[HLSL] Add support for root constant generation from llvm IR. #127932
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 30 commits
February 19, 2025 21:38
… a virtual register. (llvm#129017) NVPTX needs to be able to get the Dwarf register number for a virtual register. The interface we have for this today is on MCRegisterInfo and take a MCRegister argument. It shouldn't be legal to convert a Register containing a virtual register to an MCRegister. This patch adds a getDwarfRegNumForVirtReg function that takes a Register to TRI and splits the NVPTX override of getDwarfRegNum.
…auses of taskloop construct (llvm#128477) Added modifier(strict) support to the granularity(grainsize and num_tasks) clauses of taskloop construct.
…m#129066) When we need to reclone other callees of a caller node during function assignment due to the creation of a new function clone, we need to skip recursive edges on that caller. We don't want to reclone the callee in that case (which is the caller), which isn't necessary and also isn't correct from a graph update perspective. It resulted in an assertion and in an NDEBUG build caused an infinite loop.
…ars with multi uses If the vectorized scalars has multiple uses, need to check if it is safe to truncate the vectorized value, before actually trying doing it. Otherwise, the compiler may loose some important bits, which may lead to a miscompilation. Fixes llvm#129057
This set usage of operator& instead of std::addressof seems not be easy to "abuse". Some seem easy to misuse, like basic_ostream::operator<<, trying to do that results in compilation errors since the `widen` function is not specialized for the hijacking character type. Hence there are no tests.
…lvm#128897) This patch changes the codegen for non-precise erfc calls to generate math.erfc ops. This wasn't done before because the math dialect did not have a erfc operation at the time.
…vm#126068) Multiple improvements to make the messages more concrete, actionable and less confusing when multiple prefixes are used in `-verify=`. The common theme among these was that prior to the patch all error messages would use the alphabetically first prefix, even if the error was associated with a different one. - Mention the actual expected but unseen directive: Prior to this change when reporting expected but unseen directive, the alphabetically first one would be used to report the error even if that's not the one present in the source. Reword the diagnostic if multiple prefixes are active and include the real spelling of the expected directive for each expected but not seen line in the output. - Reword the seen but not expected error message if multiple directives are active to avoid having to pick an arbitrary (the first) prefix for it. - Include the full spelling of the directive when reporting a directive following the no-diagnostics directive. For example "'foo-error' directive cannot follow 'foo-no-diagnostics' directive" - Use the first appearing `-no-diagnostics` directive, in the above message instead of the first one alphabetically. The new wording > diagnostics with '(error|warning|remark|note)' severity seen but not expected instead of > '<prefix>-(error|warning|remark|note)' diagnostics seen but not expected is only used when multiple prefixes are present, the error messages stay the same with a single prefix only.
For GFX10+ the destination reg of v_cmpx instructions is implicitly EXEC, which is encoded as 0x7E. However, the disassembler does not check this field, thus allowing any value. With this patch, if the field is not EXEC a warning is issued.
…lvm#126621)" (llvm#129078) This reverts commit 1559a65. Fixed test (I suspect broken by unrelated change in the merge)
Previously checker objects were created by raw `new` calls, which necessitated managing and calling their destructors explicitly. This commit refactors this convoluted logic by introducing `unique_ptr`s that to manage the ownership of these objects automatically. This change can be thought of as stand-alone code quality improvement; but I also have a secondary motivation that I'm planning further changes in the checker registration/initialization process (to formalize our tradition of multi-part checker) and this commit "prepares the ground" for those changes.
) Avoid negations for normalized blends by reordering operands. PR: llvm#128375
In several cases the flags entries in ompt_frame_t are not initialized. According to @jdelsign the address provided as reenter and exit address is the canonical frame address (cfa) rather than a "framepointer". This patch makes sure that the flags entry is always initialized and changes the value from ompt_frame_framepointer to ompt_frame_cfa. The assertion in the tests makes sure that the flags are always set, when a tool (callback.h in this case) looks at the value. Fixes llvm#89058
PeepholeOpt has a nicer version of this which handles more cases.
…lvm#128413) Fixes llvm#113359 --------- Signed-off-by: krishna2803 <[email protected]>
Shared state between dependency scanning workers is managed by the dependency scanning service. Right now, the members are individually threaded through the worker, action, and collector. This makes any change to the service and its members a very laborious process. Moreover, this situation causes frequent merge conflicts in our downstream repo where the service does have some extra members that need to be passed around. To ease the maintenance burden, this PR starts passing a reference to the entire service.
This is similar to llvm#127317
…-back (llvm#128092) This patch fixes the behavior of the scheduler by making sure the instrs that are part of a SchedBundle are scheduled back-to-back.
joaosaffran
force-pushed
the
root-signature/root-constants
branch
from
d6ace04 to
79cdfff
Compare
|
Please open a new PR after verifying that it only contains your changes. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds support for root constant metadata extraction and generation in dxcontainer. It's changes include: