LLVM and SPIRV-LLVM-Translator pulldown (WW08 2026)#21335
Draft
LLVM and SPIRV-LLVM-Translator pulldown (WW08 2026)#21335
Conversation
Compressing to a single shuffle doesn't remove any information and the backend can better apply specific optimizations to a single shuffle. Addresses #176218. --------- Co-authored-by: Luke Lau <luke_lau@igalia.com>
…9985) Extend operands when computing ub - lb to avoid overflow in signed arithmetic. E.g., i8: ub=127, lb=-128 yields 255, which overflows without extension.
Adds missing test coverage for reductions with intermediate stores, including partial reductions with intermediate stores, as well as chained min/max reductions with intermediate stores.
Clinger fast path bloats baremetal targets which are constrained in binary size. Disabling it for baremetal libc builds.
The code seems to have considered the potential problem but did not quite succeed in solving it ;)
Use new UTC support to re-generate check lines.
An AI told me these were missing and helped me add them.
…296) The logic of `is_x86_64_non_windows` looks unnecessarily complicated and is only used at one site... clean up the unused targets and refactor x86_64 BLAKE3 asm sources into a separate filegroup. And then `is_x86_64_non_windows` can be put inside a default condition.
Add set of FindLast tests where the selected expression is based on an IV and could be sunk.
…serves X15 (#179738) The target function to be checked by the Control Flow Guard Check function is stored in `X15` on AArch64. This register is guaranteed to be preserved by that function (on success), thus after it returns `X15` can be used to branch to the target function instead of having to load it from another register or the stack.
…180218) Fixes error of handing constant integers with width in (64; 128) range. Found during review of llvm/llvm-project#180182
…n (#180347) Update stale links and remove duplication in table.
CONFLICT (content): Merge conflict in llvm/lib/Transforms/IPO/IPO.cpp
…ared-libsan` (#164842)
This PR contains two commits:
- Add required dependencies when using `-shared-libsan` and fuzzer.
Since libFuzzer is a static library we need to make sure that we add its
dependencies when building with `-shared-libsan`. E.g libFuzzer uses
`ceilf()` from `libm.so` when building on Gnu toolchain.
Previously, the resulting command did not contain the required link
libraries, giving build failures
(only a static sanitizer runtime would trigger the call to
`linkSanitizerRuntimeDeps`).
- Correcting dependency order when using fuzzer.
When building using `-shared-libsan` the sanitizer library needs to be
first in link order.
Since the fuzzer requires `-lstdc++` we have to make sure that the
sanitizer library is added before `-lstdc++`.
---------
Signed-off-by: Björn Svensson <bjorn.a.svensson@est.tech>
Get the shared cache filepath and uuid that the inferior process is using from debugserver, try to open that shared cache on the lldb host mac and if the UUID matches, index all of the binaries in that shared cache. When looking for binaries loaded in the process, get them from the already-indexed shared cache. Every time a binary is loaded, PlatformMacOSX may query the shared cache filepath and uuid from the Process, and pass that to HostInfo::GetSharedCacheImageInfo() if available (else fall back to the old HostInfo::GetSharedCacheImageInfo method which only looks at lldb's own shared cache), to get the file being requested. ProcessGDBRemote caches the shared cache filepath and uuid from the inferior, once it has a non-zero UUID. I added a lock for this ivar specifically, so I don't have 20 threads all asking for the shared cache information from debugserver and updating the cached answer. If we never get back a non-zero UUID shared cache reply, we will re-query at every library loaded notification. debugserver has been providing the shared cache UUID since 2013, although I only added the shared cache filepath field last November. Note that a process will not report its shared cache filepath or uuid at initial launch. As dyld gets a chance to execute a bit, it will start returning binaries -- it will be available at the point when libraries start loading. (it won't be available yet when the binary & dyld are the only two binaries loaded in the process) I tested this by disabling lldb's scan of its own shared cache pre-execution -- only loading the system shared cache when the inferior process reports that it is using that. I got 6-7 additional testsuite failures running lldb like that, because no system binaries were loaded before exeuction start, and the tests assumed they would be. rdar://148939795 --------- Co-authored-by: Jonas Devlieghere <jonas@devlieghere.com>
…se notes (#180299) (#180650) We were using one token for both pushing to the llvmbot fork and for creating a pull request against the www-releases repository, since the fork and the repository have different owners, we were using a classic access token which has very coarse-grained permissions. By using two separate tokens, we limit the permissions to just what we need to do the task. This is a re-commit of b6ee085 minus the environment changes which were causing the workflow to fail.
This adds atomicrmw `uinc_wrap` and `udec_wrap` operations support for SPIR-V. Since SPIR-V doesn't provide dedicated instructions for those two operations, we have to use the `AtomicExpand` pass to expand the operations into CAS forms. Closes #177204.
The test was manually generated and out-of-sync with `cross-project-tests/debuginfo-tests/clang_llvm_roundtrip/Inputs/simplified_template_names.cpp`. We update the test such that: 1. Automate the test generation process by using `llvm/utils/update_test_body.py` 2. Remove host machine info when updating the tests Predecessor of #178986, the PR was split per reviewer's request, since that change disturbed this test a lot.
