if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
return NewVSel;
+ // Narrow a suitable binary operation with a constant operand by moving it
+ // ahead of the truncate. This is limited to pre-legalization because targets
+ // may prefer a wider type during later combines and invert this transform.
+ switch (N0.getOpcode()) {
+ // TODO: Add case for ADD - that will likely require a change in logic here
+ // or target-specific changes to avoid regressions.
+ case ISD::SUB:
+ case ISD::MUL:
+ case ISD::AND:
+ case ISD::OR:
+ case ISD::XOR:
+ // TODO: This should allow vector constants/types too.
+ if (!LegalOperations && N0.hasOneUse() &&
+ (isa<ConstantSDNode>(N0.getOperand(0)) ||
+ isa<ConstantSDNode>(N0.getOperand(1)))) {
+ SDLoc DL(N);
+ SDValue NarrowL = DAG.getNode(ISD::TRUNCATE, DL, VT, N0.getOperand(0));
+ SDValue NarrowR = DAG.getNode(ISD::TRUNCATE, DL, VT, N0.getOperand(1));
+ return DAG.getNode(N0.getOpcode(), DL, VT, NarrowL, NarrowR);
+ }
+ }
+
return SDValue();
}
; GCN: s_cbranch_scc1
; SI: s_bfe_u32 s{{[0-9]+}}, s{{[0-9]+}}, 0x80004
-; VI: s_and_b32 s{{[0-9]+}}, [[BFE]], 0xff
+; VI: v_mov_b32_e32 v{{[0-9]+}}, 0xff
; GCN: BB2_2:
; SI: s_bfe_u32 s{{[0-9]+}}, s{{[0-9]+}}, 0x70004
-; VI: s_and_b32 s{{[0-9]+}}, [[BFE]], 0x7f
+; VI: v_mov_b32_e32 v{{[0-9]+}}, 0x7f
; GCN: BB2_3:
; GCN: buffer_store_short
; SOURCE-SCHED-NEXT: setg %cl
; SOURCE-SCHED-NEXT: movb g_73, %dl
; SOURCE-SCHED-NEXT: xorl %eax, %eax
-; SOURCE-SCHED-NEXT: subl {{[0-9]+}}(%esp), %eax
+; SOURCE-SCHED-NEXT: subb {{[0-9]+}}(%esp), %al
; SOURCE-SCHED-NEXT: testb %dl, %dl
; SOURCE-SCHED-NEXT: jne .LBB0_2
; SOURCE-SCHED-NEXT: # %bb.1: # %bb11
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: cmpq {{.*}}(%rip), %rax
-; CHECK-NEXT: sbbl %eax, %eax
+; CHECK-NEXT: sbbb %al, %al
; CHECK-NEXT: testb $-106, %al
; CHECK-NEXT: jle .LBB0_1
; CHECK-NEXT: # %bb.2: # %if.then
; CHECK-LABEL: PR30841:
; CHECK: ## %bb.0: ## %entry
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT: negl %eax
+; CHECK-NEXT: negb %al
; CHECK-NEXT: ## kill: def $al killed $al killed $eax
; CHECK-NEXT: retl
entry:
; CHECK-LABEL: sub_zext_cmp_mask_narrower_result:
; CHECK: # %bb.0:
; CHECK-NEXT: movl %edi, %eax
-; CHECK-NEXT: andl $1, %eax
+; CHECK-NEXT: andb $1, %al
; CHECK-NEXT: orb $46, %al
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: retq
; CHECK-LABEL: add_zext_cmp_mask_narrower_result:
; CHECK: # %bb.0:
; CHECK-NEXT: movl %edi, %eax
-; CHECK-NEXT: andl $1, %eax
+; CHECK-NEXT: andb $1, %al
; CHECK-NEXT: xorb $43, %al
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: retq
; CHECK-LABEL: low_bit_select_constants_bigger_true_narrower_result:
; CHECK: # %bb.0:
; CHECK-NEXT: movl %edi, %eax
-; CHECK-NEXT: andl $1, %eax
+; CHECK-NEXT: andb $1, %al
; CHECK-NEXT: xorb $41, %al
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: retq
; X86-NOBMI2-LABEL: clear_lowbits16_ic0:
; X86-NOBMI2: # %bb.0:
; X86-NOBMI2-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; X86-NOBMI2-NEXT: movw $16, %cx
-; X86-NOBMI2-NEXT: subw {{[0-9]+}}(%esp), %cx
+; X86-NOBMI2-NEXT: movb $16, %cl
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: shrl %cl, %eax
-; X86-NOBMI2-NEXT: # kill: def $cl killed $cl killed $cx
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-NOBMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NOBMI2-NEXT: retl
; X86-BMI2-LABEL: clear_lowbits16_ic0:
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; X86-BMI2-NEXT: movw $16, %cx
-; X86-BMI2-NEXT: subw {{[0-9]+}}(%esp), %cx
+; X86-BMI2-NEXT: movb $16, %cl
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: shrxl %ecx, %eax, %eax
; X86-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X86-BMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NOBMI2-LABEL: clear_lowbits16_ic0:
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movzwl %di, %eax
-; X64-NOBMI2-NEXT: movl $16, %ecx
-; X64-NOBMI2-NEXT: subl %esi, %ecx
+; X64-NOBMI2-NEXT: movb $16, %cl
+; X64-NOBMI2-NEXT: subb %sil, %cl
; X64-NOBMI2-NEXT: shrl %cl, %eax
-; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI2-NEXT: shll %cl, %eax
; X64-NOBMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NOBMI2-NEXT: retq
; X64-BMI2-LABEL: clear_lowbits16_ic0:
; X64-BMI2: # %bb.0:
; X64-BMI2-NEXT: movzwl %di, %eax
-; X64-BMI2-NEXT: movl $16, %ecx
-; X64-BMI2-NEXT: subl %esi, %ecx
+; X64-BMI2-NEXT: movb $16, %cl
+; X64-BMI2-NEXT: subb %sil, %cl
; X64-BMI2-NEXT: shrxl %ecx, %eax, %eax
; X64-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X64-BMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NOBMI2: # %bb.0:
; X86-NOBMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI2-NEXT: movzwl (%eax), %eax
-; X86-NOBMI2-NEXT: movw $16, %cx
-; X86-NOBMI2-NEXT: subw {{[0-9]+}}(%esp), %cx
+; X86-NOBMI2-NEXT: movb $16, %cl
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: shrl %cl, %eax
-; X86-NOBMI2-NEXT: # kill: def $cl killed $cl killed $cx
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-NOBMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NOBMI2-NEXT: retl
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI2-NEXT: movzwl (%eax), %eax
-; X86-BMI2-NEXT: movw $16, %cx
-; X86-BMI2-NEXT: subw {{[0-9]+}}(%esp), %cx
+; X86-BMI2-NEXT: movb $16, %cl
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: shrxl %ecx, %eax, %eax
; X86-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X86-BMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NOBMI2-LABEL: clear_lowbits16_ic2_load:
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movzwl (%rdi), %eax
-; X64-NOBMI2-NEXT: movl $16, %ecx
-; X64-NOBMI2-NEXT: subl %esi, %ecx
+; X64-NOBMI2-NEXT: movb $16, %cl
+; X64-NOBMI2-NEXT: subb %sil, %cl
; X64-NOBMI2-NEXT: shrl %cl, %eax
-; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI2-NEXT: shll %cl, %eax
; X64-NOBMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NOBMI2-NEXT: retq
; X64-BMI2-LABEL: clear_lowbits16_ic2_load:
; X64-BMI2: # %bb.0:
; X64-BMI2-NEXT: movzwl (%rdi), %eax
-; X64-BMI2-NEXT: movl $16, %ecx
-; X64-BMI2-NEXT: subl %esi, %ecx
+; X64-BMI2-NEXT: movb $16, %cl
+; X64-BMI2-NEXT: subb %sil, %cl
; X64-BMI2-NEXT: shrxl %ecx, %eax, %eax
; X64-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X64-BMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NOBMI2-LABEL: clear_lowbits16_ic4_commutative:
; X86-NOBMI2: # %bb.0:
; X86-NOBMI2-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; X86-NOBMI2-NEXT: movw $16, %cx
-; X86-NOBMI2-NEXT: subw {{[0-9]+}}(%esp), %cx
+; X86-NOBMI2-NEXT: movb $16, %cl
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: shrl %cl, %eax
-; X86-NOBMI2-NEXT: # kill: def $cl killed $cl killed $cx
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-NOBMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NOBMI2-NEXT: retl
; X86-BMI2-LABEL: clear_lowbits16_ic4_commutative:
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; X86-BMI2-NEXT: movw $16, %cx
-; X86-BMI2-NEXT: subw {{[0-9]+}}(%esp), %cx
+; X86-BMI2-NEXT: movb $16, %cl
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: shrxl %ecx, %eax, %eax
; X86-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X86-BMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NOBMI2-LABEL: clear_lowbits16_ic4_commutative:
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movzwl %di, %eax
-; X64-NOBMI2-NEXT: movl $16, %ecx
-; X64-NOBMI2-NEXT: subl %esi, %ecx
+; X64-NOBMI2-NEXT: movb $16, %cl
+; X64-NOBMI2-NEXT: subb %sil, %cl
; X64-NOBMI2-NEXT: shrl %cl, %eax
-; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI2-NEXT: shll %cl, %eax
; X64-NOBMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NOBMI2-NEXT: retq
; X64-BMI2-LABEL: clear_lowbits16_ic4_commutative:
; X64-BMI2: # %bb.0:
; X64-BMI2-NEXT: movzwl %di, %eax
-; X64-BMI2-NEXT: movl $16, %ecx
-; X64-BMI2-NEXT: subl %esi, %ecx
+; X64-BMI2-NEXT: movb $16, %cl
+; X64-BMI2-NEXT: subb %sil, %cl
; X64-BMI2-NEXT: shrxl %ecx, %eax, %eax
; X64-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X64-BMI2-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NOBMI2: # %bb.0:
; X86-NOBMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI2-NEXT: xorl %ecx, %ecx
-; X86-NOBMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: shrl %cl, %eax
; X86-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-BMI2-LABEL: clear_lowbits32_ic0:
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: xorl %eax, %eax
-; X86-BMI2-NEXT: subl {{[0-9]+}}(%esp), %eax
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %al
; X86-BMI2-NEXT: shrxl %eax, {{[0-9]+}}(%esp), %ecx
; X86-BMI2-NEXT: shlxl %eax, %ecx, %eax
; X86-BMI2-NEXT: retl
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movl %esi, %ecx
; X64-NOBMI2-NEXT: movl %edi, %eax
-; X64-NOBMI2-NEXT: negl %ecx
+; X64-NOBMI2-NEXT: negb %cl
; X64-NOBMI2-NEXT: shrl %cl, %eax
; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI2-NEXT: shll %cl, %eax
;
; X64-BMI2-LABEL: clear_lowbits32_ic0:
; X64-BMI2: # %bb.0:
-; X64-BMI2-NEXT: negl %esi
+; X64-BMI2-NEXT: negb %sil
; X64-BMI2-NEXT: shrxl %esi, %edi, %eax
; X64-BMI2-NEXT: shlxl %esi, %eax, %eax
; X64-BMI2-NEXT: retq
; X86-NOBMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI2-NEXT: movl (%eax), %eax
