Skip to content

Commit 5842dfe

Browse files
RKSimonRKSimon
committed
[CostModel][X86] Update SSSE3/AVX1 BSWAP costs
Drop atom/slm costs from the default bswap costs, and update the avx1 latency costs based off latest codegen. Based off analysis report from https://github.com/RKSimon/llvm-scripts/check_cost_tables.py Fixes #62659
1 parent c1764a7 commit 5842dfe

File tree

3 files changed

+52
-49
lines changed

3 files changed

+52
-49
lines changed

‎llvm/lib/Target/X86/X86TargetTransformInfo.cpp

Lines changed: 10 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -3733,10 +3733,10 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
37333733
{ ISD::BITREVERSE, MVT::v8i16, { 8, 13, 10, 16 } },
37343734
{ ISD::BITREVERSE, MVT::v32i8, { 13, 15, 17, 26 } }, // 2 x 128-bit Op + extract/insert
37353735
{ ISD::BITREVERSE, MVT::v16i8, { 7, 7, 9, 13 } },
3736-
{ ISD::BSWAP, MVT::v4i64, { 5, 7, 5, 10 } },
3737-
{ ISD::BSWAP, MVT::v2i64, { 2, 3, 1, 3 } },
3738-
{ ISD::BSWAP, MVT::v8i32, { 5, 7, 5, 10 } },
3739-
{ ISD::BSWAP, MVT::v4i32, { 2, 3, 1, 3 } },
3736+
{ ISD::BSWAP, MVT::v4i64, { 5, 6, 5, 10 } },
3737+
{ ISD::BSWAP, MVT::v2i64, { 2, 2, 1, 3 } },
3738+
{ ISD::BSWAP, MVT::v8i32, { 5, 6, 5, 10 } },
3739+
{ ISD::BSWAP, MVT::v4i32, { 2, 2, 1, 3 } },
37403740
{ ISD::BSWAP, MVT::v16i16, { 5, 6, 5, 10 } },
37413741
{ ISD::BSWAP, MVT::v8i16, { 2, 2, 1, 3 } },
37423742
{ ISD::CTLZ, MVT::v4i64, { 29, 33, 49, 58 } }, // 2 x 128-bit Op + extract/insert
@@ -3813,6 +3813,9 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
38133813
{ ISD::FSQRT, MVT::v2f64, { 67, 71, 1, 5 } }, // sqrtpd
38143814
};
38153815
static const CostKindTblEntry SLMCostTbl[] = {
3816+
{ ISD::BSWAP, MVT::v2i64, { 5, 5, 1, 5 } },
3817+
{ ISD::BSWAP, MVT::v4i32, { 5, 5, 1, 5 } },
3818+
{ ISD::BSWAP, MVT::v8i16, { 5, 5, 1, 5 } },
38163819
{ ISD::FSQRT, MVT::f32, { 20, 20, 1, 1 } }, // sqrtss
38173820
{ ISD::FSQRT, MVT::v4f32, { 40, 41, 1, 5 } }, // sqrtps
38183821
{ ISD::FSQRT, MVT::f64, { 35, 35, 1, 1 } }, // sqrtsd
@@ -3851,9 +3854,9 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
38513854
{ ISD::BITREVERSE, MVT::v4i32, { 16, 20, 11, 21 } },
38523855
{ ISD::BITREVERSE, MVT::v8i16, { 16, 20, 11, 21 } },
38533856
{ ISD::BITREVERSE, MVT::v16i8, { 11, 12, 10, 16 } },
3854-
{ ISD::BSWAP, MVT::v2i64, { 5, 5, 1, 5 } },
3855-
{ ISD::BSWAP, MVT::v4i32, { 5, 5, 1, 5 } },
3856-
{ ISD::BSWAP, MVT::v8i16, { 5, 5, 1, 5 } },
3857+
{ ISD::BSWAP, MVT::v2i64, { 2, 3, 1, 5 } },
3858+
{ ISD::BSWAP, MVT::v4i32, { 2, 3, 1, 5 } },
3859+
{ ISD::BSWAP, MVT::v8i16, { 2, 3, 1, 5 } },
38573860
{ ISD::CTLZ, MVT::v2i64, { 18, 28, 28, 35 } },
38583861
{ ISD::CTLZ, MVT::v4i32, { 15, 20, 22, 28 } },
38593862
{ ISD::CTLZ, MVT::v8i16, { 13, 17, 16, 22 } },

‎llvm/test/Analysis/CostModel/X86/bswap-latency.ll

Lines changed: 24 additions & 24 deletions
Original file line numberDiff line numberDiff line change
@@ -24,23 +24,23 @@ define void @cost_bswap_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
2424
;
2525
; SSSE3-LABEL: 'cost_bswap_i64'
2626
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.bswap.i64(i64 %a64)
27-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
28-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
29-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
27+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
28+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
29+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
3030
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
3131
;
3232
; SSE42-LABEL: 'cost_bswap_i64'
3333
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.bswap.i64(i64 %a64)
34-
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
35-
; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
36-
; SSE42-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
34+
; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
35+
; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
36+
; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
3737
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
3838
;
3939
; AVX1-LABEL: 'cost_bswap_i64'
4040
; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I64 = call i64 @llvm.bswap.i64(i64 %a64)
41-
; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
42-
; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
43-
; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
41+
; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
42+
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
43+
; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
4444
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
4545
;
4646
; AVX2-LABEL: 'cost_bswap_i64'
@@ -95,23 +95,23 @@ define void @cost_bswap_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
9595
;
9696
; SSSE3-LABEL: 'cost_bswap_i32'
9797
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.bswap.i32(i32 %a32)
98-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
99-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
100-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
98+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
99+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
100+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
101101
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
102102
;
103103
; SSE42-LABEL: 'cost_bswap_i32'
104104
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.bswap.i32(i32 %a32)
105-
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
106-
; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
107-
; SSE42-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
105+
; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
106+
; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
107+
; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
108108
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
109109
;
110110
; AVX1-LABEL: 'cost_bswap_i32'
111111
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.bswap.i32(i32 %a32)
112-
; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
113-
; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
114-
; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
112+
; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
113+
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
114+
; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
115115
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
116116
;
117117
; AVX2-LABEL: 'cost_bswap_i32'
@@ -166,16 +166,16 @@ define void @cost_bswap_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
166166
;
167167
; SSSE3-LABEL: 'cost_bswap_i16'
168168
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.bswap.i16(i16 %a16)
169-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
170-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
171-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
169+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
170+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
171+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
172172
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
173173
;
174174
; SSE42-LABEL: 'cost_bswap_i16'
175175
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.bswap.i16(i16 %a16)
176-
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
177-
; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
178-
; SSE42-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
176+
; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
177+
; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
178+
; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
179179
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
180180
;
181181
; AVX1-LABEL: 'cost_bswap_i16'

‎llvm/test/Analysis/CostModel/X86/bswap.ll

Lines changed: 18 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -24,16 +24,16 @@ define void @cost_bswap_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
2424
;
2525
; SSSE3-LABEL: 'cost_bswap_i64'
2626
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.bswap.i64(i64 %a64)
27-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
28-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
29-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
27+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
28+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
29+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
3030
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
3131
;
3232
; SSE42-LABEL: 'cost_bswap_i64'
3333
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.bswap.i64(i64 %a64)
34-
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
35-
; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
36-
; SSE42-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
34+
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I64 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a128)
35+
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4I64 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a256)
36+
; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8I64 = call <8 x i64> @llvm.bswap.v8i64(<8 x i64> %a512)
3737
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
3838
;
3939
; AVX1-LABEL: 'cost_bswap_i64'
@@ -95,16 +95,16 @@ define void @cost_bswap_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
9595
;
9696
; SSSE3-LABEL: 'cost_bswap_i32'
9797
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.bswap.i32(i32 %a32)
98-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
99-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
100-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
98+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
99+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
100+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
101101
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
102102
;
103103
; SSE42-LABEL: 'cost_bswap_i32'
104104
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.bswap.i32(i32 %a32)
105-
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
106-
; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
107-
; SSE42-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
105+
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2I32 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a128)
106+
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4I32 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a256)
107+
; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8I32 = call <16 x i32> @llvm.bswap.v16i32(<16 x i32> %a512)
108108
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
109109
;
110110
; AVX1-LABEL: 'cost_bswap_i32'
@@ -166,16 +166,16 @@ define void @cost_bswap_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
166166
;
167167
; SSSE3-LABEL: 'cost_bswap_i16'
168168
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.bswap.i16(i16 %a16)
169-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
170-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
171-
; SSSE3-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
169+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
170+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
171+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
172172
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
173173
;
174174
; SSE42-LABEL: 'cost_bswap_i16'
175175
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.bswap.i16(i16 %a16)
176-
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
177-
; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
178-
; SSE42-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
176+
; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a128)
177+
; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a256)
178+
; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.bswap.v32i16(<32 x i16> %a512)
179179
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
180180
;
181181
; AVX1-LABEL: 'cost_bswap_i16'

0 commit comments

Comments
 (0)