[llvm] 854a459 - [CostModel][X86] getArithmeticInstrCost - move GLM/SLM custom costs AFTER constant shift -> multiply canonicalization
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 14 03:46:47 PDT 2022
Author: Simon Pilgrim
Date: 2022-09-14T11:46:26+01:00
New Revision: 854a4595b6e14ea178152c44585a42bb6e1a779e
URL: https://github.com/llvm/llvm-project/commit/854a4595b6e14ea178152c44585a42bb6e1a779e
DIFF: https://github.com/llvm/llvm-project/commit/854a4595b6e14ea178152c44585a42bb6e1a779e.diff
LOG: [CostModel][X86] getArithmeticInstrCost - move GLM/SLM custom costs AFTER constant shift -> multiply canonicalization
Corrects the shift by constant costs to better account for them being converted to multiples for lowering - which demonstrates that we should probably be trying harder NOT to convert these to multiplies for some CPUs (v4i32 in particular).
Added:
Modified:
llvm/lib/Target/X86/X86TargetTransformInfo.cpp
llvm/test/Analysis/CostModel/X86/fshl.ll
llvm/test/Analysis/CostModel/X86/fshr.ll
llvm/test/Analysis/CostModel/X86/mul.ll
llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index 8a1a52fdeee2..d7e00a1ae8db 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -328,47 +328,6 @@ InstructionCost X86TTIImpl::getArithmeticInstrCost(
Op1Info.getNoProps(), Op2Info.getNoProps());
}
- static const CostKindTblEntry GLMCostTable[] = {
- { ISD::FDIV, MVT::f32, { 18, 19, 1, 1 } }, // divss
- { ISD::FDIV, MVT::v4f32, { 35, 36, 1, 1 } }, // divps
- { ISD::FDIV, MVT::f64, { 33, 34, 1, 1 } }, // divsd
- { ISD::FDIV, MVT::v2f64, { 65, 66, 1, 1 } }, // divpd
- };
-
- if (ST->useGLMDivSqrtCosts())
- if (const auto *Entry = CostTableLookup(GLMCostTable, ISD, LT.second))
- if (auto KindCost = Entry->Cost[CostKind])
- return LT.first * KindCost.value();
-
- static const CostKindTblEntry SLMCostTable[] = {
- { ISD::MUL, MVT::v4i32, { 11, 11, 1, 7 } }, // pmulld
- { ISD::MUL, MVT::v8i16, { 2, 5, 1, 1 } }, // pmullw
- { ISD::FMUL, MVT::f64, { 2, 5, 1, 1 } }, // mulsd
- { ISD::FMUL, MVT::f32, { 1, 4, 1, 1 } }, // mulss
- { ISD::FMUL, MVT::v2f64, { 4, 7, 1, 1 } }, // mulpd
- { ISD::FMUL, MVT::v4f32, { 2, 5, 1, 1 } }, // mulps
- { ISD::FDIV, MVT::f32, { 17, 19, 1, 1 } }, // divss
- { ISD::FDIV, MVT::v4f32, { 39, 39, 1, 6 } }, // divps
- { ISD::FDIV, MVT::f64, { 32, 34, 1, 1 } }, // divsd
- { ISD::FDIV, MVT::v2f64, { 69, 69, 1, 6 } }, // divpd
- { ISD::FADD, MVT::v2f64, { 2, 4, 1, 1 } }, // addpd
- { ISD::FSUB, MVT::v2f64, { 2, 4, 1, 1 } }, // subpd
- // v2i64/v4i64 mul is custom lowered as a series of long:
- // multiplies(3), shifts(3) and adds(2)
- // slm muldq version throughput is 2 and addq throughput 4
- // thus: 3X2 (muldq throughput) + 3X1 (shift throughput) +
- // 3X4 (addq throughput) = 17
- { ISD::MUL, MVT::v2i64, { 17, 22, 9, 9 } },
- // slm addq\subq throughput is 4
- { ISD::ADD, MVT::v2i64, { 4, 2, 1, 2 } },
- { ISD::SUB, MVT::v2i64, { 4, 2, 1, 2 } },
- };
-
- if (ST->useSLMArithCosts())
- if (const auto *Entry = CostTableLookup(SLMCostTable, ISD, LT.second))
- if (auto KindCost = Entry->Cost[CostKind])
- return LT.first * KindCost.value();
-
static const CostKindTblEntry AVX512BWUniformConstCostTable[] = {
{ ISD::SHL, MVT::v64i8, { 2 } }, // psllw + pand.
{ ISD::SRL, MVT::v64i8, { 2 } }, // psrlw + pand.
@@ -874,6 +833,47 @@ InstructionCost X86TTIImpl::getArithmeticInstrCost(
ISD = ISD::MUL;
}
+ static const CostKindTblEntry GLMCostTable[] = {
+ { ISD::FDIV, MVT::f32, { 18, 19, 1, 1 } }, // divss
+ { ISD::FDIV, MVT::v4f32, { 35, 36, 1, 1 } }, // divps
+ { ISD::FDIV, MVT::f64, { 33, 34, 1, 1 } }, // divsd
+ { ISD::FDIV, MVT::v2f64, { 65, 66, 1, 1 } }, // divpd
+ };
+
+ if (ST->useGLMDivSqrtCosts())
+ if (const auto *Entry = CostTableLookup(GLMCostTable, ISD, LT.second))
+ if (auto KindCost = Entry->Cost[CostKind])
+ return LT.first * KindCost.value();
+
+ static const CostKindTblEntry SLMCostTable[] = {
+ { ISD::MUL, MVT::v4i32, { 11, 11, 1, 7 } }, // pmulld
+ { ISD::MUL, MVT::v8i16, { 2, 5, 1, 1 } }, // pmullw
+ { ISD::FMUL, MVT::f64, { 2, 5, 1, 1 } }, // mulsd
+ { ISD::FMUL, MVT::f32, { 1, 4, 1, 1 } }, // mulss
+ { ISD::FMUL, MVT::v2f64, { 4, 7, 1, 1 } }, // mulpd
+ { ISD::FMUL, MVT::v4f32, { 2, 5, 1, 1 } }, // mulps
+ { ISD::FDIV, MVT::f32, { 17, 19, 1, 1 } }, // divss
+ { ISD::FDIV, MVT::v4f32, { 39, 39, 1, 6 } }, // divps
+ { ISD::FDIV, MVT::f64, { 32, 34, 1, 1 } }, // divsd
+ { ISD::FDIV, MVT::v2f64, { 69, 69, 1, 6 } }, // divpd
+ { ISD::FADD, MVT::v2f64, { 2, 4, 1, 1 } }, // addpd
+ { ISD::FSUB, MVT::v2f64, { 2, 4, 1, 1 } }, // subpd
+ // v2i64/v4i64 mul is custom lowered as a series of long:
+ // multiplies(3), shifts(3) and adds(2)
+ // slm muldq version throughput is 2 and addq throughput 4
+ // thus: 3X2 (muldq throughput) + 3X1 (shift throughput) +
+ // 3X4 (addq throughput) = 17
+ { ISD::MUL, MVT::v2i64, { 17, 22, 9, 9 } },
+ // slm addq\subq throughput is 4
+ { ISD::ADD, MVT::v2i64, { 4, 2, 1, 2 } },
+ { ISD::SUB, MVT::v2i64, { 4, 2, 1, 2 } },
+ };
+
+ if (ST->useSLMArithCosts())
+ if (const auto *Entry = CostTableLookup(SLMCostTable, ISD, LT.second))
+ if (auto KindCost = Entry->Cost[CostKind])
+ return LT.first * KindCost.value();
+
static const CostKindTblEntry AVX2CostTable[] = {
{ ISD::SHL, MVT::v16i8, { 6 } }, // vpblendvb sequence.
