[llvm] [X86] Add basic folds for CLMUL nodes (PR #176961)
Vishruth Thimmaiah via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 20 08:31:56 PST 2026
https://github.com/vishruth-thimmaiah created https://github.com/llvm/llvm-project/pull/176961
Closes #176783
Adds support for folding `ISD::CMUL`/`CMULH`/`CMULR` nodes.
>From 7835d61911548a26371679614191ae81f45352b3 Mon Sep 17 00:00:00 2001
From: vishruth-thimmaiah <vishruththimmaiah at gmail.com>
Date: Tue, 20 Jan 2026 21:35:42 +0530
Subject: [PATCH] [X86] Add basic folds for CLMUL nodes
Signed-off-by: vishruth-thimmaiah <vishruththimmaiah at gmail.com>
---
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 29 +
llvm/test/CodeGen/X86/clmul-vector.ll | 834 ++++++++++++++++++
2 files changed, 863 insertions(+)
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 4567b4017a7e1..31c863ba57bb3 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -477,6 +477,7 @@ namespace {
SDValue visitSHLSAT(SDNode *N);
SDValue visitRotate(SDNode *N);
SDValue visitABS(SDNode *N);
+ SDValue visitCLMUL(SDNode *N);
SDValue visitBSWAP(SDNode *N);
SDValue visitBITREVERSE(SDNode *N);
SDValue visitCTLZ(SDNode *N);
@@ -1979,6 +1980,9 @@ SDValue DAGCombiner::visit(SDNode *N) {
case ISD::SSHLSAT:
case ISD::USHLSAT: return visitSHLSAT(N);
case ISD::ABS: return visitABS(N);
+ case ISD::CLMUL:
+ case ISD::CLMULR:
+ case ISD::CLMULH: return visitCLMUL(N);
case ISD::BSWAP: return visitBSWAP(N);
case ISD::BITREVERSE: return visitBITREVERSE(N);
case ISD::CTLZ: return visitCTLZ(N);
@@ -11767,6 +11771,31 @@ SDValue DAGCombiner::visitABS(SDNode *N) {
return SDValue();
}
+SDValue DAGCombiner::visitCLMUL(SDNode *N) {
+ unsigned Opcode = N->getOpcode();
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ EVT VT = N->getValueType(0);
+ SDLoc DL(N);
+
+ // fold (clmul c1, c2)
+ if (SDValue C = DAG.FoldConstantArithmetic(Opcode, DL, VT, {N0, N1}))
+ return C;
+
+ // canonicalize constant to RHS
+ if (DAG.isConstantIntBuildVectorOrConstantInt(N0) &&
+ !DAG.isConstantIntBuildVectorOrConstantInt(N1))
+ return DAG.getNode(Opcode, DL, VT, N1, N0);
+
+ // fold (clmul x, 0) -> 0
+ if (isNullConstant(N1) || ISD::isConstantSplatVectorAllZeros(N1.getNode()))
+ return DAG.getConstant(0, DL, VT);
+
+ return SDValue();
+}
+
+
+
SDValue DAGCombiner::visitBSWAP(SDNode *N) {
SDValue N0 = N->getOperand(0);
EVT VT = N->getValueType(0);
diff --git a/llvm/test/CodeGen/X86/clmul-vector.ll b/llvm/test/CodeGen/X86/clmul-vector.ll
index a7483dca32a02..db3ad486dd19c 100644
--- a/llvm/test/CodeGen/X86/clmul-vector.ll
+++ b/llvm/test/CodeGen/X86/clmul-vector.ll
@@ -7932,3 +7932,837 @@ define void @mul_use_commutative_clmul_v2i64(<2 x i64> %x, <2 x i64> %y, ptr %p0
store <2 x i64> %yx, ptr %p1
ret void
}
+
+define <2 x i64> @pclmul(<2 x i64> %v0, <2 x i64> %v1) {
+; SSE2-NOPCLMUL-LABEL: pclmul:
+; SSE2-NOPCLMUL: # %bb.0:
+; SSE2-NOPCLMUL-NEXT: pushq %rbp
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 16
+; SSE2-NOPCLMUL-NEXT: pushq %r15
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 24
+; SSE2-NOPCLMUL-NEXT: pushq %r14
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 32
+; SSE2-NOPCLMUL-NEXT: pushq %r13
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 40
+; SSE2-NOPCLMUL-NEXT: pushq %r12
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 48
+; SSE2-NOPCLMUL-NEXT: pushq %rbx
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 56
+; SSE2-NOPCLMUL-NEXT: subq $560, %rsp # imm = 0x230
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 616
+; SSE2-NOPCLMUL-NEXT: .cfi_offset %rbx, -56
+; SSE2-NOPCLMUL-NEXT: .cfi_offset %r12, -48
+; SSE2-NOPCLMUL-NEXT: .cfi_offset %r13, -40
+; SSE2-NOPCLMUL-NEXT: .cfi_offset %r14, -32
+; SSE2-NOPCLMUL-NEXT: .cfi_offset %r15, -24
+; SSE2-NOPCLMUL-NEXT: .cfi_offset %rbp, -16
+; SSE2-NOPCLMUL-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE2-NOPCLMUL-NEXT: movq %xmm1, %r12
+; SSE2-NOPCLMUL-NEXT: movabsq $1085102592571150095, %rdx # imm = 0xF0F0F0F0F0F0F0F
+; SSE2-NOPCLMUL-NEXT: movabsq $3689348814741910323, %rdi # imm = 0x3333333333333333
+; SSE2-NOPCLMUL-NEXT: movabsq $6148914691236517205, %rsi # imm = 0x5555555555555555
+; SSE2-NOPCLMUL-NEXT: movq %r12, %rax
+; SSE2-NOPCLMUL-NEXT: bswapq %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: shrq $4, %rcx
+; SSE2-NOPCLMUL-NEXT: andq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: andq %rdx, %rax
+; SSE2-NOPCLMUL-NEXT: shlq $4, %rax
