replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[CostModel][X86] Include the cost of 256-bit upper subvector extract/insertion in AVX1 v4i64 MUL 

Matches other MUL/ADD/SUB 256-bit case on AVX1

llvm-svn: 291149
This commit is contained in:
Simon Pilgrim 2017-01-05 18:20:25 +00:00
parent e9987a1d2f
commit b01e844241
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/X86/X86TargetTransformInfo.cpp
View File

@ -556,9 +556,9 @@ int X86TTIImpl::getArithmeticInstrCost(
// A v4i64 multiply is custom lowered as two split v2i64 vectors that then
// 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 16
// extract+insert in the cost table. Therefore, the cost here is 18
// instead of 8.
{ ISD::MUL, MVT::v4i64, 16 },
{ ISD::MUL, MVT::v4i64, 18 },
};
// Look for AVX1 lowering tricks.

4
llvm/test/Analysis/CostModel/X86/arith.ll
View File

@ -436,7 +436,7 @@ define i32 @mul(i32 %arg) {
%A = mul <2 x i64> undef, undef
; SSSE3: cost of 16 {{.*}} %B = mul
; SSE42: cost of 16 {{.*}} %B = mul
; AVX: cost of 16 {{.*}} %B = mul
; AVX: cost of 18 {{.*}} %B = mul
; AVX2: cost of 8 {{.*}} %B = mul
; AVX512F: cost of 8 {{.*}} %B = mul
; AVX512BW: cost of 8 {{.*}} %B = mul
@ -444,7 +444,7 @@ define i32 @mul(i32 %arg) {
%B = mul <4 x i64> undef, undef
; SSSE3: cost of 32 {{.*}} %C = mul
; SSE42: cost of 32 {{.*}} %C = mul
; AVX: cost of 32 {{.*}} %C = mul
; AVX: cost of 36 {{.*}} %C = mul
; AVX2: cost of 16 {{.*}} %C = mul
; AVX512F: cost of 8 {{.*}} %C = mul
; AVX512BW: cost of 8 {{.*}} %C = mul