; X86-NOBMI2-NEXT: xorl %ecx, %ecx
-; X86-NOBMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: shrl %cl, %eax
; X86-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI2-NEXT: xorl %ecx, %ecx
-; X86-BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: shrxl %ecx, (%eax), %eax
; X86-BMI2-NEXT: shlxl %ecx, %eax, %eax
; X86-BMI2-NEXT: retl
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movl %esi, %ecx
; X64-NOBMI2-NEXT: movl (%rdi), %eax
-; X64-NOBMI2-NEXT: negl %ecx
+; X64-NOBMI2-NEXT: negb %cl
; X64-NOBMI2-NEXT: shrl %cl, %eax
; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI2-NEXT: shll %cl, %eax
;
; X64-BMI2-LABEL: clear_lowbits32_ic2_load:
; X64-BMI2: # %bb.0:
-; X64-BMI2-NEXT: negl %esi
+; X64-BMI2-NEXT: negb %sil
; X64-BMI2-NEXT: shrxl %esi, (%rdi), %eax
; X64-BMI2-NEXT: shlxl %esi, %eax, %eax
; X64-BMI2-NEXT: retq
; X86-NOBMI2: # %bb.0:
; X86-NOBMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI2-NEXT: xorl %ecx, %ecx
-; X86-NOBMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: shrl %cl, %eax
; X86-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-BMI2-LABEL: clear_lowbits32_ic4_commutative:
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: xorl %eax, %eax
-; X86-BMI2-NEXT: subl {{[0-9]+}}(%esp), %eax
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %al
; X86-BMI2-NEXT: shrxl %eax, {{[0-9]+}}(%esp), %ecx
; X86-BMI2-NEXT: shlxl %eax, %ecx, %eax
; X86-BMI2-NEXT: retl
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movl %esi, %ecx
; X64-NOBMI2-NEXT: movl %edi, %eax
-; X64-NOBMI2-NEXT: negl %ecx
+; X64-NOBMI2-NEXT: negb %cl
; X64-NOBMI2-NEXT: shrl %cl, %eax
; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI2-NEXT: shll %cl, %eax
;
; X64-BMI2-LABEL: clear_lowbits32_ic4_commutative:
; X64-BMI2: # %bb.0:
-; X64-BMI2-NEXT: negl %esi
+; X64-BMI2-NEXT: negb %sil
; X64-BMI2-NEXT: shrxl %esi, %edi, %eax
; X64-BMI2-NEXT: shlxl %esi, %eax, %eax
; X64-BMI2-NEXT: retq
define i64 @clear_lowbits64_ic0(i64 %val, i64 %numlowbits) nounwind {
; X86-NOBMI2-LABEL: clear_lowbits64_ic0:
; X86-NOBMI2: # %bb.0:
-; X86-NOBMI2-NEXT: movl $64, %ecx
-; X86-NOBMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI2-NEXT: movb $64, %cl
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: movl $-1, %edx
; X86-NOBMI2-NEXT: movl $-1, %eax
; X86-NOBMI2-NEXT: shll %cl, %eax
;
; X86-BMI2-LABEL: clear_lowbits64_ic0:
; X86-BMI2: # %bb.0:
-; X86-BMI2-NEXT: movl $64, %ecx
-; X86-BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI2-NEXT: movb $64, %cl
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: movl $-1, %edx
; X86-BMI2-NEXT: shlxl %ecx, %edx, %eax
; X86-BMI2-NEXT: shldl %cl, %edx, %edx
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movq %rsi, %rcx
; X64-NOBMI2-NEXT: movq %rdi, %rax
-; X64-NOBMI2-NEXT: negl %ecx
+; X64-NOBMI2-NEXT: negb %cl
; X64-NOBMI2-NEXT: shrq %cl, %rax
; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI2-NEXT: shlq %cl, %rax
;
; X64-BMI2-LABEL: clear_lowbits64_ic0:
; X64-BMI2: # %bb.0:
-; X64-BMI2-NEXT: negl %esi
+; X64-BMI2-NEXT: negb %sil
; X64-BMI2-NEXT: shrxq %rsi, %rdi, %rax
; X64-BMI2-NEXT: shlxq %rsi, %rax, %rax
; X64-BMI2-NEXT: retq
; X86-NOBMI2: # %bb.0:
; X86-NOBMI2-NEXT: pushl %esi
; X86-NOBMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-NOBMI2-NEXT: movl $64, %ecx
-; X86-NOBMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI2-NEXT: movb $64, %cl
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: movl $-1, %edx
; X86-NOBMI2-NEXT: movl $-1, %eax
; X86-NOBMI2-NEXT: shll %cl, %eax
; X86-BMI2: # %bb.0:
; X86-BMI2-NEXT: pushl %esi
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-BMI2-NEXT: movl $64, %ecx
-; X86-BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI2-NEXT: movb $64, %cl
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: movl $-1, %edx
; X86-BMI2-NEXT: shlxl %ecx, %edx, %eax
; X86-BMI2-NEXT: shldl %cl, %edx, %edx
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movq %rsi, %rcx
; X64-NOBMI2-NEXT: movq (%rdi), %rax
-; X64-NOBMI2-NEXT: negl %ecx
+; X64-NOBMI2-NEXT: negb %cl
; X64-NOBMI2-NEXT: shrq %cl, %rax
; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI2-NEXT: shlq %cl, %rax
;
; X64-BMI2-LABEL: clear_lowbits64_ic2_load:
; X64-BMI2: # %bb.0:
-; X64-BMI2-NEXT: negl %esi
+; X64-BMI2-NEXT: negb %sil
; X64-BMI2-NEXT: shrxq %rsi, (%rdi), %rax
; X64-BMI2-NEXT: shlxq %rsi, %rax, %rax
; X64-BMI2-NEXT: retq
define i64 @clear_lowbits64_ic4_commutative(i64 %val, i64 %numlowbits) nounwind {
; X86-NOBMI2-LABEL: clear_lowbits64_ic4_commutative:
; X86-NOBMI2: # %bb.0:
-; X86-NOBMI2-NEXT: movl $64, %ecx
-; X86-NOBMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI2-NEXT: movb $64, %cl
+; X86-NOBMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI2-NEXT: movl $-1, %edx
; X86-NOBMI2-NEXT: movl $-1, %eax
; X86-NOBMI2-NEXT: shll %cl, %eax
;
; X86-BMI2-LABEL: clear_lowbits64_ic4_commutative:
; X86-BMI2: # %bb.0:
-; X86-BMI2-NEXT: movl $64, %ecx
-; X86-BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI2-NEXT: movb $64, %cl
+; X86-BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI2-NEXT: movl $-1, %edx
; X86-BMI2-NEXT: shlxl %ecx, %edx, %eax
; X86-BMI2-NEXT: shldl %cl, %edx, %edx
; X64-NOBMI2: # %bb.0:
; X64-NOBMI2-NEXT: movq %rsi, %rcx
; X64-NOBMI2-NEXT: movq %rdi, %rax
-; X64-NOBMI2-NEXT: negl %ecx
+; X64-NOBMI2-NEXT: negb %cl
; X64-NOBMI2-NEXT: shrq %cl, %rax
; X64-NOBMI2-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI2-NEXT: shlq %cl, %rax
;
; X64-BMI2-LABEL: clear_lowbits64_ic4_commutative:
; X64-BMI2: # %bb.0:
-; X64-BMI2-NEXT: negl %esi
+; X64-BMI2-NEXT: negb %sil
; X64-BMI2-NEXT: shrxq %rsi, %rdi, %rax
; X64-BMI2-NEXT: shlxq %rsi, %rax, %rax
; X64-BMI2-NEXT: retq
; CHECK-NEXT: movsbl {{.*}}(%rip), %edx
; CHECK-NEXT: movzbl %dl, %ecx
; CHECK-NEXT: shrl $7, %ecx
-; CHECK-NEXT: xorl $1, %ecx
+; CHECK-NEXT: xorb $1, %cl
; CHECK-NEXT: # kill: def $cl killed $cl killed $ecx
; CHECK-NEXT: sarl %cl, %edx
; CHECK-NEXT: movb {{.*}}(%rip), %al
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-NOBMI-NEXT: shrl %cl, %edi
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-BMI1NOTBM-NEXT: shrl %cl, %edi
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %esi
;
; X86-BMI1BMI2-LABEL: bextr32_c0:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: pushl %edi
+; X86-BMI1BMI2-NEXT: pushl %ebx
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: pushl %eax
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %bl
; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
-; X86-BMI1BMI2-NEXT: shrxl %eax, {{[0-9]+}}(%esp), %edi
-; X86-BMI1BMI2-NEXT: movl %esi, %eax
-; X86-BMI1BMI2-NEXT: negl %eax
+; X86-BMI1BMI2-NEXT: shrxl %eax, {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
+; X86-BMI1BMI2-NEXT: negb %al
; X86-BMI1BMI2-NEXT: movl $-1, %ecx
; X86-BMI1BMI2-NEXT: shrxl %eax, %ecx, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
-; X86-BMI1BMI2-NEXT: bzhil %esi, %edi, %eax
+; X86-BMI1BMI2-NEXT: bzhil %ebx, %esi, %eax
; X86-BMI1BMI2-NEXT: addl $4, %esp
; X86-BMI1BMI2-NEXT: popl %esi
-; X86-BMI1BMI2-NEXT: popl %edi
+; X86-BMI1BMI2-NEXT: popl %ebx
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI-LABEL: bextr32_c0:
; X64-NOBMI-NEXT: movl %edi, %ebx
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %ebx
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl $-1, %ebp
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrl %cl, %ebp
; X64-BMI1NOTBM-NEXT: movl %edi, %ebx
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebx
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movl $-1, %ebp
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebp
; X64-BMI1BMI2-NEXT: pushq %rax
; X64-BMI1BMI2-NEXT: movl %edx, %ebx
; X64-BMI1BMI2-NEXT: shrxl %esi, %edi, %ebp
-; X64-BMI1BMI2-NEXT: movl %edx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: movl %ebx, %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movl $-1, %ecx
; X64-BMI1BMI2-NEXT: shrxl %eax, %ecx, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: movl (%eax), %edi
; X86-NOBMI-NEXT: shrl %cl, %edi
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM-NEXT: movl (%eax), %edi
; X86-BMI1NOTBM-NEXT: shrl %cl, %edi
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %esi
;
; X86-BMI1BMI2-LABEL: bextr32_c2_load:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: pushl %edi
+; X86-BMI1BMI2-NEXT: pushl %ebx
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: pushl %eax
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %bl
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1BMI2-NEXT: shrxl %ecx, (%eax), %edi
-; X86-BMI1BMI2-NEXT: movl %esi, %eax
-; X86-BMI1BMI2-NEXT: negl %eax
+; X86-BMI1BMI2-NEXT: shrxl %ecx, (%eax), %esi
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
+; X86-BMI1BMI2-NEXT: negb %al
; X86-BMI1BMI2-NEXT: movl $-1, %ecx
; X86-BMI1BMI2-NEXT: shrxl %eax, %ecx, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
-; X86-BMI1BMI2-NEXT: bzhil %esi, %edi, %eax
+; X86-BMI1BMI2-NEXT: bzhil %ebx, %esi, %eax
; X86-BMI1BMI2-NEXT: addl $4, %esp
; X86-BMI1BMI2-NEXT: popl %esi
-; X86-BMI1BMI2-NEXT: popl %edi
+; X86-BMI1BMI2-NEXT: popl %ebx
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI-LABEL: bextr32_c2_load:
; X64-NOBMI-NEXT: movl (%rdi), %ebp
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %ebp
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl $-1, %ebx
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrl %cl, %ebx
; X64-BMI1NOTBM-NEXT: movl (%rdi), %ebp
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebp
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movl $-1, %ebx
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebx
; X64-BMI1BMI2-NEXT: pushq %rax
; X64-BMI1BMI2-NEXT: movl %edx, %ebx
; X64-BMI1BMI2-NEXT: shrxl %esi, (%rdi), %ebp
-; X64-BMI1BMI2-NEXT: movl %edx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: movl %ebx, %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movl $-1, %ecx
; X64-BMI1BMI2-NEXT: shrxl %eax, %ecx, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-NOBMI-NEXT: shrl %cl, %edi
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-BMI1NOTBM-NEXT: shrl %cl, %edi
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %esi
;
; X86-BMI1BMI2-LABEL: bextr32_c4_commutative:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: pushl %edi
+; X86-BMI1BMI2-NEXT: pushl %ebx
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: pushl %eax
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %bl
; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
-; X86-BMI1BMI2-NEXT: shrxl %eax, {{[0-9]+}}(%esp), %edi
-; X86-BMI1BMI2-NEXT: movl %esi, %eax
-; X86-BMI1BMI2-NEXT: negl %eax
+; X86-BMI1BMI2-NEXT: shrxl %eax, {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
+; X86-BMI1BMI2-NEXT: negb %al
; X86-BMI1BMI2-NEXT: movl $-1, %ecx
; X86-BMI1BMI2-NEXT: shrxl %eax, %ecx, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
-; X86-BMI1BMI2-NEXT: bzhil %esi, %edi, %eax
+; X86-BMI1BMI2-NEXT: bzhil %ebx, %esi, %eax
; X86-BMI1BMI2-NEXT: addl $4, %esp
; X86-BMI1BMI2-NEXT: popl %esi
-; X86-BMI1BMI2-NEXT: popl %edi
+; X86-BMI1BMI2-NEXT: popl %ebx
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI-LABEL: bextr32_c4_commutative:
; X64-NOBMI-NEXT: movl %edi, %ebx
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %ebx
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl $-1, %ebp
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrl %cl, %ebp
; X64-BMI1NOTBM-NEXT: movl %edi, %ebx
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebx
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movl $-1, %ebp
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebp
; X64-BMI1BMI2-NEXT: pushq %rax
; X64-BMI1BMI2-NEXT: movl %edx, %ebx
; X64-BMI1BMI2-NEXT: shrxl %esi, %edi, %ebp
-; X64-BMI1BMI2-NEXT: movl %edx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: movl %ebx, %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movl $-1, %ecx
; X64-BMI1BMI2-NEXT: shrxl %eax, %ecx, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: movl %ebx, %ecx
; X86-NOBMI-NEXT: shrl %cl, %edi
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM-NEXT: movl %ebx, %ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %edi
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %esi
; X86-BMI1BMI2-NEXT: pushl %edi
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: subl $16, %esp
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %bl
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %edi
-; X86-BMI1BMI2-NEXT: shrxl %edi, {{[0-9]+}}(%esp), %ebx
-; X86-BMI1BMI2-NEXT: movl %esi, %eax
-; X86-BMI1BMI2-NEXT: negl %eax
+; X86-BMI1BMI2-NEXT: shrxl %edi, {{[0-9]+}}(%esp), %esi
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
+; X86-BMI1BMI2-NEXT: negb %al
; X86-BMI1BMI2-NEXT: movl $-1, %ecx
; X86-BMI1BMI2-NEXT: shrxl %eax, %ecx, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
-; X86-BMI1BMI2-NEXT: bzhil %esi, %ebx, %esi
+; X86-BMI1BMI2-NEXT: bzhil %ebx, %esi, %esi
; X86-BMI1BMI2-NEXT: movl %edi, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
; X86-BMI1BMI2-NEXT: movl %esi, %eax
; X64-NOBMI-NEXT: movl %edi, %ebp
; X64-NOBMI-NEXT: movl %r14d, %ecx
; X64-NOBMI-NEXT: shrl %cl, %ebp
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl $-1, %ebx
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrl %cl, %ebx
; X64-BMI1NOTBM-NEXT: movl %edi, %ebp
; X64-BMI1NOTBM-NEXT: movl %r14d, %ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebp
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movl $-1, %ebx
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebx
; X64-BMI1BMI2-NEXT: movl %edx, %ebx
; X64-BMI1BMI2-NEXT: movl %esi, %ebp
; X64-BMI1BMI2-NEXT: shrxl %esi, %edi, %r14d
-; X64-BMI1BMI2-NEXT: movl %edx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: movl %ebx, %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movl $-1, %ecx
; X64-BMI1BMI2-NEXT: shrxl %eax, %ecx, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: movl %edi, %esi
; X86-NOBMI-NEXT: xorl %edi, %edi
; X86-NOBMI-NEXT: .LBB32_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %ebp
; X86-NOBMI-NEXT: movl $-1, %ebx
; X86-NOBMI-NEXT: shrl %cl, %ebx
; X86-BMI1NOTBM-NEXT: movl %edi, %esi
; X86-BMI1NOTBM-NEXT: xorl %edi, %edi
; X86-BMI1NOTBM-NEXT: .LBB32_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %ebp
; X86-BMI1NOTBM-NEXT: movl $-1, %ebx
; X86-BMI1NOTBM-NEXT: shrl %cl, %ebx
; X86-BMI1BMI2-NEXT: movl %edi, %esi
; X86-BMI1BMI2-NEXT: xorl %edi, %edi
; X86-BMI1BMI2-NEXT: .LBB32_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: movl $-1, %ebx
; X86-BMI1BMI2-NEXT: shrxl %ecx, %ebx, %ebp
; X86-BMI1BMI2-NEXT: shrdl %cl, %ebx, %ebx
; X64-NOBMI-NEXT: movq %rdi, %r14
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %r14
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movq $-1, %rbx
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X64-BMI1NOTBM-NEXT: movq %rdi, %r14
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-BMI1NOTBM-NEXT: shrq %cl, %r14
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movq $-1, %rbx
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rbx
; X64-BMI1BMI2-NEXT: movq %rdx, %rbx
; X64-BMI1BMI2-NEXT: shrxq %rsi, %rdi, %r14
; X64-BMI1BMI2-NEXT: movl %ebx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movq $-1, %rcx
; X64-BMI1BMI2-NEXT: shrxq %rax, %rcx, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI-NEXT: movl %edi, %esi
; X86-NOBMI-NEXT: xorl %edi, %edi
; X86-NOBMI-NEXT: .LBB34_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %ebp
; X86-NOBMI-NEXT: movl $-1, %ebx
; X86-NOBMI-NEXT: shrl %cl, %ebx
; X86-BMI1NOTBM-NEXT: movl %edi, %esi
; X86-BMI1NOTBM-NEXT: xorl %edi, %edi
; X86-BMI1NOTBM-NEXT: .LBB34_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %ebp
; X86-BMI1NOTBM-NEXT: movl $-1, %ebx
; X86-BMI1NOTBM-NEXT: shrl %cl, %ebx
; X86-BMI1BMI2-NEXT: movl %edi, %esi
; X86-BMI1BMI2-NEXT: xorl %edi, %edi
; X86-BMI1BMI2-NEXT: .LBB34_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: movl $-1, %ebx
; X86-BMI1BMI2-NEXT: shrxl %ecx, %ebx, %ebp
; X86-BMI1BMI2-NEXT: shrdl %cl, %ebx, %ebx
; X64-NOBMI-NEXT: movq (%rdi), %r14
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %r14
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movq $-1, %rbx
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X64-BMI1NOTBM-NEXT: movq (%rdi), %r14
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-BMI1NOTBM-NEXT: shrq %cl, %r14
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movq $-1, %rbx
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rbx
; X64-BMI1BMI2-NEXT: movq %rdx, %rbx
; X64-BMI1BMI2-NEXT: shrxq %rsi, (%rdi), %r14
; X64-BMI1BMI2-NEXT: movl %ebx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movq $-1, %rcx
; X64-BMI1BMI2-NEXT: shrxq %rax, %rcx, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI-NEXT: movl %edi, %esi
; X86-NOBMI-NEXT: xorl %edi, %edi
; X86-NOBMI-NEXT: .LBB36_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %ebp
; X86-NOBMI-NEXT: movl $-1, %ebx
; X86-NOBMI-NEXT: shrl %cl, %ebx
; X86-BMI1NOTBM-NEXT: movl %edi, %esi
; X86-BMI1NOTBM-NEXT: xorl %edi, %edi
; X86-BMI1NOTBM-NEXT: .LBB36_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %ebp
; X86-BMI1NOTBM-NEXT: movl $-1, %ebx
; X86-BMI1NOTBM-NEXT: shrl %cl, %ebx
; X86-BMI1BMI2-NEXT: movl %edi, %esi
; X86-BMI1BMI2-NEXT: xorl %edi, %edi
; X86-BMI1BMI2-NEXT: .LBB36_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: movl $-1, %ebx
; X86-BMI1BMI2-NEXT: shrxl %ecx, %ebx, %ebp
; X86-BMI1BMI2-NEXT: shrdl %cl, %ebx, %ebx
; X64-NOBMI-NEXT: movq %rdi, %r14
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %r14
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movq $-1, %rbx
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X64-BMI1NOTBM-NEXT: movq %rdi, %r14
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-BMI1NOTBM-NEXT: shrq %cl, %r14
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movq $-1, %rbx
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rbx
; X64-BMI1BMI2-NEXT: movq %rdx, %rbx
; X64-BMI1BMI2-NEXT: shrxq %rsi, %rdi, %r14
; X64-BMI1BMI2-NEXT: movl %ebx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movq $-1, %rcx
; X64-BMI1BMI2-NEXT: shrxq %rax, %rcx, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI-NEXT: movl %edi, %esi
; X86-NOBMI-NEXT: xorl %edi, %edi
; X86-NOBMI-NEXT: .LBB37_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %ebx
; X86-NOBMI-NEXT: movl $-1, %ebp
