[clang] VectorExprEvaluator::VisitCallExpr - add constant folding for X86 pslldqi/psrldqi infrinsics

[kimyounhoex1]

These X86 builtins (__builtin_ia32_pslldqi128/256_byteshift and
__builtin_ia32_psrldqi128/256_byteshift) can now be evaluated at
compile time when the shift amount is a constant integer.

This improves consistency with other vector shift intrinsics and
reduces unnecessary runtime evaluation.

Fixes #156494

[github-actions]

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

[llvmbot]

@llvm/pr-subscribers-clang

Author: None (kimyounhoex1)

Changes

These X86 builtins (__builtin_ia32_pslldqi128/256_byteshift and
__builtin_ia32_psrldqi128/256_byteshift) can now be evaluated at
compile time when the shift amount is a constant integer.

This improves consistency with other vector shift intrinsics and
reduces unnecessary runtime evaluation.

Fixes #156494

---

Full diff: https://github.com/llvm/llvm-project/pull/157403.diff

1 Files Affected:

• (modified) clang/lib/AST/ExprConstant.cpp (+65)

diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp
index b4f1e76187e25..2b06705a4870c 100644
--- a/clang/lib/AST/ExprConstant.cpp
+++ b/clang/lib/AST/ExprConstant.cpp
@@ -12039,6 +12039,71 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) {
     }
     return Success(APValue(ResultElements.data(), ResultElements.size()), E);
   }
+  case X86::BI__builtin_ia32_pslldqi128_byteshift:
+  case X86::BI__builtin_ia32_psrldqi128_byteshift: {
+    unsigned BuiltinID = E->getBuiltinCallee();
+
+    APSInt Amt;
+    if (!EvaluateInteger(E->getArg(1), Amt, Info))
+      break;
+    unsigned Shift = (unsigned)Amt.getZExtValue();
+
+    APValue Vec;
+    if (!Evaluate(Vec, Info, E->getArg(0)) || !Vec.isVector())
+      break;
+
+    SmallVector<APValue, 16> ResultElements;
+    ResultElements.reserve(16);
+
+    bool isLeft = (BuiltinID == X86::BI__builtin_ia32_pslldqi128_byteshift);
+
+    for (unsigned i = 0; i < 16; i++) {
+      int SrcIdx = -1;
+      if (isLeft)
+        SrcIdx = i + Shift;
+      else if (i >= Shift)
+        SrcIdx = i - Shift;
+
+      if (SrcIdx >= 0 && (unsigned)SrcIdx < 16)
+        ResultElements.push_back(Vec.getVectorElt(SrcIdx));
+      else
+        ResultElements.push_back(APValue(0));
+    }
+    return Success(APValue(ResultElements.data(), ResultElements.size()), E);
+  }
+
+  case X86::BI__builtin_ia32_pslldqi256_byteshift:
+  case X86::BI__builtin_ia32_psrldqi256_byteshift: {
+    unsigned BuiltinID = E->getBuiltinCallee();
+
+    APSInt Amt;
+    if (!EvaluateInteger(E->getArg(1), Amt, Info))
+      break;
+    unsigned Shift = (unsigned)Amt.getZExtValue();
+
+    APValue Vec;
+    if (!Evaluate(Vec, Info, E->getArg(0)) || !Vec.isVector())
+      break;
+
+    SmallVector<APValue, 32> ResultElements;
+    ResultElements.reserve(32);
+
+    bool isLeft = (BuiltinID == X86::BI__builtin_ia32_pslldqi256_byteshift);
+
+    for (unsigned i = 0; i < 32; i++) {
+      int SrcIdx = -1;
+      if (isLeft)
+        SrcIdx = i + Shift;
+      else if (i >= Shift)
+        SrcIdx = i - Shift;
+
+      if (SrcIdx >= 0 && (unsigned)SrcIdx < 32)
+        ResultElements.push_back(Vec.getVectorElt(SrcIdx));
+      else
+        ResultElements.push_back(APValue(0));
+    }
+    return Success(APValue(ResultElements.data(), ResultElements.size()), E);
+  }
   }
 }
 

[kimyounhoex1]

Hi @RKSimon, just kindly pinging this PR
The workflows are pending approval and the required build checks haven’t started yet.
Could a maintainer please take a look? Thanks a lot!

[RKSimon]

Hi @RKSimon, just kindly pinging this PR The workflows are pending approval and the required build checks haven’t started yet. Could a maintainer please take a look? Thanks a lot!

Will take a look - sorry I missed this (unfortunately github is terrible with notifications for contributions from non project members) - CC'ing me is the best way so it appears in my github mentioned tab

[kimyounhoex1]

Hi @RKSimon, just kindly pinging this PR The workflows are pending approval and the required build checks haven’t started yet. Could a maintainer please take a look? Thanks a lot!

Will take a look - sorry I missed this (unfortunately github is terrible with notifications for contributions from non project members) - CC'ing me is the best way so it appears in my github mentioned tab

Understood, I’ll keep that in mind next time.

[RKSimon]

test coverage? they need adding to sse2-builtins.c and avx2-builtins.c

[RKSimon]

This isn't going to work as currently the intrinsics take <X x long long int> types - we're going to have to change these to <8X x char> types to make this a lot easier to deal with - see the palignr builtins for an example

[kimyounhoex1]

I’m done with this one, could you please check one more?

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.