…terType (#178345) This patch introduces `get(T)` and `set(T, Val)` functions for Waitcnt and removes getCounterRef() and getWait(). For this to work we also need to move InstrCounterType to AMDGPUBaseInfo.h. Please note that the member variables are still public to keep this patch small. They will be replaced in the follow-up patch.
When a type test has two phases and is used by llvm.cond.loop to implement a conditional trap, it is more efficient for two infinite loops to be generated. Arrange for this by having the pass detect the typical IR pattern used for conditional CFI traps and generate the second llvm.cond.loop if found. Part of this RFC: https://discourse.llvm.org/t/rfc-optimizing-conditional-traps/89456 Reviewers: fmayer, vitalybuka Reviewed By: vitalybuka Pull Request: llvm/llvm-project#177687
This PR is mainly to address review suggestions in #179705.
…paces (#180661) Since #171876, -amdgpu-to-rocdl (the pass) is now set up to handle address spaces like `#gpu.address_space<global>`. Update the tests accordingly.
Add S16 rules for G_FSQRT. S32 and S64 are expanded by the legalizer.
…e (#180496) It should be a tune feature just like others.
…exception-escape` (#168324) As of AI Usage: Gemini 3 was used for rephrasing the documentation. Closes llvm/llvm-project#164795 --------- Co-authored-by: EugeneZelenko <eugene.zelenko@gmail.com> Co-authored-by: Baranov Victor <bar.victor.2002@gmail.com>
Regression caused by 0dd21ad which removed address space casts from CreateIRTemp assuming temporaries are only used for loads/stores. However, on SPIR targets, regcall functions (like invoke_simd helpers) may call spir_func functions with different calling conventions. When the callee uses sret with generic address space (ptr addrspace(4)), the ReturnValue temporary must have the correct address space: - invoke_simd's simd_func_call_helper uses x86_regcallcc - It calls user functions with spir_func that expect sret with ptr addrspace(4) - Without address space cast, ReturnValue is ptr (private AS 0) - This causes: "Calling a function with a bad signature!" This patch adds a targeted fix: only apply address space cast for regcall functions on SPIR/SPIRV targets, implementing CreateIRTemp inline to avoid modifying headers. Fixes: sycl/test/invoke_simd/return-type-struct.cpp
jsji
force-pushed
the
llvmspirv_pulldown
branch
from
1e821b9 to
267c7fa
Compare
Should use l32 similar to clang/test/Driver/sycl-libspirv-toolchain.cpp " // Select remangled libclc variant. // Decide long size based on host triple, because offloading targets are going // to match that. **// All known windows environments except Cygwin use 32-bit long.** " This is a bug when resolving conflicts. commit ed037fb Merge: 5934308 c5cb48c Author: Wenju He <wenju.he@intel.com> Date: Tue Jan 27 05:24:14 2026 +0100 Merge from 'main' to 'sycl-web' (45 commits) CONFLICT (content): Merge conflict in libclc/cmake/modules/AddLibclc.cmake diff --git a/clang/test/Driver/sycl-nvptx-link.cpp b/clang/test/Driver/sycl-nvptx-link.cpp index 9f3d0bf..b51401d 100644 --- a/clang/test/Driver/sycl-nvptx-link.cpp +++ b/clang/test/Driver/sycl-nvptx-link.cpp @@ -42,7 +42,7 @@ // CHECK: llvm-link // CHECK-SAME: -only-needed -// CHECK-SAME: libspirv-nvptx64-nvidia-cuda.bc +// CHECK-SAME: remangled-l64-signed_char.libspirv.bc // LIBDEVICE10-SAME: libdevice.10.bc // LIBDEVICE30-SAME: libdevice.compute_30.10.bc // LIBDEVICE35-SAME: libdevice.compute_35.10.bc
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
jsji
force-pushed
the
llvmspirv_pulldown
branch
from
de4d278 to
346832c
Compare
Contributor
|
This is ready for review.
Others are cherry-picks. |
sarnex
reviewed
Feb 25, 2026
| CGM.getTarget().getTriple().isSPIRV()) && | ||
| CurFnInfo->getCallingConvention() == llvm::CallingConv::X86_RegCall; | ||
|
|
||
| if (NeedsAddrSpaceCast) { |
Contributor
There was a problem hiding this comment.
The fix itself seems fine but is this a bug in the original upstream change?
Contributor
There was a problem hiding this comment.
The problem only happens with ESIMD tests here, so not sure whether upstream would consider such case.
Contributor
There was a problem hiding this comment.
If the ESIMD IR is valid then I would argue they have to consider such a case. If it is invalid or breaking some rule then of course they shouldn't.
Contributor
There was a problem hiding this comment.
Yes, I agree. We probably can try to upstream this fix later.
Contributor
|
The two CFE items that @jsji asked me to look at seem okay to me. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
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.
LLVM: llvm/llvm-project@8b9fd48
SPIRV-LLVM-Translator: KhronosGroup/SPIRV-LLVM-Translator@8dfdbae11aaf8ba