[llvm-branch-commits] [llvm] 385a27d - [CostModel][X86] Refresh ISD::ABS costs
Simon Pilgrim via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Wed Nov 25 10:45:18 PST 2020
Author: Simon Pilgrim
Date: 2020-11-25T18:40:19Z
New Revision: 385a27d6cdaddbb99da5518da3cb6713d9677df3
URL: https://github.com/llvm/llvm-project/commit/385a27d6cdaddbb99da5518da3cb6713d9677df3
DIFF: https://github.com/llvm/llvm-project/commit/385a27d6cdaddbb99da5518da3cb6713d9677df3.diff
LOG: [CostModel][X86] Refresh ISD::ABS costs
Update costs now that D92095 and D92102 have tweaked the SSE2 implementation
The SSE42 BLENDVPD cost can actually be used on SSE41 as we don't attempt to generate PCMPGT anymore
Add scalar i16/i32/i64 costs as we can do this cheaply with CMOV
Added:
Modified:
llvm/lib/Target/X86/X86TargetTransformInfo.cpp
llvm/test/Analysis/CostModel/X86/abs.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index 2c35f74ab95e..770317a9a8b5 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -2483,7 +2483,7 @@ int X86TTIImpl::getTypeBasedIntrinsicInstrCost(
{ ISD::FSQRT, MVT::v4f64, 28 }, // Haswell from http://www.agner.org/
};
static const CostTblEntry AVX1CostTbl[] = {
- { ISD::ABS, MVT::v4i64, 6 }, // VBLENDVPD(X,VPSUBQ(0,X),X)
+ { ISD::ABS, MVT::v4i64, 5 }, // VBLENDVPD(X,VPSUBQ(0,X),X)
{ ISD::ABS, MVT::v8i32, 3 },
{ ISD::ABS, MVT::v16i16, 3 },
{ ISD::ABS, MVT::v32i8, 3 },
@@ -2554,13 +2554,13 @@ int X86TTIImpl::getTypeBasedIntrinsicInstrCost(
{ ISD::FSQRT, MVT::v2f64, 70 }, // sqrtpd
};
static const CostTblEntry SSE42CostTbl[] = {
- { ISD::ABS, MVT::v2i64, 3 }, // BLENDVPD(X,PSUBQ(0,X),X)
{ ISD::USUBSAT, MVT::v4i32, 2 }, // pmaxud + psubd
{ ISD::UADDSAT, MVT::v4i32, 3 }, // not + pminud + paddd
{ ISD::FSQRT, MVT::f32, 18 }, // Nehalem from http://www.agner.org/
{ ISD::FSQRT, MVT::v4f32, 18 }, // Nehalem from http://www.agner.org/
};
static const CostTblEntry SSE41CostTbl[] = {
+ { ISD::ABS, MVT::v2i64, 2 }, // BLENDVPD(X,PSUBQ(0,X),X)
{ ISD::SMAX, MVT::v4i32, 1 },
{ ISD::SMAX, MVT::v16i8, 1 },
{ ISD::SMIN, MVT::v4i32, 1 },
@@ -2597,8 +2597,8 @@ int X86TTIImpl::getTypeBasedIntrinsicInstrCost(
static const CostTblEntry SSE2CostTbl[] = {
{ ISD::ABS, MVT::v2i64, 4 },
{ ISD::ABS, MVT::v4i32, 3 },
- { ISD::ABS, MVT::v8i16, 3 },
- { ISD::ABS, MVT::v16i8, 3 },
+ { ISD::ABS, MVT::v8i16, 2 },
+ { ISD::ABS, MVT::v16i8, 2 },
{ ISD::BITREVERSE, MVT::v2i64, 29 },
{ ISD::BITREVERSE, MVT::v4i32, 27 },
{ ISD::BITREVERSE, MVT::v8i16, 27 },
@@ -2668,6 +2668,7 @@ int X86TTIImpl::getTypeBasedIntrinsicInstrCost(
{ ISD::CTPOP, MVT::i8, 1 },
};
static const CostTblEntry X64CostTbl[] = { // 64-bit targets
+ { ISD::ABS, MVT::i64, 2 }, // SUB+CMOV
{ ISD::BITREVERSE, MVT::i64, 14 },
{ ISD::CTLZ, MVT::i64, 4 }, // BSR+XOR or BSR+XOR+CMOV
{ ISD::CTTZ, MVT::i64, 3 }, // TEST+BSF+CMOV/BRANCH
@@ -2677,6 +2678,8 @@ int X86TTIImpl::getTypeBasedIntrinsicInstrCost(
{ ISD::UMULO, MVT::i64, 2 }, // mulq + seto
};
static const CostTblEntry X86CostTbl[] = { // 32 or 64-bit targets
+ { ISD::ABS, MVT::i32, 2 }, // SUB+CMOV
+ { ISD::ABS, MVT::i16, 2 }, // SUB+CMOV
{ ISD::BITREVERSE, MVT::i32, 14 },
{ ISD::BITREVERSE, MVT::i16, 14 },
{ ISD::BITREVERSE, MVT::i8, 11 },
diff --git a/llvm/test/Analysis/CostModel/X86/abs.ll b/llvm/test/Analysis/CostModel/X86/abs.ll
index 636160201342..dbfcb20779dc 100644
--- a/llvm/test/Analysis/CostModel/X86/abs.ll
+++ b/llvm/test/Analysis/CostModel/X86/abs.ll
@@ -14,42 +14,42 @@
define void @cost_abs_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512) {
; SSE2-LABEL: 'cost_abs_i64'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i64'
-; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSE42-LABEL: 'cost_abs_i64'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
-; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
-; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX1-LABEL: 'cost_abs_i64'
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
-; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
-; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX2-LABEL: 'cost_abs_i64'
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512-LABEL: 'cost_abs_i64'
-; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
+; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 false)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 false)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 false)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 false)
@@ -64,42 +64,42 @@ define void @cost_abs_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64>
define void @cost_abs_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512) {
; SSE2-LABEL: 'cost_abs_i32'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i32'
-; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSE42-LABEL: 'cost_abs_i32'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX1-LABEL: 'cost_abs_i32'
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX2-LABEL: 'cost_abs_i32'
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512-LABEL: 'cost_abs_i32'
-; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
+; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 false)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 false)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 false)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 false)
@@ -114,56 +114,56 @@ define void @cost_abs_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32>
define void @cost_abs_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512) {
; SSE2-LABEL: 'cost_abs_i16'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i16'
-; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSE42-LABEL: 'cost_abs_i16'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX1-LABEL: 'cost_abs_i16'
