maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/avx2-vbroadcast.ll

574 lines
21 KiB
LLVM
Raw Normal View History

Add AVX2 vpbroadcast support llvm-svn: 144967
2011-11-18 10:49:55 +08:00
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core-avx2 -mattr=+avx2 | FileCheck %s
; CHECK: vpbroadcastb (%
define <16 x i8> @BB16(i8* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i8* %ptr, align 4
%q0 = insertelement <16 x i8> undef, i8 %q, i32 0
%q1 = insertelement <16 x i8> %q0, i8 %q, i32 1
%q2 = insertelement <16 x i8> %q1, i8 %q, i32 2
%q3 = insertelement <16 x i8> %q2, i8 %q, i32 3
%q4 = insertelement <16 x i8> %q3, i8 %q, i32 4
%q5 = insertelement <16 x i8> %q4, i8 %q, i32 5
%q6 = insertelement <16 x i8> %q5, i8 %q, i32 6
%q7 = insertelement <16 x i8> %q6, i8 %q, i32 7
%q8 = insertelement <16 x i8> %q7, i8 %q, i32 8
%q9 = insertelement <16 x i8> %q8, i8 %q, i32 9
%qa = insertelement <16 x i8> %q9, i8 %q, i32 10
%qb = insertelement <16 x i8> %qa, i8 %q, i32 11
%qc = insertelement <16 x i8> %qb, i8 %q, i32 12
%qd = insertelement <16 x i8> %qc, i8 %q, i32 13
%qe = insertelement <16 x i8> %qd, i8 %q, i32 14
%qf = insertelement <16 x i8> %qe, i8 %q, i32 15
ret <16 x i8> %qf
}
; CHECK: vpbroadcastb (%
define <32 x i8> @BB32(i8* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i8* %ptr, align 4
%q0 = insertelement <32 x i8> undef, i8 %q, i32 0
%q1 = insertelement <32 x i8> %q0, i8 %q, i32 1
%q2 = insertelement <32 x i8> %q1, i8 %q, i32 2
%q3 = insertelement <32 x i8> %q2, i8 %q, i32 3
%q4 = insertelement <32 x i8> %q3, i8 %q, i32 4
%q5 = insertelement <32 x i8> %q4, i8 %q, i32 5
%q6 = insertelement <32 x i8> %q5, i8 %q, i32 6
%q7 = insertelement <32 x i8> %q6, i8 %q, i32 7
%q8 = insertelement <32 x i8> %q7, i8 %q, i32 8
%q9 = insertelement <32 x i8> %q8, i8 %q, i32 9
%qa = insertelement <32 x i8> %q9, i8 %q, i32 10
%qb = insertelement <32 x i8> %qa, i8 %q, i32 11
%qc = insertelement <32 x i8> %qb, i8 %q, i32 12
%qd = insertelement <32 x i8> %qc, i8 %q, i32 13
%qe = insertelement <32 x i8> %qd, i8 %q, i32 14
%qf = insertelement <32 x i8> %qe, i8 %q, i32 15
%q20 = insertelement <32 x i8> %qf, i8 %q, i32 16
%q21 = insertelement <32 x i8> %q20, i8 %q, i32 17
%q22 = insertelement <32 x i8> %q21, i8 %q, i32 18
%q23 = insertelement <32 x i8> %q22, i8 %q, i32 19
%q24 = insertelement <32 x i8> %q23, i8 %q, i32 20
%q25 = insertelement <32 x i8> %q24, i8 %q, i32 21
%q26 = insertelement <32 x i8> %q25, i8 %q, i32 22
%q27 = insertelement <32 x i8> %q26, i8 %q, i32 23
%q28 = insertelement <32 x i8> %q27, i8 %q, i32 24
%q29 = insertelement <32 x i8> %q28, i8 %q, i32 25
%q2a = insertelement <32 x i8> %q29, i8 %q, i32 26
%q2b = insertelement <32 x i8> %q2a, i8 %q, i32 27
%q2c = insertelement <32 x i8> %q2b, i8 %q, i32 28
%q2d = insertelement <32 x i8> %q2c, i8 %q, i32 29
%q2e = insertelement <32 x i8> %q2d, i8 %q, i32 30
%q2f = insertelement <32 x i8> %q2e, i8 %q, i32 31
ret <32 x i8> %q2f
}
; CHECK: vpbroadcastw (%
define <8 x i16> @W16(i16* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i16* %ptr, align 4
%q0 = insertelement <8 x i16> undef, i16 %q, i32 0
%q1 = insertelement <8 x i16> %q0, i16 %q, i32 1
%q2 = insertelement <8 x i16> %q1, i16 %q, i32 2
%q3 = insertelement <8 x i16> %q2, i16 %q, i32 3
%q4 = insertelement <8 x i16> %q3, i16 %q, i32 4
%q5 = insertelement <8 x i16> %q4, i16 %q, i32 5
%q6 = insertelement <8 x i16> %q5, i16 %q, i32 6
%q7 = insertelement <8 x i16> %q6, i16 %q, i32 7
ret <8 x i16> %q7
}
; CHECK: vpbroadcastw (%
define <16 x i16> @WW16(i16* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i16* %ptr, align 4
%q0 = insertelement <16 x i16> undef, i16 %q, i32 0
%q1 = insertelement <16 x i16> %q0, i16 %q, i32 1
%q2 = insertelement <16 x i16> %q1, i16 %q, i32 2
%q3 = insertelement <16 x i16> %q2, i16 %q, i32 3
%q4 = insertelement <16 x i16> %q3, i16 %q, i32 4
%q5 = insertelement <16 x i16> %q4, i16 %q, i32 5
%q6 = insertelement <16 x i16> %q5, i16 %q, i32 6
%q7 = insertelement <16 x i16> %q6, i16 %q, i32 7
%q8 = insertelement <16 x i16> %q7, i16 %q, i32 8
%q9 = insertelement <16 x i16> %q8, i16 %q, i32 9
%qa = insertelement <16 x i16> %q9, i16 %q, i32 10
%qb = insertelement <16 x i16> %qa, i16 %q, i32 11
%qc = insertelement <16 x i16> %qb, i16 %q, i32 12
%qd = insertelement <16 x i16> %qc, i16 %q, i32 13
%qe = insertelement <16 x i16> %qd, i16 %q, i32 14
%qf = insertelement <16 x i16> %qe, i16 %q, i32 15
ret <16 x i16> %qf
}
[X86] Add broadcast instructions to the table used by ExeDepsFix pass. Adds the different broadcast instructions to the ReplaceableInstrsAVX2 table. That way the ExeDepsFix pass can take better decisions when AVX2 broadcasts are across domain (int <-> float). In particular, prior to this patch we were generating: vpbroadcastd LCPI1_0(%rip), %ymm2 vpand %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 ## <- domain change penalty Now, we generate the following nice sequence where everything is in the float domain: vbroadcastss LCPI1_0(%rip), %ymm2 vandps %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 <rdar://problem/16354675> llvm-svn: 204770
2014-03-26 08:10:22 +08:00
; CHECK: vbroadcastss (%
Add AVX2 vpbroadcast support llvm-svn: 144967
2011-11-18 10:49:55 +08:00
define <4 x i32> @D32(i32* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i32* %ptr, align 4
%q0 = insertelement <4 x i32> undef, i32 %q, i32 0
%q1 = insertelement <4 x i32> %q0, i32 %q, i32 1
%q2 = insertelement <4 x i32> %q1, i32 %q, i32 2
%q3 = insertelement <4 x i32> %q2, i32 %q, i32 3
ret <4 x i32> %q3
}
[X86] Add broadcast instructions to the table used by ExeDepsFix pass. Adds the different broadcast instructions to the ReplaceableInstrsAVX2 table. That way the ExeDepsFix pass can take better decisions when AVX2 broadcasts are across domain (int <-> float). In particular, prior to this patch we were generating: vpbroadcastd LCPI1_0(%rip), %ymm2 vpand %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 ## <- domain change penalty Now, we generate the following nice sequence where everything is in the float domain: vbroadcastss LCPI1_0(%rip), %ymm2 vandps %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 <rdar://problem/16354675> llvm-svn: 204770
2014-03-26 08:10:22 +08:00
; CHECK: vbroadcastss (%
Add AVX2 vpbroadcast support llvm-svn: 144967
2011-11-18 10:49:55 +08:00
define <8 x i32> @DD32(i32* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i32* %ptr, align 4
%q0 = insertelement <8 x i32> undef, i32 %q, i32 0
%q1 = insertelement <8 x i32> %q0, i32 %q, i32 1
%q2 = insertelement <8 x i32> %q1, i32 %q, i32 2
%q3 = insertelement <8 x i32> %q2, i32 %q, i32 3
%q4 = insertelement <8 x i32> %q3, i32 %q, i32 4
%q5 = insertelement <8 x i32> %q4, i32 %q, i32 5
%q6 = insertelement <8 x i32> %q5, i32 %q, i32 6
%q7 = insertelement <8 x i32> %q6, i32 %q, i32 7
ret <8 x i32> %q7
}
; CHECK: vpbroadcastq (%
define <2 x i64> @Q64(i64* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i64* %ptr, align 4
%q0 = insertelement <2 x i64> undef, i64 %q, i32 0
%q1 = insertelement <2 x i64> %q0, i64 %q, i32 1
ret <2 x i64> %q1
}
[X86] Add broadcast instructions to the table used by ExeDepsFix pass. Adds the different broadcast instructions to the ReplaceableInstrsAVX2 table. That way the ExeDepsFix pass can take better decisions when AVX2 broadcasts are across domain (int <-> float). In particular, prior to this patch we were generating: vpbroadcastd LCPI1_0(%rip), %ymm2 vpand %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 ## <- domain change penalty Now, we generate the following nice sequence where everything is in the float domain: vbroadcastss LCPI1_0(%rip), %ymm2 vandps %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 <rdar://problem/16354675> llvm-svn: 204770
2014-03-26 08:10:22 +08:00
; CHECK: vbroadcastsd (%
Add AVX2 vpbroadcast support llvm-svn: 144967
2011-11-18 10:49:55 +08:00
