From 3a15c6a80b7bdae4296f7c964d160c52548e42eb Mon Sep 17 00:00:00 2001
From: Tobias Decking <Tobias.Decking@gmail.com>
Date: Fri, 11 Jul 2025 23:37:04 +0200
Subject: [PATCH 1/5] Fix #148238

---
 llvm/lib/Target/X86/X86ISelLowering.cpp    |   4 +-
 llvm/test/CodeGen/X86/vector-bitreverse.ll | 100 +++++++++++----------
 2 files changed, 56 insertions(+), 48 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 5e35d5630d667..73e487bb3b206 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1328,6 +1328,7 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::BITREVERSE, MVT::i16, Custom);
     setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
     setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
+    setOperationAction(ISD::BITREVERSE, MVT::v16i8, Custom);
   }
 
   if (!Subtarget.useSoftFloat() && Subtarget.hasSSSE3()) {
@@ -32694,7 +32695,8 @@ static SDValue LowerBITREVERSE(SDValue Op, const X86Subtarget &Subtarget,
   if (Subtarget.hasXOP() && !VT.is512BitVector())
     return LowerBITREVERSE_XOP(Op, DAG);
 
-  assert(Subtarget.hasSSSE3() && "SSSE3 required for BITREVERSE");
+  assert((Subtarget.hasSSSE3() || Subtarget.hasGFNI()) &&
+         "SSSE3 or GFNI required for BITREVERSE");
 
   SDValue In = Op.getOperand(0);
   SDLoc DL(Op);
diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index 5dcf19013f0b7..4ba2ec393d5d4 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -8,6 +8,7 @@
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=ALL,AVX,AVX512,AVX512BW
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx | FileCheck %s --check-prefixes=ALL,XOP,XOPAVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=ALL,XOP,XOPAVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+gfni | FileCheck %s --check-prefixes=ALL,GFNIPURE
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3,+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX2
@@ -60,13 +61,13 @@ define i8 @test_bitreverse_i8(i8 %a) nounwind {
 ; XOP-NEXT:    # kill: def $al killed $al killed $eax
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_i8:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movd %edi, %xmm0
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    movd %xmm0, %eax
-; GFNISSE-NEXT:    # kill: def $al killed $al killed $eax
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_i8:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movd %edi, %xmm0
+; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNIPURE-NEXT:    movd %xmm0, %eax
+; GFNIPURE-NEXT:    # kill: def $al killed $al killed $eax
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i8:
 ; GFNIAVX:       # %bb.0:
@@ -134,14 +135,14 @@ define i16 @test_bitreverse_i16(i16 %a) nounwind {
 ; XOP-NEXT:    # kill: def $ax killed $ax killed $eax
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_i16:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movd %edi, %xmm0
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    movd %xmm0, %eax
-; GFNISSE-NEXT:    rolw $8, %ax
-; GFNISSE-NEXT:    # kill: def $ax killed $ax killed $eax
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_i16:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movd %edi, %xmm0
+; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNIPURE-NEXT:    movd %xmm0, %eax
+; GFNIPURE-NEXT:    rolw $8, %ax
+; GFNIPURE-NEXT:    # kill: def $ax killed $ax killed $eax
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i16:
 ; GFNIAVX:       # %bb.0:
@@ -207,13 +208,13 @@ define i32 @test_bitreverse_i32(i32 %a) nounwind {
 ; XOP-NEXT:    vmovd %xmm0, %eax
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_i32:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movd %edi, %xmm0
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    movd %xmm0, %eax
-; GFNISSE-NEXT:    bswapl %eax
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_i32:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movd %edi, %xmm0
+; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNIPURE-NEXT:    movd %xmm0, %eax
+; GFNIPURE-NEXT:    bswapl %eax
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i32:
 ; GFNIAVX:       # %bb.0:
@@ -282,13 +283,13 @@ define i64 @test_bitreverse_i64(i64 %a) nounwind {
 ; XOP-NEXT:    vmovq %xmm0, %rax
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_i64:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movq %rdi, %xmm0
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    movq %xmm0, %rax
-; GFNISSE-NEXT:    bswapq %rax
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_i64:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movq %rdi, %xmm0
+; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNIPURE-NEXT:    movq %xmm0, %rax
+; GFNIPURE-NEXT:    bswapq %rax
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i64:
 ; GFNIAVX:       # %bb.0:
@@ -386,10 +387,10 @@ define <16 x i8> @test_bitreverse_v16i8(<16 x i8> %a) nounwind {
 ; XOP-NEXT:    vpperm {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v16i8:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_v16i8:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v16i8:
 ; GFNIAVX:       # %bb.0:
@@ -873,12 +874,12 @@ define <32 x i8> @test_bitreverse_v32i8(<32 x i8> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v32i8:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_v32i8:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
+; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v32i8:
 ; GFNIAVX:       # %bb.0:
@@ -1724,14 +1725,14 @@ define <64 x i8> @test_bitreverse_v64i8(<64 x i8> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v64i8:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
-; GFNISSE-NEXT:    retq
+; GFNIPURE-LABEL: test_bitreverse_v64i8:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
+; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
+; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
+; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
+; GFNIPURE-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v64i8:
 ; GFNIAVX1:       # %bb.0:
@@ -2849,6 +2850,11 @@ define <16 x i8> @fold_bitreverse_v16i8() nounwind {
 ; XOP-NEXT:    vmovaps {{.*#+}} xmm0 = [0,255,64,191,32,223,96,159,16,239,80,175,48,207,112,143]
 ; XOP-NEXT:    retq
 ;
+; GFNIPURE-LABEL: fold_bitreverse_v16i8:
+; GFNIPURE:       # %bb.0:
+; GFNIPURE-NEXT:    movaps {{.*#+}} xmm0 = [0,255,64,191,32,223,96,159,16,239,80,175,48,207,112,143]
+; GFNIPURE-NEXT:    retq
+;
 ; GFNISSE-LABEL: fold_bitreverse_v16i8:
 ; GFNISSE:       # %bb.0:
 ; GFNISSE-NEXT:    movaps {{.*#+}} xmm0 = [0,255,64,191,32,223,96,159,16,239,80,175,48,207,112,143]

From d4194ec78bc8960b4225c07c0bac7225acae139a Mon Sep 17 00:00:00 2001
From: Tobias Decking <Tobias.Decking@gmail.com>
Date: Fri, 18 Jul 2025 13:37:53 +0200
Subject: [PATCH 2/5] Update the test file

---
 llvm/test/CodeGen/X86/vector-bitreverse.ll | 905 ++++++++++++++++++---
 1 file changed, 770 insertions(+), 135 deletions(-)

diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index 4ba2ec393d5d4..c8394d590cb8f 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -8,8 +8,8 @@
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=ALL,AVX,AVX512,AVX512BW
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx | FileCheck %s --check-prefixes=ALL,XOP,XOPAVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=ALL,XOP,XOPAVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+gfni | FileCheck %s --check-prefixes=ALL,GFNIPURE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3,+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE,GFNISSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3,+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE,GFNISSSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX512,GFNIAVX512F
@@ -61,13 +61,13 @@ define i8 @test_bitreverse_i8(i8 %a) nounwind {
 ; XOP-NEXT:    # kill: def $al killed $al killed $eax
 ; XOP-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_i8:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movd %edi, %xmm0
-; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNIPURE-NEXT:    movd %xmm0, %eax
-; GFNIPURE-NEXT:    # kill: def $al killed $al killed $eax
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_i8:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    movd %edi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movd %xmm0, %eax
+; GFNISSE-NEXT:    # kill: def $al killed $al killed $eax
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i8:
 ; GFNIAVX:       # %bb.0:
@@ -135,14 +135,14 @@ define i16 @test_bitreverse_i16(i16 %a) nounwind {
 ; XOP-NEXT:    # kill: def $ax killed $ax killed $eax
 ; XOP-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_i16:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movd %edi, %xmm0
-; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNIPURE-NEXT:    movd %xmm0, %eax
-; GFNIPURE-NEXT:    rolw $8, %ax
-; GFNIPURE-NEXT:    # kill: def $ax killed $ax killed $eax
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_i16:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    movd %edi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movd %xmm0, %eax
+; GFNISSE-NEXT:    rolw $8, %ax
+; GFNISSE-NEXT:    # kill: def $ax killed $ax killed $eax
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i16:
 ; GFNIAVX:       # %bb.0:
@@ -208,13 +208,13 @@ define i32 @test_bitreverse_i32(i32 %a) nounwind {
 ; XOP-NEXT:    vmovd %xmm0, %eax
 ; XOP-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_i32:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movd %edi, %xmm0
-; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNIPURE-NEXT:    movd %xmm0, %eax
-; GFNIPURE-NEXT:    bswapl %eax
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_i32:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    movd %edi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movd %xmm0, %eax
+; GFNISSE-NEXT:    bswapl %eax
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i32:
 ; GFNIAVX:       # %bb.0:
@@ -283,13 +283,13 @@ define i64 @test_bitreverse_i64(i64 %a) nounwind {
 ; XOP-NEXT:    vmovq %xmm0, %rax
 ; XOP-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_i64:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movq %rdi, %xmm0
-; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNIPURE-NEXT:    movq %xmm0, %rax
-; GFNIPURE-NEXT:    bswapq %rax
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_i64:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    movq %rdi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movq %xmm0, %rax
+; GFNISSE-NEXT:    bswapq %rax
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i64:
 ; GFNIAVX:       # %bb.0:
@@ -387,10 +387,10 @@ define <16 x i8> @test_bitreverse_v16i8(<16 x i8> %a) nounwind {
 ; XOP-NEXT:    vpperm {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_v16i8:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_v16i8:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v16i8:
 ; GFNIAVX:       # %bb.0:
@@ -493,11 +493,69 @@ define <8 x i16> @test_bitreverse_v8i16(<8 x i16> %a) nounwind {
 ; XOP-NEXT:    vpperm {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v8i16:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v8i16:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    pextrw $7, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
+; GFNISSE2-NEXT:    movd %xmm2, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    pextrw $6, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
+; GFNISSE2-NEXT:    movd %xmm2, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    pextrw $4, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
+; GFNISSE2-NEXT:    movd %xmm2, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    pextrw $2, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm0, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm0
+; GFNISSE2-NEXT:    movd %xmm0, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm0
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm1
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v8i16:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v8i16:
 ; GFNIAVX:       # %bb.0:
@@ -606,11 +664,38 @@ define <4 x i32> @test_bitreverse_v4i32(<4 x i32> %a) nounwind {
 ; XOP-NEXT:    vpperm {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v4i32:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v4i32:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNISSE2-NEXT:    movd %xmm1, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm1
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
+; GFNISSE2-NEXT:    movd %xmm0, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNISSE2-NEXT:    movd %xmm1, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm1
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v4i32:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v4i32:
 ; GFNIAVX:       # %bb.0:
@@ -721,11 +806,26 @@ define <2 x i64> @test_bitreverse_v2i64(<2 x i64> %a) nounwind {
 ; XOP-NEXT:    vpperm {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v2i64:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
-; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v2i64:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm0
+; GFNISSE2-NEXT:    movq %xmm0, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
+; GFNISSE2-NEXT:    movq %xmm2, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm1
