define an enum type for code object version number.

Update based on reviewers' comments. Thanks.

In D143293#4106251, @kosarev wrote:

Do we still need getHsaAbiVersion() and the ELFABIVERSION_AMDGPU_HSA_* constants?

Update based on Matt's comments.

Update based on Matt's comments:

1. Add a test case for out of range code object version;
2. Move code object version into AMDGPUInformationCache
3. Pass code object version as an argument in MetadataStreamerMsgPackV3::getHSAKernelProps

4 remove unnecessary parens

update test

Add a test.

Updated based on arsenm's comments.

Should the module flag name be amdgpu_code_object_version or amdhsa_code_object_version?

Updated based on arsenm's comment to merge two cases.

Update based on arsenm's comments.

Merge two ProcessUse functions, and do selection based on code object version.

Still use the existing AMDGPU::getAmdhsaCodeObjectVersion() to check code object version.
This is for consistency in the backend. Plan to use module flag in a later patch for all cases.

Do the pattern matching optimizations based on their existence with code object version.
Get code object version from amdhsa-code-object-version module flag. Note that now
I just declare a static function to do as a proof of concept because we don't know what is the default
code object version if the module flag does not exists.
Also, it is beyond this work to get code object version from module flag everywhere in the compiler.

In D131276#3703686, @bcahoon wrote:

In D131276#3703399, @cfang wrote:

In D131276#3703034, @b-sumner wrote:

In D131276#3702934, @arsenm wrote:

I also think we're still missing a module flag to indicate the code object version

The module flag was implemented quite a while back. See D119026 from February. From a recent compile:

!llvm.module.flags = !{!0, !1, !2}
!0 = !{i32 1, !"amdgpu_code_object_version", i32 500}

We were using --amdhsa-code-object-version=5 to run LIT tests. Do you mean this flag will no longer take effect if we
switch to module flag for code object version?

I think the flag is distinct from the metadata. The module metadata just records which version is being used. You can access it in the code:

auto Ver = mdconst::dyn_extract_or_null<ConstantInt>(M.getModuleFlag("amdgpu_code_object_version"))

and, then if Ver != 5, then there is no need to execute the loop with the calls to processImplicitArgUse().

In D131276#3703034, @b-sumner wrote:

In D131276#3702934, @arsenm wrote:

I also think we're still missing a module flag to indicate the code object version

The module flag was implemented quite a while back. See D119026 from February. From a recent compile:

!llvm.module.flags = !{!0, !1, !2}
!0 = !{i32 1, !"amdgpu_code_object_version", i32 500}

Merge the new test into frame-index-elimination.ll

Add a mir test, and remove the use of undef in the test.

Rename the LIT test to frame-index-elimination-tied-operand.ll as suggested.

Update https://github.com/llvm/llvm-project/blob/main/clang/include/clang/Driver/Options.td#L3626
as well as the corresponding LIT tests.

This patch triggered a correctness issue in running mixbench-ocl-alt.
I am not familiar with the CFG in SCCP pass at all. But the comments
in the code seems suggest we should not change the CFG:

In D123548#3444956, @foad wrote:

Typos in description: "and also and also", "WE".

1. Correct the function funcRetrievesMultigridSyncArg;
2. update after rebase.

Ping
Now that we agreed that we have to use alignTo to align the Offset to what the implicitarg_ptr requires
and the LIT tests have been updated to show the alignment related layout. Thanks.

Did correct "git diff" to include deleted and added files in the diff

In D123346#3440166, @arsenm wrote:

These test changes don't show a change in behavior

Use alignTo to force the alignment for the implicit kernarg segment:
Offset = alignTo(Offset, ST.getAlignmentForImplicitArgPtr());

[AMD Official Use Only]

For the LIT test, add back the GCN check prefix. For the GCN checks that are common to code object version 2 and MESA but different from code object version 5, we split the checks to HSA (version 2) and MESA.

update LIT test

In D123262#3434565, @arsenm wrote:

This can be tested in the existing llvm/test/CodeGen/AMDGPU/llvm.amdgcn.implicitarg.ptr.ll

A minor change: add suffix to the enum itself instead of the individual field.
Also remove the "Fixes" field in the summary (commit message).

Update based on Matt's comments:

1. Use buildPtrAdd
2. Remove a space
3. Add suffix for the enum definition and also wrap with a namespace
4. Remove the redundant def of ST (SubTarget)
5. Updated according to clang-format

Ping!

1. Introduce a common function, SITargetLowering::loadImplicitKernelArgument, which is used

in both getSegmentAperture and lowerTrapHsaQueuePtr.

Rebase and update LIT tests.