LLVM and SPIRV-LLVM-Translator pulldown (WW36) #6748

dbudanov-cmplr · 2022-09-09T01:45:48Z

LLVM: llvm/llvm-project@e1ebe47
SPIRV-LLVM-Translator: KhronosGroup/SPIRV-LLVM-Translator@49b08e8

…numbers for vectorization and parallelization options. The SparseTensor passes currently use opaque numbers for the CLI, despite using an enum internally. This patch exposes the enums instead of numbered items that are matched back to the enum. Fixes llvm/llvm-project#53389 Differential Revision: https://reviews.llvm.org/D123876 Please also see: https://reviews.llvm.org/D118379 https://reviews.llvm.org/D117919

GCC from 11 onwards defaults to -std=gnu++17 for C++ source files. We want to do the same (https://discourse.llvm.org/t/c-objc-switch-to-gnu-17-as-the-default-dialect/64360). Split RUN lines, adjust `-verify`, or add `__cplusplus < 201703L` or `-Wno-dynamic-exception-spec`, so that tests will pass regardless of gnu++14/gnu++17 default. We have a desire to mark a test compatible with multiple language standards. There are ongoing discussions how to add markers in the long term: * https://discourse.llvm.org/t/iterating-lit-run-lines/62596 * https://discourse.llvm.org/t/lit-run-a-run-line-multiple-times-with-different-replacements/64932 As a workaround in the short term, add lit substitutions `%std_cxx98-`, `%std_cxx11-14`, etc. They can be used for tests which work across multiple language standards. If a range has `n` standards, run lit multiple times, with `LIT_CLANG_STD_GROUP=0`, `LIT_CLANG_STD_GROUP=1`, etc to cover all `n` standards. Reviewed By: #clang-language-wg, aaron.ballman Differential Revision: https://reviews.llvm.org/D131464

Differential Revision: https://reviews.llvm.org/D132813

Differential Revision: https://reviews.llvm.org/D132814

Identified with modernize-loop-convert.

…size-latency cost kinds

…p (NFC)

…aceImpl.cpp (NFC)

This was achieved using the 'cost-tables vs llvm-mca' script D103695 Also fix a missing pmullw v16i16 half-rate throughput as znver1 double-pumps - matches numbers from AMD SoG + Agner

These are handled earlier (and more accurately) in AVX512BWShiftCostTable

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp:3397:27: warning: unused variable 'SV' [-Werror] for (const llvm::Value *SV : StoreValues) ^

…TblEntry Begin the refactoring to use CostKindTblEntry and return real latency/codesize/sizelatency costs instead of reusing the throughput numbers

Or else compiler-rt/lib/fuzzer/FuzzerUtilLinux.cpp fails to compile on FreeBSD.

…etTypeBasedIntrinsicInstrCost - convert to CostKindTblEntry" Some arm buildbots are complaining about a phase ordering test failure in unsigned-multiply-overflow-check.ll - I guess this test needs making x86 specific first

…er X86

…TblEntry Begin the refactoring to use CostKindTblEntry and return real latency/codesize/sizelatency costs instead of reusing the throughput numbers

This was achieved using the 'cost-tables vs llvm-mca' script from D103695 Some of the znver1/znver2 latency/throughput numbers were really weird (some copy+paste afaict) - I've used the numbers from the AMD SoG, which roughly match the 'worst case' range value from Agner

…ruct This patch adds private/firstprivate support for sections construct. For a source like the following: ``` !$omp sections private(x) firstprivate(y) !$omp section <block of code> !$omp section <block of code> !$omp end sections ``` ...privatization proceeds to privatize `x` and `y` accordingly inside each of the generated `omp.section` operations. Reviewed By: peixin Differential Revision: https://reviews.llvm.org/D131463

This patch introduces a new checker, called NewArraySize checker, which detects if the expression that yields the element count of the array in new[], results in an Undefined value. Differential Revision: https://reviews.llvm.org/D131299

bader · 2022-09-13T08:28:28Z

llvm-spirv/.clang-tidy has a little bit different history due to these issues introduced. We checked that currently it's matching upstream.

I think you mean here is that "it will match the upstream after WW36 pulldown will be merged". Right?

This should address the failure introduced in: https://lab.llvm.org/buildbot/#/builders/92/builds/32377

pvchupin · 2022-09-13T14:58:56Z

@dbudanov-cmplr, I cherry-picked documentation generation fix
There are these issues left:

SYCL :: ESIMD/aot_mixed.cpp
--shared-libs build with clang

pvchupin · 2022-09-13T15:43:48Z

@v-klochkov, @kbobrovs, can you take a look into SYCL :: ESIMD/aot_mixed.cpp one?

dbudanov-cmplr · 2022-09-13T15:49:29Z

@v-klochkov, @kbobrovs, can you take a look into SYCL :: ESIMD/aot_mixed.cpp one?