define <4 x i64> @QQ64(i64* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load i64* %ptr, align 4
%q0 = insertelement <4 x i64> undef, i64 %q, i32 0
%q1 = insertelement <4 x i64> %q0, i64 %q, i32 1
%q2 = insertelement <4 x i64> %q1, i64 %q, i32 2
%q3 = insertelement <4 x i64> %q2, i64 %q, i32 3
ret <4 x i64> %q3
}
Fix a crash in AVX2 when trying to broadcast a double into a 128-bit vector. There is no vbroadcastsd xmm, but we do need to support 64-bit integers broadcasted into xmm. Also factor the AVX check into the isVectorBroadcast function. This makes more sense since the AVX2 check was already inside. llvm-svn: 147844
2012-01-10 16:23:59 +08:00
; make sure that we still don't support broadcast double into 128-bit vector
; this used to crash
define <2 x double> @I(double* %ptr) nounwind uwtable readnone ssp {
entry:
%q = load double* %ptr, align 4
%vecinit.i = insertelement <2 x double> undef, double %q, i32 0
%vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1
ret <2 x double> %vecinit2.i
}
AVX2: Build splat vectors by broadcasting a scalar from the constant pool. Previously we used three instructions to broadcast an immediate value into a vector register. On Sandybridge we continue to load the broadcasted value from the constant pool. llvm-svn: 154284
2012-04-08 20:54:54 +08:00
; CHECK: V111
; CHECK: vpbroadcastd
; CHECK: ret
define <8 x i32> @V111(<8 x i32> %in) nounwind uwtable readnone ssp {
entry:
%g = add <8 x i32> %in, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
ret <8 x i32> %g
}
AVX: We lower VECTOR_SHUFFLE and BUILD_VECTOR nodes into vbroadcast instructions using the pattern (vbroadcast (i32load src)). In some cases, after we generate this pattern new users are added to the load node, which prevent the selection of the blend pattern. This commit provides fallback patterns which perform in-vector broadcast (using in-vector vbroadcast in AVX2 and pshufd on AVX1). llvm-svn: 155437
2012-04-24 19:07:03 +08:00
; CHECK: V113
; CHECK: vbroadcastss
; CHECK: ret
define <8 x float> @V113(<8 x float> %in) nounwind uwtable readnone ssp {
entry:
%g = fadd <8 x float> %in, <float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000>
ret <8 x float> %g
}
AVX2: Build splat vectors by broadcasting a scalar from the constant pool. Previously we used three instructions to broadcast an immediate value into a vector register. On Sandybridge we continue to load the broadcasted value from the constant pool. llvm-svn: 154284
2012-04-08 20:54:54 +08:00
; CHECK: _e2
; CHECK: vbroadcastss
; CHECK: ret
define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
%vecinit.i = insertelement <4 x float> undef, float 0xbf80000000000000, i32 0
%vecinit2.i = insertelement <4 x float> %vecinit.i, float 0xbf80000000000000, i32 1
%vecinit4.i = insertelement <4 x float> %vecinit2.i, float 0xbf80000000000000, i32 2
%vecinit6.i = insertelement <4 x float> %vecinit4.i, float 0xbf80000000000000, i32 3
ret <4 x float> %vecinit6.i
}
Fix a bug in the lowering of broadcasts: ConstantPools need to use the target pointer type. Move NormalizeVectorShuffle and LowerVectorBroadcast into X86TargetLowering. llvm-svn: 154310
2012-04-09 15:45:58 +08:00
; CHECK: _e4
; CHECK-NOT: broadcast
; CHECK: ret
define <8 x i8> @_e4(i8* %ptr) nounwind uwtable readnone ssp {
%vecinit0.i = insertelement <8 x i8> undef, i8 52, i32 0
%vecinit1.i = insertelement <8 x i8> %vecinit0.i, i8 52, i32 1
%vecinit2.i = insertelement <8 x i8> %vecinit1.i, i8 52, i32 2
%vecinit3.i = insertelement <8 x i8> %vecinit2.i, i8 52, i32 3
AVX: We lower VECTOR_SHUFFLE and BUILD_VECTOR nodes into vbroadcast instructions using the pattern (vbroadcast (i32load src)). In some cases, after we generate this pattern new users are added to the load node, which prevent the selection of the blend pattern. This commit provides fallback patterns which perform in-vector broadcast (using in-vector vbroadcast in AVX2 and pshufd on AVX1). llvm-svn: 155437
