[PATCH] D99244: [X86] Pass to transform tdpbsud&tdpbusd&tdpbuud intrinsics to scalar operation

Sat Mar 27 21:19:22 PDT 2021

pengfei added inline comments.

================
Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:357

-  if (IntrID == Intrinsic::x86_tdpbssd_internal) {
+  if constexpr (IntrID != Intrinsic::x86_tdpbf16ps_internal) {
     // tiledpbssd.scalarize.inner.body:
----------------
What's this used for? Would it fails when some buildbot only has lower C++ libraries?

================
Comment at: llvm/test/CodeGen/X86/AMX/amx-low-intrinsics.ll:409
+; CHECK-NEXT:    br label [[TILEDPBF16PS_SCALARIZE_ROWS_HEADER:%.*]]
+; CHECK:       tiledpbf16ps.scalarize.rows.header:
+; CHECK-NEXT:    [[TILEDPBF16PS_SCALARIZE_ROWS_IV:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ], [ [[TILEDPBF16PS_SCALARIZE_ROWS_STEP:%.*]], [[TILEDPBF16PS_SCALARIZE_ROWS_LATCH:%.*]] ]
----------------
This change doesn't seem affected by this patch. Do you know which one affects it? Why it still passes without this change?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99244/new/

https://reviews.llvm.org/D99244