; X86-NOBMI-NEXT: shrl %cl, %ebp
; X86-BMI1NOTBM-NEXT: movl %edi, %esi
; X86-BMI1NOTBM-NEXT: xorl %edi, %edi
; X86-BMI1NOTBM-NEXT: .LBB37_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %ebx
; X86-BMI1NOTBM-NEXT: movl $-1, %ebp
; X86-BMI1NOTBM-NEXT: shrl %cl, %ebp
; X86-BMI1BMI2-NEXT: movl %edi, %esi
; X86-BMI1BMI2-NEXT: xorl %edi, %edi
; X86-BMI1BMI2-NEXT: .LBB37_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: movl $-1, %ebp
; X86-BMI1BMI2-NEXT: shrxl %ecx, %ebp, %ebx
; X86-BMI1BMI2-NEXT: shrdl %cl, %ebp, %ebp
; X64-NOBMI-NEXT: movq %rdi, %r15
; X64-NOBMI-NEXT: movl %r14d, %ecx
; X64-NOBMI-NEXT: shrq %cl, %r15
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movq $-1, %rbx
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X64-BMI1NOTBM-NEXT: movq %rdi, %r15
; X64-BMI1NOTBM-NEXT: movl %r14d, %ecx
; X64-BMI1NOTBM-NEXT: shrq %cl, %r15
-; X64-BMI1NOTBM-NEXT: negl %edx
+; X64-BMI1NOTBM-NEXT: negb %dl
; X64-BMI1NOTBM-NEXT: movq $-1, %rbx
; X64-BMI1NOTBM-NEXT: movl %edx, %ecx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rbx
; X64-BMI1BMI2-NEXT: movq %rsi, %r14
; X64-BMI1BMI2-NEXT: shrxq %rsi, %rdi, %r15
; X64-BMI1BMI2-NEXT: movl %ebx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movq $-1, %rcx
; X64-BMI1BMI2-NEXT: shrxq %rax, %rcx, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI-NEXT: shrl %cl, %eax
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shll %cl, %eax
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %eax
;
; X86-BMI1NOTBM-LABEL: bextr32_d0:
; X86-BMI1NOTBM: # %bb.0:
-; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT: movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI1NOTBM-NEXT: shll $8, %eax
-; X86-BMI1NOTBM-NEXT: orl %ecx, %eax
-; X86-BMI1NOTBM-NEXT: bextrl %eax, {{[0-9]+}}(%esp), %eax
+; X86-BMI1NOTBM-NEXT: movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT: bextrl %ecx, {{[0-9]+}}(%esp), %eax
; X86-BMI1NOTBM-NEXT: retl
;
; X86-BMI1BMI2-LABEL: bextr32_d0:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: shrxl %ecx, {{[0-9]+}}(%esp), %ecx
; X86-BMI1BMI2-NEXT: bzhil %eax, %ecx, %eax
; X64-NOBMI-NEXT: movl %edi, %eax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %eax
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shll %cl, %eax
; X64-NOBMI-NEXT: shrl %cl, %eax
; X86-NOBMI-NEXT: movl (%eax), %eax
; X86-NOBMI-NEXT: shrl %cl, %eax
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shll %cl, %eax
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %eax
; X86-BMI1NOTBM-LABEL: bextr32_d2_load:
; X86-BMI1NOTBM: # %bb.0:
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1NOTBM-NEXT: movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: shll $8, %ecx
-; X86-BMI1NOTBM-NEXT: orl %edx, %ecx
-; X86-BMI1NOTBM-NEXT: bextrl %ecx, (%eax), %eax
+; X86-BMI1NOTBM-NEXT: movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT: orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT: bextrl %edx, (%eax), %eax
; X86-BMI1NOTBM-NEXT: retl
;
; X86-BMI1BMI2-LABEL: bextr32_d2_load:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %dl
; X86-BMI1BMI2-NEXT: shrxl %edx, (%ecx), %ecx
; X64-NOBMI-NEXT: movl (%rdi), %eax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %eax
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shll %cl, %eax
; X64-NOBMI-NEXT: shrl %cl, %eax
; X86-NOBMI-NEXT: movl %eax, %ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shll %cl, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM: # %bb.0:
; X86-BMI1NOTBM-NEXT: pushl %esi
; X86-BMI1NOTBM-NEXT: subl $8, %esp
-; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI1NOTBM-NEXT: shll $8, %ecx
; X86-BMI1NOTBM-NEXT: movzbl %al, %edx
; X86-BMI1BMI2: # %bb.0:
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: subl $8, %esp
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-BMI1BMI2-NEXT: shrxl %ecx, {{[0-9]+}}(%esp), %edx
; X86-BMI1BMI2-NEXT: bzhil %eax, %edx, %esi
; X64-NOBMI-NEXT: movl %edi, %ebx
; X64-NOBMI-NEXT: movl %esi, %ecx
; X64-NOBMI-NEXT: shrl %cl, %ebx
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shll %cl, %ebx
; X64-NOBMI-NEXT: shrl %cl, %ebx
; X86-NOBMI-NEXT: movl %eax, %edi
; X86-NOBMI-NEXT: xorl %eax, %eax
; X86-NOBMI-NEXT: .LBB43_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shldl %cl, %edi, %eax
; X86-NOBMI-NEXT: shll %cl, %edi
; X86-NOBMI-NEXT: testb $32, %cl
; X86-BMI1NOTBM-NEXT: movl %eax, %edi
; X86-BMI1NOTBM-NEXT: xorl %eax, %eax
; X86-BMI1NOTBM-NEXT: .LBB43_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: shldl %cl, %edi, %eax
; X86-BMI1NOTBM-NEXT: shll %cl, %edi
; X86-BMI1NOTBM-NEXT: testb $32, %cl
; X86-BMI1BMI2-NEXT: movl %esi, %eax
; X86-BMI1BMI2-NEXT: xorl %esi, %esi
; X86-BMI1BMI2-NEXT: .LBB43_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: shldl %cl, %eax, %esi
; X86-BMI1BMI2-NEXT: shlxl %ecx, %eax, %edi
; X86-BMI1BMI2-NEXT: testb $32, %cl
; X64-NOBMI-NEXT: movq %rdi, %rax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rax
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shlq %cl, %rax
; X64-NOBMI-NEXT: shrq %cl, %rax
; X86-NOBMI-NEXT: movl %eax, %edi
; X86-NOBMI-NEXT: xorl %eax, %eax
; X86-NOBMI-NEXT: .LBB45_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shldl %cl, %edi, %eax
; X86-NOBMI-NEXT: shll %cl, %edi
; X86-NOBMI-NEXT: testb $32, %cl
; X86-BMI1NOTBM-NEXT: movl %eax, %edi
; X86-BMI1NOTBM-NEXT: xorl %eax, %eax
; X86-BMI1NOTBM-NEXT: .LBB45_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: shldl %cl, %edi, %eax
; X86-BMI1NOTBM-NEXT: shll %cl, %edi
; X86-BMI1NOTBM-NEXT: testb $32, %cl
; X86-BMI1BMI2-NEXT: movl %esi, %eax
; X86-BMI1BMI2-NEXT: xorl %esi, %esi
; X86-BMI1BMI2-NEXT: .LBB45_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: shldl %cl, %eax, %esi
; X86-BMI1BMI2-NEXT: shlxl %ecx, %eax, %edi
; X86-BMI1BMI2-NEXT: testb $32, %cl
; X64-NOBMI-NEXT: movq (%rdi), %rax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rax
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shlq %cl, %rax
; X64-NOBMI-NEXT: shrq %cl, %rax
; X86-NOBMI-NEXT: movl %esi, %ebx
; X86-NOBMI-NEXT: xorl %esi, %esi
; X86-NOBMI-NEXT: .LBB47_2:
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shldl %cl, %ebx, %esi
; X86-NOBMI-NEXT: shll %cl, %ebx
; X86-NOBMI-NEXT: testb $32, %cl
; X86-BMI1NOTBM-NEXT: movl %esi, %ebx
; X86-BMI1NOTBM-NEXT: xorl %esi, %esi
; X86-BMI1NOTBM-NEXT: .LBB47_2:
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: shldl %cl, %ebx, %esi
; X86-BMI1NOTBM-NEXT: shll %cl, %ebx
; X86-BMI1NOTBM-NEXT: testb $32, %cl
; X86-BMI1BMI2-NEXT: movl %edx, %edi
; X86-BMI1BMI2-NEXT: xorl %edx, %edx
; X86-BMI1BMI2-NEXT: .LBB47_2:
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: shldl %cl, %edi, %edx
; X86-BMI1BMI2-NEXT: shlxl %ecx, %edi, %ebx
; X86-BMI1BMI2-NEXT: testb $32, %cl
; X64-NOBMI-NEXT: movq %rdi, %rbx
; X64-NOBMI-NEXT: movl %esi, %ecx
; X64-NOBMI-NEXT: shrq %cl, %rbx
-; X64-NOBMI-NEXT: negl %edx
+; X64-NOBMI-NEXT: negb %dl
; X64-NOBMI-NEXT: movl %edx, %ecx
; X64-NOBMI-NEXT: shlq %cl, %rbx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X86-NOBMI-NEXT: pushl %esi
; X86-NOBMI-NEXT: subl $8, %esp
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM-NEXT: pushl %esi
; X86-BMI1NOTBM-NEXT: subl $8, %esp
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %esi
;
; X86-BMI1BMI2-LABEL: bzhi32_c0:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: pushl %esi
+; X86-BMI1BMI2-NEXT: pushl %ebx
; X86-BMI1BMI2-NEXT: subl $8, %esp
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-BMI1BMI2-NEXT: movl %esi, %eax
-; X86-BMI1BMI2-NEXT: negl %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %bl
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
+; X86-BMI1BMI2-NEXT: negb %al
; X86-BMI1BMI2-NEXT: movl $-1, %ecx
; X86-BMI1BMI2-NEXT: shrxl %eax, %ecx, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
-; X86-BMI1BMI2-NEXT: bzhil %esi, {{[0-9]+}}(%esp), %eax
+; X86-BMI1BMI2-NEXT: bzhil %ebx, {{[0-9]+}}(%esp), %eax
; X86-BMI1BMI2-NEXT: addl $8, %esp
-; X86-BMI1BMI2-NEXT: popl %esi
+; X86-BMI1BMI2-NEXT: popl %ebx
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI-LABEL: bzhi32_c0:
; X64-NOBMI-NEXT: pushq %rax
; X64-NOBMI-NEXT: movl %esi, %ecx
; X64-NOBMI-NEXT: movl %edi, %ebx
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: movl $-1, %ebp
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %ebp
; X64-BMI1NOTBM-NEXT: pushq %rax
; X64-BMI1NOTBM-NEXT: movl %esi, %ecx
; X64-BMI1NOTBM-NEXT: movl %edi, %ebx
-; X64-BMI1NOTBM-NEXT: negl %ecx
+; X64-BMI1NOTBM-NEXT: negb %cl
; X64-BMI1NOTBM-NEXT: movl $-1, %ebp
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebp
; X64-BMI1BMI2-NEXT: pushq %rax
; X64-BMI1BMI2-NEXT: movl %esi, %ebx
; X64-BMI1BMI2-NEXT: movl %edi, %ebp
-; X64-BMI1BMI2-NEXT: movl %esi, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: movl %ebx, %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movl $-1, %ecx
; X64-BMI1BMI2-NEXT: shrxl %eax, %ecx, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: subl $8, %esp
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %edx
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %edx
; X86-BMI1NOTBM-NEXT: subl $8, %esp
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %edx