2012-04-24 19:07:03 +08:00
%vecinit4.i = insertelement <8 x i8> %vecinit3.i, i8 52, i32 4
%vecinit5.i = insertelement <8 x i8> %vecinit4.i, i8 52, i32 5
%vecinit6.i = insertelement <8 x i8> %vecinit5.i, i8 52, i32 6
%vecinit7.i = insertelement <8 x i8> %vecinit6.i, i8 52, i32 7
Fix a bug in the lowering of broadcasts: ConstantPools need to use the target pointer type. Move NormalizeVectorShuffle and LowerVectorBroadcast into X86TargetLowering. llvm-svn: 154310
2012-04-09 15:45:58 +08:00
ret <8 x i8> %vecinit7.i
}
AVX: We lower VECTOR_SHUFFLE and BUILD_VECTOR nodes into vbroadcast instructions using the pattern (vbroadcast (i32load src)). In some cases, after we generate this pattern new users are added to the load node, which prevent the selection of the blend pattern. This commit provides fallback patterns which perform in-vector broadcast (using in-vector vbroadcast in AVX2 and pshufd on AVX1). llvm-svn: 155437
2012-04-24 19:07:03 +08:00
define void @crash() nounwind alwaysinline {
WGLoopsEntry:
br i1 undef, label %ret, label %footer329VF
footer329VF:
%A.0.inVF = fmul float undef, 6.553600e+04
%B.0.in407VF = fmul <8 x float> undef, <float 6.553600e+04, float 6.553600e+04, float 6.553600e+04, float 6.553600e+04, float 6.553600e+04, float 6.553600e+04, float 6.553600e+04, float 6.553600e+04>
%A.0VF = fptosi float %A.0.inVF to i32
%B.0408VF = fptosi <8 x float> %B.0.in407VF to <8 x i32>
%0 = and <8 x i32> %B.0408VF, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
%1 = and i32 %A.0VF, 65535
%temp1098VF = insertelement <8 x i32> undef, i32 %1, i32 0
%vector1099VF = shufflevector <8 x i32> %temp1098VF, <8 x i32> undef, <8 x i32> zeroinitializer
br i1 undef, label %preload1201VF, label %footer349VF
preload1201VF:
br label %footer349VF
footer349VF:
%2 = mul nsw <8 x i32> undef, %0
%3 = mul nsw <8 x i32> undef, %vector1099VF
br label %footer329VF
ret:
ret void
}
Add support for additional in-reg vbroadcast patterns llvm-svn: 157127
2012-05-20 03:57:37 +08:00
; CHECK: _inreg0
; CHECK: broadcastss
; CHECK: ret
define <8 x i32> @_inreg0(i32 %scalar) nounwind uwtable readnone ssp {
%in = insertelement <8 x i32> undef, i32 %scalar, i32 0
%wide = shufflevector <8 x i32> %in, <8 x i32> undef, <8 x i32> zeroinitializer
ret <8 x i32> %wide
}
; CHECK: _inreg1
; CHECK: broadcastss
; CHECK: ret
define <8 x float> @_inreg1(float %scalar) nounwind uwtable readnone ssp {
%in = insertelement <8 x float> undef, float %scalar, i32 0
%wide = shufflevector <8 x float> %in, <8 x float> undef, <8 x i32> zeroinitializer
ret <8 x float> %wide
}
; CHECK: _inreg2
; CHECK: broadcastss
; CHECK: ret
define <4 x float> @_inreg2(float %scalar) nounwind uwtable readnone ssp {
%in = insertelement <4 x float> undef, float %scalar, i32 0
%wide = shufflevector <4 x float> %in, <4 x float> undef, <4 x i32> zeroinitializer
ret <4 x float> %wide
}
; CHECK: _inreg3
; CHECK: broadcastsd
; CHECK: ret
define <4 x double> @_inreg3(double %scalar) nounwind uwtable readnone ssp {
%in = insertelement <4 x double> undef, double %scalar, i32 0
%wide = shufflevector <4 x double> %in, <4 x double> undef, <4 x i32> zeroinitializer
ret <4 x double> %wide
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg8xfloat:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vbroadcastss
;CHECK: ret
define <8 x float> @_inreg8xfloat(<8 x float> %a) {
%b = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> zeroinitializer
ret <8 x float> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg4xfloat:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vbroadcastss
;CHECK: ret
define <4 x float> @_inreg4xfloat(<4 x float> %a) {
%b = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> zeroinitializer
ret <4 x float> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg16xi16:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vpbroadcastw
;CHECK: ret
define <16 x i16> @_inreg16xi16(<16 x i16> %a) {