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX2-LABEL: 'cost_abs_i16'
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512F-LABEL: 'cost_abs_i16'
-; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512DQ-LABEL: 'cost_abs_i16'
-; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512BW-LABEL: 'cost_abs_i16'
-; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 false)
; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 false)
; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 false)
; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 false)
@@ -179,9 +179,9 @@ define void @cost_abs_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16
define void @cost_abs_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512) {
; SSE2-LABEL: 'cost_abs_i8'
; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 %a8, i1 false)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> %a128, i1 false)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> %a256, i1 false)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> %a512, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> %a128, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> %a256, i1 false)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> %a512, i1 false)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i8'
@@ -246,42 +246,42 @@ define void @cost_abs_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a5
define void @cost_abs_i64_poison(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512) {
; SSE2-LABEL: 'cost_abs_i64_poison'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i64_poison'
-; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSE42-LABEL: 'cost_abs_i64_poison'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
-; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
-; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX1-LABEL: 'cost_abs_i64_poison'
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
-; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
-; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX2-LABEL: 'cost_abs_i64_poison'
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512-LABEL: 'cost_abs_i64_poison'
-; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
+; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.abs.i64(i64 %a64, i1 true)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %a128, i1 true)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %a256, i1 true)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %a512, i1 true)
@@ -296,42 +296,42 @@ define void @cost_abs_i64_poison(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
define void @cost_abs_i32_poison(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512) {
; SSE2-LABEL: 'cost_abs_i32_poison'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i32_poison'
-; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSE42-LABEL: 'cost_abs_i32_poison'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX1-LABEL: 'cost_abs_i32_poison'
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX2-LABEL: 'cost_abs_i32_poison'
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512-LABEL: 'cost_abs_i32_poison'
-; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
+; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = call i32 @llvm.abs.i32(i32 %a32, i1 true)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %a128, i1 true)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %a256, i1 true)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %a512, i1 true)
@@ -346,56 +346,56 @@ define void @cost_abs_i32_poison(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
define void @cost_abs_i16_poison(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512) {
; SSE2-LABEL: 'cost_abs_i16_poison'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i16_poison'
-; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSE42-LABEL: 'cost_abs_i16_poison'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX1-LABEL: 'cost_abs_i16_poison'
-; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX2-LABEL: 'cost_abs_i16_poison'
-; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512F-LABEL: 'cost_abs_i16_poison'
-; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512DQ-LABEL: 'cost_abs_i16_poison'
-; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; AVX512BW-LABEL: 'cost_abs_i16_poison'
-; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.abs.i16(i16 %a16, i1 true)
; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %a128, i1 true)
; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %a256, i1 true)
; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %a512, i1 true)
@@ -411,9 +411,9 @@ define void @cost_abs_i16_poison(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
define void @cost_abs_i8_poison(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512) {
; SSE2-LABEL: 'cost_abs_i8_poison'
; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 %a8, i1 true)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> %a128, i1 true)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> %a256, i1 true)
-; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> %a512, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I8 = call <16 x i8> @llvm.abs.v16i8(<16 x i8> %a128, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.abs.v32i8(<32 x i8> %a256, i1 true)
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.abs.v64i8(<64 x i8> %a512, i1 true)
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; SSSE3-LABEL: 'cost_abs_i8_poison'
More information about the llvm-branch-commits
mailing list