+; SSE2-NOPCLMUL-NEXT: orq %rcx, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: andq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: shrq $2, %rax
+; SSE2-NOPCLMUL-NEXT: andq %rdi, %rax
+; SSE2-NOPCLMUL-NEXT: leaq (%rax,%rcx,4), %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: andq %rsi, %rcx
+; SSE2-NOPCLMUL-NEXT: shrq %rax
+; SSE2-NOPCLMUL-NEXT: andq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: leaq (%rax,%rcx,2), %rsi
+; SSE2-NOPCLMUL-NEXT: movabsq $4294967296, %rcx # imm = 0x100000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $8589934592, %rdi # imm = 0x200000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %rdi, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $17179869184, %rdx # imm = 0x400000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %rdx, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $34359738368, %r11 # imm = 0x800000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r11, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $68719476736, %r8 # imm = 0x1000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $137438953472, %r8 # imm = 0x2000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $274877906944, %r8 # imm = 0x4000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $549755813888, %r8 # imm = 0x8000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $1099511627776, %r8 # imm = 0x10000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $2199023255552, %r8 # imm = 0x20000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $4398046511104, %r8 # imm = 0x40000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $8796093022208, %r8 # imm = 0x80000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $17592186044416, %r8 # imm = 0x100000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $35184372088832, %r8 # imm = 0x200000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $70368744177664, %r8 # imm = 0x400000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $140737488355328, %r8 # imm = 0x800000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $281474976710656, %r8 # imm = 0x1000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $562949953421312, %r8 # imm = 0x2000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $1125899906842624, %r8 # imm = 0x4000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $2251799813685248, %r8 # imm = 0x8000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $4503599627370496, %r8 # imm = 0x10000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $9007199254740992, %r8 # imm = 0x20000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $18014398509481984, %r8 # imm = 0x40000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $36028797018963968, %r8 # imm = 0x80000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $72057594037927936, %r8 # imm = 0x100000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $144115188075855872, %r8 # imm = 0x200000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $288230376151711744, %r8 # imm = 0x400000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $576460752303423488, %r8 # imm = 0x800000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $1152921504606846976, %r8 # imm = 0x1000000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $2305843009213693952, %r8 # imm = 0x2000000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $4611686018427387904, %r8 # imm = 0x4000000000000000
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: andq %r8, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $2, %r8d
+; SSE2-NOPCLMUL-NEXT: movq %xmm0, %rax
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $1, %r8d
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $4, %r8d
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $8, %r8d
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $16, %r8d
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $32, %r8d
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %ebp
+; SSE2-NOPCLMUL-NEXT: andl $64, %ebp
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rbp
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $128, %r8d
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $256, %r8d # imm = 0x100
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $512, %r8d # imm = 0x200
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r15d
+; SSE2-NOPCLMUL-NEXT: andl $1024, %r15d # imm = 0x400
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r15
