[llvm] a751084 - [X86][Costmodel] `trunc v16i8 to v8i1` can appear after legalization, cost is same as for `trunc v8i8 to v8i1`

[Roman Lebedev via llvm-commits]

Author: Roman Lebedev
Date: 2021-11-19T15:57:32+03:00
New Revision: a751084bb4521f0847f39e9eb564e225ccce311e

URL: https://github.com/llvm/llvm-project/commit/a751084bb4521f0847f39e9eb564e225ccce311e
DIFF: https://github.com/llvm/llvm-project/commit/a751084bb4521f0847f39e9eb564e225ccce311e.diff

LOG: [X86][Costmodel] `trunc v16i8 to v8i1` can appear after legalization, cost is same as for `trunc v8i8 to v8i1`

Note that there are many other missing costs, i'm *only* adding the ones that are queried
from `getReplicationShuffleCost()` for the existing (quite exhaustive) test coverage.

Reviewed By: RKSimon

Differential Revision: https://reviews.llvm.org/D114070

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86TargetTransformInfo.cpp
    llvm/test/Analysis/CostModel/X86/trunc.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index f61d315f01c7..a68e6a9117ab 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -1620,6 +1620,7 @@ InstructionCost X86TTIImpl::getCastInstrCost(unsigned Opcode, Type *Dst,
     { ISD::TRUNCATE,    MVT::v4i1,   MVT::v4i16,  2 }, // widen to zmm
     { ISD::TRUNCATE,    MVT::v4i8,   MVT::v4i16,  2 }, // vpmovwb
     { ISD::TRUNCATE,    MVT::v8i1,   MVT::v8i8,   2 }, // widen to zmm
+    { ISD::TRUNCATE,    MVT::v8i1,   MVT::v16i8,  2 }, // widen to zmm
     { ISD::TRUNCATE,    MVT::v8i1,   MVT::v8i16,  2 }, // widen to zmm
     { ISD::TRUNCATE,    MVT::v8i8,   MVT::v8i16,  2 }, // vpmovwb
     { ISD::TRUNCATE,    MVT::v16i1,  MVT::v16i8,  2 }, // widen to zmm

diff  --git a/llvm/test/Analysis/CostModel/X86/trunc.ll b/llvm/test/Analysis/CostModel/X86/trunc.ll
index dcf70b94eda1..c0b295bd7c23 100644
--- a/llvm/test/Analysis/CostModel/X86/trunc.ll
+++ b/llvm/test/Analysis/CostModel/X86/trunc.ll
@@ -2490,9 +2490,9 @@ define i32 @trunc_vXi1() {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i8 = trunc <2 x i8> undef to <2 x i1>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V3i8 = trunc <3 x i8> undef to <3 x i1>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i8 = trunc <4 x i8> undef to <4 x i1>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V5i8 = trunc <5 x i8> undef to <5 x i1>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V6i8 = trunc <6 x i8> undef to <6 x i1>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V7i8 = trunc <7 x i8> undef to <7 x i1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V5i8 = trunc <5 x i8> undef to <5 x i1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V6i8 = trunc <6 x i8> undef to <6 x i1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V7i8 = trunc <7 x i8> undef to <7 x i1>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i8 = trunc <8 x i8> undef to <8 x i1>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V10i8 = trunc <10 x i8> undef to <10 x i1>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V12i8 = trunc <12 x i8> undef to <12 x i1>