[llvm] r291354 - [CostModel][AVX512BW] Add v32i16 vector shift costs for avx512bw targets.
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sat Jan 7 09:54:11 PST 2017
Author: rksimon
Date: Sat Jan 7 11:54:10 2017
New Revision: 291354
URL: http://llvm.org/viewvc/llvm-project?rev=291354&view=rev
Log:
[CostModel][AVX512BW] Add v32i16 vector shift costs for avx512bw targets.
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=291354&r1=291353&r2=291354&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp Sat Jan 7 11:54:10 2017
@@ -219,6 +219,10 @@ int X86TTIImpl::getArithmeticInstrCost(
return LT.first * Entry->Cost;
static const CostTblEntry AVX512BWCostTable[] = {
+ { ISD::SHL, MVT::v32i16, 1 }, // vpsllvw
+ { ISD::SRL, MVT::v32i16, 1 }, // vpsrlvw
+ { ISD::SRA, MVT::v32i16, 1 }, // vpsravw
+
{ ISD::MUL, MVT::v64i8, 11 }, // extend/pmullw/trunc sequence.
{ ISD::MUL, MVT::v32i8, 4 }, // extend/pmullw/trunc sequence.
{ ISD::MUL, MVT::v16i8, 4 }, // extend/pmullw/trunc 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=291354&r1=291353&r2=291354&view=diff
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll (original)
+++ llvm/trunk/test/Analysis/CostModel/X86/vshift-ashr-cost.ll Sat Jan 7 11:54:10 2017
@@ -120,7 +120,7 @@ define <32 x i16> @var_shift_v32i16(<32
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%shift = ashr <32 x i16> %a, %b
ret <32 x i16> %shift
@@ -282,7 +282,7 @@ define <32 x i16> @splatvar_shift_v32i16
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%splat = shufflevector <32 x i16> %b, <32 x i16> undef, <32 x i32> zeroinitializer
%shift = ashr <32 x i16> %a, %splat
@@ -439,7 +439,7 @@ define <32 x i16> @constant_shift_v32i16
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%shift = ashr <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>
ret <32 x i16> %shift
@@ -592,7 +592,7 @@ define <32 x i16> @splatconstant_shift_v
; AVX: Found an estimated cost of 4 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%shift = ashr <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>
ret <32 x i16> %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=291354&r1=291353&r2=291354&view=diff
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll (original)
+++ llvm/trunk/test/Analysis/CostModel/X86/vshift-lshr-cost.ll Sat Jan 7 11:54:10 2017
@@ -123,7 +123,7 @@ define <32 x i16> @var_shift_v32i16(<32
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%shift = lshr <32 x i16> %a, %b
ret <32 x i16> %shift
@@ -287,7 +287,7 @@ define <32 x i16> @splatvar_shift_v32i16
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%splat = shufflevector <32 x i16> %b, <32 x i16> undef, <32 x i32> zeroinitializer
%shift = lshr <32 x i16> %a, %splat
@@ -447,7 +447,7 @@ define <32 x i16> @constant_shift_v32i16
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%shift = lshr <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>
ret <32 x i16> %shift
@@ -603,7 +603,7 @@ define <32 x i16> @splatconstant_shift_v
; AVX: Found an estimated cost of 4 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 8 for instruction: %shift
%shift = lshr <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>
ret <32 x i16> %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=291354&r1=291353&r2=291354&view=diff
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll (original)
+++ llvm/trunk/test/Analysis/CostModel/X86/vshift-shl-cost.ll Sat Jan 7 11:54:10 2017
@@ -124,7 +124,7 @@ define <32 x i16> @var_shift_v32i16(<32
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 4 for instruction: %shift
%shift = shl <32 x i16> %a, %b
ret <32 x i16> %shift
@@ -288,7 +288,7 @@ define <32 x i16> @splatvar_shift_v32i16
; AVX: Found an estimated cost of 56 for instruction: %shift
; AVX2: Found an estimated cost of 20 for instruction: %shift
; AVX512F: Found an estimated cost of 20 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOP: Found an estimated cost of 4 for instruction: %shift
%splat = shufflevector <32 x i16> %b, <32 x i16> undef, <32 x i32> zeroinitializer
%shift = shl <32 x i16> %a, %splat
@@ -449,7 +449,7 @@ define <32 x i16> @constant_shift_v32i16
; AVX: Found an estimated cost of 8 for instruction: %shift
; AVX2: Found an estimated cost of 2 for instruction: %shift
; AVX512F: Found an estimated cost of 2 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOPAVX: Found an estimated cost of 4 for instruction: %shift
; XOPAVX2: Found an estimated cost of 2 for instruction: %shift
%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>
@@ -607,7 +607,7 @@ define <32 x i16> @splatconstant_shift_v
; AVX: Found an estimated cost of 4 for instruction: %shift
; AVX2: Found an estimated cost of 2 for instruction: %shift
; AVX512F: Found an estimated cost of 2 for instruction: %shift
-; AVX512BW: Found an estimated cost of 2 for instruction: %shift
+; AVX512BW: Found an estimated cost of 1 for instruction: %shift
; XOPAVX: Found an estimated cost of 4 for instruction: %shift
; XOPAVX2: Found an estimated cost of 2 for instruction: %shift
%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