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v2i64:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v2i64:
 ; GFNIAVX:       # %bb.0:
@@ -874,12 +974,12 @@ define <32 x i8> @test_bitreverse_v32i8(<32 x i8> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_v32i8:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
-; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_v32i8:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v32i8:
 ; GFNIAVX:       # %bb.0:
@@ -1043,15 +1143,126 @@ define <16 x i16> @test_bitreverse_v16i16(<16 x i16> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v16i16:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm2 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; GFNISSE-NEXT:    pshufb %xmm2, %xmm0
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
-; GFNISSE-NEXT:    pshufb %xmm2, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v16i16:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    pextrw $7, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    pextrw $6, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pextrw $4, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    pextrw $2, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm0, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
+; GFNISSE2-NEXT:    movd %xmm0, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm0
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; GFNISSE2-NEXT:    pextrw $7, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    pextrw $6, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pextrw $4, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    pextrw $2, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm1, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNISSE2-NEXT:    movd %xmm1, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm1
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v16i16:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm0
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm1
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v16i16:
 ; GFNIAVX1:       # %bb.0:
@@ -1242,15 +1453,64 @@ define <8 x i32> @test_bitreverse_v8i32(<8 x i32> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v8i32:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm2 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; GFNISSE-NEXT:    pshufb %xmm2, %xmm0
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
-; GFNISSE-NEXT:    pshufb %xmm2, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v8i32:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
+; GFNISSE2-NEXT:    movd %xmm0, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,3,3,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
+; GFNISSE2-NEXT:    movd %xmm4, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNISSE2-NEXT:    movd %xmm1, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v8i32:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm2 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm0
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm1
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v8i32:
 ; GFNIAVX1:       # %bb.0:
@@ -1445,15 +1705,40 @@ define <4 x i64> @test_bitreverse_v4i64(<4 x i64> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v4i64:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm2 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
-; GFNISSE-NEXT:    pshufb %xmm2, %xmm0
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
-; GFNISSE-NEXT:    pshufb %xmm2, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v4i64:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
+; GFNISSE2-NEXT:    movq %xmm0, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movq %xmm3, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm3
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
+; GFNISSE2-NEXT:    movq %xmm1, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
+; GFNISSE2-NEXT:    movq %xmm3, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm2
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v4i64:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm2 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
+; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm0
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm1
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v4i64:
 ; GFNIAVX1:       # %bb.0:
@@ -1725,14 +2010,14 @@ define <64 x i8> @test_bitreverse_v64i8(<64 x i8> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNIPURE-LABEL: test_bitreverse_v64i8:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
-; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
-; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
-; GFNIPURE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
-; GFNIPURE-NEXT:    retq
+; GFNISSE-LABEL: test_bitreverse_v64i8:
+; GFNISSE:       # %bb.0:
+; GFNISSE-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
+; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
+; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
+; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v64i8:
 ; GFNIAVX1:       # %bb.0:
@@ -2036,19 +2321,236 @@ define <32 x i16> @test_bitreverse_v32i16(<32 x i16> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v32i16:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm4 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm0
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm2
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm3
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v32i16:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    pextrw $7, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pextrw $6, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    pextrw $4, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    pextrw $2, %xmm0, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
+; GFNISSE2-NEXT:    movd %xmm7, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm0, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
+; GFNISSE2-NEXT:    movd %xmm0, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm0
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]
+; GFNISSE2-NEXT:    pextrw $7, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pextrw $6, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    pextrw $4, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    pextrw $2, %xmm1, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
+; GFNISSE2-NEXT:    movd %xmm7, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm1, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
+; GFNISSE2-NEXT:    movd %xmm1, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm1
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm5[0]