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $2048, %r8d # imm = 0x800
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $4096, %r8d # imm = 0x1000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $8192, %r8d # imm = 0x2000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $16384, %r8d # imm = 0x4000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r10d
+; SSE2-NOPCLMUL-NEXT: andl $32768, %r10d # imm = 0x8000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r10
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $65536, %r8d # imm = 0x10000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $131072, %r8d # imm = 0x20000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $262144, %r8d # imm = 0x40000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $524288, %r8d # imm = 0x80000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $1048576, %r8d # imm = 0x100000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r9d
+; SSE2-NOPCLMUL-NEXT: andl $2097152, %r9d # imm = 0x200000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r9
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $4194304, %r8d # imm = 0x400000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $8388608, %r8d # imm = 0x800000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $16777216, %r8d # imm = 0x1000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $33554432, %r8d # imm = 0x2000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $67108864, %r8d # imm = 0x4000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r13d
+; SSE2-NOPCLMUL-NEXT: andl $134217728, %r13d # imm = 0x8000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r13
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r8d
+; SSE2-NOPCLMUL-NEXT: andl $268435456, %r8d # imm = 0x10000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r8
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %ebx
+; SSE2-NOPCLMUL-NEXT: andl $536870912, %ebx # imm = 0x20000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rbx
+; SSE2-NOPCLMUL-NEXT: movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %r14d
+; SSE2-NOPCLMUL-NEXT: andl $1073741824, %r14d # imm = 0x40000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r14
+; SSE2-NOPCLMUL-NEXT: movl %r12d, %ebx
+; SSE2-NOPCLMUL-NEXT: andl $-2147483648, %ebx # imm = 0x80000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rbx
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rdi
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rdx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: andq %r12, %r11
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r11
+; SSE2-NOPCLMUL-NEXT: movq %r11, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $68719476736, %rcx # imm = 0x1000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $137438953472, %rcx # imm = 0x2000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $274877906944, %rcx # imm = 0x4000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $549755813888, %rcx # imm = 0x8000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $1099511627776, %rcx # imm = 0x10000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $2199023255552, %rcx # imm = 0x20000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $4398046511104, %rcx # imm = 0x40000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $8796093022208, %rcx # imm = 0x80000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $17592186044416, %rcx # imm = 0x100000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $35184372088832, %rcx # imm = 0x200000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $70368744177664, %rcx # imm = 0x400000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, (%rsp) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $140737488355328, %rcx # imm = 0x800000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $281474976710656, %rcx # imm = 0x1000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $562949953421312, %rcx # imm = 0x2000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $1125899906842624, %rcx # imm = 0x4000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $2251799813685248, %rcx # imm = 0x8000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $4503599627370496, %rcx # imm = 0x10000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $9007199254740992, %rcx # imm = 0x20000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $18014398509481984, %rcx # imm = 0x40000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $36028797018963968, %rcx # imm = 0x80000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $72057594037927936, %rcx # imm = 0x100000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $144115188075855872, %rcx # imm = 0x200000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $288230376151711744, %rcx # imm = 0x400000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $576460752303423488, %rcx # imm = 0x800000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $1152921504606846976, %rcx # imm = 0x1000000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $2305843009213693952, %rcx # imm = 0x2000000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $4611686018427387904, %rcx # imm = 0x4000000000000000
+; SSE2-NOPCLMUL-NEXT: andq %r12, %rcx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NOPCLMUL-NEXT: movabsq $-9223372036854775808, %rcx # imm = 0x8000000000000000
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %r12
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %r12
+; SSE2-NOPCLMUL-NEXT: bswapq %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: shrq $4, %rdi
+; SSE2-NOPCLMUL-NEXT: movabsq $1085102592571150095, %rcx # imm = 0xF0F0F0F0F0F0F0F
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rax
+; SSE2-NOPCLMUL-NEXT: shlq $4, %rax
+; SSE2-NOPCLMUL-NEXT: orq %rdi, %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: movabsq $3689348814741910323, %rcx # imm = 0x3333333333333333
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: shrq $2, %rax
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rax
+; SSE2-NOPCLMUL-NEXT: leaq (%rax,%rdi,4), %rax
+; SSE2-NOPCLMUL-NEXT: movq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: movabsq $6148914691236517205, %r11 # imm = 0x5555555555555555
+; SSE2-NOPCLMUL-NEXT: andq %r11, %rdi
+; SSE2-NOPCLMUL-NEXT: shrq %rax
+; SSE2-NOPCLMUL-NEXT: andq %r11, %rax
+; SSE2-NOPCLMUL-NEXT: leaq (%rax,%rdi,2), %rax
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $2, %edi
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $1, %ecx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $4, %edi
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $8, %edx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $16, %ecx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $32, %edi
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $64, %ecx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $128, %edx
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $256, %edi # imm = 0x100
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $512, %edx # imm = 0x200
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $1024, %edi # imm = 0x400
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $2048, %ecx # imm = 0x800
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $4096, %edx # imm = 0x1000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $8192, %ecx # imm = 0x2000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $16384, %edx # imm = 0x4000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $32768, %ecx # imm = 0x8000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $65536, %edx # imm = 0x10000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $131072, %edi # imm = 0x20000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $262144, %edx # imm = 0x40000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $524288, %edi # imm = 0x80000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $1048576, %edx # imm = 0x100000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $2097152, %edi # imm = 0x200000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $4194304, %ecx # imm = 0x400000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $8388608, %edx # imm = 0x800000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $16777216, %ecx # imm = 0x1000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $33554432, %edx # imm = 0x2000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $67108864, %ecx # imm = 0x4000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $134217728, %edx # imm = 0x8000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %ecx
+; SSE2-NOPCLMUL-NEXT: andl $268435456, %ecx # imm = 0x10000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $536870912, %edx # imm = 0x20000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edi
+; SSE2-NOPCLMUL-NEXT: andl $1073741824, %edi # imm = 0x40000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movl %esi, %edx