@aelovikov-intel could you check if this is the same issues as in intel/llvm-test-suite#1256?

aelovikov-intel · 2022-09-13T16:12:37Z

@v-klochkov, @kbobrovs, can you take a look into SYCL :: ESIMD/aot_mixed.cpp one?

@aelovikov-intel could you check if this is the same issues as in intel/llvm-test-suite#1256?

Where is the link to the fail output?

dbudanov-cmplr · 2022-09-13T16:15:37Z

@v-klochkov, @kbobrovs, can you take a look into SYCL :: ESIMD/aot_mixed.cpp one?

@aelovikov-intel could you check if this is the same issues as in intel/llvm-test-suite#1256?

Where is the link to the fail output?

it was https://github.com/intel/llvm/actions/runs/3041366082/jobs/4898814666

aelovikov-intel · 2022-09-13T16:46:28Z

That looks different to what I'm trying to fix in intel/llvm-test-suite#1256

I'm trying to address runtime exception when we try to use gen9 binary on a newer gen device (TGL in our Windows CI).

dbudanov-cmplr · 2022-09-13T17:27:13Z

@v-klochkov, @kbobrovs, can you take a look into SYCL :: ESIMD/aot_mixed.cpp one?
Talked with @v-klochkov, he suggested to create a JIRA and mark the test as XFAIL and continue pulldown.

dbudanov-cmplr · 2022-09-13T17:41:23Z

Created issue for aot_mixed

This might be an FE bug. Will follow up separately. Workaround should be fine to unblock pulldown.

Mutex changes could be reverted if/when this change lands: https://reviews.llvm.org/D133425

dbudanov-cmplr · 2022-09-14T04:26:31Z

@pvchupin do you have an idea who can solve the problem with ESIMD Emu llvm test suite?
It seems like CI/configuration issue.

Error: Unable to find an artifact with the name: sycl_linux_default
Run actions/upload-artifact@v1
Error: Path does not exist /localdisk/github/_work/llvm/llvm/build/results_esimd_emu_default.json
Error: Exit code 1 returned from process: file name '/localdisk/github/bin.2.296.2/Runner.PluginHost', arguments 'action "GitHub.Runner.Plugins.Artifact.PublishArtifact, Runner.Plugins"'.

Do you think we can move forward and merge the PR since this is the only issue which remains.

sycl/plugins/level_zero/pi_level_zero.cpp

pvchupin · 2022-09-14T18:39:47Z

@pvchupin do you have an idea who can solve the problem with ESIMD Emu llvm test suite? It seems like CI/configuration issue.

Error: Unable to find an artifact with the name: sycl_linux_default Run actions/upload-artifact@v1 Error: Path does not exist /localdisk/github/_work/llvm/llvm/build/results_esimd_emu_default.json Error: Exit code 1 returned from process: file name '/localdisk/github/bin.2.296.2/Runner.PluginHost', arguments 'action "GitHub.Runner.Plugins.Artifact.PublishArtifact, Runner.Plugins"'.

Do you think we can move forward and merge the PR since this is the only issue which remains.

@dbudanov-cmplr, yes, looks like network glitch. I've restarted.

pvchupin · 2022-09-14T19:45:03Z

New fails on ESIMD Emu target:

********************
Failed Tests (5):
  SYCL :: dword_atomic_smoke.cpp
  SYCL :: lsc/lsc_fence.cpp
  SYCL :: lsc/lsc_slm.cpp
  SYCL :: lsc/lsc_surf.cpp
  SYCL :: lsc/lsc_usm.cpp

********************
Unexpectedly Passed Tests (1):
  SYCL :: spec_const/spec_const_double.cpp

@dongkyunahn-intel, @v-klochkov can you address please?

dongkyunahn-intel · 2022-09-14T19:46:52Z

New fails on ESIMD Emu target:

********************
Failed Tests (5):
  SYCL :: dword_atomic_smoke.cpp
  SYCL :: lsc/lsc_fence.cpp
  SYCL :: lsc/lsc_slm.cpp
  SYCL :: lsc/lsc_surf.cpp
  SYCL :: lsc/lsc_usm.cpp

********************
Unexpectedly Passed Tests (1):
  SYCL :: spec_const/spec_const_double.cpp

@dongkyunahn-intel, @v-klochkov can you address please?

