[empty])
filled = linalg_dialect.fill(arith.constant(0, type=T.i8()), outs=[empty])
return linalg.matmul(A, B, filled)

@module(attrs={"transform.target_tag": StringAttr.get("payload")})
Expand Down Expand Up @@ -997,13 +997,13 @@ def main(variant_op: any_op_t()):
module {
module attributes {transform.target_tag = "payload"} {
func.func @matmul_i8_i8(%arg0: tensor<16x256xi8>, %arg1: tensor<256x256xi8>) -> tensor<16x256xi8> {
%c0_i32 = arith.constant 0 : i32
%c0_i8 = arith.constant 0 : i8
%0 = tensor.empty() : tensor<16x256xi8>
%1 = tensor.empty() : tensor<1x4x16x64xi8>
%pack = linalg.pack %arg0 inner_dims_pos = [0, 1] inner_tiles = [16, 64] into %1 : tensor<16x256xi8> -> tensor<1x4x16x64xi8>
%2 = tensor.empty() : tensor<4x1x64x64xi8>
%3 = tensor.empty() : tensor<1x1x16x64xi8>
%4 = linalg.fill ins(%c0_i32 : i32) outs(%3 : tensor<1x1x16x64xi8>) -> tensor<1x1x16x64xi8>
%4 = linalg.fill ins(%c0_i8 : i8) outs(%3 : tensor<1x1x16x64xi8>) -> tensor<1x1x16x64xi8>
%5 = scf.forall (%arg2, %arg3) in (1, 4) shared_outs(%arg4 = %0) -> (tensor<16x256xi8>) {
%6 = affine.apply #map(%arg3)
%extracted_slice = tensor.extract_slice %arg1[0, %6] [256, 64] [1, 1] : tensor<256x256xi8> to tensor<256x64xi8>
Expand Down
2 changes: 1 addition & 1 deletion third_party/llvm-project
Open in desktop
Submodule llvm-project updated 44 files
+4 −2 clang/include/clang/Basic/arm_mve.td
+5 −4 clang/include/clang/Basic/arm_mve_defs.td
+4 −5 clang/lib/AST/RecordLayoutBuilder.cpp
+692 −320 clang/test/CodeGen/arm-mve-intrinsics/ternary.c
+11 −4 clang/utils/TableGen/MveEmitter.cpp
+47 −4 flang/lib/Lower/Runtime.cpp
+26 −2 flang/test/Lower/pause-statement.f90
+70 −87 lld/ELF/SyntheticSections.cpp
+11 −19 lld/ELF/SyntheticSections.h
+54 −56 llvm/docs/LangRef.rst
+2 −1 llvm/include/llvm/Analysis/TargetTransformInfo.h
+12 −1 llvm/include/llvm/CodeGen/BasicTTIImpl.h
+5 −15 llvm/include/llvm/CodeGen/ISDOpcodes.h
+3 −0 llvm/include/llvm/IR/IntrinsicsARM.td
+0 −7 llvm/include/llvm/MC/MCObjectFileInfo.h
+3 −3 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+0 −4 llvm/lib/MC/MCObjectFileInfo.cpp
+12 −0 llvm/lib/Target/ARM/ARMInstrMVE.td
+16 −0 llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+4 −0 llvm/lib/Target/RISCV/RISCVISelLowering.h
+19 −0 llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+4 −0 llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
+2 −1 llvm/lib/Transforms/Vectorize/VPlanUtils.cpp
+65 −65 llvm/test/Analysis/CostModel/RISCV/vp-intrinsics.ll
+112 −12 llvm/test/CodeGen/Thumb2/mve-intrinsics/strict-intrinsics.ll
+3 −12 llvm/test/CodeGen/X86/addcarry.ll
+89 −0 llvm/test/Transforms/LoopVectorize/select-umin-first-index.ll
+2 −4 llvm/tools/llc/llc.cpp
+2 −0 llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+9 −8 mlir/docs/Dialects/Linalg/OpDSL.md
+6 −12 mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
+1 −0 mlir/lib/Dialect/ControlFlow/IR/CMakeLists.txt
+33 −1 mlir/lib/Dialect/ControlFlow/IR/ControlFlowOps.cpp
+21 −2 mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp
+4 −4 mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
+2 −2 mlir/test/Dialect/Affine/value-bounds-reification.mlir
+22 −0 mlir/test/Dialect/ControlFlow/canonicalize.mlir
+1 −25 mlir/test/Dialect/Linalg/fusion-elementwise-ops.mlir
+4 −4 mlir/test/Dialect/Linalg/generalize-named-polymorphic-ops.mlir
+18 −0 mlir/test/Dialect/Linalg/invalid.mlir
+2 −2 mlir/test/Integration/Dialect/Linalg/CPU/test-matmul-masked-vec.mlir
+2 −2 mlir/test/Integration/Dialect/Transform/match_matmul.mlir
+12 −14 mlir/test/python/integration/dialects/linalg/opsrun.py
+1 −0 utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
Loading