[llvm] r303012 - [X86][SSE] Account for cost of extract/insert of v32i8 vector shifts
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sun May 14 10:36:08 PDT 2017
Author: rksimon
Date: Sun May 14 12:36:07 2017
New Revision: 303012
URL: http://llvm.org/viewvc/llvm-project?rev=303012&view=rev
Log:
[X86][SSE] Account for cost of extract/insert of v32i8 vector shifts
Modified:
llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp
llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll
Modified: llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp?rev=303012&r1=303011&r2=303012&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp Sun May 14 12:36:07 2017
@@ -251,9 +251,9 @@ int X86TTIImpl::getArithmeticInstrCost(
{ ISD::SRL, MVT::v16i8, 2 }, // psrlw + pand.
{ ISD::SRA, MVT::v16i8, 4 }, // psrlw, pand, pxor, psubb.
- { ISD::SHL, MVT::v32i8, 4 }, // 2*(psllw + pand).
- { ISD::SRL, MVT::v32i8, 4 }, // 2*(psrlw + pand).
- { ISD::SRA, MVT::v32i8, 8 }, // 2*(psrlw, pand, pxor, psubb).
+ { ISD::SHL, MVT::v32i8, 4+2 }, // 2*(psllw + pand) + split.
+ { ISD::SRL, MVT::v32i8, 4+2 }, // 2*(psrlw + pand) + split.
+ { ISD::SRA, MVT::v32i8, 8+2 }, // 2*(psrlw, pand, pxor, psubb) + split.
{ ISD::SDIV, MVT::v16i16, 12 }, // pmulhw sequence
{ ISD::SDIV, MVT::v8i16, 6 }, // pmulhw sequence
Modified: llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll?rev=303012&r1=303011&r2=303012&view=diff
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll (original)
+++ llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll Sun May 14 12:36:07 2017
@@ -637,10 +637,10 @@ define <32 x i8> @splatconstant_shift_v3
; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i8':
; SSE2: Found an estimated cost of 8 for instruction: %shift
; SSE41: Found an estimated cost of 8 for instruction: %shift
-; AVX: Found an estimated cost of 8 for instruction: %shift
+; AVX: Found an estimated cost of 10 for instruction: %shift
; AVX2: Found an estimated cost of 4 for instruction: %shift
; AVX512: Found an estimated cost of 4 for instruction: %shift
-; XOPAVX: Found an estimated cost of 8 for instruction: %shift
+; XOPAVX: Found an estimated cost of 10 for instruction: %shift
; XOPAVX2: Found an estimated cost of 4 for instruction: %shift
%shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <32 x i8> %shift
@@ -650,11 +650,11 @@ define <64 x i8> @splatconstant_shift_v6
; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v64i8':
; SSE2: Found an estimated cost of 16 for instruction: %shift
; SSE41: Found an estimated cost of 16 for instruction: %shift
-; AVX: Found an estimated cost of 16 for instruction: %shift
+; AVX: Found an estimated cost of 20 for instruction: %shift
; AVX2: Found an estimated cost of 8 for instruction: %shift
; AVX512F: Found an estimated cost of 8 for instruction: %shift
; AVX512BW: Found an estimated cost of 4 for instruction: %shift
-; XOPAVX: Found an estimated cost of 16 for instruction: %shift
+; XOPAVX: Found an estimated cost of 20 for instruction: %shift
; XOPAVX2: Found an estimated cost of 8 for instruction: %shift
%shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <64 x i8> %shift
Modified: llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll?rev=303012&r1=303011&r2=303012&view=diff
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll (original)
+++ llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll Sun May 14 12:36:07 2017
@@ -644,10 +644,10 @@ define <32 x i8> @splatconstant_shift_v3
; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i8':
; SSE2: Found an estimated cost of 4 for instruction: %shift
; SSE41: Found an estimated cost of 4 for instruction: %shift
-; AVX: Found an estimated cost of 4 for instruction: %shift
+; AVX: Found an estimated cost of 6 for instruction: %shift
; AVX2: Found an estimated cost of 2 for instruction: %shift
; AVX512: Found an estimated cost of 2 for instruction: %shift
-; XOPAVX: Found an estimated cost of 4 for instruction: %shift
+; XOPAVX: Found an estimated cost of 6 for instruction: %shift
; XOPAVX2: Found an estimated cost of 2 for instruction: %shift
%shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <32 x i8> %shift
@@ -657,11 +657,11 @@ define <64 x i8> @splatconstant_shift_v6
; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v64i8':
; SSE2: Found an estimated cost of 8 for instruction: %shift
; SSE41: Found an estimated cost of 8 for instruction: %shift
-; AVX: Found an estimated cost of 8 for instruction: %shift
+; AVX: Found an estimated cost of 12 for instruction: %shift
; AVX2: Found an estimated cost of 4 for instruction: %shift
; AVX512F: Found an estimated cost of 4 for instruction: %shift
; AVX512BW: Found an estimated cost of 2 for instruction: %shift
-; XOPAVX: Found an estimated cost of 8 for instruction: %shift
+; XOPAVX: Found an estimated cost of 12 for instruction: %shift
; XOPAVX2: Found an estimated cost of 4 for instruction: %shift
%shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <64 x i8> %shift
Modified: llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll?rev=303012&r1=303011&r2=303012&view=diff
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll (original)
+++ llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll Sun May 14 12:36:07 2017
@@ -647,10 +647,10 @@ define <32 x i8> @splatconstant_shift_v3
; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i8':
; SSE2: Found an estimated cost of 4 for instruction: %shift
; SSE41: Found an estimated cost of 4 for instruction: %shift
-; AVX: Found an estimated cost of 4 for instruction: %shift
+; AVX: Found an estimated cost of 6 for instruction: %shift
; AVX2: Found an estimated cost of 2 for instruction: %shift
; AVX512: Found an estimated cost of 2 for instruction: %shift
-; XOPAVX: Found an estimated cost of 4 for instruction: %shift
+; XOPAVX: Found an estimated cost of 6 for instruction: %shift
; XOPAVX2: Found an estimated cost of 2 for instruction: %shift
%shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <32 x i8> %shift
@@ -660,11 +660,11 @@ define <64 x i8> @splatconstant_shift_v6
; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v64i8':
; SSE2: Found an estimated cost of 8 for instruction: %shift
; SSE41: Found an estimated cost of 8 for instruction: %shift
-; AVX: Found an estimated cost of 8 for instruction: %shift
+; AVX: Found an estimated cost of 12 for instruction: %shift
; AVX2: Found an estimated cost of 4 for instruction: %shift
; AVX512F: Found an estimated cost of 4 for instruction: %shift
; AVX512BW: Found an estimated cost of 2 for instruction: %shift
-; XOPAVX: Found an estimated cost of 8 for instruction: %shift
+; XOPAVX: Found an estimated cost of 12 for instruction: %shift
; XOPAVX2: Found an estimated cost of 4 for instruction: %shift
%shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <64 x i8> %shift
More information about the llvm-commits
mailing list