{ ISD::SHL, MVT::v32i8, { 6 } }, // vpblendvb sequence.
diff --git a/llvm/test/Analysis/CostModel/X86/fshl.ll b/llvm/test/Analysis/CostModel/X86/fshl.ll
index f0d70c06cf30..011b77f94d2a 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl.ll
@@ -752,9 +752,9 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
;
; SLM-LABEL: 'constant_funnel_i32'
; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 46 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 92 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_funnel_i32'
@@ -830,9 +830,9 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
;
; SLM-LABEL: 'constant_funnel_i16'
; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 84 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_funnel_i16'
@@ -1915,9 +1915,9 @@ define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
;
; SLM-LABEL: 'constant_rotate_i32'
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 29 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 58 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 116 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_rotate_i32'
@@ -1993,9 +1993,9 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
;
; SLM-LABEL: 'constant_rotate_i16'
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_rotate_i16'
diff --git a/llvm/test/Analysis/CostModel/X86/fshr.ll b/llvm/test/Analysis/CostModel/X86/fshr.ll
index e081faa0fa0f..f9237e09f670 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr.ll
@@ -752,9 +752,9 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
;
; SLM-LABEL: 'constant_funnel_i32'
; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 46 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 92 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_funnel_i32'
@@ -830,9 +830,9 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
;
; SLM-LABEL: 'constant_funnel_i16'
; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 84 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_funnel_i16'
@@ -1916,9 +1916,9 @@ define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
;
; SLM-LABEL: 'constant_rotate_i32'
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 29 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 58 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 116 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_rotate_i32'
@@ -1994,9 +1994,9 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
;
; SLM-LABEL: 'constant_rotate_i16'
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
-; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SLM-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; GLM-LABEL: 'constant_rotate_i16'
diff --git a/llvm/test/Analysis/CostModel/X86/mul.ll b/llvm/test/Analysis/CostModel/X86/mul.ll
index 3cfccbbae20a..05033916259c 100644
--- a/llvm/test/Analysis/CostModel/X86/mul.ll
+++ b/llvm/test/Analysis/CostModel/X86/mul.ll
@@ -151,17 +151,17 @@ define i32 @mul_constpow2() {
; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i64 = mul <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, 16
-; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, 16
-; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, 16
-; SLM-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLM-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
;
%I64 = mul i64 undef, 16
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
index 13cd5dfcbb09..da1019ba0934 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
@@ -12,7 +12,7 @@
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512VL
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BWVL
;
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,BTVER2
@@ -142,6 +142,10 @@ define <4 x i32> @var_shift_v4i32(<4 x i32> %a, <4 x i32> %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %b
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
;
+; SLM-LABEL: 'var_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i32> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
; BTVER2-LABEL: 'var_shift_v4i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <4 x i32> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
@@ -179,6 +183,10 @@ define <8 x i32> @var_shift_v8i32(<8 x i32> %a, <8 x i32> %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %b
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'var_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i32> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'var_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %shift = shl <8 x i32> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
@@ -216,6 +224,10 @@ define <16 x i32> @var_shift_v16i32(<16 x i32> %a, <16 x i32> %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %b
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'var_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <16 x i32> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'var_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %shift = shl <16 x i32> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
@@ -261,6 +273,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
;
+; SLM-LABEL: 'var_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
; BTVER2-LABEL: 'var_shift_v8i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <8 x i16> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
@@ -306,6 +322,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'var_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'var_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <16 x i16> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
@@ -351,6 +371,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'var_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'var_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %shift = shl <32 x i16> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
@@ -396,6 +420,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
;
+; SLM-LABEL: 'var_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
; BTVER2-LABEL: 'var_shift_v16i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
@@ -441,6 +469,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'var_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i8> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'var_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
@@ -486,6 +518,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'var_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <64 x i8> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'var_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
@@ -528,6 +564,12 @@ define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> poison, <2 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v2i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> poison, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> poison, <2 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
;
%insert = insertelement <2 x i64> poison, i64 %b, i32 0
%splat = shufflevector <2 x i64> %insert, <2 x i64> poison, <2 x i32> zeroinitializer
@@ -578,6 +620,12 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
;
+; SLM-LABEL: 'splatvar_shift_v4i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v4i64'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
@@ -633,6 +681,12 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
;
+; SLM-LABEL: 'splatvar_shift_v8i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v8i64'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
@@ -675,6 +729,12 @@ define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> poison, <4 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i32> poison, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> poison, <4 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
;
%insert = insertelement <4 x i32> poison, i32 %b, i32 0
%splat = shufflevector <4 x i32> %insert, <4 x i32> poison, <4 x i32> zeroinitializer
@@ -725,6 +785,12 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'splatvar_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
@@ -780,6 +846,12 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
@@ -822,6 +894,12 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
;
%insert = insertelement <8 x i16> poison, i16 %b, i32 0
%splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
@@ -872,6 +950,12 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
@@ -927,6 +1011,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'splatvar_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -994,6 +1084,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %splat
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
@@ -1067,6 +1163,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %splat
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
@@ -1140,6 +1242,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %splat
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <64 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1271,6 +1379,10 @@ define <4 x i32> @constant_shift_v4i32(<4 x i32> %a) {
; AVX512-LABEL: 'constant_shift_v4i32'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SLM-LABEL: 'constant_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
;
%shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
ret <4 x i32> %shift
@@ -1305,6 +1417,10 @@ define <8 x i32> @constant_shift_v8i32(<8 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
@@ -1342,6 +1458,10 @@ define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
@@ -1351,18 +1471,42 @@ define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
}
define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
-; CHECK-LABEL: 'constant_shift_v8i16'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+; SSE2-LABEL: 'constant_shift_v8i16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE42-LABEL: 'constant_shift_v8i16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'constant_shift_v8i16'
+; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v8i16'
+; XOP-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i16'
+; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SLM-LABEL: 'constant_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
;
%shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
ret <8 x i16> %shift
}
define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
-; SSE-LABEL: 'constant_shift_v16i16'
-; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+; SSE2-LABEL: 'constant_shift_v16i16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE42-LABEL: 'constant_shift_v16i16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
; AVX1-LABEL: 'constant_shift_v16i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
@@ -1396,6 +1540,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'constant_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
@@ -1405,9 +1553,13 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
}
define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
-; SSE-LABEL: 'constant_shift_v32i16'
-; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+; SSE2-LABEL: 'constant_shift_v32i16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE42-LABEL: 'constant_shift_v32i16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
; AVX1-LABEL: 'constant_shift_v32i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
@@ -1441,6 +1593,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'constant_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'constant_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
@@ -1486,6 +1642,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
;
+; SLM-LABEL: 'constant_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
@@ -1531,6 +1691,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'constant_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'constant_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
@@ -1576,6 +1740,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'constant_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'constant_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
@@ -1932,18 +2100,58 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
; non-uniform constant.
define <8 x i16> @test1(<8 x i16> %a) {
-; CHECK-LABEL: 'test1'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
-; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+; SSE2-LABEL: 'test1'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SSE42-LABEL: 'test1'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX-LABEL: 'test1'
+; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; XOP-LABEL: 'test1'
+; XOP-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; XOP-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX512-LABEL: 'test1'
+; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SLM-LABEL: 'test1'
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
;
%shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
ret <8 x i16> %shl
}
define <8 x i16> @test2(<8 x i16> %a) {
-; CHECK-LABEL: 'test2'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
-; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+; SSE2-LABEL: 'test2'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SSE42-LABEL: 'test2'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX-LABEL: 'test2'
+; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; XOP-LABEL: 'test2'
+; XOP-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; XOP-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX512-LABEL: 'test2'
+; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SLM-LABEL: 'test2'
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
;
%shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
ret <8 x i16> %shl
@@ -1974,6 +2182,10 @@ define <4 x i32> @test3(<4 x i32> %a) {
; AVX512-LABEL: 'test3'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; SLM-LABEL: 'test3'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
;
%shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
ret <4 x i32> %shl
@@ -1999,6 +2211,10 @@ define <4 x i32> @test4(<4 x i32> %a) {
; AVX512-LABEL: 'test4'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; SLM-LABEL: 'test4'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
;
%shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
ret <4 x i32> %shl
@@ -2051,9 +2267,13 @@ define <2 x i64> @test5(<2 x i64> %a) {
; simply emits 2 pmullw with no extract/insert.
define <16 x i16> @test6(<16 x i16> %a) {
-; SSE-LABEL: 'test6'
-; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+; SSE2-LABEL: 'test6'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; SSE42-LABEL: 'test6'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
;
; AVX1-LABEL: 'test6'
; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
@@ -2087,6 +2307,10 @@ define <16 x i16> @test6(<16 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
;
+; SLM-LABEL: 'test6'
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
; BTVER2-LABEL: 'test6'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
@@ -2128,6 +2352,10 @@ define <8 x i32> @test7(<8 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
;
+; SLM-LABEL: 'test7'
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
; BTVER2-LABEL: 'test7'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
@@ -2177,9 +2405,13 @@ define <4 x i64> @test8(<4 x i64> %a) {
; Same as 'test6', with the
diff erence that the cost is double.
define <32 x i16> @test9(<32 x i16> %a) {
-; SSE-LABEL: 'test9'
-; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+; SSE2-LABEL: 'test9'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; SSE42-LABEL: 'test9'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
;
; AVX1-LABEL: 'test9'
; AVX1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
@@ -2213,6 +2445,10 @@ define <32 x i16> @test9(<32 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
;
+; SLM-LABEL: 'test9'
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
; BTVER2-LABEL: 'test9'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
@@ -2252,6 +2488,10 @@ define <16 x i32> @test10(<16 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
;
+; SLM-LABEL: 'test10'
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
; BTVER2-LABEL: 'test10'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
index cdf35945821d..98202fd018fe 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
@@ -12,7 +12,7 @@
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512VL
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BWVL
;
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,BTVER2
@@ -142,6 +142,10 @@ define <4 x i32> @var_shift_v4i32(<4 x i32> %a, <4 x i32> %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %b
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
;
+; SLM-LABEL: 'var_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <4 x i32> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
; BTVER2-LABEL: 'var_shift_v4i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <4 x i32> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
@@ -179,6 +183,10 @@ define <8 x i32> @var_shift_v8i32(<8 x i32> %a, <8 x i32> %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %b
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'var_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <8 x i32> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'var_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %shift = shl <8 x i32> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
@@ -216,6 +224,10 @@ define <16 x i32> @var_shift_v16i32(<16 x i32> %a, <16 x i32> %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %b
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'var_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <16 x i32> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'var_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %shift = shl <16 x i32> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
@@ -261,6 +273,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
;
+; SLM-LABEL: 'var_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
; BTVER2-LABEL: 'var_shift_v8i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <8 x i16> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
@@ -306,6 +322,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'var_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'var_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <16 x i16> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
@@ -351,6 +371,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'var_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'var_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %shift = shl <32 x i16> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
@@ -396,6 +420,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
;
+; SLM-LABEL: 'var_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
; BTVER2-LABEL: 'var_shift_v16i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
@@ -441,6 +469,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'var_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i8> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'var_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
@@ -486,6 +518,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %b
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'var_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <64 x i8> %a, %b
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'var_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, %b
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
@@ -528,6 +564,12 @@ define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v2i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %shift
;
%insert = insertelement <2 x i64> undef, i64 %b, i32 0
%splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
@@ -578,6 +620,12 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
;
+; SLM-LABEL: 'splatvar_shift_v4i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v4i64'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -633,6 +681,12 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
;
+; SLM-LABEL: 'splatvar_shift_v8i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <8 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v8i64'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
@@ -675,6 +729,12 @@ define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i32> undef, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
;
%insert = insertelement <4 x i32> undef, i32 %b, i32 0
%splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -725,6 +785,12 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'splatvar_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -780,6 +846,12 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
@@ -822,6 +894,12 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
;
%insert = insertelement <8 x i16> undef, i16 %b, i32 0
%splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
@@ -872,6 +950,12 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -927,6 +1011,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'splatvar_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -994,6 +1084,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %splat
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -1067,6 +1163,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %splat
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -1140,6 +1242,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %splat
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <64 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1271,6 +1379,10 @@ define <4 x i32> @constant_shift_v4i32(<4 x i32> %a) {
; AVX512-LABEL: 'constant_shift_v4i32'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
+;
+; SLM-LABEL: 'constant_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shift
;
%shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
ret <4 x i32> %shift
@@ -1305,6 +1417,10 @@ define <8 x i32> @constant_shift_v8i32(<8 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shift
@@ -1342,6 +1458,10 @@ define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shift
@@ -1351,18 +1471,42 @@ define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
}
define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
-; CHECK-LABEL: 'constant_shift_v8i16'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+; SSE2-LABEL: 'constant_shift_v8i16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SSE42-LABEL: 'constant_shift_v8i16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX-LABEL: 'constant_shift_v8i16'
+; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; XOP-LABEL: 'constant_shift_v8i16'
+; XOP-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; XOP-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512-LABEL: 'constant_shift_v8i16'
+; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; SLM-LABEL: 'constant_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
;
%shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
ret <8 x i16> %shift
}
define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
-; SSE-LABEL: 'constant_shift_v16i16'
-; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+; SSE2-LABEL: 'constant_shift_v16i16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; SSE42-LABEL: 'constant_shift_v16i16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
; AVX1-LABEL: 'constant_shift_v16i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
@@ -1396,6 +1540,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'constant_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
@@ -1405,9 +1553,13 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
}
define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
-; SSE-LABEL: 'constant_shift_v32i16'
-; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+; SSE2-LABEL: 'constant_shift_v32i16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; SSE42-LABEL: 'constant_shift_v32i16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
; AVX1-LABEL: 'constant_shift_v32i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
@@ -1441,6 +1593,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'constant_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'constant_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
@@ -1486,6 +1642,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
;
+; SLM-LABEL: 'constant_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
@@ -1531,6 +1691,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'constant_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'constant_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
@@ -1576,6 +1740,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'constant_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'constant_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
@@ -1932,18 +2100,58 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
; non-uniform constant.
define <8 x i16> @test1(<8 x i16> %a) {
-; CHECK-LABEL: 'test1'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
-; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+; SSE2-LABEL: 'test1'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SSE42-LABEL: 'test1'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX-LABEL: 'test1'
+; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; XOP-LABEL: 'test1'
+; XOP-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; XOP-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX512-LABEL: 'test1'
+; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SLM-LABEL: 'test1'
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
;
%shl = shl <8 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
ret <8 x i16> %shl
}
define <8 x i16> @test2(<8 x i16> %a) {
-; CHECK-LABEL: 'test2'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
-; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+; SSE2-LABEL: 'test2'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SSE42-LABEL: 'test2'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX-LABEL: 'test2'
+; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; XOP-LABEL: 'test2'
+; XOP-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; XOP-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; AVX512-LABEL: 'test2'
+; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
+;
+; SLM-LABEL: 'test2'
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shl
;
%shl = shl <8 x i16> %a, <i16 0, i16 undef, i16 0, i16 0, i16 1, i16 undef, i16 -1, i16 1>
ret <8 x i16> %shl
@@ -1974,6 +2182,10 @@ define <4 x i32> @test3(<4 x i32> %a) {
; AVX512-LABEL: 'test3'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; SLM-LABEL: 'test3'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
;
%shl = shl <4 x i32> %a, <i32 1, i32 -1, i32 2, i32 -3>
ret <4 x i32> %shl
@@ -1999,6 +2211,10 @@ define <4 x i32> @test4(<4 x i32> %a) {
; AVX512-LABEL: 'test4'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
+;
+; SLM-LABEL: 'test4'
+; SLM-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %shl
;
%shl = shl <4 x i32> %a, <i32 0, i32 0, i32 1, i32 1>
ret <4 x i32> %shl
@@ -2051,9 +2267,13 @@ define <2 x i64> @test5(<2 x i64> %a) {
; simply emits 2 pmullw with no extract/insert.
define <16 x i16> @test6(<16 x i16> %a) {
-; SSE-LABEL: 'test6'
-; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+; SSE2-LABEL: 'test6'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; SSE42-LABEL: 'test6'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
;
; AVX1-LABEL: 'test6'
; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
@@ -2087,6 +2307,10 @@ define <16 x i16> @test6(<16 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
;
+; SLM-LABEL: 'test6'
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
; BTVER2-LABEL: 'test6'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
@@ -2128,6 +2352,10 @@ define <8 x i32> @test7(<8 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
;
+; SLM-LABEL: 'test7'
+; SLM-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
+;
; BTVER2-LABEL: 'test7'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %shl = shl <8 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %shl
@@ -2177,9 +2405,13 @@ define <4 x i64> @test8(<4 x i64> %a) {
; Same as 'test6', with the
diff erence that the cost is double.
define <32 x i16> @test9(<32 x i16> %a) {
-; SSE-LABEL: 'test9'
-; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
-; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+; SSE2-LABEL: 'test9'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; SSE42-LABEL: 'test9'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
;
; AVX1-LABEL: 'test9'
; AVX1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
@@ -2213,6 +2445,10 @@ define <32 x i16> @test9(<32 x i16> %a) {
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; AVX512BWVL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
;
+; SLM-LABEL: 'test9'
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
; BTVER2-LABEL: 'test9'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
@@ -2252,6 +2488,10 @@ define <16 x i32> @test10(<16 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
;
+; SLM-LABEL: 'test10'
+; SLM-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
+;
; BTVER2-LABEL: 'test10'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shl = shl <16 x i32> %a, <i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3, i32 1, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %shl
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
index 9247d1c1c249..ff6c245124c1 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+xop,+avx | FileCheck %s --check-prefixes=CHECK,XOP,XOPAVX1
@@ -12,8 +12,8 @@
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
;
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,BTVER2
; Verify the cost of vector logical shift right instructions.
@@ -164,6 +164,12 @@ define <2 x i64> @splatvar_shift_v2i64(<2 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v2i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <2 x i64> undef, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <2 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %shift
;
%insert = insertelement <2 x i64> undef, i64 %b, i32 0
%splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
@@ -214,6 +220,12 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %shift
;
+; SLM-LABEL: 'splatvar_shift_v4i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v4i64'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -269,6 +281,12 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %shift
;
+; SLM-LABEL: 'splatvar_shift_v8i64'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i64> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v8i64'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
@@ -311,6 +329,12 @@ define <4 x i32> @splatvar_shift_v4i32(<4 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <4 x i32> undef, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <4 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %shift
;
%insert = insertelement <4 x i32> undef, i32 %b, i32 0
%splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -361,6 +385,12 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'splatvar_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -416,6 +446,12 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
@@ -458,6 +494,12 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v8i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
;
%insert = insertelement <8 x i16> undef, i16 %b, i32 0
%splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
@@ -508,6 +550,12 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
;
+; SLM-LABEL: 'splatvar_shift_v16i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v16i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -563,6 +611,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
;
+; SLM-LABEL: 'splatvar_shift_v32i16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v32i16'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -605,6 +659,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i8> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; SLM-LABEL: 'splatvar_shift_v16i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
;
%insert = insertelement <16 x i8> undef, i8 %b, i32 0
%splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -655,6 +715,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v32i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v32i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -710,6 +776,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, %splat
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
;
+; SLM-LABEL: 'splatvar_shift_v64i8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, %splat
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
; BTVER2-LABEL: 'splatvar_shift_v64i8'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -782,6 +854,10 @@ define <4 x i32> @constant_shift_v4i32(<4 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v4i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v4i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <4 x i32> %a, <i32 4, i32 5, i32 6, i32 7>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %shift
@@ -819,6 +895,10 @@ define <8 x i32> @constant_shift_v8i32(<8 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v8i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v8i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %shift = shl <8 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %shift
@@ -856,6 +936,10 @@ define <16 x i32> @constant_shift_v16i32(<16 x i32> %a) {
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %shift
;
+; SLM-LABEL: 'constant_shift_v16i32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %shift
+;
; BTVER2-LABEL: 'constant_shift_v16i32'
; BTVER2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %shift = shl <16 x i32> %a, <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %shift
@@ -874,13 +958,9 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
}
define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
-; SSE2-LABEL: 'constant_shift_v16i16'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
-;
-; SSE42-LABEL: 'constant_shift_v16i16'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+; SSE-LABEL: 'constant_shift_v16i16'
+; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
;
; AVX1-LABEL: 'constant_shift_v16i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
@@ -911,13 +991,9 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
}
define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
-; SSE2-LABEL: 'constant_shift_v32i16'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
-;
-; SSE42-LABEL: 'constant_shift_v32i16'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
-; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+; SSE-LABEL: 'constant_shift_v32i16'
+; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
;
; AVX1-LABEL: 'constant_shift_v32i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
@@ -1051,13 +1127,9 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
}
define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
-; SSE2-LABEL: 'splatconstant_shift_v32i16'
-; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
-; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
-;
-; SSE42-LABEL: 'splatconstant_shift_v32i16'
-; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
-; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+; SSE-LABEL: 'splatconstant_shift_v32i16'
+; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
;
; AVX1-LABEL: 'splatconstant_shift_v32i16'
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
More information about the llvm-commits
mailing list