LLVM and SPIRV-LLVM-Translator pulldown (WW13 2025) #17710

iclsrc · 2025-03-28T14:49:52Z

LLVM: llvm/llvm-project@a6c09d4
SPIRV-LLVM-Translator: KhronosGroup/SPIRV-LLVM-Translator@22a520d

…/packed-fp32.ll` (#131629)

…1486) Fixes #131359 After #129645, a first-order recurrence will no longer have it's splice costed if the VPInstruction::FirstOrderRecurrenceSplice has no users and is dead. The legacy cost model didn't account for this, so this accounts for it in planContainsAdditionalSimplifications to avoid the "VPlan cost model and legacy cost model disagreed" assertion.

…(#131401) Until recently checkFeature was quite slow. #130936 I was curious where we use checkFeature and noticed these. I thought we could use hasFeature instead of going through strings.

…0176) When negating an FMA opcode, a new node created for a negated FMA operand may be deleted while recursively negating another FMA operand. This causes the following assertion to fail: ``` llc: /root/llvm-project/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:7686: llvm::SDValue llvm::SelectionDAG::getNode(unsigned int, const llvm::SDLoc&, llvm::EVT, llvm::SDValue, llvm::SDValue, llvm::SDValue, llvm::SDNodeFlags): Assertion `N1.getOpcode() != ISD::DELETED_NODE && N2.getOpcode() != ISD::DELETED_NODE && N3.getOpcode() != ISD::DELETED_NODE && "Operand is DELETED_NODE!"' failed. ``` This patch adds a temporary handle on the new negated nodes to prevent them from being deleted. For eg. see https://godbolt.org/z/Tq4PvnKM4 . Co-authored-by: Vineet Kumar <[email protected]>

This is split off from #131300. A VPReductionRecipe will never have a AnyOf or FindLastIV recurrence, so when it calls createReduction it always calls createSimpleReduction. If we replace the call then it leaves createReduction with one user in VPInstruction::ComputeReductionResult, which we can inline and then remove.

This commit resolves an issue in the LLVMIR export that caused the duplication of the "Debug Info Version" module flag, when it was already in MLIR.

Fixes #104862

@nickdesaulniers

…ists (#126923) These changes tracks `glob.h` for the implementation status of functions and macros, with respect to the issue ( #122006 ) . cc @nickdesaulniers

…ponse. (#130090) This is a work in progress refactor to add explicit types instead of generic 'llvm::json::Value' types to the DAP protocol. This updates RequestHandler to have take the type of the arguments and response body for serialization for requests. The 'source' and 'disconnect' request is updated to show how the new flow works and includes serialization handling for optional arguments and 'void' responses. This is built on top of #130026 --------- Co-authored-by: Adrian Vogelsgesang <[email protected]>

CONFLICT (content): Merge conflict in llvm/tools/sycl-post-link/sycl-post-link.cpp

And diagnose if we're trying to destroy an inactive member of a union.

Need to check VectorizableTree is not empty before accessing elements. Fixes #131635

This reverts commit f9146cc ([libc][bazel] explicitly use system-provided errno in Bazel builds. (#130663)) This change causes problems in Bazel builds where system errno is set to non-zero before the tests even begin to run - see PR #131650 for the disucssion on how to address this.

Add additional vector scalar overloads for clamp using templates Add Tests fixup tests which have changed. Closes #128230

Implements GCC's constexpr string ASM extension https://gcc.gnu.org/onlinedocs/gcc/Asm-constexprs.html

…27354) This patch adds benchmarks for all the remaining algorithms in [alg.modifying.operations] that we didn't already have a benchmark for.

…12-bit types

We need to add commutated variants (to match the original bug report), some of which are still failing....

I messed this up in #130800.

… (#131409) Rename RISCV32GPRPair and RISCV32Only_ to RV32Only. This gives a more natural home for the P extension RV32 conflicts. While I was there I made some improvements to the debug messages.