%b = shufflevector <16 x i16> %a, <16 x i16> undef, <16 x i32> zeroinitializer
ret <16 x i16> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg8xi16:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vpbroadcastw
;CHECK: ret
define <8 x i16> @_inreg8xi16(<8 x i16> %a) {
%b = shufflevector <8 x i16> %a, <8 x i16> undef, <8 x i32> zeroinitializer
ret <8 x i16> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg4xi64:
[X86] Add broadcast instructions to the table used by ExeDepsFix pass. Adds the different broadcast instructions to the ReplaceableInstrsAVX2 table. That way the ExeDepsFix pass can take better decisions when AVX2 broadcasts are across domain (int <-> float). In particular, prior to this patch we were generating: vpbroadcastd LCPI1_0(%rip), %ymm2 vpand %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 ## <- domain change penalty Now, we generate the following nice sequence where everything is in the float domain: vbroadcastss LCPI1_0(%rip), %ymm2 vandps %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 <rdar://problem/16354675> llvm-svn: 204770
2014-03-26 08:10:22 +08:00
;CHECK: vbroadcastsd
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: ret
define <4 x i64> @_inreg4xi64(<4 x i64> %a) {
%b = shufflevector <4 x i64> %a, <4 x i64> undef, <4 x i32> zeroinitializer
ret <4 x i64> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg2xi64:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vpbroadcastq
;CHECK: ret
define <2 x i64> @_inreg2xi64(<2 x i64> %a) {
%b = shufflevector <2 x i64> %a, <2 x i64> undef, <2 x i32> zeroinitializer
ret <2 x i64> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg4xdouble:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vbroadcastsd
;CHECK: ret
define <4 x double> @_inreg4xdouble(<4 x double> %a) {
%b = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> zeroinitializer
ret <4 x double> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg2xdouble:
[x86] Enable the new vector shuffle lowering by default. Update the entire regression test suite for the new shuffles. Remove most of the old testing which was devoted to the old shuffle lowering path and is no longer relevant really. Also remove a few other random tests that only really exercised shuffles and only incidently or without any interesting aspects to them. Benchmarking that I have done shows a few small regressions with this on LNT, zero measurable regressions on real, large applications, and for several benchmarks where the loop vectorizer fires in the hot path it shows 5% to 40% improvements for SSE2 and SSE3 code running on Sandy Bridge machines. Running on AMD machines shows even more dramatic improvements. When using newer ISA vector extensions the gains are much more modest, but the code is still better on the whole. There are a few regressions being tracked (PR21137, PR21138, PR21139) but by and large this is expected to be a win for x86 generated code performance. It is also more correct than the code it replaces. I have fuzz tested this extensively with ISA extensions up through AVX2 and found no crashes or miscompiles (yet...). The old lowering had a few miscompiles and crashers after a somewhat smaller amount of fuzz testing. There is one significant area where the new code path lags behind and that is in AVX-512 support. However, there was *extremely little* support for that already and so this isn't a significant step backwards and the new framework will probably make it easier to implement lowering that uses the full power of AVX-512's table-based shuffle+blend (IMO). Many thanks to Quentin, Andrea, Robert, and others for benchmarking assistance. Thanks to Adam and others for help with AVX-512. Thanks to Hal, Eric, and *many* others for answering my incessant questions about how the backend actually works. =] I will leave the old code path in the tree until the 3 PRs above are at least resolved to folks' satisfaction. Then I will rip it (and 1000s of lines of code) out. =] I don't expect this flag to stay around for very long. It may not survive next week. llvm-svn: 219046
2014-10-04 11:52:55 +08:00
;CHECK: vunpcklpd
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: ret
define <2 x double> @_inreg2xdouble(<2 x double> %a) {
%b = shufflevector <2 x double> %a, <2 x double> undef, <2 x i32> zeroinitializer
ret <2 x double> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg8xi32:
[X86] Add broadcast instructions to the table used by ExeDepsFix pass. Adds the different broadcast instructions to the ReplaceableInstrsAVX2 table. That way the ExeDepsFix pass can take better decisions when AVX2 broadcasts are across domain (int <-> float). In particular, prior to this patch we were generating: vpbroadcastd LCPI1_0(%rip), %ymm2 vpand %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 ## <- domain change penalty Now, we generate the following nice sequence where everything is in the float domain: vbroadcastss LCPI1_0(%rip), %ymm2 vandps %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 <rdar://problem/16354675> llvm-svn: 204770
2014-03-26 08:10:22 +08:00
;CHECK: vbroadcastss
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: ret
define <8 x i32> @_inreg8xi32(<8 x i32> %a) {
%b = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> zeroinitializer
ret <8 x i32> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg4xi32:
[X86] Add broadcast instructions to the table used by ExeDepsFix pass. Adds the different broadcast instructions to the ReplaceableInstrsAVX2 table. That way the ExeDepsFix pass can take better decisions when AVX2 broadcasts are across domain (int <-> float). In particular, prior to this patch we were generating: vpbroadcastd LCPI1_0(%rip), %ymm2 vpand %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 ## <- domain change penalty Now, we generate the following nice sequence where everything is in the float domain: vbroadcastss LCPI1_0(%rip), %ymm2 vandps %ymm2, %ymm0, %ymm0 vmaxps %ymm1, %ymm0, %ymm0 <rdar://problem/16354675> llvm-svn: 204770
2014-03-26 08:10:22 +08:00
;CHECK: vbroadcastss
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: ret
define <4 x i32> @_inreg4xi32(<4 x i32> %a) {
%b = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> zeroinitializer
ret <4 x i32> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg32xi8:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vpbroadcastb
;CHECK: ret
define <32 x i8> @_inreg32xi8(<32 x i8> %a) {
%b = shufflevector <32 x i8> %a, <32 x i8> undef, <32 x i32> zeroinitializer
ret <32 x i8> %b
}
Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change. All changes were made by the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" grep -q "^; *RUN: *llc.*debug" $NAME && continue grep -q "^; *RUN:.*llvm-objdump" $NAME && continue grep -q "^; *RUN: *opt.*" $NAME && continue TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME done This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621. llvm-svn: 186624
2013-07-19 06:47:09 +08:00
;CHECK-LABEL: _inreg16xi8:
Optimization of shuffle node that can fit to the register form of VBROADCAST instruction on AVX2. llvm-svn: 159504
2012-07-01 14:12:26 +08:00
;CHECK: vpbroadcastb
;CHECK: ret
define <16 x i8> @_inreg16xi8(<16 x i8> %a) {
%b = shufflevector <16 x i8> %a, <16 x i8> undef, <16 x i32> zeroinitializer
ret <16 x i8> %b
}
Teach the DAGCombiner how to fold concat_vector nodes when the input is two BUILD_VECTOR nodes, e.g.: (concat_vectors (BUILD_VECTOR a1, a2, a3, a4), (BUILD_VECTOR b1, b2, b3, b4)) -> (BUILD_VECTOR a1, a2, a3, a4, b1, b2, b3, b4) This fixes an issue with AVX, where a sequence was not recognized as a 256-bit vbroadcast due to the concat_vectors. llvm-svn: 201158
2014-02-11 23:42:46 +08:00
; These tests check that a vbroadcast instruction is used when we have a splat
; formed from a concat_vectors (via the shufflevector) of two BUILD_VECTORs
; (via the insertelements).
; CHECK-LABEL: splat_concat1
; CHECK-NOT: vinsertf128
; CHECK: vbroadcastss
; CHECK-NEXT: ret
define <8 x float> @splat_concat1(float %f) {
%1 = insertelement <4 x float> undef, float %f, i32 0
%2 = insertelement <4 x float> %1, float %f, i32 1
%3 = insertelement <4 x float> %2, float %f, i32 2
%4 = insertelement <4 x float> %3, float %f, i32 3
%5 = shufflevector <4 x float> %4, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
ret <8 x float> %5
}
; CHECK-LABEL: splat_concat2
; CHECK-NOT: vinsertf128
; CHECK: vbroadcastss
; CHECK-NEXT: ret
define <8 x float> @splat_concat2(float %f) {
%1 = insertelement <4 x float> undef, float %f, i32 0
%2 = insertelement <4 x float> %1, float %f, i32 1
%3 = insertelement <4 x float> %2, float %f, i32 2
%4 = insertelement <4 x float> %3, float %f, i32 3
%5 = insertelement <4 x float> undef, float %f, i32 0
%6 = insertelement <4 x float> %5, float %f, i32 1
%7 = insertelement <4 x float> %6, float %f, i32 2
%8 = insertelement <4 x float> %7, float %f, i32 3
%9 = shufflevector <4 x float> %4, <4 x float> %8, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
ret <8 x float> %9
}
; CHECK-LABEL: splat_concat3
; CHECK-NOT: vinsertf128
; CHECK: vbroadcastsd
; CHECK-NEXT: ret
define <4 x double> @splat_concat3(double %d) {
%1 = insertelement <2 x double> undef, double %d, i32 0
%2 = insertelement <2 x double> %1, double %d, i32 1
%3 = shufflevector <2 x double> %2, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
ret <4 x double> %3
}
; CHECK-LABEL: splat_concat4
; CHECK-NOT: vinsertf128
; CHECK: vbroadcastsd
; CHECK-NEXT: ret
define <4 x double> @splat_concat4(double %d) {
%1 = insertelement <2 x double> undef, double %d, i32 0
%2 = insertelement <2 x double> %1, double %d, i32 1
%3 = insertelement <2 x double> undef, double %d, i32 0
%4 = insertelement <2 x double> %3, double %d, i32 1
%5 = shufflevector <2 x double> %2, <2 x double> %4, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
ret <4 x double> %5
}
[X86][ISelDAG] Add missing fallback patterns for avx2 broadcast instructions. Those patterns are used when the load cannot be folded into the related broadcast during the select phase. This happens when the load gets additional uses that were not anticipated during the previous lowering phases (constant vector to constant load, then constant load reused) or when selection DAG is not able to prove that folding the load will not create a cycle in the DAG. <rdar://problem/16074331> llvm-svn: 204631
2014-03-25 01:54:19 +08:00
; Test cases for <rdar://problem/16074331>.
; Instruction selection for broacast instruction fails if
; the load cannot be folded into the broadcast.
; This happens if the load has initial one use but other uses are
; created later, or if selection DAG cannot prove that folding the
; load will not create a cycle in the DAG.
; Those test cases exerce the latter.
; CHECK-LABEL: isel_crash_16b
; CHECK: vpbroadcastb {{[^,]+}}, %xmm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_16b(i8* %cV_R.addr) {
eintry:
%__a.addr.i = alloca <2 x i64>, align 16
%__b.addr.i = alloca <2 x i64>, align 16
%vCr = alloca <2 x i64>, align 16
store <2 x i64> zeroinitializer, <2 x i64>* %vCr, align 16
%tmp = load <2 x i64>* %vCr, align 16
%tmp2 = load i8* %cV_R.addr, align 4
%splat.splatinsert = insertelement <16 x i8> undef, i8 %tmp2, i32 0
%splat.splat = shufflevector <16 x i8> %splat.splatinsert, <16 x i8> undef, <16 x i32> zeroinitializer
%tmp3 = bitcast <16 x i8> %splat.splat to <2 x i64>
store <2 x i64> %tmp, <2 x i64>* %__a.addr.i, align 16
store <2 x i64> %tmp3, <2 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_32b
; CHECK: vpbroadcastb {{[^,]+}}, %ymm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_32b(i8* %cV_R.addr) {
eintry:
%__a.addr.i = alloca <4 x i64>, align 16
%__b.addr.i = alloca <4 x i64>, align 16
%vCr = alloca <4 x i64>, align 16
store <4 x i64> zeroinitializer, <4 x i64>* %vCr, align 16
%tmp = load <4 x i64>* %vCr, align 16
%tmp2 = load i8* %cV_R.addr, align 4
%splat.splatinsert = insertelement <32 x i8> undef, i8 %tmp2, i32 0
%splat.splat = shufflevector <32 x i8> %splat.splatinsert, <32 x i8> undef, <32 x i32> zeroinitializer
%tmp3 = bitcast <32 x i8> %splat.splat to <4 x i64>
store <4 x i64> %tmp, <4 x i64>* %__a.addr.i, align 16
store <4 x i64> %tmp3, <4 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_8w
; CHECK: vpbroadcastw {{[^,]+}}, %xmm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_8w(i16* %cV_R.addr) {
entry:
%__a.addr.i = alloca <2 x i64>, align 16
%__b.addr.i = alloca <2 x i64>, align 16
%vCr = alloca <2 x i64>, align 16
store <2 x i64> zeroinitializer, <2 x i64>* %vCr, align 16
%tmp = load <2 x i64>* %vCr, align 16
%tmp2 = load i16* %cV_R.addr, align 4
%splat.splatinsert = insertelement <8 x i16> undef, i16 %tmp2, i32 0
%splat.splat = shufflevector <8 x i16> %splat.splatinsert, <8 x i16> undef, <8 x i32> zeroinitializer
%tmp3 = bitcast <8 x i16> %splat.splat to <2 x i64>
store <2 x i64> %tmp, <2 x i64>* %__a.addr.i, align 16
store <2 x i64> %tmp3, <2 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_16w
; CHECK: vpbroadcastw {{[^,]+}}, %ymm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_16w(i16* %cV_R.addr) {
eintry:
%__a.addr.i = alloca <4 x i64>, align 16
%__b.addr.i = alloca <4 x i64>, align 16
%vCr = alloca <4 x i64>, align 16
store <4 x i64> zeroinitializer, <4 x i64>* %vCr, align 16
%tmp = load <4 x i64>* %vCr, align 16
%tmp2 = load i16* %cV_R.addr, align 4
%splat.splatinsert = insertelement <16 x i16> undef, i16 %tmp2, i32 0
%splat.splat = shufflevector <16 x i16> %splat.splatinsert, <16 x i16> undef, <16 x i32> zeroinitializer
%tmp3 = bitcast <16 x i16> %splat.splat to <4 x i64>
store <4 x i64> %tmp, <4 x i64>* %__a.addr.i, align 16
store <4 x i64> %tmp3, <4 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_4d
; CHECK: vbroadcastss {{[^,]+}}, %xmm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_4d(i32* %cV_R.addr) {
entry:
%__a.addr.i = alloca <2 x i64>, align 16
%__b.addr.i = alloca <2 x i64>, align 16
%vCr = alloca <2 x i64>, align 16
store <2 x i64> zeroinitializer, <2 x i64>* %vCr, align 16
%tmp = load <2 x i64>* %vCr, align 16
%tmp2 = load i32* %cV_R.addr, align 4
%splat.splatinsert = insertelement <4 x i32> undef, i32 %tmp2, i32 0
%splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
%tmp3 = bitcast <4 x i32> %splat.splat to <2 x i64>
store <2 x i64> %tmp, <2 x i64>* %__a.addr.i, align 16
store <2 x i64> %tmp3, <2 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_8d
; CHECK: vbroadcastss {{[^,]+}}, %ymm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_8d(i32* %cV_R.addr) {
eintry:
%__a.addr.i = alloca <4 x i64>, align 16
%__b.addr.i = alloca <4 x i64>, align 16
%vCr = alloca <4 x i64>, align 16
store <4 x i64> zeroinitializer, <4 x i64>* %vCr, align 16
%tmp = load <4 x i64>* %vCr, align 16
%tmp2 = load i32* %cV_R.addr, align 4
%splat.splatinsert = insertelement <8 x i32> undef, i32 %tmp2, i32 0
%splat.splat = shufflevector <8 x i32> %splat.splatinsert, <8 x i32> undef, <8 x i32> zeroinitializer
%tmp3 = bitcast <8 x i32> %splat.splat to <4 x i64>
store <4 x i64> %tmp, <4 x i64>* %__a.addr.i, align 16
store <4 x i64> %tmp3, <4 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_2q
; CHECK: vpbroadcastq {{[^,]+}}, %xmm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_2q(i64* %cV_R.addr) {
entry:
%__a.addr.i = alloca <2 x i64>, align 16
%__b.addr.i = alloca <2 x i64>, align 16
%vCr = alloca <2 x i64>, align 16
store <2 x i64> zeroinitializer, <2 x i64>* %vCr, align 16
%tmp = load <2 x i64>* %vCr, align 16
%tmp2 = load i64* %cV_R.addr, align 4
%splat.splatinsert = insertelement <2 x i64> undef, i64 %tmp2, i32 0
%splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
store <2 x i64> %tmp, <2 x i64>* %__a.addr.i, align 16
store <2 x i64> %splat.splat, <2 x i64>* %__b.addr.i, align 16
ret void
}
; CHECK-LABEL: isel_crash_4q
; CHECK: vbroadcastsd {{[^,]+}}, %ymm{{[0-9]+}}
; CHECK: ret
define void @isel_crash_4q(i64* %cV_R.addr) {
eintry:
%__a.addr.i = alloca <4 x i64>, align 16
%__b.addr.i = alloca <4 x i64>, align 16
%vCr = alloca <4 x i64>, align 16
store <4 x i64> zeroinitializer, <4 x i64>* %vCr, align 16
%tmp = load <4 x i64>* %vCr, align 16
%tmp2 = load i64* %cV_R.addr, align 4
%splat.splatinsert = insertelement <4 x i64> undef, i64 %tmp2, i32 0
%splat.splat = shufflevector <4 x i64> %splat.splatinsert, <4 x i64> undef, <4 x i32> zeroinitializer
store <4 x i64> %tmp, <4 x i64>* %__a.addr.i, align 16
store <4 x i64> %splat.splat, <4 x i64>* %__b.addr.i, align 16
ret void
}