+; SSE2-NOPCLMUL-NEXT: andl $-2147483648, %edx # imm = 0x80000000
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq %rdx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rdi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rcx
+; SSE2-NOPCLMUL-NEXT: movabsq $-9223372036854775808, %rdi # imm = 0x8000000000000000
+; SSE2-NOPCLMUL-NEXT: andq %rdi, %rsi
+; SSE2-NOPCLMUL-NEXT: imulq %rax, %rsi
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: bswapq %rsi
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: shrq $4, %rax
+; SSE2-NOPCLMUL-NEXT: movabsq $1085102592571150095, %rcx # imm = 0xF0F0F0F0F0F0F0F
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rax
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: shlq $4, %rsi
+; SSE2-NOPCLMUL-NEXT: orq %rax, %rsi
+; SSE2-NOPCLMUL-NEXT: movq %rsi, %rax
+; SSE2-NOPCLMUL-NEXT: movabsq $3689348814741910323, %rcx # imm = 0x3333333333333333
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rax
+; SSE2-NOPCLMUL-NEXT: shrq $2, %rsi
+; SSE2-NOPCLMUL-NEXT: andq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: leaq (%rsi,%rax,4), %rax
+; SSE2-NOPCLMUL-NEXT: andq %rax, %r11
+; SSE2-NOPCLMUL-NEXT: shrq %rax
+; SSE2-NOPCLMUL-NEXT: movabsq $6148914691236517204, %rcx # imm = 0x5555555555555554
+; SSE2-NOPCLMUL-NEXT: andq %rax, %rcx
+; SSE2-NOPCLMUL-NEXT: leaq (%rcx,%r11,2), %rax
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rbp
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rbp
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %r15
+; SSE2-NOPCLMUL-NEXT: xorq %rbp, %r15
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %r10
+; SSE2-NOPCLMUL-NEXT: xorq %r15, %r10
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %r9
+; SSE2-NOPCLMUL-NEXT: xorq %r10, %r9
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %r13
+; SSE2-NOPCLMUL-NEXT: xorq %r13, %r8
+; SSE2-NOPCLMUL-NEXT: xorq %r9, %r8
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %r14 # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: xorq %r14, %rbx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rbx, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %r8, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rdi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdi, %rdx
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq (%rsp), %rcx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rcx
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rcx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %rdx
+; SSE2-NOPCLMUL-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; SSE2-NOPCLMUL-NEXT: xorq %rdx, %rsi
+; SSE2-NOPCLMUL-NEXT: xorq %rsi, %r12
+; SSE2-NOPCLMUL-NEXT: xorq %rcx, %r12
+; SSE2-NOPCLMUL-NEXT: movq %rax, %xmm1
+; SSE2-NOPCLMUL-NEXT: movq %r12, %xmm0
+; SSE2-NOPCLMUL-NEXT: psrlq $1, %xmm1
+; SSE2-NOPCLMUL-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE2-NOPCLMUL-NEXT: addq $560, %rsp # imm = 0x230
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 56
+; SSE2-NOPCLMUL-NEXT: popq %rbx
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 48
+; SSE2-NOPCLMUL-NEXT: popq %r12
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 40
+; SSE2-NOPCLMUL-NEXT: popq %r13
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 32
+; SSE2-NOPCLMUL-NEXT: popq %r14
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 24
+; SSE2-NOPCLMUL-NEXT: popq %r15
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 16
+; SSE2-NOPCLMUL-NEXT: popq %rbp
+; SSE2-NOPCLMUL-NEXT: .cfi_def_cfa_offset 8
+; SSE2-NOPCLMUL-NEXT: retq
+;
+; SSE-PCLMUL-LABEL: pclmul:
+; SSE-PCLMUL: # %bb.0:
+; SSE-PCLMUL-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE-PCLMUL-NEXT: pclmulqdq $0, %xmm1, %xmm0
+; SSE-PCLMUL-NEXT: retq
+;
+; AVX2-LABEL: pclmul:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-NEXT: vpclmulqdq $0, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: pclmul:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX512-NEXT: vpclmulqdq $0, %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: retq
+ %i0 = zext i1 0 to i64 ; constant time lo/hi select
+ %i1 = zext i1 1 to i64 ; constant time lo/hi select
+ %a0 = extractelement <2 x i64> %v0, i64 %i0
+ %a1 = extractelement <2 x i64> %v1, i64 %i1
+ %x0 = zext i64 %a0 to i128
+ %x1 = zext i64 %a1 to i128
+ %cl = call i128 @llvm.clmul.i128(i128 %x0, i128 %x1)
+ %r = bitcast i128 %cl to <2 x i64>
+ ret <2 x i64> %r
+}
More information about the llvm-commits
mailing list