[X86][SSE] SimplifyDemandedBitsForTargetNode - Add initial PACKSS support

In the case where we only want the sign bit (e.g. when using PACKSS truncation of comparison results for MOVMSK) then we can just demand the sign bit of the source operands.

This makes use of the fact that PACKSS saturates out of range values to the min/max int values - so the sign bit is always preserved.

Differential Revision: https://reviews.llvm.org/D60333

llvm-svn: 357859

6 files changed, 67 insertions, 187 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index dd694281521..217c0e6ce48 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -33508,6 +33508,25 @@ bool X86TargetLowering::SimplifyDemandedBitsForTargetNode(
     }
     break;
   }
+  case X86ISD::PACKSS:
+    // PACKSS saturates to MIN/MAX integer values. So if we just want the
+    // sign bit then we can just ask for the source operands sign bit.
+    // TODO - add known bits handling.
+    if (OriginalDemandedBits.isSignMask()) {
+      APInt DemandedLHS, DemandedRHS;
+      getPackDemandedElts(VT, OriginalDemandedElts, DemandedLHS, DemandedRHS);
+
+      KnownBits KnownLHS, KnownRHS;
+      APInt SignMask = APInt::getSignMask(BitWidth * 2);
+      if (SimplifyDemandedBits(Op.getOperand(0), SignMask, DemandedLHS,
+                               KnownLHS, TLO, Depth + 1))
+        return true;
+      if (SimplifyDemandedBits(Op.getOperand(1), SignMask, DemandedRHS,
+                               KnownRHS, TLO, Depth + 1))
+        return true;
+    }
+    // TODO - add general PACKSS/PACKUS SimplifyDemandedBits support.
+    break;
   case X86ISD::PCMPGT:
     // icmp sgt(0, R) == ashr(R, BitWidth-1).
     // iff we only need the sign bit then we can use R directly.
diff --git a/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll b/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll
index bab37c34888..2abcdee7adb 100644
--- a/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll
+++ b/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll
@@ -74,7 +74,6 @@ define i8 @v8i64(<8 x i64> %a, <8 x i64> %b, <8 x i64> %c, <8 x i64> %d) {
 ; AVX1-NEXT:    vpackssdw %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    # kill: def $al killed $al killed $eax
@@ -99,7 +98,6 @@ define i8 @v8i64(<8 x i64> %a, <8 x i64> %b, <8 x i64> %c, <8 x i64> %d) {
 ; AVX2-NEXT:    vpackssdw %xmm1, %xmm2, %xmm1
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX2-NEXT:    vpsraw $15, %xmm0, %xmm0
 ; AVX2-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
@@ -191,7 +189,6 @@ define i8 @v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x double>
 ; AVX12-NEXT:    vpackssdw %xmm1, %xmm2, %xmm1
 ; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX12-NEXT:    vpsraw $15, %xmm0, %xmm0
 ; AVX12-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX12-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX12-NEXT:    # kill: def $al killed $al killed $eax
diff --git a/llvm/test/CodeGen/X86/bitcast-setcc-128.ll b/llvm/test/CodeGen/X86/bitcast-setcc-128.ll
index e9dea85e14d..0a44e791b92 100644
--- a/llvm/test/CodeGen/X86/bitcast-setcc-128.ll
+++ b/llvm/test/CodeGen/X86/bitcast-setcc-128.ll
@@ -765,17 +765,13 @@ define void @bitcast_16i8_store(i16* %p, <16 x i8> %a0) {
 define void @bitcast_8i16_store(i8* %p, <8 x i16> %a0) {
 ; SSE2-SSSE3-LABEL: bitcast_8i16_store:
 ; SSE2-SSSE3:       # %bb.0:
-; SSE2-SSSE3-NEXT:    pxor %xmm1, %xmm1
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    pmovmskb %xmm1, %eax
+; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-SSSE3-NEXT:    movb %al, (%rdi)
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: bitcast_8i16_store:
 ; AVX12:       # %bb.0:
-; AVX12-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX12-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
 ; AVX12-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX12-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX12-NEXT:    movb %al, (%rdi)
diff --git a/llvm/test/CodeGen/X86/bitcast-setcc-256.ll b/llvm/test/CodeGen/X86/bitcast-setcc-256.ll
index 7489c785f14..41635f37528 100644
--- a/llvm/test/CodeGen/X86/bitcast-setcc-256.ll
+++ b/llvm/test/CodeGen/X86/bitcast-setcc-256.ll
@@ -359,21 +359,14 @@ define void @bitcast_32i8_store(i32* %p, <32 x i8> %a0) {
 define void @bitcast_16i16_store(i16* %p, <16 x i16> %a0) {
 ; SSE2-SSSE3-LABEL: bitcast_16i16_store:
 ; SSE2-SSSE3:       # %bb.0:
-; SSE2-SSSE3-NEXT:    pxor %xmm2, %xmm2
-; SSE2-SSSE3-NEXT:    pxor %xmm3, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    packsswb %xmm3, %xmm2
-; SSE2-SSSE3-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-SSSE3-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-SSSE3-NEXT:    movw %ax, (%rdi)
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: bitcast_16i16_store:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    movw %ax, (%rdi)
@@ -416,13 +409,9 @@ define void @bitcast_16i16_store(i16* %p, <16 x i16> %a0) {
 define void @bitcast_8i32_store(i8* %p, <8 x i32> %a0) {
 ; SSE2-SSSE3-LABEL: bitcast_8i32_store:
 ; SSE2-SSSE3:       # %bb.0:
-; SSE2-SSSE3-NEXT:    pxor %xmm2, %xmm2
-; SSE2-SSSE3-NEXT:    pxor %xmm3, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    packssdw %xmm3, %xmm2
-; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-SSSE3-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-SSSE3-NEXT:    movb %al, (%rdi)
 ; SSE2-SSSE3-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/bitcast-setcc-512.ll b/llvm/test/CodeGen/X86/bitcast-setcc-512.ll
index b6a38aa132b..3c294345dd5 100644
--- a/llvm/test/CodeGen/X86/bitcast-setcc-512.ll
+++ b/llvm/test/CodeGen/X86/bitcast-setcc-512.ll
@@ -481,18 +481,10 @@ define void @bitcast_64i8_store(i64* %p, <64 x i8> %a0) {
 define void @bitcast_32i16_store(i32* %p, <32 x i16> %a0) {
 ; SSE-LABEL: bitcast_32i16_store:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    pxor %xmm5, %xmm5
-; SSE-NEXT:    pcmpgtw %xmm1, %xmm5
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE-NEXT:    packsswb %xmm5, %xmm1
-; SSE-NEXT:    pmovmskb %xmm1, %eax
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    pcmpgtw %xmm3, %xmm0
-; SSE-NEXT:    pcmpgtw %xmm2, %xmm4
-; SSE-NEXT:    packsswb %xmm0, %xmm4
-; SSE-NEXT:    pmovmskb %xmm4, %ecx
+; SSE-NEXT:    packsswb %xmm1, %xmm0
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    packsswb %xmm3, %xmm2
+; SSE-NEXT:    pmovmskb %xmm2, %ecx
 ; SSE-NEXT:    shll $16, %ecx
 ; SSE-NEXT:    orl %eax, %ecx
 ; SSE-NEXT:    movl %ecx, (%rdi)
@@ -501,14 +493,9 @@ define void @bitcast_32i16_store(i32* %p, <32 x i16> %a0) {
 ; AVX1-LABEL: bitcast_32i16_store:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpacksswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
 ; AVX1-NEXT:    shll $16, %ecx
@@ -519,9 +506,6 @@ define void @bitcast_32i16_store(i32* %p, <32 x i16> %a0) {
 ;
 ; AVX2-LABEL: bitcast_32i16_store:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX2-NEXT:    vpmovmskb %ymm0, %eax
@@ -558,31 +542,18 @@ define void @bitcast_32i16_store(i32* %p, <32 x i16> %a0) {
 define void @bitcast_16i32_store(i16* %p, <16 x i32> %a0) {
 ; SSE-LABEL: bitcast_16i32_store:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    pxor %xmm5, %xmm5
-; SSE-NEXT:    pcmpgtd %xmm3, %xmm5
-; SSE-NEXT:    pxor %xmm3, %xmm3
-; SSE-NEXT:    pcmpgtd %xmm2, %xmm3
-; SSE-NEXT:    packssdw %xmm5, %xmm3
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    pcmpgtd %xmm1, %xmm2
-; SSE-NEXT:    pcmpgtd %xmm0, %xmm4
-; SSE-NEXT:    packssdw %xmm2, %xmm4
-; SSE-NEXT:    packsswb %xmm3, %xmm4
-; SSE-NEXT:    pmovmskb %xmm4, %eax
+; SSE-NEXT:    packssdw %xmm3, %xmm2
+; SSE-NEXT:    packssdw %xmm1, %xmm0
+; SSE-NEXT:    packsswb %xmm2, %xmm0
+; SSE-NEXT:    pmovmskb %xmm0, %eax
 ; SSE-NEXT:    movw %ax, (%rdi)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bitcast_16i32_store:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
diff --git a/llvm/test/CodeGen/X86/movmsk-cmp.ll b/llvm/test/CodeGen/X86/movmsk-cmp.ll
index 2b13c85a961..01d5252c231 100644
--- a/llvm/test/CodeGen/X86/movmsk-cmp.ll
+++ b/llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -316,18 +316,14 @@ define i1 @allzeros_v64i8_sign(<64 x i8> %arg) {
 define i1 @allones_v8i16_sign(<8 x i16> %arg) {
 ; SSE2-LABEL: allones_v8i16_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE2-NEXT:    packsswb %xmm0, %xmm1
-; SSE2-NEXT:    pmovmskb %xmm1, %eax
+; SSE2-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    cmpb $-1, %al
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
 ;
 ; AVX-LABEL: allones_v8i16_sign:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX-NEXT:    cmpb $-1, %al
@@ -361,18 +357,14 @@ define i1 @allones_v8i16_sign(<8 x i16> %arg) {
 define i1 @allzeros_v8i16_sign(<8 x i16> %arg) {
 ; SSE2-LABEL: allzeros_v8i16_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE2-NEXT:    packsswb %xmm0, %xmm1
-; SSE2-NEXT:    pmovmskb %xmm1, %eax
+; SSE2-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    testb %al, %al
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
 ;
 ; AVX-LABEL: allzeros_v8i16_sign:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX-NEXT:    testb %al, %al
@@ -406,12 +398,8 @@ define i1 @allzeros_v8i16_sign(<8 x i16> %arg) {
 define i1 @allones_v16i16_sign(<16 x i16> %arg) {
 ; SSE2-LABEL: allones_v16i16_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtw %xmm1, %xmm3
-; SSE2-NEXT:    pcmpgtw %xmm0, %xmm2
-; SSE2-NEXT:    packsswb %xmm3, %xmm2
-; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    cmpw $-1, %ax
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
@@ -419,9 +407,6 @@ define i1 @allones_v16i16_sign(<16 x i16> %arg) {
 ; AVX1-LABEL: allones_v16i16_sign:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    cmpw $-1, %ax
@@ -468,12 +453,8 @@ define i1 @allones_v16i16_sign(<16 x i16> %arg) {
 define i1 @allzeros_v16i16_sign(<16 x i16> %arg) {
 ; SSE2-LABEL: allzeros_v16i16_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtw %xmm1, %xmm3
-; SSE2-NEXT:    pcmpgtw %xmm0, %xmm2
-; SSE2-NEXT:    packsswb %xmm3, %xmm2
-; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    testw %ax, %ax
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
@@ -481,9 +462,6 @@ define i1 @allzeros_v16i16_sign(<16 x i16> %arg) {
 ; AVX1-LABEL: allzeros_v16i16_sign:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    testw %ax, %ax
@@ -530,18 +508,10 @@ define i1 @allzeros_v16i16_sign(<16 x i16> %arg) {
 define i1 @allones_v32i16_sign(<32 x i16> %arg) {
 ; SSE2-LABEL: allones_v32i16_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pxor %xmm5, %xmm5
-; SSE2-NEXT:    pcmpgtw %xmm1, %xmm5
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE2-NEXT:    packsswb %xmm5, %xmm1
-; SSE2-NEXT:    pmovmskb %xmm1, %eax
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pcmpgtw %xmm3, %xmm0
-; SSE2-NEXT:    pcmpgtw %xmm2, %xmm4
-; SSE2-NEXT:    packsswb %xmm0, %xmm4
-; SSE2-NEXT:    pmovmskb %xmm4, %ecx
+; SSE2-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-NEXT:    packsswb %xmm3, %xmm2
+; SSE2-NEXT:    pmovmskb %xmm2, %ecx
 ; SSE2-NEXT:    shll $16, %ecx
 ; SSE2-NEXT:    orl %eax, %ecx
 ; SSE2-NEXT:    cmpl $-1, %ecx
@@ -551,14 +521,9 @@ define i1 @allones_v32i16_sign(<32 x i16> %arg) {
 ; AVX1-LABEL: allones_v32i16_sign:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpacksswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
 ; AVX1-NEXT:    shll $16, %ecx
@@ -570,9 +535,6 @@ define i1 @allones_v32i16_sign(<32 x i16> %arg) {
 ;
 ; AVX2-LABEL: allones_v32i16_sign:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX2-NEXT:    vpmovmskb %ymm0, %eax
@@ -615,18 +577,10 @@ define i1 @allones_v32i16_sign(<32 x i16> %arg) {
 define i1 @allzeros_v32i16_sign(<32 x i16> %arg) {
 ; SSE2-LABEL: allzeros_v32i16_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pxor %xmm5, %xmm5
-; SSE2-NEXT:    pcmpgtw %xmm1, %xmm5
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE2-NEXT:    packsswb %xmm5, %xmm1
-; SSE2-NEXT:    pmovmskb %xmm1, %eax
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pcmpgtw %xmm3, %xmm0
-; SSE2-NEXT:    pcmpgtw %xmm2, %xmm4
-; SSE2-NEXT:    packsswb %xmm0, %xmm4
-; SSE2-NEXT:    pmovmskb %xmm4, %ecx
+; SSE2-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-NEXT:    packsswb %xmm3, %xmm2
+; SSE2-NEXT:    pmovmskb %xmm2, %ecx
 ; SSE2-NEXT:    shll $16, %ecx
 ; SSE2-NEXT:    orl %eax, %ecx
 ; SSE2-NEXT:    sete %al
@@ -635,14 +589,9 @@ define i1 @allzeros_v32i16_sign(<32 x i16> %arg) {
 ; AVX1-LABEL: allzeros_v32i16_sign:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpacksswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
 ; AVX1-NEXT:    shll $16, %ecx
@@ -653,9 +602,6 @@ define i1 @allzeros_v32i16_sign(<32 x i16> %arg) {
 ;
 ; AVX2-LABEL: allzeros_v32i16_sign:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX2-NEXT:    vpmovmskb %ymm0, %eax
@@ -777,13 +723,9 @@ define i1 @allzeros_v4i32_sign(<4 x i32> %arg) {
 define i1 @allones_v8i32_sign(<8 x i32> %arg) {
 ; SSE2-LABEL: allones_v8i32_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    packssdw %xmm3, %xmm2
-; SSE2-NEXT:    packsswb %xmm0, %xmm2
-; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    cmpb $-1, %al
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
@@ -823,13 +765,9 @@ define i1 @allones_v8i32_sign(<8 x i32> %arg) {
 define i1 @allzeros_v8i32_sign(<8 x i32> %arg) {
 ; SSE2-LABEL: allzeros_v8i32_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    packssdw %xmm3, %xmm2
-; SSE2-NEXT:    packsswb %xmm0, %xmm2
-; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    testb %al, %al
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
@@ -869,18 +807,10 @@ define i1 @allzeros_v8i32_sign(<8 x i32> %arg) {
 define i1 @allones_v16i32_sign(<16 x i32> %arg) {
 ; SSE2-LABEL: allones_v16i32_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pxor %xmm5, %xmm5
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm5
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm2, %xmm3
-; SSE2-NEXT:    packssdw %xmm5, %xmm3
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm4
-; SSE2-NEXT:    packssdw %xmm2, %xmm4
-; SSE2-NEXT:    packsswb %xmm3, %xmm4
-; SSE2-NEXT:    pmovmskb %xmm4, %eax
+; SSE2-NEXT:    packssdw %xmm3, %xmm2
+; SSE2-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-NEXT:    packsswb %xmm2, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    cmpw $-1, %ax
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
@@ -888,13 +818,8 @@ define i1 @allones_v16i32_sign(<16 x i32> %arg) {
 ; AVX1-LABEL: allones_v16i32_sign:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
@@ -943,18 +868,10 @@ define i1 @allones_v16i32_sign(<16 x i32> %arg) {
 define i1 @allzeros_v16i32_sign(<16 x i32> %arg) {
 ; SSE2-LABEL: allzeros_v16i32_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pxor %xmm5, %xmm5
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm5
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm2, %xmm3
-; SSE2-NEXT:    packssdw %xmm5, %xmm3
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm4
-; SSE2-NEXT:    packssdw %xmm2, %xmm4
-; SSE2-NEXT:    packsswb %xmm3, %xmm4
-; SSE2-NEXT:    pmovmskb %xmm4, %eax
+; SSE2-NEXT:    packssdw %xmm3, %xmm2
+; SSE2-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-NEXT:    packsswb %xmm2, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    testw %ax, %ax
 ; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
@@ -962,13 +879,8 @@ define i1 @allzeros_v16i32_sign(<16 x i32> %arg) {
 ; AVX1-LABEL: allzeros_v16i32_sign:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
@@ -4663,13 +4575,9 @@ define i32 @movmskpd256(<4 x double> %x) {
 define i32 @movmskps256(<8 x float> %x) {
 ; SSE2-LABEL: movmskps256:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    packssdw %xmm3, %xmm2
-; SSE2-NEXT:    packsswb %xmm0, %xmm2
-; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
 ; SSE2-NEXT:    movzbl %al, %eax
 ; SSE2-NEXT:    retq
 ;