… (#131055) Report the range in diagnostics, in addition to the location in case the range helps disambiguate a little in chained `->` expressions. ``` b->a->f->x = 1; ^~~~~~~ ``` instead of just: ``` b->a->f->x = 1; ^ ``` As a followup we should probably also report the location/range of an `->` if that operator is used. Like: ``` b->a->f->x = 1; ^~ ```

…urn types (#131035) Fixes #131024. - Fixes template for scalar and vector `fmod` intrinsic overloads - Fixes `fmod` Sema test

jsji · 2025-04-06T18:40:34Z

This is ready for review. (Failures in https://github.com/intel/llvm/actions/runs/14293545650/job/40057838391?pr=17710 is CI bug will be fixed in #17875)

Issues opened, needs follow up:

Need reviews:

Revert "[LLVM] Only enable -fno-lifetime-dse in LTO mode (#131381)" @intel/dpcpp-tools-reviewers
[NFC][SYCL] Regen test after sret addressspace change @intel/dpcpp-cfe-reviewers
[NFC][HIP] Update __clang_hip_math after sret addressspace change @intel/dpcpp-cfe-reviewers

Other fixes are cherry-pick of already reviewed changes.

asudarsa

About 4604246

It is interesting that such a change results in errors in sycl-ls. Reverting seems ok to unblock pulldown. But, issue should be triaged soon.

Thanks

premanandrao · 2025-04-07T15:05:32Z

sycl/test/check_device_code/extensions/address_cast.cpp

 // CHECK-NEXT:    ret void
 //
 SYCL_EXTERNAL auto to_global_decorated(decorated_generic_ptr<int> p) {
  return static_address_cast<access::address_space::global_space>(p);
 }
 // CHECK-LABEL: define dso_local spir_func void @_ZN14static_as_cast23to_global_not_decoratedEPi(
-// CHECK-SAME: ptr addrspace(4) dead_on_unwind noalias writable writeonly sret(%"class.sycl::_V1::multi_ptr.1") align 8 captures(none) initializes((0, 8)) [[AGG_RESULT:%.*]], ptr addrspace(4) noundef [[P:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] !srcloc [[META20:![0-9]+]] !sycl_fixed_targets [[META7]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable writeonly sret(%"class.sycl::_V1::multi_ptr.1") align 8 captures(none) [[AGG_RESULT:%.*]], ptr addrspace(4) noundef [[P:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] {{.*}}{


Are we losing the "initializes" info with this change?

Yes, that looks like the side effect of addspace change

The CHECK lines are autogened by script, and pass the test.

Okay, I think it may be fine.

jsji · 2025-04-07T15:17:36Z

About 4604246

It is interesting that such a change results in errors in sycl-ls. Reverting seems ok to unblock pulldown. But, issue should be triaged soon.

Thanks

I believe it is exactly the same problem with llvm/llvm-project#24952, but sure,
#17888 opened for team to follow up to double confirm.

jsji · 2025-04-07T15:18:59Z

@intel/llvm-gatekeepers I think this is read for merge. Unless we want to wait for CI fix.

dkhaldi · 2025-04-07T15:40:39Z

sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_unroll.cpp

@@ -8,6 +8,8 @@
 // UNSUPPORTED: target-nvidia, target-amd
 // UNSUPPORTED-INTENDED: aspect-ext_intel_matrix isn't currently supported for
 // other triples
+// XFAIL: run-mode && arch-intel_gpu_bmg_g21


Does this pass on PVC? only BMG is affected?

There are NO PVC run in CI for now.