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %edx
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: subl $8, %esp
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: bzhil %ecx, (%eax), %esi
-; X86-BMI1BMI2-NEXT: negl %ecx
+; X86-BMI1BMI2-NEXT: # kill: def $cl killed $cl killed $ecx def $ecx
+; X86-BMI1BMI2-NEXT: negb %cl
; X86-BMI1BMI2-NEXT: movl $-1, %eax
; X86-BMI1BMI2-NEXT: shrxl %ecx, %eax, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X64-NOBMI: # %bb.0:
; X64-NOBMI-NEXT: pushq %rbx
; X64-NOBMI-NEXT: movl %esi, %ecx
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: movl $-1, %eax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %eax
; X64-BMI1NOTBM: # %bb.0:
; X64-BMI1NOTBM-NEXT: pushq %rbx
; X64-BMI1NOTBM-NEXT: movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT: negl %ecx
+; X64-BMI1NOTBM-NEXT: negb %cl
; X64-BMI1NOTBM-NEXT: movl $-1, %eax
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %eax
; X64-BMI1BMI2: # %bb.0:
; X64-BMI1BMI2-NEXT: pushq %rbx
; X64-BMI1BMI2-NEXT: bzhil %esi, (%rdi), %ebx
-; X64-BMI1BMI2-NEXT: negl %esi
+; X64-BMI1BMI2-NEXT: # kill: def $sil killed $sil killed $esi def $esi
+; X64-BMI1BMI2-NEXT: negb %sil
; X64-BMI1BMI2-NEXT: movl $-1, %eax
; X64-BMI1BMI2-NEXT: shrxl %esi, %eax, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: pushl %esi
; X86-NOBMI-NEXT: subl $8, %esp
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %esi
; X86-BMI1NOTBM-NEXT: pushl %esi
; X86-BMI1NOTBM-NEXT: subl $8, %esp
; X86-BMI1NOTBM-NEXT: xorl %ecx, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-BMI1NOTBM-NEXT: shrl %cl, %esi
;
; X86-BMI1BMI2-LABEL: bzhi32_c4_commutative:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: pushl %esi
+; X86-BMI1BMI2-NEXT: pushl %ebx
; X86-BMI1BMI2-NEXT: subl $8, %esp
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-BMI1BMI2-NEXT: movl %esi, %eax
-; X86-BMI1BMI2-NEXT: negl %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %bl
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
+; X86-BMI1BMI2-NEXT: negb %al
; X86-BMI1BMI2-NEXT: movl $-1, %ecx
; X86-BMI1BMI2-NEXT: shrxl %eax, %ecx, %eax
; X86-BMI1BMI2-NEXT: movl %eax, (%esp)
; X86-BMI1BMI2-NEXT: calll use32
-; X86-BMI1BMI2-NEXT: bzhil %esi, {{[0-9]+}}(%esp), %eax
+; X86-BMI1BMI2-NEXT: bzhil %ebx, {{[0-9]+}}(%esp), %eax
; X86-BMI1BMI2-NEXT: addl $8, %esp
-; X86-BMI1BMI2-NEXT: popl %esi
+; X86-BMI1BMI2-NEXT: popl %ebx
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI-LABEL: bzhi32_c4_commutative:
; X64-NOBMI-NEXT: pushq %rax
; X64-NOBMI-NEXT: movl %esi, %ecx
; X64-NOBMI-NEXT: movl %edi, %ebx
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: movl $-1, %ebp
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %ebp
; X64-BMI1NOTBM-NEXT: pushq %rax
; X64-BMI1NOTBM-NEXT: movl %esi, %ecx
; X64-BMI1NOTBM-NEXT: movl %edi, %ebx
-; X64-BMI1NOTBM-NEXT: negl %ecx
+; X64-BMI1NOTBM-NEXT: negb %cl
; X64-BMI1NOTBM-NEXT: movl $-1, %ebp
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1NOTBM-NEXT: shrl %cl, %ebp
; X64-BMI1BMI2-NEXT: pushq %rax
; X64-BMI1BMI2-NEXT: movl %esi, %ebx
; X64-BMI1BMI2-NEXT: movl %edi, %ebp
-; X64-BMI1BMI2-NEXT: movl %esi, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: movl %ebx, %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movl $-1, %ecx
; X64-BMI1BMI2-NEXT: shrxl %eax, %ecx, %edi
; X64-BMI1BMI2-NEXT: callq use32
; X86-NOBMI-NEXT: pushl %edi
; X86-NOBMI-NEXT: pushl %esi
; X86-NOBMI-NEXT: pushl %eax
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: movl $-1, %edi
; X86-NOBMI-NEXT: shrl %cl, %edi
; X86-BMI1NOTBM-NEXT: pushl %edi
; X86-BMI1NOTBM-NEXT: pushl %esi
; X86-BMI1NOTBM-NEXT: pushl %eax
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: movl $-1, %edi
; X86-BMI1NOTBM-NEXT: shrl %cl, %edi
; X86-BMI1BMI2-NEXT: pushl %edi
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: pushl %eax
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: movl $-1, %esi
; X86-BMI1BMI2-NEXT: shrxl %ecx, %esi, %edi
; X86-BMI1BMI2-NEXT: shrdl %cl, %esi, %esi
; X64-NOBMI-NEXT: pushq %rax
; X64-NOBMI-NEXT: movq %rsi, %rcx
; X64-NOBMI-NEXT: movq %rdi, %r14
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: movq $-1, %rbx
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X64-BMI1NOTBM-NEXT: pushq %rax
; X64-BMI1NOTBM-NEXT: movq %rsi, %rcx
; X64-BMI1NOTBM-NEXT: movq %rdi, %r14
-; X64-BMI1NOTBM-NEXT: negl %ecx
+; X64-BMI1NOTBM-NEXT: negb %cl
; X64-BMI1NOTBM-NEXT: movq $-1, %rbx
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rbx
; X64-BMI1BMI2-NEXT: movq %rsi, %rbx
; X64-BMI1BMI2-NEXT: movq %rdi, %r14
; X64-BMI1BMI2-NEXT: movl %ebx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movq $-1, %rcx
; X64-BMI1BMI2-NEXT: shrxq %rax, %rcx, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI-NEXT: pushl %ebx
; X86-NOBMI-NEXT: pushl %edi
; X86-NOBMI-NEXT: pushl %esi
-; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
-; X86-NOBMI-NEXT: movl $-1, %edx
+; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %edx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
+; X86-NOBMI-NEXT: movl $-1, %eax
; X86-NOBMI-NEXT: movl $-1, %ebx
; X86-NOBMI-NEXT: shrl %cl, %ebx
-; X86-NOBMI-NEXT: shrdl %cl, %edx, %edx
+; X86-NOBMI-NEXT: shrdl %cl, %eax, %eax
; X86-NOBMI-NEXT: testb $32, %cl
; X86-NOBMI-NEXT: je .LBB27_2
; X86-NOBMI-NEXT: # %bb.1:
-; X86-NOBMI-NEXT: movl %ebx, %edx
+; X86-NOBMI-NEXT: movl %ebx, %eax
; X86-NOBMI-NEXT: xorl %ebx, %ebx
; X86-NOBMI-NEXT: .LBB27_2:
-; X86-NOBMI-NEXT: movl (%eax), %esi
-; X86-NOBMI-NEXT: andl %edx, %esi
-; X86-NOBMI-NEXT: movl 4(%eax), %edi
+; X86-NOBMI-NEXT: movl (%edx), %esi
+; X86-NOBMI-NEXT: andl %eax, %esi
+; X86-NOBMI-NEXT: movl 4(%edx), %edi
; X86-NOBMI-NEXT: andl %ebx, %edi
; X86-NOBMI-NEXT: subl $8, %esp
; X86-NOBMI-NEXT: pushl %ebx
-; X86-NOBMI-NEXT: pushl %edx
+; X86-NOBMI-NEXT: pushl %eax
; X86-NOBMI-NEXT: calll use64
; X86-NOBMI-NEXT: addl $16, %esp
; X86-NOBMI-NEXT: movl %esi, %eax
; X86-BMI1NOTBM-NEXT: pushl %ebx
; X86-BMI1NOTBM-NEXT: pushl %edi
; X86-BMI1NOTBM-NEXT: pushl %esi
-; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1NOTBM-NEXT: movl $-1, %edx
+; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
+; X86-BMI1NOTBM-NEXT: movl $-1, %eax
; X86-BMI1NOTBM-NEXT: movl $-1, %ebx
; X86-BMI1NOTBM-NEXT: shrl %cl, %ebx
-; X86-BMI1NOTBM-NEXT: shrdl %cl, %edx, %edx
+; X86-BMI1NOTBM-NEXT: shrdl %cl, %eax, %eax
; X86-BMI1NOTBM-NEXT: testb $32, %cl
; X86-BMI1NOTBM-NEXT: je .LBB27_2
; X86-BMI1NOTBM-NEXT: # %bb.1:
-; X86-BMI1NOTBM-NEXT: movl %ebx, %edx
+; X86-BMI1NOTBM-NEXT: movl %ebx, %eax
; X86-BMI1NOTBM-NEXT: xorl %ebx, %ebx
; X86-BMI1NOTBM-NEXT: .LBB27_2:
-; X86-BMI1NOTBM-NEXT: movl (%eax), %esi
-; X86-BMI1NOTBM-NEXT: andl %edx, %esi
-; X86-BMI1NOTBM-NEXT: movl 4(%eax), %edi
+; X86-BMI1NOTBM-NEXT: movl (%edx), %esi
+; X86-BMI1NOTBM-NEXT: andl %eax, %esi
+; X86-BMI1NOTBM-NEXT: movl 4(%edx), %edi
; X86-BMI1NOTBM-NEXT: andl %ebx, %edi
; X86-BMI1NOTBM-NEXT: subl $8, %esp
; X86-BMI1NOTBM-NEXT: pushl %ebx
-; X86-BMI1NOTBM-NEXT: pushl %edx
+; X86-BMI1NOTBM-NEXT: pushl %eax
; X86-BMI1NOTBM-NEXT: calll use64
; X86-BMI1NOTBM-NEXT: addl $16, %esp
; X86-BMI1NOTBM-NEXT: movl %esi, %eax
; X86-BMI1BMI2-NEXT: pushl %ebx
; X86-BMI1BMI2-NEXT: pushl %edi
; X86-BMI1BMI2-NEXT: pushl %esi
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1BMI2-NEXT: movl $-1, %edx
-; X86-BMI1BMI2-NEXT: shrxl %ecx, %edx, %ebx
-; X86-BMI1BMI2-NEXT: shrdl %cl, %edx, %edx
+; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %edx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
+; X86-BMI1BMI2-NEXT: movl $-1, %eax
+; X86-BMI1BMI2-NEXT: shrxl %ecx, %eax, %ebx
+; X86-BMI1BMI2-NEXT: shrdl %cl, %eax, %eax
; X86-BMI1BMI2-NEXT: testb $32, %cl
; X86-BMI1BMI2-NEXT: je .LBB27_2
; X86-BMI1BMI2-NEXT: # %bb.1:
-; X86-BMI1BMI2-NEXT: movl %ebx, %edx
+; X86-BMI1BMI2-NEXT: movl %ebx, %eax
; X86-BMI1BMI2-NEXT: xorl %ebx, %ebx
; X86-BMI1BMI2-NEXT: .LBB27_2:
-; X86-BMI1BMI2-NEXT: movl (%eax), %esi
-; X86-BMI1BMI2-NEXT: andl %edx, %esi
-; X86-BMI1BMI2-NEXT: movl 4(%eax), %edi
+; X86-BMI1BMI2-NEXT: movl (%edx), %esi
+; X86-BMI1BMI2-NEXT: andl %eax, %esi
+; X86-BMI1BMI2-NEXT: movl 4(%edx), %edi
; X86-BMI1BMI2-NEXT: andl %ebx, %edi
; X86-BMI1BMI2-NEXT: subl $8, %esp
; X86-BMI1BMI2-NEXT: pushl %ebx
-; X86-BMI1BMI2-NEXT: pushl %edx
+; X86-BMI1BMI2-NEXT: pushl %eax
; X86-BMI1BMI2-NEXT: calll use64
; X86-BMI1BMI2-NEXT: addl $16, %esp
; X86-BMI1BMI2-NEXT: movl %esi, %eax
; X64-NOBMI: # %bb.0:
; X64-NOBMI-NEXT: pushq %rbx
; X64-NOBMI-NEXT: movq %rsi, %rcx
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: movq $-1, %rax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rax
; X64-BMI1NOTBM: # %bb.0:
; X64-BMI1NOTBM-NEXT: pushq %rbx
; X64-BMI1NOTBM-NEXT: movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT: negl %ecx
+; X64-BMI1NOTBM-NEXT: negb %cl
; X64-BMI1NOTBM-NEXT: movq $-1, %rax
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rax
; X64-BMI1BMI2: # %bb.0:
; X64-BMI1BMI2-NEXT: pushq %rbx
; X64-BMI1BMI2-NEXT: bzhiq %rsi, (%rdi), %rbx
-; X64-BMI1BMI2-NEXT: # kill: def $esi killed $esi killed $rsi def $rsi
-; X64-BMI1BMI2-NEXT: negl %esi
+; X64-BMI1BMI2-NEXT: # kill: def $sil killed $sil killed $rsi def $rsi
+; X64-BMI1BMI2-NEXT: negb %sil
; X64-BMI1BMI2-NEXT: movq $-1, %rax