PR is ready : intel/llvm-test-suite#1264

Add deduction guide to kernel_bundle

pvchupin · 2022-09-15T14:24:13Z

@dongkyunahn-intel, @v-klochkov, this one is not covered by the intel/llvm-test-suite#1264

Unexpectedly Passed Tests (1):
  SYCL :: spec_const/spec_const_double.cpp

Can you take a look?

pvchupin · 2022-09-15T14:30:29Z

Remaining fails are unrelated.

ESIMD Emu issue fails in other PRs (mentioned above).
Windows issue is addressed in [SYCL][Windows] Fix DataMovement test #6790
opaque pointers are expected to fail today

pvchupin · 2022-09-15T14:31:20Z

/merge

bb-sycl · 2022-09-15T14:32:17Z

Thu 15 Sep 2022 02:32:17 PM UTC --- Start to merge the commit into sycl branch. It will take several minutes.

bb-sycl · 2022-09-15T14:36:17Z

Thu 15 Sep 2022 02:36:16 PM UTC --- Merge the branch in this PR to base automatically. Will close the PR later.

dongkyunahn-intel · 2022-09-15T17:53:01Z

@dongkyunahn-intel, @v-klochkov, this one is not covered by the intel/llvm-test-suite#1264
Unexpectedly Passed Tests (1):
  SYCL :: spec_const/spec_const_double.cpp
Can you take a look?

The unexpected pass was because the test returns '0' for backends that do not support FP64/double type. While ESIMD_EMULATOR can support FP64 type as native double type, the emulator Plug-in implementation answers request for extension info as if it does not support FP64 type. I fixed the problem in following PR.

#6796

With this PR, the test will fail as expected.

yanfeng3721 · 2022-09-16T01:54:23Z

Pre-ci RHEL Build fail(http://icl-jenkins.sc.intel.com:8080/blue/organizations/jenkins/SYCL_CI%2Fintel%2FPrecommit_Check_User/detail/Precommit_Check_User/12893/pipeline/89/), impact Linux nightly build.
The related error log is as below:

[2022-09-15T14:53:12.539Z] [868/2957] Building CXX object lib/CodeGen/AsmPrinter/CMakeFiles/LLVMAsmPrinter.dir/DebugLocStream.cpp.o

[2022-09-15T14:53:12.539Z] FAILED: lib/CodeGen/AsmPrinter/CMakeFiles/LLVMAsmPrinter.dir/DebugLocStream.cpp.o

pvchupin · 2022-09-16T19:12:33Z

@yanfeng3721, thanks! I also noticed that. 01ffe31 is the reason.
It looks like llorg build is broken right now with gcc7.3. I've asked for revert at https://reviews.llvm.org/D132311

pvchupin · 2022-09-17T00:11:55Z

It seems gcc7.5 works fine. Should we migrate?

pvchupin · 2022-09-19T18:50:27Z

@yanfeng3721, community kindly reverted the change at llvm/llvm-project@29c841c
It should be pulled in with the next pulldown.
My question on 7.5 migration stays though...

MaskRay and others added 30 commits September 3, 2022 14:29

[SanitizerCoverage] Simplify pc-table and improve test. NFC

9fc679b

[Support] Simplify isInt and isUInt with constexpr if (NFC)

ee40ef7

Differential Revision: https://reviews.llvm.org/D132813

[Support] Simplify reverseBits with constexpr if (NFC)

2423cf4

Differential Revision: https://reviews.llvm.org/D132814

[ADT] Use std::add_lvalue_reference_t (NFC)

9dc6223

[ADT] Use std::add_pointer_t (NFC)

230e57d

[mlir] Use std::enable_if_t (NFC)

f11925e

[llvm] Use std::remove_reference_t (NFC)

03c3c2d

[ADT] Use std::tuple_element_t (NFC)

2bb43d7

[llvm] Use range-based for loops (NFC)

7d8c2d1

Identified with modernize-loop-convert.