I believe it should affect PVC as well.

dkhaldi · 2025-04-07T15:41:15Z

sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_unroll_init.cpp

@@ -9,6 +9,9 @@
 // UNSUPPORTED-INTENDED: aspect-ext_intel_matrix isn't currently supported for
 // other triples

+// XFAIL: run-mode && arch-intel_gpu_bmg_g21


Does this pass on PVC? only BMG is affected?

jsji · 2025-04-08T00:01:43Z

@intel/llvm-gatekeepers Can someone help to issue a /merge? Thanks!

uditagarwal97 · 2025-04-08T00:03:42Z

/merge

jsji · 2025-04-08T00:17:37Z

/merge

:(, stop working again... @DoyleLi

bb-sycl · 2025-04-08T00:34:58Z

Tue 08 Apr 2025 12:34:58 AM UTC --- Start to merge the commit into sycl branch. It will take several minutes.

bb-sycl · 2025-04-08T00:45:13Z

Tue 08 Apr 2025 12:45:12 AM UTC --- Merge the branch in this PR to base automatically. Will close the PR later.

DoyleLi · 2025-04-08T01:08:35Z

/merge

:(, stop working again... @DoyleLi

Hi @jsji should be infrastructure issue from node. Now the system is back online.

jsji · 2025-04-08T02:00:35Z

/merge

:(, stop working again... @DoyleLi

Hi @jsji should be infrastructure issue from node. Now the system is back online.

Great. Thanks!

shiltian and others added 30 commits March 17, 2025 11:42

[NFC][AMDGPU] Auto generate check lines for `llvm/test/CodeGen/AMDGPU…

e2c43ba

…/packed-fp32.ll` (#131629)

[RISCV] Use hasFeature instead of checkFeature in llvm-exegesis. NFC …

b00ad36

…(#131401) Until recently checkFeature was quite slow. #130936 I was curious where we use checkFeature and noticed these. I thought we could use hasFeature instead of going through strings.

[MLIR][LLVM] Avoid duplicated module flags in the export (#131627)

800593a

This commit resolves an issue in the LLVMIR export that caused the duplication of the "Debug Info Version" module flag, when it was already in MLIR.

[X86] add test coverage for concatenation to 512-bit VPERMILPD nodes

561f1d0

[X86] Add additional test coverage for #109272

9a92fe0

[HLSL] Remove HLSLResource attribute (#130342)

b3c5031

Fixes #104862

[libc][docs] Add glob implementation status doc and include in CMakeL…

47f7daa

…ists (#126923) These changes tracks `glob.h` for the implementation status of functions and macros, with respect to the issue ( #122006 ) . cc @nickdesaulniers

[gn build] Port fbb8929

681b241

[clang][driver][NFC] Remove else after return (#131182)

0460418

[libc++][NFC] Remove dead link in comment

c53caae

Merge from 'sycl' to 'sycl-web' (2 commits)

7a46cca

CONFLICT (content): Merge conflict in llvm/tools/sycl-post-link/sycl-post-link.cpp

[clang][bytecode] Check dtor instance pointers for active-ness (#128732)

ca1bde0

And diagnose if we're trying to destroy an inactive member of a union.

[clang][bytecode] Fix builtin_memchr with non-0 start index (#131633)

cfa07cc

[SLP]Check VectorizableTree is not empty before accessing elements

ead9d6a

Need to check VectorizableTree is not empty before accessing elements. Fixes #131635

[clang][CIR] Add missing dependency on MLIR headers (#131057)

f6a7306

[HLSL] add extra scalar vector overloads for clamp (#129939)

af5abd9

Add additional vector scalar overloads for clamp using templates Add Tests fixup tests which have changed. Closes #128230

[gn build] Port af5abd9

2443fe5

[Clang] Constant Expressions inside of GCC' asm strings (#131003)

911b200

Implements GCC's constexpr string ASM extension https://gcc.gnu.org/onlinedocs/gcc/Asm-constexprs.html

[libc++] Add remaining benchmarks from [alg.modifying.operations] (#1…

24e88b0

…27354) This patch adds benchmarks for all the remaining algorithms in [alg.modifying.operations] that we didn't already have a benchmark for.

[X86] combineConcatVectorOps - extend VPERMILPD handling to support 5…

20cdffb

…12-bit types

[X86] Cleanup test coverage for #109272

bbaf743

We need to add commutated variants (to match the original bug report), some of which are still failing....

[RISCV] Put CV_ELW back in XCV DecoderNamespace.

9eb6b37

I messed this up in #130800.

[RISCV] Rename some DecoderNamespaces and cleanup debug messages. NFC…

d2e1e30

… (#131409) Rename RISCV32GPRPair and RISCV32Only_ to RV32Only. This gives a more natural home for the P extension RV32 conflicts. While I was there I made some improvements to the debug messages.

[HLSL] Add bounds checks for the HLSL fmod vector arguments and ret…

e3ef5f2

…urn types (#131035) Fixes #131024. - Fixes template for scalar and vector `fmod` intrinsic overloads - Fixes `fmod` Sema test

jsji temporarily deployed to WindowsCILock April 6, 2025 14:34 — with GitHub Actions Inactive

jsji temporarily deployed to WindowsCILock April 6, 2025 15:01 — with GitHub Actions Inactive

jsji temporarily deployed to WindowsCILock April 6, 2025 15:22 — with GitHub Actions Inactive

asudarsa approved these changes Apr 7, 2025

View reviewed changes

premanandrao reviewed Apr 7, 2025

View reviewed changes

premanandrao approved these changes Apr 7, 2025

View reviewed changes

dkhaldi reviewed Apr 7, 2025

View reviewed changes

Merge remote-tracking branch 'origin/sycl' into llvmspirv_pulldown

62f1b55

jsji temporarily deployed to WindowsCILock April 7, 2025 16:59 — with GitHub Actions Inactive

jsji temporarily deployed to WindowsCILock April 7, 2025 17:00 — with GitHub Actions Inactive

jsji temporarily deployed to WindowsCILock April 7, 2025 17:53 — with GitHub Actions Inactive

jsji temporarily deployed to WindowsCILock April 7, 2025 18:14 — with GitHub Actions Inactive

bb-sycl approved these changes Apr 8, 2025

View reviewed changes

bb-sycl merged commit 22c6c11 into sycl Apr 8, 2025
52 of 54 checks passed

jsji deleted the llvmspirv_pulldown branch April 8, 2025 02:00

LLVM and SPIRV-LLVM-Translator pulldown (WW13 2025) #17710

LLVM and SPIRV-LLVM-Translator pulldown (WW13 2025) #17710

Uh oh!

Conversation

iclsrc commented Mar 28, 2025

Uh oh!

jsji commented Apr 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

asudarsa left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

premanandrao Apr 7, 2025

Choose a reason for hiding this comment

Uh oh!

jsji Apr 7, 2025

Choose a reason for hiding this comment

Uh oh!

jsji Apr 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

premanandrao Apr 7, 2025

Choose a reason for hiding this comment

Uh oh!

jsji commented Apr 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jsji commented Apr 7, 2025

Uh oh!

dkhaldi Apr 7, 2025

Choose a reason for hiding this comment

Uh oh!

jsji Apr 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jsji Apr 7, 2025

Choose a reason for hiding this comment

Uh oh!

dkhaldi Apr 7, 2025

Choose a reason for hiding this comment

Uh oh!

jsji commented Apr 8, 2025

Uh oh!

uditagarwal97 commented Apr 8, 2025

Uh oh!

jsji commented Apr 8, 2025

Uh oh!

bb-sycl commented Apr 8, 2025

Uh oh!

bb-sycl commented Apr 8, 2025

Uh oh!

Uh oh!

DoyleLi commented Apr 8, 2025

Uh oh!

jsji commented Apr 8, 2025

Uh oh!

Uh oh!

jsji commented Apr 6, 2025 •

edited

Loading

asudarsa left a comment •

edited

Loading

jsji Apr 7, 2025 •

edited

Loading

jsji commented Apr 7, 2025 •

edited

Loading

jsji Apr 7, 2025 •

edited

Loading