; X64-BMI1BMI2-NEXT: shrxq %rsi, %rax, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI-NEXT: pushl %edi
; X86-NOBMI-NEXT: pushl %esi
; X86-NOBMI-NEXT: pushl %eax
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl $-1, %esi
; X86-NOBMI-NEXT: movl $-1, %edi
; X86-NOBMI-NEXT: shrl %cl, %edi
; X86-BMI1NOTBM-NEXT: pushl %edi
; X86-BMI1NOTBM-NEXT: pushl %esi
; X86-BMI1NOTBM-NEXT: pushl %eax
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl $-1, %esi
; X86-BMI1NOTBM-NEXT: movl $-1, %edi
; X86-BMI1NOTBM-NEXT: shrl %cl, %edi
; X86-BMI1BMI2-NEXT: pushl %edi
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: pushl %eax
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: movl $-1, %esi
; X86-BMI1BMI2-NEXT: shrxl %ecx, %esi, %edi
; X86-BMI1BMI2-NEXT: shrdl %cl, %esi, %esi
; X64-NOBMI-NEXT: pushq %rax
; X64-NOBMI-NEXT: movq %rsi, %rcx
; X64-NOBMI-NEXT: movq %rdi, %r14
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: movq $-1, %rbx
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rbx
; X64-BMI1NOTBM-NEXT: pushq %rax
; X64-BMI1NOTBM-NEXT: movq %rsi, %rcx
; X64-BMI1NOTBM-NEXT: movq %rdi, %r14
-; X64-BMI1NOTBM-NEXT: negl %ecx
+; X64-BMI1NOTBM-NEXT: negb %cl
; X64-BMI1NOTBM-NEXT: movq $-1, %rbx
; X64-BMI1NOTBM-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-BMI1NOTBM-NEXT: shrq %cl, %rbx
; X64-BMI1BMI2-NEXT: movq %rsi, %rbx
; X64-BMI1BMI2-NEXT: movq %rdi, %r14
; X64-BMI1BMI2-NEXT: movl %ebx, %eax
-; X64-BMI1BMI2-NEXT: negl %eax
+; X64-BMI1BMI2-NEXT: negb %al
; X64-BMI1BMI2-NEXT: movq $-1, %rcx
; X64-BMI1BMI2-NEXT: shrxq %rax, %rcx, %rdi
; X64-BMI1BMI2-NEXT: callq use64
; X86-NOBMI: # %bb.0:
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shll %cl, %eax
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %eax
;
; X86-BMI1NOTBM-LABEL: bzhi32_d0:
; X86-BMI1NOTBM: # %bb.0:
-; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1NOTBM-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI1NOTBM-NEXT: shll $8, %eax
; X86-BMI1NOTBM-NEXT: bextrl %eax, {{[0-9]+}}(%esp), %eax
; X86-BMI1NOTBM-NEXT: retl
;
; X86-BMI1BMI2-LABEL: bzhi32_d0:
; X86-BMI1BMI2: # %bb.0:
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI1BMI2-NEXT: bzhil %eax, {{[0-9]+}}(%esp), %eax
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI: # %bb.0:
; X64-NOBMI-NEXT: movl %esi, %ecx
; X64-NOBMI-NEXT: movl %edi, %eax
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: shll %cl, %eax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %eax
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI-NEXT: movl (%eax), %eax
; X86-NOBMI-NEXT: xorl %ecx, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: shll %cl, %eax
; X86-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NOBMI-NEXT: shrl %cl, %eax
; X86-BMI1NOTBM-LABEL: bzhi32_d2_load:
; X86-BMI1NOTBM: # %bb.0:
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: shll $8, %ecx
; X86-BMI1NOTBM-NEXT: bextrl %ecx, (%eax), %eax
; X86-BMI1NOTBM-NEXT: retl
; X86-BMI1BMI2-LABEL: bzhi32_d2_load:
; X86-BMI1BMI2: # %bb.0:
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1BMI2-NEXT: bzhil %eax, (%ecx), %eax
+; X86-BMI1BMI2-NEXT: movb {{[0-9]+}}(%esp), %cl
+; X86-BMI1BMI2-NEXT: bzhil %ecx, (%eax), %eax
; X86-BMI1BMI2-NEXT: retl
;
; X64-NOBMI-LABEL: bzhi32_d2_load:
; X64-NOBMI: # %bb.0:
; X64-NOBMI-NEXT: movl %esi, %ecx
; X64-NOBMI-NEXT: movl (%rdi), %eax
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: shll %cl, %eax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NOBMI-NEXT: shrl %cl, %eax
; X86-NOBMI-NEXT: pushl %esi
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl %edx, %esi
; X86-NOBMI-NEXT: shll %cl, %esi
; X86-NOBMI-NEXT: shldl %cl, %edx, %eax
; X86-BMI1NOTBM-NEXT: pushl %esi
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl %edx, %esi
; X86-BMI1NOTBM-NEXT: shll %cl, %esi
; X86-BMI1NOTBM-NEXT: shldl %cl, %edx, %eax
; X86-BMI1BMI2-NEXT: pushl %esi
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: shldl %cl, %eax, %esi
; X86-BMI1BMI2-NEXT: shlxl %ecx, %eax, %edi
; X86-BMI1BMI2-NEXT: xorl %edx, %edx
; X64-NOBMI: # %bb.0:
; X64-NOBMI-NEXT: movq %rsi, %rcx
; X64-NOBMI-NEXT: movq %rdi, %rax
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: shlq %cl, %rax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rax
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NOBMI-NEXT: movl (%eax), %edx
; X86-NOBMI-NEXT: movl 4(%eax), %eax
-; X86-NOBMI-NEXT: movl $64, %ecx
-; X86-NOBMI-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-NOBMI-NEXT: movb $64, %cl
+; X86-NOBMI-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-NOBMI-NEXT: movl %edx, %esi
; X86-NOBMI-NEXT: shll %cl, %esi
; X86-NOBMI-NEXT: shldl %cl, %edx, %eax
; X86-BMI1NOTBM-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI1NOTBM-NEXT: movl (%eax), %edx
; X86-BMI1NOTBM-NEXT: movl 4(%eax), %eax
-; X86-BMI1NOTBM-NEXT: movl $64, %ecx
-; X86-BMI1NOTBM-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT: movb $64, %cl
+; X86-BMI1NOTBM-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1NOTBM-NEXT: movl %edx, %esi
; X86-BMI1NOTBM-NEXT: shll %cl, %esi
; X86-BMI1NOTBM-NEXT: shldl %cl, %edx, %eax
; X86-BMI1BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-BMI1BMI2-NEXT: movl (%eax), %edx
; X86-BMI1BMI2-NEXT: movl 4(%eax), %esi
-; X86-BMI1BMI2-NEXT: movl $64, %ecx
-; X86-BMI1BMI2-NEXT: subl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1BMI2-NEXT: movb $64, %cl
+; X86-BMI1BMI2-NEXT: subb {{[0-9]+}}(%esp), %cl
; X86-BMI1BMI2-NEXT: shldl %cl, %edx, %esi
; X86-BMI1BMI2-NEXT: shlxl %ecx, %edx, %edi
; X86-BMI1BMI2-NEXT: xorl %edx, %edx
; X64-NOBMI: # %bb.0:
; X64-NOBMI-NEXT: movq %rsi, %rcx
; X64-NOBMI-NEXT: movq (%rdi), %rax
-; X64-NOBMI-NEXT: negl %ecx
+; X64-NOBMI-NEXT: negb %cl
; X64-NOBMI-NEXT: shlq %cl, %rax
; X64-NOBMI-NEXT: # kill: def $cl killed $cl killed $rcx
; X64-NOBMI-NEXT: shrq %cl, %rax
; X86-NEXT: movl %edi, %ebp
; X86-NEXT: xorl %edi, %edi
; X86-NEXT: .LBB3_2:
-; X86-NEXT: movl $64, %ecx
-; X86-NEXT: subl %ebx, %ecx
+; X86-NEXT: movb $64, %cl
+; X86-NEXT: subb %bl, %cl
; X86-NEXT: movl %edx, %esi
; X86-NEXT: shrl %cl, %esi
; X86-NEXT: shrdl %cl, %edx, (%esp) # 4-byte Folded Spill
; X86-NEXT: pushl %eax
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl %eax, (%esp) # 4-byte Spill
+; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: andl $63, %ebx
-; X86-NEXT: movl $64, %ecx
-; X86-NEXT: subl %ebx, %ecx
+; X86-NEXT: movb $64, %cl
+; X86-NEXT: subb %bl, %cl
; X86-NEXT: movl %eax, %edi
; X86-NEXT: shll %cl, %edi
; X86-NEXT: shldl %cl, %eax, %esi
; X86-NEXT: testb $32, %cl
-; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-NEXT: je .LBB3_2
; X86-NEXT: # %bb.1:
; X86-NEXT: movl %edi, %esi
; X32-SSE2-NEXT: pushl %ebx
; X32-SSE2-NEXT: pushl %edi
; X32-SSE2-NEXT: pushl %esi
+; X32-SSE2-NEXT: movb {{[0-9]+}}(%esp), %cl
; X32-SSE2-NEXT: movl {{[0-9]+}}(%esp), %esi
; X32-SSE2-NEXT: movl {{[0-9]+}}(%esp), %edx
-; X32-SSE2-NEXT: movl {{[0-9]+}}(%esp), %ebx
-; X32-SSE2-NEXT: movl %ebx, %ecx
-; X32-SSE2-NEXT: andl $63, %ecx
; X32-SSE2-NEXT: movl %edx, %edi
; X32-SSE2-NEXT: shrl %cl, %edi
-; X32-SSE2-NEXT: movl %esi, %ebp
-; X32-SSE2-NEXT: shrdl %cl, %edx, %ebp
-; X32-SSE2-NEXT: xorl %eax, %eax
+; X32-SSE2-NEXT: movl %esi, %ebx
+; X32-SSE2-NEXT: shrdl %cl, %edx, %ebx
+; X32-SSE2-NEXT: xorl %ebp, %ebp
; X32-SSE2-NEXT: testb $32, %cl
-; X32-SSE2-NEXT: cmovnel %edi, %ebp
-; X32-SSE2-NEXT: cmovnel %eax, %edi
-; X32-SSE2-NEXT: negl %ebx
-; X32-SSE2-NEXT: andl $63, %ebx
+; X32-SSE2-NEXT: cmovnel %edi, %ebx
+; X32-SSE2-NEXT: cmovnel %ebp, %edi
+; X32-SSE2-NEXT: negb %cl
; X32-SSE2-NEXT: movl %esi, %eax
-; X32-SSE2-NEXT: movl %ebx, %ecx
; X32-SSE2-NEXT: shll %cl, %eax
; X32-SSE2-NEXT: shldl %cl, %esi, %edx
-; X32-SSE2-NEXT: testb $32, %bl
+; X32-SSE2-NEXT: testb $32, %cl
; X32-SSE2-NEXT: cmovnel %eax, %edx
-; X32-SSE2-NEXT: movl $0, %ecx
-; X32-SSE2-NEXT: cmovnel %ecx, %eax
-; X32-SSE2-NEXT: orl %ebp, %eax
+; X32-SSE2-NEXT: cmovnel %ebp, %eax
+; X32-SSE2-NEXT: orl %ebx, %eax
; X32-SSE2-NEXT: orl %edi, %edx
; X32-SSE2-NEXT: popl %esi
; X32-SSE2-NEXT: popl %edi
; X32-SSE2-NEXT: cmovnel %ebp, %eax
; X32-SSE2-NEXT: cmovnel %ecx, %ebp
; X32-SSE2-NEXT: xorl %edx, %edx
-; X32-SSE2-NEXT: movl $37, %ecx
-; X32-SSE2-NEXT: subl %ebx, %ecx
+; X32-SSE2-NEXT: movb $37, %cl
+; X32-SSE2-NEXT: subb %bl, %cl
; X32-SSE2-NEXT: shrdl %cl, %esi, %edi
; X32-SSE2-NEXT: shrl %cl, %esi
; X32-SSE2-NEXT: testb $32, %cl
; X32-SSE2-NEXT: calll __umoddi3
; X32-SSE2-NEXT: addl $16, %esp
; X32-SSE2-NEXT: movl %eax, %ebx
-; X32-SSE2-NEXT: movl $37, %ecx
-; X32-SSE2-NEXT: subl %eax, %ecx
+; X32-SSE2-NEXT: movb $37, %cl
+; X32-SSE2-NEXT: subb %bl, %cl
; X32-SSE2-NEXT: movl %ebp, %eax
; X32-SSE2-NEXT: shll %cl, %ebp
; X32-SSE2-NEXT: shldl %cl, %eax, %edi
; 686-NEXT: movzbl c, %eax
; 686-NEXT: xorl %ecx, %ecx
; 686-NEXT: testl %eax, %eax
-; 686-NEXT: setne %cl
-; 686-NEXT: testb %al, %al
; 686-NEXT: setne {{[0-9]+}}(%esp)
+; 686-NEXT: setne %cl
; 686-NEXT: xorl %edx, %edx
; 686-NEXT: cmpl %eax, %ecx
; 686-NEXT: setle %dl
; CHECK: ## %bb.0:
; CHECK-NEXT: movq (%rax), %rax
; CHECK-NEXT: vcvtsi2sdq %rax, %xmm0, %xmm1
-; CHECK-NEXT: ## kill: def $eax killed $eax killed $rax
-; CHECK-NEXT: andl $1, %eax
; CHECK-NEXT: kmovd %eax, %k1
; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-NEXT: vmovsd %xmm1, %xmm0, %xmm0 {%k1}
; X64-NEXT: movl %edx, %ecx
; X64-NEXT: movl %edi, %eax
; X64-NEXT: # kill: def $cl killed $cl killed $ecx
-; X64-NEXT: shldw %cl, %ax, %ax
+; X64-NEXT: rolw %cl, %ax
; X64-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NEXT: retq
%t0 = shl i16 %x, %z
; X64-NEXT: movl %edx, %ecx
; X64-NEXT: movl %edi, %eax
; X64-NEXT: # kill: def $cl killed $cl killed $ecx
-; X64-NEXT: shrdw %cl, %ax, %ax
+; X64-NEXT: rorw %cl, %ax
; X64-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NEXT: retq
%t0 = lshr i16 %x, %z
; X86-NEXT: pushl %ebx
; X86-NEXT: pushl %edi
; X86-NEXT: pushl %esi
+; X86-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: movl %esi, %eax
; X86-NEXT: shll %cl, %eax
; X86-NEXT: testb $32, %cl
-; X86-NEXT: movl $0, %ebx
+; X86-NEXT: movl $0, %edi
; X86-NEXT: jne .LBB28_2
; X86-NEXT: # %bb.1: # %entry
-; X86-NEXT: movl %eax, %ebx
+; X86-NEXT: movl %eax, %edi
; X86-NEXT: .LBB28_2: # %entry
-; X86-NEXT: movl $64, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: movl %edi, %eax
+; X86-NEXT: movb $64, %dl
+; X86-NEXT: subb %cl, %dl
+; X86-NEXT: movl %ebx, %eax
; X86-NEXT: movl %edx, %ecx
; X86-NEXT: shrl %cl, %eax
-; X86-NEXT: shrdl %cl, %edi, %esi
+; X86-NEXT: shrdl %cl, %ebx, %esi
; X86-NEXT: testb $32, %dl
; X86-NEXT: jne .LBB28_4
; X86-NEXT: # %bb.3: # %entry
; X86-NEXT: movl %esi, %eax
; X86-NEXT: .LBB28_4: # %entry
-; X86-NEXT: orl %ebx, %eax
+; X86-NEXT: orl %edi, %eax
; X86-NEXT: xorl %edx, %edx
; X86-NEXT: popl %esi
; X86-NEXT: popl %edi
; X86-NEXT: .cfi_offset %esi, -16
; X86-NEXT: .cfi_offset %edi, -12
; X86-NEXT: .cfi_offset %ebx, -8
+; X86-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl %esi, %eax
; X86-NEXT: shll %cl, %eax
; X86-NEXT: movl %edi, %edx
; X86-NEXT: movl %eax, %edx
; X86-NEXT: xorl %eax, %eax
; X86-NEXT: .LBB2_2:
-; X86-NEXT: negl %ecx
+; X86-NEXT: negb %cl
; X86-NEXT: movl %edi, %ebx
; X86-NEXT: shrl %cl, %ebx
; X86-NEXT: shrdl %cl, %edi, %esi
; X86-NEXT: .cfi_offset %esi, -16
; X86-NEXT: .cfi_offset %edi, -12
; X86-NEXT: .cfi_offset %ebx, -8
+; X86-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl %esi, %edx
; X86-NEXT: shrl %cl, %edx
; X86-NEXT: movl %edi, %eax
; X86-NEXT: movl %edx, %eax
; X86-NEXT: xorl %edx, %edx
; X86-NEXT: .LBB3_2:
-; X86-NEXT: negl %ecx
+; X86-NEXT: negb %cl
; X86-NEXT: movl %edi, %ebx
; X86-NEXT: shll %cl, %ebx
; X86-NEXT: shldl %cl, %edi, %esi
; X86-NEXT: .cfi_offset %edi, -16
; X86-NEXT: .cfi_offset %ebx, -12
; X86-NEXT: .cfi_offset %ebp, -8
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl (%eax), %edx
; X86-NEXT: movl 4(%eax), %ebx
; X86-NEXT: movl %esi, %edi
; X86-NEXT: xorl %esi, %esi
; X86-NEXT: .LBB6_2:
-; X86-NEXT: negl %ecx
+; X86-NEXT: negb %cl
; X86-NEXT: movl %ebx, %ebp
; X86-NEXT: shrl %cl, %ebp
; X86-NEXT: shrdl %cl, %ebx, %edx
; X86-NEXT: .cfi_offset %edi, -16
; X86-NEXT: .cfi_offset %ebx, -12
; X86-NEXT: .cfi_offset %ebp, -8
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl (%eax), %ebx
-; X86-NEXT: movl 4(%eax), %edx
-; X86-NEXT: movl %edx, %esi
-; X86-NEXT: shrl %cl, %esi
+; X86-NEXT: movl 4(%eax), %esi
+; X86-NEXT: movl %esi, %edx
+; X86-NEXT: shrl %cl, %edx
; X86-NEXT: movl %ebx, %edi
-; X86-NEXT: shrdl %cl, %edx, %edi
+; X86-NEXT: shrdl %cl, %esi, %edi
; X86-NEXT: testb $32, %cl
; X86-NEXT: je .LBB7_2
; X86-NEXT: # %bb.1:
-; X86-NEXT: movl %esi, %edi
-; X86-NEXT: xorl %esi, %esi
+; X86-NEXT: movl %edx, %edi
+; X86-NEXT: xorl %edx, %edx
; X86-NEXT: .LBB7_2:
-; X86-NEXT: negl %ecx
+; X86-NEXT: negb %cl
; X86-NEXT: movl %ebx, %ebp
; X86-NEXT: shll %cl, %ebp
-; X86-NEXT: shldl %cl, %ebx, %edx
+; X86-NEXT: shldl %cl, %ebx, %esi
; X86-NEXT: testb $32, %cl
; X86-NEXT: je .LBB7_4
; X86-NEXT: # %bb.3:
-; X86-NEXT: movl %ebp, %edx
+; X86-NEXT: movl %ebp, %esi
; X86-NEXT: xorl %ebp, %ebp
; X86-NEXT: .LBB7_4:
-; X86-NEXT: orl %edx, %esi
+; X86-NEXT: orl %esi, %edx
; X86-NEXT: orl %ebp, %edi
-; X86-NEXT: movl %esi, 4(%eax)
+; X86-NEXT: movl %edx, 4(%eax)
; X86-NEXT: movl %edi, (%eax)
; X86-NEXT: popl %esi
; X86-NEXT: .cfi_def_cfa_offset 16
; BDVER12-NEXT: movq %rdx, %rcx # sched: [1:0.50]
; BDVER12-NEXT: movq %rsi, %rax # sched: [1:0.50]
; BDVER12-NEXT: shlq %cl, %rdi # sched: [1:0.50]
-; BDVER12-NEXT: negl %ecx # sched: [1:0.50]
+; BDVER12-NEXT: negb %cl # sched: [1:0.50]
; BDVER12-NEXT: # kill: def $cl killed $cl killed $rcx
; BDVER12-NEXT: shrq %cl, %rax # sched: [1:0.50]
; BDVER12-NEXT: orq %rdi, %rax # sched: [1:0.50]
; BTVER2-NEXT: movq %rdx, %rcx # sched: [1:0.50]
; BTVER2-NEXT: movq %rsi, %rax # sched: [1:0.50]
; BTVER2-NEXT: shlq %cl, %rdi # sched: [1:0.50]
-; BTVER2-NEXT: negl %ecx # sched: [1:0.50]
+; BTVER2-NEXT: negb %cl # sched: [1:0.50]
; BTVER2-NEXT: # kill: def $cl killed $cl killed $rcx
; BTVER2-NEXT: shrq %cl, %rax # sched: [1:0.50]
; BTVER2-NEXT: orq %rdi, %rax # sched: [1:0.50]
; BDVER12-NEXT: movq %rdx, %rcx # sched: [1:0.50]
; BDVER12-NEXT: movq %rsi, %rax # sched: [1:0.50]
; BDVER12-NEXT: shrq %cl, %rdi # sched: [1:0.50]
-; BDVER12-NEXT: negl %ecx # sched: [1:0.50]
+; BDVER12-NEXT: negb %cl # sched: [1:0.50]
; BDVER12-NEXT: # kill: def $cl killed $cl killed $rcx
; BDVER12-NEXT: shlq %cl, %rax # sched: [1:0.50]
; BDVER12-NEXT: orq %rdi, %rax # sched: [1:0.50]
; BTVER2-NEXT: movq %rdx, %rcx # sched: [1:0.50]
; BTVER2-NEXT: movq %rsi, %rax # sched: [1:0.50]
; BTVER2-NEXT: shrq %cl, %rdi # sched: [1:0.50]
-; BTVER2-NEXT: negl %ecx # sched: [1:0.50]
+; BTVER2-NEXT: negb %cl # sched: [1:0.50]
; BTVER2-NEXT: # kill: def $cl killed $cl killed $rcx
; BTVER2-NEXT: shlq %cl, %rax # sched: [1:0.50]
; BTVER2-NEXT: orq %rdi, %rax # sched: [1:0.50]
; BDVER12-NEXT: movq {{.*}}(%rip), %rax # sched: [5:0.50]
; BDVER12-NEXT: movq %rsi, %rcx # sched: [1:0.50]
; BDVER12-NEXT: shlq %cl, %rax # sched: [1:0.50]
-; BDVER12-NEXT: negl %ecx # sched: [1:0.50]
+; BDVER12-NEXT: negb %cl # sched: [1:0.50]
; BDVER12-NEXT: # kill: def $cl killed $cl killed $rcx
; BDVER12-NEXT: shrq %cl, %rdi # sched: [1:0.50]
; BDVER12-NEXT: orq %rax, %rdi # sched: [1:0.50]
; BTVER2-NEXT: movq {{.*}}(%rip), %rax # sched: [5:1.00]
; BTVER2-NEXT: movq %rsi, %rcx # sched: [1:0.50]
; BTVER2-NEXT: shlq %cl, %rax # sched: [1:0.50]
-; BTVER2-NEXT: negl %ecx # sched: [1:0.50]
+; BTVER2-NEXT: negb %cl # sched: [1:0.50]
; BTVER2-NEXT: # kill: def $cl killed $cl killed $rcx
; BTVER2-NEXT: shrq %cl, %rdi # sched: [1:0.50]
; BTVER2-NEXT: orq %rax, %rdi # sched: [1:0.50]
; ILP-NEXT: pushq %rbx
; ILP-NEXT: movq %rdi, %rax
; ILP-NEXT: xorl %r8d, %r8d
-; ILP-NEXT: leal 3(%rsi,%rsi), %r11d
+; ILP-NEXT: incl %esi
+; ILP-NEXT: addb %sil, %sil
+; ILP-NEXT: orb $1, %sil
; ILP-NEXT: movl $1, %r9d
; ILP-NEXT: xorl %r14d, %r14d
-; ILP-NEXT: movl %r11d, %ecx
+; ILP-NEXT: movl %esi, %ecx
; ILP-NEXT: shldq %cl, %r9, %r14
+; ILP-NEXT: movl $1, %edx
+; ILP-NEXT: shlq %cl, %rdx
+; ILP-NEXT: movl %esi, %r11d
+; ILP-NEXT: addb $-128, %r11b
+; ILP-NEXT: movb $-128, %r10b
+; ILP-NEXT: xorl %ebx, %ebx
+; ILP-NEXT: movl %r11d, %ecx
+; ILP-NEXT: shldq %cl, %r9, %rbx
+; ILP-NEXT: testb $64, %sil
+; ILP-NEXT: cmovneq %rdx, %r14
+; ILP-NEXT: cmovneq %r8, %rdx
; ILP-NEXT: movl $1, %edi
; ILP-NEXT: shlq %cl, %rdi
-; ILP-NEXT: movb $-128, %r10b
-; ILP-NEXT: subb %r11b, %r10b
-; ILP-NEXT: movl %r11d, %edx
-; ILP-NEXT: addb $-128, %dl
-; ILP-NEXT: xorl %esi, %esi
-; ILP-NEXT: movl %edx, %ecx
-; ILP-NEXT: shldq %cl, %r9, %rsi
-; ILP-NEXT: movl $1, %ebx
-; ILP-NEXT: shlq %cl, %rbx
+; ILP-NEXT: subb %sil, %r10b
; ILP-NEXT: movl %r10d, %ecx
; ILP-NEXT: shrdq %cl, %r8, %r9
-; ILP-NEXT: testb $64, %r11b
-; ILP-NEXT: cmovneq %rdi, %r14
-; ILP-NEXT: cmovneq %r8, %rdi
; ILP-NEXT: testb $64, %r10b
; ILP-NEXT: cmovneq %r8, %r9
-; ILP-NEXT: testb $64, %dl
-; ILP-NEXT: cmovneq %rbx, %rsi
-; ILP-NEXT: cmovneq %r8, %rbx
-; ILP-NEXT: testb %r11b, %r11b
+; ILP-NEXT: testb $64, %r11b
+; ILP-NEXT: cmovneq %rdi, %rbx
+; ILP-NEXT: cmovneq %r8, %rdi
+; ILP-NEXT: testb %sil, %sil
; ILP-NEXT: cmovsq %r8, %r14
-; ILP-NEXT: cmovsq %r8, %rdi
+; ILP-NEXT: cmovsq %r8, %rdx
; ILP-NEXT: movq %r14, 8(%rax)
-; ILP-NEXT: movq %rdi, (%rax)
-; ILP-NEXT: cmovnsq %r8, %rsi
-; ILP-NEXT: cmoveq %r8, %rsi
-; ILP-NEXT: movq %rsi, 24(%rax)
-; ILP-NEXT: cmovnsq %r9, %rbx
+; ILP-NEXT: movq %rdx, (%rax)
+; ILP-NEXT: cmovnsq %r8, %rbx
; ILP-NEXT: cmoveq %r8, %rbx
-; ILP-NEXT: movq %rbx, 16(%rax)
+; ILP-NEXT: movq %rbx, 24(%rax)
+; ILP-NEXT: cmovnsq %r9, %rdi
+; ILP-NEXT: cmoveq %r8, %rdi
+; ILP-NEXT: movq %rdi, 16(%rax)
; ILP-NEXT: popq %rbx
; ILP-NEXT: popq %r14
; ILP-NEXT: retq
; HYBRID-LABEL: test1:
; HYBRID: # %bb.0:
; HYBRID-NEXT: movq %rdi, %rax
-; HYBRID-NEXT: leal 3(%rsi,%rsi), %r10d
+; HYBRID-NEXT: incl %esi
+; HYBRID-NEXT: addb %sil, %sil
+; HYBRID-NEXT: orb $1, %sil
; HYBRID-NEXT: movb $-128, %cl
-; HYBRID-NEXT: subb %r10b, %cl
+; HYBRID-NEXT: subb %sil, %cl
; HYBRID-NEXT: xorl %r8d, %r8d
-; HYBRID-NEXT: movl $1, %esi
+; HYBRID-NEXT: movl $1, %r11d
; HYBRID-NEXT: movl $1, %r9d
; HYBRID-NEXT: shrdq %cl, %r8, %r9
; HYBRID-NEXT: testb $64, %cl
; HYBRID-NEXT: cmovneq %r8, %r9
-; HYBRID-NEXT: xorl %r11d, %r11d
-; HYBRID-NEXT: movl %r10d, %ecx
-; HYBRID-NEXT: shldq %cl, %rsi, %r11
+; HYBRID-NEXT: xorl %r10d, %r10d
+; HYBRID-NEXT: movl %esi, %ecx
