From e1ebc3762f006981359014edcb20fe0f42058850 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Sat, 24 Feb 2018 14:06:39 +0000
Subject: [PATCH] [X86][SSE] combineSubToSubus - support v8i64 handling from
 SSSE3

Our UMIN/UMAX, vector truncation and shuffle combining is good enough to efficiently handle v8i64 with the number of leading zeros that are necessary for PSUBUS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@326034 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86ISelLowering.cpp |   2 +-
 test/CodeGen/X86/psubus.ll         | 404 +++++++++++++++----------------------
 2 files changed, 164 insertions(+), 242 deletions(-)

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 14362c6392e..2cce3669719 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -37638,7 +37638,7 @@ static SDValue combineSubToSubus(SDNode *N, SelectionDAG &DAG,
   // PSUBUS is supported, starting from SSE2, but truncation for v8i32
   // is only worth it with SSSE3 (PSHUFB).
   if (!(Subtarget.hasSSE2() && (VT == MVT::v16i8 || VT == MVT::v8i16)) &&
-      !(Subtarget.hasSSSE3() && (VT == MVT::v8i32)) &&
+      !(Subtarget.hasSSSE3() && (VT == MVT::v8i32 || VT == MVT::v8i64)) &&
       !(Subtarget.hasAVX() && (VT == MVT::v32i8 || VT == MVT::v16i16)) &&
       !(Subtarget.useBWIRegs() && (VT == MVT::v64i8 || VT == MVT::v32i16 ||
                                    VT == MVT::v16i32 || VT == MVT::v8i64)))
diff --git a/test/CodeGen/X86/psubus.ll b/test/CodeGen/X86/psubus.ll
index 6e7aa0d8b7d..7b375b494a6 100644
--- a/test/CodeGen/X86/psubus.ll
+++ b/test/CodeGen/X86/psubus.ll
@@ -1496,272 +1496,194 @@ define <8 x i16> @psubus_8i64_max(<8 x i16> %x, <8 x i64> %y) nounwind {
 ;
 ; SSSE3-LABEL: psubus_8i64_max:
 ; SSSE3:       # %bb.0: # %vector.ph
-; SSSE3-NEXT:    pxor %xmm5, %xmm5
-; SSSE3-NEXT:    movdqa %xmm0, %xmm10
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
-; SSSE3-NEXT:    movdqa %xmm10, %xmm8
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
-; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm5[2],xmm10[3],xmm5[3]
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm9
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
-; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm11 = [2147483648,2147483648,2147483648,2147483648]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm6
-; SSSE3-NEXT:    pxor %xmm11, %xmm6
-; SSSE3-NEXT:    movdqa %xmm0, %xmm7
-; SSSE3-NEXT:    por %xmm11, %xmm7
-; SSSE3-NEXT:    movdqa %xmm7, %xmm5
-; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm5
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm12 = xmm5[0,0,2,2]
-; SSSE3-NEXT:    pcmpeqd %xmm6, %xmm7
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
-; SSSE3-NEXT:    pand %xmm12, %xmm7
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm13 = xmm5[1,1,3,3]
-; SSSE3-NEXT:    por %xmm7, %xmm13
-; SSSE3-NEXT:    pand %xmm13, %xmm0
-; SSSE3-NEXT:    pandn %xmm2, %xmm13
-; SSSE3-NEXT:    por %xmm0, %xmm13
-; SSSE3-NEXT:    movdqa %xmm1, %xmm0
-; SSSE3-NEXT:    pxor %xmm11, %xmm0
-; SSSE3-NEXT:    movdqa %xmm9, %xmm5
-; SSSE3-NEXT:    por %xmm11, %xmm5
-; SSSE3-NEXT:    movdqa %xmm5, %xmm7
-; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm7
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2]
-; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm5
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSSE3-NEXT:    pand %xmm12, %xmm5
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
-; SSSE3-NEXT:    por %xmm5, %xmm0
-; SSSE3-NEXT:    pand %xmm0, %xmm9
-; SSSE3-NEXT:    pandn %xmm1, %xmm0
-; SSSE3-NEXT:    por %xmm9, %xmm0
-; SSSE3-NEXT:    movdqa %xmm4, %xmm5
-; SSSE3-NEXT:    pxor %xmm11, %xmm5
-; SSSE3-NEXT:    movdqa %xmm10, %xmm7
-; SSSE3-NEXT:    por %xmm11, %xmm7
-; SSSE3-NEXT:    movdqa %xmm7, %xmm6
-; SSSE3-NEXT:    pcmpgtd %xmm5, %xmm6
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2]
-; SSSE3-NEXT:    pcmpeqd %xmm5, %xmm7
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
-; SSSE3-NEXT:    pand %xmm9, %xmm5
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[1,1,3,3]
-; SSSE3-NEXT:    por %xmm5, %xmm7
-; SSSE3-NEXT:    pand %xmm7, %xmm10
-; SSSE3-NEXT:    pandn %xmm4, %xmm7
-; SSSE3-NEXT:    por %xmm10, %xmm7
-; SSSE3-NEXT:    movdqa %xmm3, %xmm5
-; SSSE3-NEXT:    pxor %xmm11, %xmm5
-; SSSE3-NEXT:    por %xmm8, %xmm11
-; SSSE3-NEXT:    movdqa %xmm11, %xmm6
-; SSSE3-NEXT:    pcmpgtd %xmm5, %xmm6
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648]
+; SSSE3-NEXT:    movdqa %xmm2, %xmm7
+; SSSE3-NEXT:    pxor %xmm5, %xmm7
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [9223372039002324991,9223372039002324991]
+; SSSE3-NEXT:    movdqa %xmm8, %xmm6
+; SSSE3-NEXT:    pcmpgtd %xmm7, %xmm6
 ; SSSE3-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2]
-; SSSE3-NEXT:    pcmpeqd %xmm5, %xmm11
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm11[1,1,3,3]
-; SSSE3-NEXT:    pand %xmm9, %xmm5
+; SSSE3-NEXT:    pcmpeqd %xmm8, %xmm7
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm9, %xmm7
 ; SSSE3-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
-; SSSE3-NEXT:    por %xmm5, %xmm6
-; SSSE3-NEXT:    pand %xmm6, %xmm8
-; SSSE3-NEXT:    pandn %xmm3, %xmm6
-; SSSE3-NEXT:    por %xmm8, %xmm6
-; SSSE3-NEXT:    psubq %xmm3, %xmm6
-; SSSE3-NEXT:    psubq %xmm4, %xmm7
-; SSSE3-NEXT:    psubq %xmm1, %xmm0
-; SSSE3-NEXT:    psubq %xmm2, %xmm13
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,2,3]
+; SSSE3-NEXT:    por %xmm7, %xmm6
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535]
+; SSSE3-NEXT:    pand %xmm6, %xmm2
+; SSSE3-NEXT:    pandn %xmm9, %xmm6
+; SSSE3-NEXT:    por %xmm2, %xmm6
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
+; SSSE3-NEXT:    pshuflw {{.*#+}} xmm10 = xmm2[0,2,2,3,4,5,6,7]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm6
+; SSSE3-NEXT:    pxor %xmm5, %xmm6
+; SSSE3-NEXT:    movdqa %xmm8, %xmm7
+; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm7
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm8, %xmm6
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm2, %xmm6
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[1,1,3,3]
+; SSSE3-NEXT:    por %xmm6, %xmm2
+; SSSE3-NEXT:    pand %xmm2, %xmm1
+; SSSE3-NEXT:    pandn %xmm9, %xmm2
+; SSSE3-NEXT:    por %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
 ; SSSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7]
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,2,2,3]
-; SSSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,1,0,2,4,5,6,7]
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3]
-; SSSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSSE3-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; SSSE3-NEXT:    movdqa %xmm4, %xmm2
+; SSSE3-NEXT:    pxor %xmm5, %xmm2
+; SSSE3-NEXT:    movdqa %xmm8, %xmm6
+; SSSE3-NEXT:    pcmpgtd %xmm2, %xmm6
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm8, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm7, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
+; SSSE3-NEXT:    por %xmm2, %xmm6
+; SSSE3-NEXT:    pand %xmm6, %xmm4
+; SSSE3-NEXT:    pandn %xmm9, %xmm6
+; SSSE3-NEXT:    por %xmm4, %xmm6
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
+; SSSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
+; SSSE3-NEXT:    pxor %xmm3, %xmm5
+; SSSE3-NEXT:    movdqa %xmm8, %xmm4
+; SSSE3-NEXT:    pcmpgtd %xmm5, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm8, %xmm5
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm6, %xmm5
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSSE3-NEXT:    por %xmm5, %xmm4
+; SSSE3-NEXT:    pand %xmm4, %xmm3
+; SSSE3-NEXT:    pandn %xmm9, %xmm4
+; SSSE3-NEXT:    por %xmm3, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
+; SSSE3-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,0,2,4,5,6,7]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSSE3-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSSE3-NEXT:    psubusw %xmm3, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: psubus_8i64_max:
 ; SSE41:       # %bb.0: # %vector.ph
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,2,3]
-; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm8 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero
-; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm10 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[2,3,0,1]
-; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm11 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm12 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; SSE41-NEXT:    movdqa %xmm0, %xmm10
 ; SSE41-NEXT:    movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648]
 ; SSE41-NEXT:    movdqa %xmm4, %xmm0
 ; SSE41-NEXT:    pxor %xmm6, %xmm0
-; SSE41-NEXT:    movdqa %xmm12, %xmm7
-; SSE41-NEXT:    por %xmm6, %xmm7
-; SSE41-NEXT:    movdqa %xmm7, %xmm5
-; SSE41-NEXT:    pcmpgtd %xmm0, %xmm5
-; SSE41-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
-; SSE41-NEXT:    pand %xmm9, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,3,3]
-; SSE41-NEXT:    por %xmm7, %xmm0
-; SSE41-NEXT:    movdqa %xmm4, %xmm9
-; SSE41-NEXT:    blendvpd %xmm0, %xmm12, %xmm9
-; SSE41-NEXT:    movdqa %xmm3, %xmm0
-; SSE41-NEXT:    pxor %xmm6, %xmm0
-; SSE41-NEXT:    movdqa %xmm11, %xmm5
-; SSE41-NEXT:    por %xmm6, %xmm5
-; SSE41-NEXT:    movdqa %xmm5, %xmm7
+; SSE41-NEXT:    movdqa {{.*#+}} xmm8 = [9223372039002324991,9223372039002324991]
+; SSE41-NEXT:    movdqa %xmm8, %xmm7
 ; SSE41-NEXT:    pcmpgtd %xmm0, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2]
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm5
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSE41-NEXT:    pand %xmm12, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm8, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm9, %xmm5
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
 ; SSE41-NEXT:    por %xmm5, %xmm0
-; SSE41-NEXT:    movdqa %xmm3, %xmm12
-; SSE41-NEXT:    blendvpd %xmm0, %xmm11, %xmm12
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
+; SSE41-NEXT:    movapd {{.*#+}} xmm7 = [65535,65535]
+; SSE41-NEXT:    movapd %xmm7, %xmm11
+; SSE41-NEXT:    blendvpd %xmm0, %xmm4, %xmm11
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
 ; SSE41-NEXT:    pxor %xmm6, %xmm0
-; SSE41-NEXT:    movdqa %xmm10, %xmm5
-; SSE41-NEXT:    por %xmm6, %xmm5
-; SSE41-NEXT:    movdqa %xmm5, %xmm7
-; SSE41-NEXT:    pcmpgtd %xmm0, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm11 = xmm7[0,0,2,2]
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm5
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSE41-NEXT:    pand %xmm11, %xmm5
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
+; SSE41-NEXT:    movdqa %xmm8, %xmm4
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm4
+; SSE41-NEXT:    pshufd {{.*#+}} xmm9 = xmm4[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm8, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm9, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,3,3]
 ; SSE41-NEXT:    por %xmm5, %xmm0
-; SSE41-NEXT:    movdqa %xmm1, %xmm5
-; SSE41-NEXT:    blendvpd %xmm0, %xmm10, %xmm5
+; SSE41-NEXT:    movapd %xmm7, %xmm4
+; SSE41-NEXT:    blendvpd %xmm0, %xmm3, %xmm4
+; SSE41-NEXT:    packusdw %xmm11, %xmm4
 ; SSE41-NEXT:    movdqa %xmm2, %xmm0
 ; SSE41-NEXT:    pxor %xmm6, %xmm0
-; SSE41-NEXT:    por %xmm8, %xmm6
-; SSE41-NEXT:    movdqa %xmm6, %xmm7
-; SSE41-NEXT:    pcmpgtd %xmm0, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm10 = xmm7[0,0,2,2]
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm6
-; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
-; SSE41-NEXT:    pand %xmm10, %xmm6
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
-; SSE41-NEXT:    por %xmm6, %xmm0
-; SSE41-NEXT:    movdqa %xmm2, %xmm6
-; SSE41-NEXT:    blendvpd %xmm0, %xmm8, %xmm6
-; SSE41-NEXT:    psubq %xmm2, %xmm6
-; SSE41-NEXT:    psubq %xmm1, %xmm5
-; SSE41-NEXT:    psubq %xmm3, %xmm12
-; SSE41-NEXT:    psubq %xmm4, %xmm9
-; SSE41-NEXT:    pxor %xmm0, %xmm0
-; SSE41-NEXT:    pblendw {{.*#+}} xmm9 = xmm9[0],xmm0[1,2,3],xmm9[4],xmm0[5,6,7]
-; SSE41-NEXT:    pblendw {{.*#+}} xmm12 = xmm12[0],xmm0[1,2,3],xmm12[4],xmm0[5,6,7]
-; SSE41-NEXT:    packusdw %xmm9, %xmm12
-; SSE41-NEXT:    pblendw {{.*#+}} xmm5 = xmm5[0],xmm0[1,2,3],xmm5[4],xmm0[5,6,7]
-; SSE41-NEXT:    pblendw {{.*#+}} xmm6 = xmm6[0],xmm0[1,2,3],xmm6[4],xmm0[5,6,7]
-; SSE41-NEXT:    packusdw %xmm6, %xmm5
-; SSE41-NEXT:    packusdw %xmm12, %xmm5
-; SSE41-NEXT:    movdqa %xmm5, %xmm0
+; SSE41-NEXT:    movdqa %xmm8, %xmm3
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm8, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm9, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE41-NEXT:    por %xmm5, %xmm0
+; SSE41-NEXT:    movapd %xmm7, %xmm3
+; SSE41-NEXT:    blendvpd %xmm0, %xmm2, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm6
+; SSE41-NEXT:    movdqa %xmm8, %xmm0
+; SSE41-NEXT:    pcmpgtd %xmm6, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm8, %xmm6
+; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3]
+; SSE41-NEXT:    pand %xmm2, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; SSE41-NEXT:    por %xmm5, %xmm0
+; SSE41-NEXT:    blendvpd %xmm0, %xmm1, %xmm7
+; SSE41-NEXT:    packusdw %xmm3, %xmm7
+; SSE41-NEXT:    packusdw %xmm4, %xmm7
+; SSE41-NEXT:    psubusw %xmm7, %xmm10
+; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm0 = xmm10[0],zero,zero,zero,xmm10[1],zero,zero,zero
+; SSE41-NEXT:    movdqa %xmm10, %xmm1
+; SSE41-NEXT:    pshufb {{.*#+}} xmm1 = xmm1[12,13],zero,zero,zero,zero,zero,zero,xmm1[14,15],zero,zero,zero,zero,zero,zero
+; SSE41-NEXT:    movdqa %xmm10, %xmm2
+; SSE41-NEXT:    pshufb {{.*#+}} xmm2 = xmm2[8,9],zero,zero,zero,zero,zero,zero,xmm2[10,11],zero,zero,zero,zero,zero,zero
+; SSE41-NEXT:    packusdw %xmm1, %xmm2
+; SSE41-NEXT:    pshufb {{.*#+}} xmm10 = xmm10[4,5],zero,zero,zero,zero,zero,zero,xmm10[6,7],zero,zero,zero,zero,zero,zero
+; SSE41-NEXT:    packusdw %xmm10, %xmm0
+; SSE41-NEXT:    packusdw %xmm2, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: psubus_8i64_max:
 ; AVX1:       # %bb.0: # %vector.ph
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[3,1,2,3]
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
-; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm8
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm6 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm6, %ymm9
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm10
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm7 = [9223372036854775808,9223372036854775808]
-; AVX1-NEXT:    vpxor %xmm7, %xmm10, %xmm5
-; AVX1-NEXT:    vpor %xmm7, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpgtq %xmm5, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm7, %xmm1, %xmm5
-; AVX1-NEXT:    vpor %xmm7, %xmm6, %xmm6
-; AVX1-NEXT:    vpcmpgtq %xmm5, %xmm6, %xmm5
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
-; AVX1-NEXT:    vblendvpd %ymm0, %ymm9, %ymm1, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm5
-; AVX1-NEXT:    vpxor %xmm7, %xmm5, %xmm6
-; AVX1-NEXT:    vpor %xmm7, %xmm4, %xmm4
-; AVX1-NEXT:    vpcmpgtq %xmm6, %xmm4, %xmm4
-; AVX1-NEXT:    vpxor %xmm7, %xmm2, %xmm6
-; AVX1-NEXT:    vpor %xmm7, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtq %xmm6, %xmm3, %xmm3
-; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-NEXT:    vblendvpd %ymm3, %ymm8, %ymm2, %ymm3
-; AVX1-NEXT:    vpsubq %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
-; AVX1-NEXT:    vpsubq %xmm5, %xmm3, %xmm3
-; AVX1-NEXT:    vpsubq %xmm1, %xmm0, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpsubq %xmm10, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808]
+; AVX1-NEXT:    vpxor %xmm4, %xmm3, %xmm3
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm5 = [9223372036854841343,9223372036854841343]
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm5, %xmm3
+; AVX1-NEXT:    vpxor %xmm4, %xmm2, %xmm6
+; AVX1-NEXT:    vpcmpgtq %xmm6, %xmm5, %xmm6
+; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm3
+; AVX1-NEXT:    vmovapd {{.*#+}} ymm6 = [65535,65535,65535,65535]
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm2, %ymm6, %ymm2
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpxor %xmm4, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm5, %xmm3
+; AVX1-NEXT:    vpxor %xmm4, %xmm1, %xmm4
+; AVX1-NEXT:    vpcmpgtq %xmm4, %xmm5, %xmm4
+; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm4, %ymm3
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm1, %ymm6, %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpackusdw %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[12,13],zero,zero,zero,zero,zero,zero,xmm0[14,15],zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[8,9],zero,zero,zero,zero,zero,zero,xmm0[10,11],zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpackusdw %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5],zero,zero,zero,zero,zero,zero,xmm0[6,7],zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0],xmm4[1,2,3],xmm3[4],xmm4[5,6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
-; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
-; AVX2-SLOW-LABEL: psubus_8i64_max:
-; AVX2-SLOW:       # %bb.0: # %vector.ph
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vpmovzxwq {{.*#+}} ymm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
-; AVX2-SLOW-NEXT:    vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm4 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
-; AVX2-SLOW-NEXT:    vpxor %ymm4, %ymm1, %ymm5
-; AVX2-SLOW-NEXT:    vpor %ymm4, %ymm0, %ymm6
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm5, %ymm6, %ymm5
-; AVX2-SLOW-NEXT:    vblendvpd %ymm5, %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpxor %ymm4, %ymm2, %ymm5
-; AVX2-SLOW-NEXT:    vpor %ymm4, %ymm3, %ymm4
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm5, %ymm4, %ymm4
-; AVX2-SLOW-NEXT:    vblendvpd %ymm4, %ymm3, %ymm2, %ymm3
-; AVX2-SLOW-NEXT:    vpsubq %ymm2, %ymm3, %ymm2
-; AVX2-SLOW-NEXT:    vpsubq %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX2-SLOW-NEXT:    vzeroupper
-; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: psubus_8i64_max:
-; AVX2-FAST:       # %bb.0: # %vector.ph
-; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vpmovzxwq {{.*#+}} ymm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
-; AVX2-FAST-NEXT:    vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm4 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
-; AVX2-FAST-NEXT:    vpxor %ymm4, %ymm1, %ymm5
-; AVX2-FAST-NEXT:    vpor %ymm4, %ymm0, %ymm6
-; AVX2-FAST-NEXT:    vpcmpgtq %ymm5, %ymm6, %ymm5
-; AVX2-FAST-NEXT:    vblendvpd %ymm5, %ymm0, %ymm1, %ymm0
-; AVX2-FAST-NEXT:    vpxor %ymm4, %ymm2, %ymm5
-; AVX2-FAST-NEXT:    vpor %ymm4, %ymm3, %ymm4
-; AVX2-FAST-NEXT:    vpcmpgtq %ymm5, %ymm4, %ymm4
-; AVX2-FAST-NEXT:    vblendvpd %ymm4, %ymm3, %ymm2, %ymm3
-; AVX2-FAST-NEXT:    vpsubq %ymm2, %ymm3, %ymm2
-; AVX2-FAST-NEXT:    vpsubq %ymm1, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
-; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
-; AVX2-FAST-NEXT:    vpermd %ymm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-FAST-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX2-FAST-NEXT:    vzeroupper
-; AVX2-FAST-NEXT:    retq
+; AVX2-LABEL: psubus_8i64_max:
+; AVX2:       # %bb.0: # %vector.ph
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm3 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
+; AVX2-NEXT:    vpxor %ymm3, %ymm2, %ymm4
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm5 = [9223372036854841343,9223372036854841343,9223372036854841343,9223372036854841343]
+; AVX2-NEXT:    vpcmpgtq %ymm4, %ymm5, %ymm4
+; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm6 = [65535,65535,65535,65535]
+; AVX2-NEXT:    vblendvpd %ymm4, %ymm2, %ymm6, %ymm2
+; AVX2-NEXT:    vpxor %ymm3, %ymm1, %ymm3
+; AVX2-NEXT:    vpcmpgtq %ymm3, %ymm5, %ymm3
+; AVX2-NEXT:    vblendvpd %ymm3, %ymm1, %ymm6, %ymm1
+; AVX2-NEXT:    vpackusdw %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,1,3]
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
+; AVX2-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: psubus_8i64_max:
 ; AVX512:       # %bb.0: # %vector.ph
-- 
2.11.0