[clang] Qualify auto in range-based for loops (NFC)

b7a7aee

[lld-macho][test] Correct a mismatched adrp-add pair

1ba8be3

[mlir:vscode] Fix bazel build after commit f7b8a70

725cdc4

[CostModel][X86] Add vector shift test coverage for codesize/latency/…

c4a174b

…size-latency cost kinds

Apply clang-tidy fixes for readability-identifier-naming in Tiling.cp…

a7a892a

…p (NFC)

Apply clang-tidy fixes for performance-for-range-copy in TilingInterf…

28e5e3d

…aceImpl.cpp (NFC)

[CostModel][X86] Add CostKinds handling for mul ops

c444af1

This was achieved using the 'cost-tables vs llvm-mca' script D103695 Also fix a missing pmullw v16i16 half-rate throughput as znver1 double-pumps - matches numbers from AMD SoG + Agner

[CostModel][X86] Remove redundant AVX512 v64i8 shift costs

59dbd6a

These are handled earlier (and more accurately) in AVX512BWShiftCostTable

[OpenMP] Silence unused variable warning in release builds

95db373

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp:3397:27: warning: unused variable 'SV' [-Werror] for (const llvm::Value *SV : StoreValues) ^

[CostModel][X86] getTypeBasedIntrinsicInstrCost - convert to CostKind…

06e7362

…TblEntry Begin the refactoring to use CostKindTblEntry and return real latency/codesize/sizelatency costs instead of reusing the throughput numbers

[Libfuzzer] Include signal.h for SIGINT

98c2754

Or else compiler-rt/lib/fuzzer/FuzzerUtilLinux.cpp fails to compile on FreeBSD.

[PhaseOrdering] Move X86 unsigned-multiply-overflow-check.ll test und…

8dc9918

…er X86

[CostModel][X86] getTypeBasedIntrinsicInstrCost - convert to CostKind…

626a84d

…TblEntry Begin the refactoring to use CostKindTblEntry and return real latency/codesize/sizelatency costs instead of reusing the throughput numbers

[test] pre-submission for the following SimpleLoopUnswitch update

07341e3

[lld-macho] Change constant std::vector to std::array (NFC)

4f688d0

Fix the Clang sphinx build bot; NFC

11c35d2

This should address the failure introduced in: https://lab.llvm.org/buildbot/#/builders/92/builds/32377

Pavel V Chupin added 2 commits September 13, 2022 12:02

[SYCL] Silence -Wctad-maybe-unsupported warning

28ba36f

This might be an FE bug. Will follow up separately. Workaround should be fine to unblock pulldown.

[SYCL] More changes to silence -Wctad-maybe-unsupported

afbc092

Mutex changes could be reverted if/when this change lands: https://reviews.llvm.org/D133425

smaslov-intel reviewed Sep 14, 2022

View reviewed changes

sycl/plugins/level_zero/pi_level_zero.cpp Outdated Show resolved Hide resolved

[SYCL] Silence -Wctad-maybe-unsupported for check-sycl

3e969cd

Add deduction guide to kernel_bundle

bb-sycl approved these changes Sep 15, 2022

View reviewed changes

bb-sycl merged commit 295b4fa into intel:sycl Sep 15, 2022

LLVM and SPIRV-LLVM-Translator pulldown (WW36) #6748

LLVM and SPIRV-LLVM-Translator pulldown (WW36) #6748

Uh oh!

Conversation

dbudanov-cmplr commented Sep 9, 2022

Uh oh!

bader commented Sep 13, 2022

Uh oh!

pvchupin commented Sep 13, 2022

Uh oh!

pvchupin commented Sep 13, 2022

Uh oh!

dbudanov-cmplr commented Sep 13, 2022

Uh oh!

aelovikov-intel commented Sep 13, 2022

Uh oh!

dbudanov-cmplr commented Sep 13, 2022

Uh oh!

aelovikov-intel commented Sep 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dbudanov-cmplr commented Sep 13, 2022

Uh oh!

dbudanov-cmplr commented Sep 13, 2022 • edited by pvchupin Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dbudanov-cmplr commented Sep 14, 2022

Uh oh!

Uh oh!

pvchupin commented Sep 14, 2022

Uh oh!

pvchupin commented Sep 14, 2022

Uh oh!

dongkyunahn-intel commented Sep 14, 2022

Uh oh!

pvchupin commented Sep 15, 2022

Uh oh!

pvchupin commented Sep 15, 2022

Uh oh!

pvchupin commented Sep 15, 2022

Uh oh!

bb-sycl commented Sep 15, 2022

Uh oh!

bb-sycl commented Sep 15, 2022

Uh oh!

dongkyunahn-intel commented Sep 15, 2022

Uh oh!

yanfeng3721 commented Sep 16, 2022

Uh oh!

pvchupin commented Sep 16, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pvchupin commented Sep 17, 2022

Uh oh!

pvchupin commented Sep 19, 2022

Uh oh!

Uh oh!

aelovikov-intel commented Sep 13, 2022 •

edited

Loading

dbudanov-cmplr commented Sep 13, 2022 •

edited by pvchupin

Loading

pvchupin commented Sep 16, 2022 •

edited

Loading