[CIR] Implement missing X86 builtins

[andykaylor]

Overview

Implement CIR handling for X86-specific builtin functions that are not yet handled in the incubator. For these builtins, the initial implementation should be done in the upstream LLVM repository and then back-ported to the incubator. The implementation should be based on the handling of the same builtins in clang\CodeGen\TargetBuiltins\X86.c.

Many people can work on this in parallel, if we can successfully coordinate the work to avoid overlap as much as possible. To that end, I am adding a partial list here of the builtins that need to be implemented. If you would like to contribute to this task, please comment saying which builtin(s) you are working on.

I will remove builtins from this list as they are upstreamed.

Please check the table below before beginning work to make sure no one else is working on the builtins you are targeting. If you have any questions, please comment here tagging @andykaylor.

Needed builtins

Builtin group	Assignee
masked rounding
vfmadd, vfmaddsub	@0dminnimda
masked expand	@cs25resch11005-bhuvan
masked compress	@cs25resch11005-bhuvan
extractf128	@badumbatish
pmovqd512_mask
permdi	@YGGkk
alignd	@YGGkk
shuff32x4, shufi32x4	@raventid
kshiftl, kshiftr	@vishruth-thimmaiah
rotate	@moar55
prord
select, selectsh	@Priyanshu3820
cmpb, ucmpb	@Priyanshu3820
vpcomb, vpcomub	@vishruth-thimmaiah
ktest, kortest	@GeneraluseAI
kadd, kor, kand, kxnor, knot, kmov	@GeneraluseAI
kunpck	@ahmednoursphinx
sqrtps	@Priyanshu3820
pmul, pmulu	@GeneraluseAI
pternlog	@medhatiwari
vpshl, vpshr, shiftleft, shiftright	@Priyanshu3820
reduce fmul, fadd, fmax, fmin	@badumbatish
rdrand_step	@fileho
fpclass	@jacquesguan
intersect	@vishruth-thimmaiah
multishift	@cs25mtech12008
convert half	@Priyanshu3820
cpuid	@rturrado
emul, mulh	@moar55
readwritebarrier	@Luhaocong
address of return address	@Luhaocong
stosb	@GeneraluseAI
t2rpntlvw
ud2, int2c	@toprakmurat
readfs, readgs	@medhatiwari
encodekey	@moar55
aesenc, aesencwide	@Luhaocong

Existing incubator tests

clang/test/CIR/CodeGen/X86/*

[llvmbot]

Hi!

This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:

1. Check that no other contributor is working on this issue. If someone is assigned to the issue or claimed to be working on it, ping the person. After one week without a response, the assignee may be changed.
2. Leave a comment indicating that you are working on the issue, or just create a pull request after following the steps below. Mention this issue in the description of the pull request.
3. Fix the issue locally.
4. Run the test suite locally. Remember that the subdirectories under test/ create fine-grained testing targets, so you can e.g. use make check-clang-ast to only run Clang's AST tests.
5. Create a Git commit.
6. Run git clang-format HEAD~1 to format your changes.
7. Open a pull request to the upstream repository on GitHub. Detailed instructions can be found in GitHub's documentation. Mention this issue in the description of the pull request.

If you have any further questions about this issue, don't hesitate to ask via a comment in the thread below.

[llvmbot]

@llvm/issue-subscribers-good-first-issue

Author: Andy Kaylor (andykaylor)

Details

### Overview Implement CIR handling for X86-specific builtin functions that are not yet handled in the incubator. For these builtins, the initial implementation should be done in the upstream LLVM repository and then back-ported to the incubator. The implementation should be based on the handling of the same builtins in `clang\CodeGen\TargetBuiltins\X86.c`.

Many people can work on this in parallel, if we can successfully coordinate the work to avoid overlap as much as possible. To that end, I am adding a partial list here of the builtins that need to be implemented. If you would like to contribute to this task, please comment saying which builtin(s) you are working on.

I will remove builtins from this list as they are upstreamed.

Please check the table below before beginning work to make sure no one else is working on the builtins you are targeting. If you have any questions, please comment here tagging @andykaylor.

Needed builtins

Builtin group	Assignee
masked rounding
vfmadd, vfmaddsub
masked expand
masked compress
extractf128
pmovqd512_mask
permdi
alignd
shuff32x4, shufi32x4
kshiftl, kshiftr
rotate
prord
select, selectsh
cmpb, ucmpb
vpcomb, vpcomub
ktest, kortest
kadd, kor, kand, kxnor, knot, kmov
kunpck
sqrtps
pmul, pmulu
pternlog
vpshl, vpshr, shiftleft, shiftright
reduce fmul, fadd, fmax, fmin
rdrand_step
fpclass
intersect
multishift
convert half
cpuid
emul, mulh
readwritebarrier
address of return address
stosb
t2rpntlvw
ud2, int2c
readfs, readgs
encodekey
aesenc, aesencwide

Existing incubator tests

clang/test/CIR/CodeGen/X86/*

[andykaylor]

I have marked this as a "good first issue" but interested contributors should be aware of some special handling needed for working on ClangIR issues.

First, in order to even reproduce the problem you will need to build clang with ClangIR enabled. By default, clang does not build ClangIR support. You can find instructions for building with ClangIR enabled here.

Second, after building but before starting any other work, you should verify that you see an error when compiling the suggested minimal test case with ClangIR enabled. You should see an error (in this case, ``) when you compile the test case using the following options: -fclangir -emit-cir. The error may be followed by other errors or a stack dump. If you do not see an error, either someone else has already implemented this feature, or you are using the wrong repository (perhaps you built the ClangIR incubator repository?).

Third, the code to implement this feature should be based on the classic codegen handling of the builtins. The task here is to find the correct code, copy it to the CIR/CodeGen/CIRGenBuiltinX86.cpp file, and update it to generate CIR instread of LLVM IR. As part of the migration process, you may need to update the code to conform to the coding standards used in CIR, which are slightly different from the usual clang standards (mostly, variable identifiers start with lowercase in CIR code). You will also need to create or migrate a LIT test that exercises the new code.

Don't hesitate to ask for help if you need it. The entire point of "good first issue" is to help new people get involved with our project.

[andykaylor]

@Thibault-Monnier @woruyu @HendrikHuebner @siddu0660 FYI

[andykaylor]

@Priyanshu3820 It looks like @Thibault-Monnier has covered most of the builtins for #167752, but there is more work to be done here.

If we end up burning through all of these quickly, there are still many builtins to be handled for other targets.

[Priyanshu3820]

@andykaylor Understood. I would start by implementing sqrtps.

[moar55]

Hello I would like to do the rotate builtin.

[Priyanshu3820]

sqrt builtins: #168320

[ahmednoursphinx]

Hey @andykaylor i would like to work on kunpck

[vishruth-thimmaiah]

@andykaylor - I can help with the kshiftl/kshiftr builtins.

[Luhaocong]

Hi @andykaylor, I would like to work on readwritebarrier