+; GFNISSE2-NEXT:    pextrw $7, %xmm2, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pextrw $6, %xmm2, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm2, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    pextrw $4, %xmm2, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm2, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    pextrw $2, %xmm2, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
+; GFNISSE2-NEXT:    movd %xmm7, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm2, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
+; GFNISSE2-NEXT:    movd %xmm2, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm2
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
+; GFNISSE2-NEXT:    pextrw $7, %xmm3, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pextrw $6, %xmm3, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; GFNISSE2-NEXT:    pextrw $5, %xmm3, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    pextrw $4, %xmm3, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; GFNISSE2-NEXT:    pextrw $3, %xmm3, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    pextrw $2, %xmm3, %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
+; GFNISSE2-NEXT:    movd %xmm7, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm7
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; GFNISSE2-NEXT:    pextrw $1, %xmm3, %eax
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %ecx
+; GFNISSE2-NEXT:    rolw $8, %cx
+; GFNISSE2-NEXT:    movd %ecx, %xmm3
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    rolw $8, %ax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v32i16:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm0
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm1
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm2
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm3
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v32i16:
 ; GFNIAVX1:       # %bb.0:
@@ -2394,19 +2896,112 @@ define <16 x i32> @test_bitreverse_v16i32(<16 x i32> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v16i32:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm4 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm0
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm2
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm3
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v16i32:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
+; GFNISSE2-NEXT:    movd %xmm0, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm6[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[3,3,3,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
+; GFNISSE2-NEXT:    movd %xmm1, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm6[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[3,3,3,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
+; GFNISSE2-NEXT:    movd %xmm2, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm2
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm6[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[3,3,3,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
+; GFNISSE2-NEXT:    movd %xmm6, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm6
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[1,1,1,1]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
+; GFNISSE2-NEXT:    movd %xmm3, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movd %xmm5, %eax
+; GFNISSE2-NEXT:    bswapl %eax
+; GFNISSE2-NEXT:    movd %eax, %xmm4
+; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm6[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v16i32:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm4 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm0
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm1
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm2
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm3
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v16i32:
 ; GFNIAVX1:       # %bb.0:
@@ -2760,19 +3355,64 @@ define <8 x i64> @test_bitreverse_v8i64(<8 x i64> %a) nounwind {
 ; XOPAVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
 ; XOPAVX2-NEXT:    retq
 ;
-; GFNISSE-LABEL: test_bitreverse_v8i64:
-; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm4 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm0
-; GFNISSE-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm1
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm2
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
-; GFNISSE-NEXT:    pshufb %xmm4, %xmm3
-; GFNISSE-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
-; GFNISSE-NEXT:    retq
+; GFNISSE2-LABEL: test_bitreverse_v8i64:
+; GFNISSE2:       # %bb.0:
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
+; GFNISSE2-NEXT:    movq %xmm0, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movq %xmm5, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm5
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
+; GFNISSE2-NEXT:    movq %xmm1, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movq %xmm5, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm5
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm5[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
+; GFNISSE2-NEXT:    movq %xmm2, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm2
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movq %xmm5, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm5
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
+; GFNISSE2-NEXT:    movq %xmm3, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
+; GFNISSE2-NEXT:    movq %xmm5, %rax
+; GFNISSE2-NEXT:    bswapq %rax
+; GFNISSE2-NEXT:    movq %rax, %xmm4
+; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; GFNISSE2-NEXT:    retq
+;
+; GFNISSSE2-LABEL: test_bitreverse_v8i64:
+; GFNISSSE2:       # %bb.0:
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm4 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm0
+; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm1
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm2
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm3
+; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
+; GFNISSSE2-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v8i64:
 ; GFNIAVX1:       # %bb.0:
@@ -2850,11 +3490,6 @@ define <16 x i8> @fold_bitreverse_v16i8() nounwind {
 ; XOP-NEXT:    vmovaps {{.*#+}} xmm0 = [0,255,64,191,32,223,96,159,16,239,80,175,48,207,112,143]
 ; XOP-NEXT:    retq
 ;
-; GFNIPURE-LABEL: fold_bitreverse_v16i8:
-; GFNIPURE:       # %bb.0:
-; GFNIPURE-NEXT:    movaps {{.*#+}} xmm0 = [0,255,64,191,32,223,96,159,16,239,80,175,48,207,112,143]
-; GFNIPURE-NEXT:    retq
-;
 ; GFNISSE-LABEL: fold_bitreverse_v16i8:
 ; GFNISSE:       # %bb.0:
 ; GFNISSE-NEXT:    movaps {{.*#+}} xmm0 = [0,255,64,191,32,223,96,159,16,239,80,175,48,207,112,143]

From c097e1215e579bec41e86ed097a7c2e9d5d39e72 Mon Sep 17 00:00:00 2001
From: Tobias Decking <Tobias.Decking@gmail.com>
Date: Fri, 18 Jul 2025 14:42:44 +0200
Subject: [PATCH 3/5] Extend the optimization towards other vectors

Also add a missing `Subtarget.useSoftFloat()`
---
 llvm/lib/Target/X86/X86ISelLowering.cpp    |   7 +-
 llvm/test/CodeGen/X86/vector-bitreverse.ll | 775 +++++----------------
 2 files changed, 184 insertions(+), 598 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 73e487bb3b206..84d288c6b0a5e 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1323,12 +1323,15 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::STRICT_FDIV,        MVT::v2f64, Legal);
   }
 
-  if (Subtarget.hasGFNI()) {
+  if (!Subtarget.useSoftFloat() && Subtarget.hasGFNI()) {
     setOperationAction(ISD::BITREVERSE, MVT::i8, Custom);
     setOperationAction(ISD::BITREVERSE, MVT::i16, Custom);
     setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
     setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
-    setOperationAction(ISD::BITREVERSE, MVT::v16i8, Custom);
+
+    for (auto VT : {MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64}) {
+      setOperationAction(ISD::BITREVERSE,       VT, Custom);
+    }
   }
 
   if (!Subtarget.useSoftFloat() && Subtarget.hasSSSE3()) {
diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index c8394d590cb8f..76dedaf648606 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -495,60 +495,11 @@ define <8 x i16> @test_bitreverse_v8i16(<8 x i16> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v8i16:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    pextrw $7, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
-; GFNISSE2-NEXT:    movd %xmm2, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    pextrw $6, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
-; GFNISSE2-NEXT:    movd %xmm2, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    pextrw $4, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
-; GFNISSE2-NEXT:    movd %xmm2, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    pextrw $2, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm0, %eax
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm0
-; GFNISSE2-NEXT:    movd %xmm0, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm0
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm1
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm1
+; GFNISSE2-NEXT:    psrlw $8, %xmm1
+; GFNISSE2-NEXT:    psllw $8, %xmm0
+; GFNISSE2-NEXT:    por %xmm1, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v8i16:
@@ -666,29 +617,16 @@ define <4 x i32> @test_bitreverse_v4i32(<4 x i32> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v4i32:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNISSE2-NEXT:    movd %xmm1, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm1
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
-; GFNISSE2-NEXT:    movd %xmm0, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm0
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNISSE2-NEXT:    movd %xmm1, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm1
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; GFNISSE2-NEXT:    pxor %xmm1, %xmm1
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm2
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm2, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v4i32:
@@ -808,17 +746,18 @@ define <2 x i64> @test_bitreverse_v2i64(<2 x i64> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v2i64:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm0
-; GFNISSE2-NEXT:    movq %xmm0, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm0
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm1, %xmm2
-; GFNISSE2-NEXT:    movq %xmm2, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm1
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; GFNISSE2-NEXT:    pxor %xmm1, %xmm1
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm2
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm2, %xmm0
+; GFNISSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v2i64:
@@ -1145,113 +1084,17 @@ define <16 x i16> @test_bitreverse_v16i16(<16 x i16> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v16i16:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    pextrw $7, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm2
+; GFNISSE2-NEXT:    psrlw $8, %xmm2
+; GFNISSE2-NEXT:    psllw $8, %xmm0
+; GFNISSE2-NEXT:    por %xmm2, %xmm0
 ; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    pextrw $6, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pextrw $4, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    pextrw $2, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm0, %eax
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
-; GFNISSE2-NEXT:    movd %xmm0, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm0
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
-; GFNISSE2-NEXT:    pextrw $7, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    pextrw $6, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pextrw $4, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    pextrw $2, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm1, %eax
+; GFNISSE2-NEXT:    movdqa %xmm1, %xmm3
+; GFNISSE2-NEXT:    psrlw $8, %xmm3
+; GFNISSE2-NEXT:    psllw $8, %xmm1
+; GFNISSE2-NEXT:    por %xmm3, %xmm1
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNISSE2-NEXT:    movd %xmm1, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm1
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v16i16:
@@ -1455,51 +1298,26 @@ define <8 x i32> @test_bitreverse_v8i32(<8 x i32> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v8i32:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
-; GFNISSE2-NEXT:    movd %xmm0, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm0
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,3,3,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm4
-; GFNISSE2-NEXT:    movd %xmm4, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNISSE2-NEXT:    movd %xmm1, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm1
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; GFNISSE2-NEXT:    pxor %xmm2, %xmm2
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm3
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm3, %xmm0
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSE2-NEXT:    movdqa %xmm1, %xmm4
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm4, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v8i32:
@@ -1707,27 +1525,30 @@ define <4 x i64> @test_bitreverse_v4i64(<4 x i64> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v4i64:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm0
-; GFNISSE2-NEXT:    movq %xmm0, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm0
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movq %xmm3, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm3
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
-; GFNISSE2-NEXT:    movq %xmm1, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm1
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm3
-; GFNISSE2-NEXT:    movq %xmm3, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm2
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; GFNISSE2-NEXT:    pxor %xmm2, %xmm2
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm3
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm3, %xmm0
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSE2-NEXT:    movdqa %xmm1, %xmm4
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm4, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v4i64:
@@ -2323,219 +2144,27 @@ define <32 x i16> @test_bitreverse_v32i16(<32 x i16> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v32i16:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    pextrw $7, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm4
+; GFNISSE2-NEXT:    psrlw $8, %xmm4
+; GFNISSE2-NEXT:    psllw $8, %xmm0
+; GFNISSE2-NEXT:    por %xmm4, %xmm0
 ; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pextrw $6, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    pextrw $4, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    pextrw $2, %xmm0, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
-; GFNISSE2-NEXT:    movd %xmm7, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm0, %eax
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
-; GFNISSE2-NEXT:    movd %xmm0, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm0
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]
-; GFNISSE2-NEXT:    pextrw $7, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pextrw $6, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    pextrw $4, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    pextrw $2, %xmm1, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
-; GFNISSE2-NEXT:    movd %xmm7, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm1, %eax
+; GFNISSE2-NEXT:    movdqa %xmm1, %xmm5
+; GFNISSE2-NEXT:    psrlw $8, %xmm5
+; GFNISSE2-NEXT:    psllw $8, %xmm1
+; GFNISSE2-NEXT:    por %xmm5, %xmm1
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
-; GFNISSE2-NEXT:    movd %xmm1, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm1
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm5[0]
-; GFNISSE2-NEXT:    pextrw $7, %xmm2, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pextrw $6, %xmm2, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm2, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    pextrw $4, %xmm2, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm2, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    pextrw $2, %xmm2, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
-; GFNISSE2-NEXT:    movd %xmm7, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm2, %eax
+; GFNISSE2-NEXT:    movdqa %xmm2, %xmm5
+; GFNISSE2-NEXT:    psrlw $8, %xmm5
+; GFNISSE2-NEXT:    psllw $8, %xmm2
+; GFNISSE2-NEXT:    por %xmm5, %xmm2
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
-; GFNISSE2-NEXT:    movd %xmm2, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm2
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
-; GFNISSE2-NEXT:    pextrw $7, %xmm3, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pextrw $6, %xmm3, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; GFNISSE2-NEXT:    pextrw $5, %xmm3, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    pextrw $4, %xmm3, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; GFNISSE2-NEXT:    pextrw $3, %xmm3, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    pextrw $2, %xmm3, %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm7
-; GFNISSE2-NEXT:    movd %xmm7, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm7
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; GFNISSE2-NEXT:    pextrw $1, %xmm3, %eax
+; GFNISSE2-NEXT:    movdqa %xmm3, %xmm5
+; GFNISSE2-NEXT:    psrlw $8, %xmm5
+; GFNISSE2-NEXT:    psllw $8, %xmm3
+; GFNISSE2-NEXT:    por %xmm5, %xmm3
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %ecx
-; GFNISSE2-NEXT:    rolw $8, %cx
-; GFNISSE2-NEXT:    movd %ecx, %xmm3
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    rolw $8, %ax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v32i16:
@@ -2898,95 +2527,44 @@ define <16 x i32> @test_bitreverse_v16i32(<16 x i32> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v16i32:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
-; GFNISSE2-NEXT:    movd %xmm0, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm0
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm6[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[3,3,3,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
-; GFNISSE2-NEXT:    movd %xmm1, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm1
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm6[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[3,3,3,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
-; GFNISSE2-NEXT:    movd %xmm2, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm2
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm6[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[3,3,3,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm5
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm6
-; GFNISSE2-NEXT:    movd %xmm6, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm6
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[1,1,1,1]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
-; GFNISSE2-NEXT:    movd %xmm3, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movd %xmm5, %eax
-; GFNISSE2-NEXT:    bswapl %eax
-; GFNISSE2-NEXT:    movd %eax, %xmm4
-; GFNISSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm6[0]
+; GFNISSE2-NEXT:    pxor %xmm4, %xmm4
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm5
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm5, %xmm0
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSE2-NEXT:    movdqa %xmm1, %xmm6
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm6, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSE2-NEXT:    movdqa %xmm2, %xmm6
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm6, %xmm2
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSE2-NEXT:    movdqa %xmm3, %xmm6
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm6, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v16i32:
@@ -3357,47 +2935,52 @@ define <8 x i64> @test_bitreverse_v8i64(<8 x i64> %a) nounwind {
 ;
 ; GFNISSE2-LABEL: test_bitreverse_v8i64:
 ; GFNISSE2:       # %bb.0:
-; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm0
-; GFNISSE2-NEXT:    movq %xmm0, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm0
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movq %xmm5, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm5
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm1
-; GFNISSE2-NEXT:    movq %xmm1, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm1
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movq %xmm5, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm5
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm5[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm2
-; GFNISSE2-NEXT:    movq %xmm2, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm2
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movq %xmm5, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm5
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
-; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
-; GFNISSE2-NEXT:    movq %xmm3, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm3
-; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm5
-; GFNISSE2-NEXT:    movq %xmm5, %rax
-; GFNISSE2-NEXT:    bswapq %rax
-; GFNISSE2-NEXT:    movq %rax, %xmm4
-; GFNISSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; GFNISSE2-NEXT:    pxor %xmm4, %xmm4
+; GFNISSE2-NEXT:    movdqa %xmm0, %xmm5
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm5, %xmm0
+; GFNISSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSE2-NEXT:    movdqa %xmm1, %xmm6
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm6, %xmm1
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSE2-NEXT:    movdqa %xmm2, %xmm6
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm6, %xmm2
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSE2-NEXT:    movdqa %xmm3, %xmm6
+; GFNISSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; GFNISSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; GFNISSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; GFNISSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; GFNISSE2-NEXT:    packuswb %xmm6, %xmm3
+; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
 ; GFNISSE2-NEXT:    retq
 ;
 ; GFNISSSE2-LABEL: test_bitreverse_v8i64:

From 718deeaca599fb9e188677d8737385ac8f136cca Mon Sep 17 00:00:00 2001
From: Tobias Decking <Tobias.Decking@gmail.com>
Date: Fri, 18 Jul 2025 16:40:06 +0200
Subject: [PATCH 4/5] Turn SSSE2 into SSSE3

---
 llvm/test/CodeGen/X86/vector-bitreverse.ll | 164 ++++++++++-----------
 1 file changed, 82 insertions(+), 82 deletions(-)

diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index 76dedaf648606..834dfd63432b0 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -9,7 +9,7 @@
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx | FileCheck %s --check-prefixes=ALL,XOP,XOPAVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=ALL,XOP,XOPAVX2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE,GFNISSE2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3,+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE,GFNISSSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3,+gfni | FileCheck %s --check-prefixes=ALL,GFNISSE,GFNISSSE3
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+gfni | FileCheck %s --check-prefixes=ALL,GFNIAVX,GFNIAVX512,GFNIAVX512F
@@ -502,11 +502,11 @@ define <8 x i16> @test_bitreverse_v8i16(<8 x i16> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v8i16:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v8i16:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v8i16:
 ; GFNIAVX:       # %bb.0:
@@ -629,11 +629,11 @@ define <4 x i32> @test_bitreverse_v4i32(<4 x i32> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v4i32:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v4i32:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v4i32:
 ; GFNIAVX:       # %bb.0:
@@ -760,11 +760,11 @@ define <2 x i64> @test_bitreverse_v2i64(<2 x i64> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v2i64:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v2i64:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_v2i64:
 ; GFNIAVX:       # %bb.0:
@@ -1097,15 +1097,15 @@ define <16 x i16> @test_bitreverse_v16i16(<16 x i16> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm2, %xmm1
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v16i16:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm0
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
-; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm1
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v16i16:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; GFNISSSE3-NEXT:    pshufb %xmm2, %xmm0
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSSE3-NEXT:    pshufb %xmm2, %xmm1
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v16i16:
 ; GFNIAVX1:       # %bb.0:
@@ -1320,15 +1320,15 @@ define <8 x i32> @test_bitreverse_v8i32(<8 x i32> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v8i32:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm2 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm0
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
-; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm1
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v8i32:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; GFNISSSE3-NEXT:    pshufb %xmm2, %xmm0
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSSE3-NEXT:    pshufb %xmm2, %xmm1
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v8i32:
 ; GFNIAVX1:       # %bb.0:
@@ -1551,15 +1551,15 @@ define <4 x i64> @test_bitreverse_v4i64(<4 x i64> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v4i64:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm2 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
-; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm0
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
-; GFNISSSE2-NEXT:    pshufb %xmm2, %xmm1
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v4i64:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
+; GFNISSSE3-NEXT:    pshufb %xmm2, %xmm0
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm3, %xmm0
+; GFNISSSE3-NEXT:    pshufb %xmm2, %xmm1
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm3, %xmm1
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v4i64:
 ; GFNIAVX1:       # %bb.0:
@@ -2167,19 +2167,19 @@ define <32 x i16> @test_bitreverse_v32i16(<32 x i16> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm4, %xmm3
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v32i16:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm4 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm0
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm1
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm2
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm3
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v32i16:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm0
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm1
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm2
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm3
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v32i16:
 ; GFNIAVX1:       # %bb.0:
@@ -2567,19 +2567,19 @@ define <16 x i32> @test_bitreverse_v16i32(<16 x i32> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v16i32:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm4 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm0
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm1
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm2
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm3
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v16i32:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm0
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm1
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm2
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm3
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v16i32:
 ; GFNIAVX1:       # %bb.0:
@@ -2983,19 +2983,19 @@ define <8 x i64> @test_bitreverse_v8i64(<8 x i64> %a) nounwind {
 ; GFNISSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
 ; GFNISSE2-NEXT:    retq
 ;
-; GFNISSSE2-LABEL: test_bitreverse_v8i64:
-; GFNISSSE2:       # %bb.0:
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm4 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm0
-; GFNISSSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm1
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm2
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
-; GFNISSSE2-NEXT:    pshufb %xmm4, %xmm3
-; GFNISSSE2-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
-; GFNISSSE2-NEXT:    retq
+; GFNISSSE3-LABEL: test_bitreverse_v8i64:
+; GFNISSSE3:       # %bb.0:
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [7,6,5,4,3,2,1,0,15,14,13,12,11,10,9,8]
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm0
+; GFNISSSE3-NEXT:    movdqa {{.*#+}} xmm5 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm0
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm1
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm1
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm2
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm2
+; GFNISSSE3-NEXT:    pshufb %xmm4, %xmm3
+; GFNISSSE3-NEXT:    gf2p8affineqb $0, %xmm5, %xmm3
+; GFNISSSE3-NEXT:    retq
 ;
 ; GFNIAVX1-LABEL: test_bitreverse_v8i64:
 ; GFNIAVX1:       # %bb.0:

From 4f27cd72f05ee2c738a420f2064b8e55d0ef37cf Mon Sep 17 00:00:00 2001
From: Tobias Decking <Tobias.Decking@gmail.com>
Date: Fri, 18 Jul 2025 18:51:11 +0200
Subject: [PATCH 5/5] Formatting

---
 llvm/lib/Target/X86/X86ISelLowering.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 84d288c6b0a5e..503aa4fba49c0 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1330,7 +1330,7 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
 
     for (auto VT : {MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64}) {
-      setOperationAction(ISD::BITREVERSE,       VT, Custom);
+      setOperationAction(ISD::BITREVERSE, VT, Custom);
     }
   }