[X86][SSE] Improve lowering of vXi64 multiplies
As mentioned on PR30845, we were performing our vXi64 multiplication as:
AloBlo = pmuludq(a, b);
AloBhi = pmuludq(a, psrlqi(b, 32));
AhiBlo = pmuludq(psrlqi(a, 32), b);
return AloBlo + psllqi(AloBhi, 32)+ psllqi(AhiBlo, 32);
when we could avoid one of the upper shifts with:
AloBlo = pmuludq(a, b);
AloBhi = pmuludq(a, psrlqi(b, 32));
AhiBlo = pmuludq(psrlqi(a, 32), b);
return AloBlo + psllqi(AloBhi + AhiBlo, 32);
This matches the lowering on gcc/icc.
Differential Revision: https://reviews.llvm.org/D27756
llvm-svn: 290267
diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index 6b5b5a1..db563c0 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -551,11 +551,11 @@
{ ISD::SUB, MVT::v4i64, 4 },
{ ISD::ADD, MVT::v4i64, 4 },
// A v4i64 multiply is custom lowered as two split v2i64 vectors that then
- // are lowered as a series of long multiplies(3), shifts(4) and adds(2)
+ // are lowered as a series of long multiplies(3), shifts(3) and adds(2)
// Because we believe v4i64 to be a legal type, we must also include the
- // split factor of two in the cost table. Therefore, the cost here is 18
- // instead of 9.
- { ISD::MUL, MVT::v4i64, 18 },
+ // split factor of two in the cost table. Therefore, the cost here is 16
+ // instead of 8.
+ { ISD::MUL, MVT::v4i64, 16 },
};
// Look for AVX1 lowering tricks.
@@ -569,10 +569,10 @@
// Custom lowering of vectors.
static const CostTblEntry CustomLowered[] = {
// A v2i64/v4i64 and multiply is custom lowered as a series of long
- // multiplies(3), shifts(4) and adds(2).
- { ISD::MUL, MVT::v2i64, 9 },
- { ISD::MUL, MVT::v4i64, 9 },
- { ISD::MUL, MVT::v8i64, 9 }
+ // multiplies(3), shifts(3) and adds(2).
+ { ISD::MUL, MVT::v2i64, 8 },
+ { ISD::MUL, MVT::v4i64, 8 },
+ { ISD::MUL, MVT::v8i64, 8 }
};
if (const auto *Entry = CostTableLookup(CustomLowered, ISD, LT.second))
return LT.first * Entry->Cost;