+; HYBRID-NEXT: shldq %cl, %r11, %r10
; HYBRID-NEXT: addb $-128, %cl
-; HYBRID-NEXT: xorl %edx, %edx
-; HYBRID-NEXT: shldq %cl, %rsi, %rdx
-; HYBRID-NEXT: movl $1, %edi
-; HYBRID-NEXT: shlq %cl, %rdi
+; HYBRID-NEXT: xorl %edi, %edi
+; HYBRID-NEXT: shldq %cl, %r11, %rdi
+; HYBRID-NEXT: movl $1, %edx
+; HYBRID-NEXT: shlq %cl, %rdx
; HYBRID-NEXT: testb $64, %cl
-; HYBRID-NEXT: cmovneq %rdi, %rdx
-; HYBRID-NEXT: cmovneq %r8, %rdi
-; HYBRID-NEXT: movl %r10d, %ecx
-; HYBRID-NEXT: shlq %cl, %rsi
-; HYBRID-NEXT: testb $64, %r10b
-; HYBRID-NEXT: cmovneq %rsi, %r11
-; HYBRID-NEXT: cmovneq %r8, %rsi
-; HYBRID-NEXT: testb %r10b, %r10b
+; HYBRID-NEXT: cmovneq %rdx, %rdi
+; HYBRID-NEXT: cmovneq %r8, %rdx
+; HYBRID-NEXT: movl %esi, %ecx
+; HYBRID-NEXT: shlq %cl, %r11
+; HYBRID-NEXT: testb $64, %sil
+; HYBRID-NEXT: cmovneq %r11, %r10
+; HYBRID-NEXT: cmovneq %r8, %r11
+; HYBRID-NEXT: testb %sil, %sil
+; HYBRID-NEXT: cmovsq %r8, %r10
+; HYBRID-NEXT: movq %r10, 8(%rax)
; HYBRID-NEXT: cmovsq %r8, %r11
-; HYBRID-NEXT: movq %r11, 8(%rax)
-; HYBRID-NEXT: cmovsq %r8, %rsi
-; HYBRID-NEXT: movq %rsi, (%rax)
-; HYBRID-NEXT: cmovnsq %r8, %rdx
-; HYBRID-NEXT: cmoveq %r8, %rdx
-; HYBRID-NEXT: movq %rdx, 24(%rax)
-; HYBRID-NEXT: cmovnsq %r9, %rdi
+; HYBRID-NEXT: movq %r11, (%rax)
+; HYBRID-NEXT: cmovnsq %r8, %rdi
; HYBRID-NEXT: cmoveq %r8, %rdi
-; HYBRID-NEXT: movq %rdi, 16(%rax)
+; HYBRID-NEXT: movq %rdi, 24(%rax)
+; HYBRID-NEXT: cmovnsq %r9, %rdx
+; HYBRID-NEXT: cmoveq %r8, %rdx
+; HYBRID-NEXT: movq %rdx, 16(%rax)
; HYBRID-NEXT: retq
;
; BURR-LABEL: test1:
; BURR: # %bb.0:
; BURR-NEXT: movq %rdi, %rax
-; BURR-NEXT: leal 3(%rsi,%rsi), %r10d
+; BURR-NEXT: incl %esi
+; BURR-NEXT: addb %sil, %sil
+; BURR-NEXT: orb $1, %sil
; BURR-NEXT: movb $-128, %cl
-; BURR-NEXT: subb %r10b, %cl
+; BURR-NEXT: subb %sil, %cl
; BURR-NEXT: xorl %r8d, %r8d
-; BURR-NEXT: movl $1, %esi
+; BURR-NEXT: movl $1, %r11d
; BURR-NEXT: movl $1, %r9d
; BURR-NEXT: shrdq %cl, %r8, %r9
; BURR-NEXT: testb $64, %cl
; BURR-NEXT: cmovneq %r8, %r9
-; BURR-NEXT: xorl %r11d, %r11d
-; BURR-NEXT: movl %r10d, %ecx
-; BURR-NEXT: shldq %cl, %rsi, %r11
+; BURR-NEXT: xorl %r10d, %r10d
+; BURR-NEXT: movl %esi, %ecx
+; BURR-NEXT: shldq %cl, %r11, %r10
; BURR-NEXT: addb $-128, %cl
-; BURR-NEXT: xorl %edx, %edx
-; BURR-NEXT: shldq %cl, %rsi, %rdx
-; BURR-NEXT: movl $1, %edi
-; BURR-NEXT: shlq %cl, %rdi
+; BURR-NEXT: xorl %edi, %edi
+; BURR-NEXT: shldq %cl, %r11, %rdi
+; BURR-NEXT: movl $1, %edx
+; BURR-NEXT: shlq %cl, %rdx
; BURR-NEXT: testb $64, %cl
-; BURR-NEXT: cmovneq %rdi, %rdx
-; BURR-NEXT: cmovneq %r8, %rdi
-; BURR-NEXT: movl %r10d, %ecx
-; BURR-NEXT: shlq %cl, %rsi
-; BURR-NEXT: testb $64, %r10b
-; BURR-NEXT: cmovneq %rsi, %r11
-; BURR-NEXT: cmovneq %r8, %rsi
-; BURR-NEXT: testb %r10b, %r10b
+; BURR-NEXT: cmovneq %rdx, %rdi
+; BURR-NEXT: cmovneq %r8, %rdx
+; BURR-NEXT: movl %esi, %ecx
+; BURR-NEXT: shlq %cl, %r11
+; BURR-NEXT: testb $64, %sil
+; BURR-NEXT: cmovneq %r11, %r10
+; BURR-NEXT: cmovneq %r8, %r11
+; BURR-NEXT: testb %sil, %sil
+; BURR-NEXT: cmovsq %r8, %r10
+; BURR-NEXT: movq %r10, 8(%rax)
; BURR-NEXT: cmovsq %r8, %r11
-; BURR-NEXT: movq %r11, 8(%rax)
-; BURR-NEXT: cmovsq %r8, %rsi
-; BURR-NEXT: movq %rsi, (%rax)
-; BURR-NEXT: cmovnsq %r8, %rdx
-; BURR-NEXT: cmoveq %r8, %rdx
-; BURR-NEXT: movq %rdx, 24(%rax)
-; BURR-NEXT: cmovnsq %r9, %rdi
+; BURR-NEXT: movq %r11, (%rax)
+; BURR-NEXT: cmovnsq %r8, %rdi
; BURR-NEXT: cmoveq %r8, %rdi
-; BURR-NEXT: movq %rdi, 16(%rax)
+; BURR-NEXT: movq %rdi, 24(%rax)
+; BURR-NEXT: cmovnsq %r9, %rdx
+; BURR-NEXT: cmoveq %r8, %rdx
+; BURR-NEXT: movq %rdx, 16(%rax)
; BURR-NEXT: retq
;
; SRC-LABEL: test1:
; SRC: # %bb.0:
; SRC-NEXT: pushq %rbx
; SRC-NEXT: movq %rdi, %rax
-; SRC-NEXT: leal 3(%rsi,%rsi), %r9d
+; SRC-NEXT: incl %esi
+; SRC-NEXT: addb %sil, %sil
+; SRC-NEXT: orb $1, %sil
; SRC-NEXT: movb $-128, %cl
-; SRC-NEXT: subb %r9b, %cl
+; SRC-NEXT: subb %sil, %cl
; SRC-NEXT: xorl %r8d, %r8d
; SRC-NEXT: movl $1, %edi
; SRC-NEXT: movl $1, %r10d
; SRC-NEXT: shrdq %cl, %r8, %r10
; SRC-NEXT: testb $64, %cl
; SRC-NEXT: cmovneq %r8, %r10
-; SRC-NEXT: movl %r9d, %r11d
-; SRC-NEXT: addb $-128, %r11b
-; SRC-NEXT: xorl %esi, %esi
-; SRC-NEXT: movl %r11d, %ecx
-; SRC-NEXT: shldq %cl, %rdi, %rsi
+; SRC-NEXT: movl %esi, %r9d
+; SRC-NEXT: addb $-128, %r9b
; SRC-NEXT: xorl %edx, %edx
; SRC-NEXT: movl %r9d, %ecx
; SRC-NEXT: shldq %cl, %rdi, %rdx
+; SRC-NEXT: xorl %r11d, %r11d
+; SRC-NEXT: movl %esi, %ecx
+; SRC-NEXT: shldq %cl, %rdi, %r11
; SRC-NEXT: movl $1, %ebx
; SRC-NEXT: shlq %cl, %rbx
-; SRC-NEXT: testb $64, %r9b
-; SRC-NEXT: cmovneq %rbx, %rdx
+; SRC-NEXT: testb $64, %sil
+; SRC-NEXT: cmovneq %rbx, %r11
; SRC-NEXT: cmovneq %r8, %rbx
-; SRC-NEXT: movl %r11d, %ecx
+; SRC-NEXT: movl %r9d, %ecx
; SRC-NEXT: shlq %cl, %rdi
-; SRC-NEXT: testb $64, %r11b
-; SRC-NEXT: cmovneq %rdi, %rsi
+; SRC-NEXT: testb $64, %r9b
+; SRC-NEXT: cmovneq %rdi, %rdx
; SRC-NEXT: cmovneq %r8, %rdi
-; SRC-NEXT: testb %r9b, %r9b
+; SRC-NEXT: testb %sil, %sil
; SRC-NEXT: cmovnsq %r10, %rdi
; SRC-NEXT: cmoveq %r8, %rdi
-; SRC-NEXT: cmovnsq %r8, %rsi
-; SRC-NEXT: cmoveq %r8, %rsi
-; SRC-NEXT: cmovsq %r8, %rdx
+; SRC-NEXT: cmovnsq %r8, %rdx
+; SRC-NEXT: cmoveq %r8, %rdx
+; SRC-NEXT: cmovsq %r8, %r11
; SRC-NEXT: cmovsq %r8, %rbx
-; SRC-NEXT: movq %rdx, 8(%rax)
+; SRC-NEXT: movq %r11, 8(%rax)
; SRC-NEXT: movq %rbx, (%rax)
-; SRC-NEXT: movq %rsi, 24(%rax)
+; SRC-NEXT: movq %rdx, 24(%rax)
; SRC-NEXT: movq %rdi, 16(%rax)
; SRC-NEXT: popq %rbx
; SRC-NEXT: retq
; LIN-NEXT: movq %rdi, %rax
; LIN-NEXT: xorl %r9d, %r9d
; LIN-NEXT: movl $1, %r8d
-; LIN-NEXT: leal 3(%rsi,%rsi), %edx
-; LIN-NEXT: movl $1, %esi
-; LIN-NEXT: movl %edx, %ecx
-; LIN-NEXT: shlq %cl, %rsi
-; LIN-NEXT: testb $64, %dl
-; LIN-NEXT: movq %rsi, %rcx
+; LIN-NEXT: incl %esi
+; LIN-NEXT: addb %sil, %sil
+; LIN-NEXT: orb $1, %sil
+; LIN-NEXT: movl $1, %edx
+; LIN-NEXT: movl %esi, %ecx
+; LIN-NEXT: shlq %cl, %rdx
+; LIN-NEXT: testb $64, %sil
+; LIN-NEXT: movq %rdx, %rcx
; LIN-NEXT: cmovneq %r9, %rcx
-; LIN-NEXT: testb %dl, %dl
+; LIN-NEXT: testb %sil, %sil
; LIN-NEXT: cmovsq %r9, %rcx
; LIN-NEXT: movq %rcx, (%rdi)
; LIN-NEXT: xorl %edi, %edi
-; LIN-NEXT: movl %edx, %ecx
+; LIN-NEXT: movl %esi, %ecx
; LIN-NEXT: shldq %cl, %r8, %rdi
-; LIN-NEXT: cmovneq %rsi, %rdi
+; LIN-NEXT: cmovneq %rdx, %rdi
; LIN-NEXT: cmovsq %r9, %rdi
; LIN-NEXT: movq %rdi, 8(%rax)
-; LIN-NEXT: movl %edx, %esi
-; LIN-NEXT: addb $-128, %sil
+; LIN-NEXT: movl %esi, %edx
+; LIN-NEXT: addb $-128, %dl
; LIN-NEXT: movl $1, %r10d
-; LIN-NEXT: movl %esi, %ecx
+; LIN-NEXT: movl %edx, %ecx
; LIN-NEXT: shlq %cl, %r10
-; LIN-NEXT: testb $64, %sil
+; LIN-NEXT: testb $64, %dl
; LIN-NEXT: movq %r10, %rdi
; LIN-NEXT: cmovneq %r9, %rdi
; LIN-NEXT: movb $-128, %cl
-; LIN-NEXT: subb %dl, %cl
-; LIN-NEXT: movl $1, %edx
-; LIN-NEXT: shrdq %cl, %r9, %rdx
+; LIN-NEXT: subb %sil, %cl
+; LIN-NEXT: movl $1, %esi
+; LIN-NEXT: shrdq %cl, %r9, %rsi
; LIN-NEXT: testb $64, %cl
-; LIN-NEXT: cmovneq %r9, %rdx
-; LIN-NEXT: cmovsq %rdi, %rdx
-; LIN-NEXT: cmoveq %r9, %rdx
-; LIN-NEXT: movq %rdx, 16(%rax)
-; LIN-NEXT: xorl %edx, %edx
-; LIN-NEXT: movl %esi, %ecx
-; LIN-NEXT: shldq %cl, %r8, %rdx
-; LIN-NEXT: cmovneq %r10, %rdx
-; LIN-NEXT: cmovnsq %r9, %rdx
-; LIN-NEXT: cmoveq %r9, %rdx
-; LIN-NEXT: movq %rdx, 24(%rax)
+; LIN-NEXT: cmovneq %r9, %rsi
+; LIN-NEXT: cmovsq %rdi, %rsi
+; LIN-NEXT: cmoveq %r9, %rsi
+; LIN-NEXT: movq %rsi, 16(%rax)
+; LIN-NEXT: xorl %esi, %esi
+; LIN-NEXT: movl %edx, %ecx
+; LIN-NEXT: shldq %cl, %r8, %rsi
+; LIN-NEXT: cmovneq %r10, %rsi
+; LIN-NEXT: cmovnsq %r9, %rsi
+; LIN-NEXT: cmoveq %r9, %rsi
+; LIN-NEXT: movq %rsi, 24(%rax)
; LIN-NEXT: retq
%b = add i256 %a, 1
%m = shl i256 %b, 1
; CHECK-WIN32-64-LABEL: and16_trunc_8_sign:
; CHECK-WIN32-64: # %bb.0:
; CHECK-WIN32-64-NEXT: subq $40, %rsp
-; CHECK-WIN32-64-NEXT: # kill: def $cx killed $cx def $ecx
; CHECK-WIN32-64-NEXT: testb $-128, %cl
; CHECK-WIN32-64-NEXT: jg .LBB13_2
; CHECK-WIN32-64-NEXT: # %bb.1: # %yes
;
; CHECK-X86-LABEL: and16_trunc_8_sign:
; CHECK-X86: # %bb.0:
-; CHECK-X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; CHECK-X86-NEXT: testb $-128, %al
+; CHECK-X86-NEXT: testb $-128, {{[0-9]+}}(%esp)
; CHECK-X86-NEXT: jg .LBB13_2
; CHECK-X86-NEXT: # %bb.1: # %yes
; CHECK-X86-NEXT: calll bar
;
; CHECK-X86-LABEL: and32_trunc_16_sign:
; CHECK-X86: # %bb.0:
-; CHECK-X86-NEXT: movl $32768, %eax # imm = 0x8000
-; CHECK-X86-NEXT: andl {{[0-9]+}}(%esp), %eax
+; CHECK-X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; CHECK-X86-NEXT: andl $32768, %eax # imm = 0x8000
; CHECK-X86-NEXT: testw %ax, %ax
; CHECK-X86-NEXT: jg .LBB16_2
; CHECK-X86-NEXT: # %bb.1: # %yes
;
; CHECK-X86-LABEL: and32_trunc_16_sign_minsize:
; CHECK-X86: # %bb.0:
-; CHECK-X86-NEXT: testw $-32768, {{[0-9]+}}(%esp) # imm = 0x8000
+; CHECK-X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; CHECK-X86-NEXT: testw $-32768, %ax # imm = 0x8000
; CHECK-X86-NEXT: jg .LBB17_2
; CHECK-X86-NEXT: # %bb.1: # %yes
; CHECK-X86-NEXT: calll bar
; CHECK-NEXT: je .LBB0_3
; CHECK-NEXT: # %bb.1:
; CHECK-NEXT: movl %edi, %edx
-; CHECK-NEXT: andl $7, %edx
+; CHECK-NEXT: andb $7, %dl
; CHECK-NEXT: cmpb %cl, %dl
; CHECK-NEXT: jge .LBB0_2
; CHECK-NEXT: .LBB0_3: