[llvm] r325696 - [X86][MMX] Regenerate MMX arithmetic tests
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Wed Feb 21 08:37:10 PST 2018
Author: rksimon
Date: Wed Feb 21 08:37:10 2018
New Revision: 325696
URL: http://llvm.org/viewvc/llvm-project?rev=325696&view=rev
Log:
[X86][MMX] Regenerate MMX arithmetic tests
Modified:
llvm/trunk/test/CodeGen/X86/mmx-arith.ll
Modified: llvm/trunk/test/CodeGen/X86/mmx-arith.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/mmx-arith.ll?rev=325696&r1=325695&r2=325696&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/mmx-arith.ll (original)
+++ llvm/trunk/test/CodeGen/X86/mmx-arith.ll Wed Feb 21 08:37:10 2018
@@ -1,12 +1,151 @@
-; RUN: llc < %s -mtriple=i686-- -mattr=+mmx,+sse2 | FileCheck -check-prefix=X32 %s
-; RUN: llc < %s -mtriple=x86_64-- -mattr=+mmx,+sse2 | FileCheck -check-prefix=X64 %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+mmx,+sse2 | FileCheck -check-prefix=X32 %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+mmx,+sse2 | FileCheck -check-prefix=X64 %s
;; A basic sanity check to make sure that MMX arithmetic actually compiles.
;; First is a straight translation of the original with bitcasts as needed.
-; X32-LABEL: test0
-; X64-LABEL: test0
define void @test0(x86_mmx* %A, x86_mmx* %B) {
+; X32-LABEL: test0:
+; X32: # %bb.0: # %entry
+; X32-NEXT: pushl %ebp
+; X32-NEXT: .cfi_def_cfa_offset 8
+; X32-NEXT: .cfi_offset %ebp, -8
+; X32-NEXT: movl %esp, %ebp
+; X32-NEXT: .cfi_def_cfa_register %ebp
+; X32-NEXT: andl $-8, %esp
+; X32-NEXT: subl $32, %esp
+; X32-NEXT: movl 12(%ebp), %ecx
+; X32-NEXT: movl 8(%ebp), %eax
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X32-NEXT: paddw %xmm0, %xmm1
+; X32-NEXT: movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
+; X32-NEXT: pand %xmm0, %xmm1
+; X32-NEXT: packuswb %xmm1, %xmm1
+; X32-NEXT: movq %xmm1, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: paddsb (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: paddusb (%ecx), %mm0
+; X32-NEXT: movq %mm0, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; X32-NEXT: psubw %xmm2, %xmm1
+; X32-NEXT: pand %xmm0, %xmm1
+; X32-NEXT: packuswb %xmm1, %xmm1
+; X32-NEXT: movq %xmm1, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: psubsb (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: psubusb (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%esp)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; X32-NEXT: pmullw %xmm1, %xmm2
+; X32-NEXT: movdqa %xmm2, %xmm1
+; X32-NEXT: pand %xmm0, %xmm1
+; X32-NEXT: packuswb %xmm1, %xmm1
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X32-NEXT: pand %xmm2, %xmm1
+; X32-NEXT: movdqa %xmm1, %xmm2
+; X32-NEXT: pand %xmm0, %xmm2
+; X32-NEXT: packuswb %xmm2, %xmm2
+; X32-NEXT: movq %xmm2, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; X32-NEXT: por %xmm1, %xmm2
+; X32-NEXT: movdqa %xmm2, %xmm1
+; X32-NEXT: pand %xmm0, %xmm1
+; X32-NEXT: packuswb %xmm1, %xmm1
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X32-NEXT: pxor %xmm2, %xmm1
+; X32-NEXT: pand %xmm0, %xmm1
+; X32-NEXT: packuswb %xmm1, %xmm1
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: emms
+; X32-NEXT: movl %ebp, %esp
+; X32-NEXT: popl %ebp
+; X32-NEXT: retl
+;
+; X64-LABEL: test0:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X64-NEXT: paddw %xmm0, %xmm1
+; X64-NEXT: movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
+; X64-NEXT: pand %xmm0, %xmm1
+; X64-NEXT: packuswb %xmm1, %xmm1
+; X64-NEXT: movq %xmm1, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq -{{[0-9]+}}(%rsp), %mm0
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: paddsb (%rsi), %mm0
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: paddusb (%rsi), %mm0
+; X64-NEXT: movq %mm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; X64-NEXT: psubw %xmm2, %xmm1
+; X64-NEXT: pand %xmm0, %xmm1
+; X64-NEXT: packuswb %xmm1, %xmm1
+; X64-NEXT: movq %xmm1, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq -{{[0-9]+}}(%rsp), %mm0
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: psubsb (%rsi), %mm0
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: psubusb (%rsi), %mm0
+; X64-NEXT: movq %mm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; X64-NEXT: pmullw %xmm1, %xmm2
+; X64-NEXT: movdqa %xmm2, %xmm1
+; X64-NEXT: pand %xmm0, %xmm1
+; X64-NEXT: packuswb %xmm1, %xmm1
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X64-NEXT: pand %xmm2, %xmm1
+; X64-NEXT: movdqa %xmm1, %xmm2
+; X64-NEXT: pand %xmm0, %xmm2
+; X64-NEXT: packuswb %xmm2, %xmm2
+; X64-NEXT: movq %xmm2, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; X64-NEXT: por %xmm1, %xmm2
+; X64-NEXT: movdqa %xmm2, %xmm1
+; X64-NEXT: pand %xmm0, %xmm1
+; X64-NEXT: packuswb %xmm1, %xmm1
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; X64-NEXT: pxor %xmm2, %xmm1
+; X64-NEXT: pand %xmm0, %xmm1
+; X64-NEXT: packuswb %xmm1, %xmm1
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: emms
+; X64-NEXT: retq
entry:
%tmp1 = load x86_mmx, x86_mmx* %A
%tmp3 = load x86_mmx, x86_mmx* %B
@@ -59,9 +198,90 @@ entry:
ret void
}
-; X32-LABEL: test1
-; X64-LABEL: test1
define void @test1(x86_mmx* %A, x86_mmx* %B) {
+; X32-LABEL: test1:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X32-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; X32-NEXT: paddq %xmm0, %xmm1
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X32-NEXT: movdqa %xmm1, %xmm2
+; X32-NEXT: psrlq $32, %xmm2
+; X32-NEXT: pmuludq %xmm0, %xmm2
+; X32-NEXT: movdqa %xmm0, %xmm3
+; X32-NEXT: psrlq $32, %xmm3
+; X32-NEXT: pmuludq %xmm1, %xmm3
+; X32-NEXT: paddq %xmm2, %xmm3
+; X32-NEXT: psllq $32, %xmm3
+; X32-NEXT: pmuludq %xmm1, %xmm0
+; X32-NEXT: paddq %xmm3, %xmm0
+; X32-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; X32-NEXT: andps %xmm0, %xmm1
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X32-NEXT: orps %xmm1, %xmm0
+; X32-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; X32-NEXT: xorps %xmm0, %xmm1
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: emms
+; X32-NEXT: retl
+;
+; X64-LABEL: test1:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; X64-NEXT: paddq %xmm0, %xmm1
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X64-NEXT: movdqa %xmm1, %xmm2
+; X64-NEXT: psrlq $32, %xmm2
+; X64-NEXT: pmuludq %xmm0, %xmm2
+; X64-NEXT: movdqa %xmm0, %xmm3
+; X64-NEXT: psrlq $32, %xmm3
+; X64-NEXT: pmuludq %xmm1, %xmm3
+; X64-NEXT: paddq %xmm2, %xmm3
+; X64-NEXT: psllq $32, %xmm3
+; X64-NEXT: pmuludq %xmm0, %xmm1
+; X64-NEXT: paddq %xmm3, %xmm1
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X64-NEXT: pand %xmm1, %xmm0
+; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; X64-NEXT: por %xmm0, %xmm1
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; X64-NEXT: pxor %xmm1, %xmm0
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: emms
+; X64-NEXT: retq
entry:
%tmp1 = load x86_mmx, x86_mmx* %A
%tmp3 = load x86_mmx, x86_mmx* %B
@@ -103,9 +323,181 @@ entry:
ret void
}
-; X32-LABEL: test2
-; X64-LABEL: test2
define void @test2(x86_mmx* %A, x86_mmx* %B) {
+; X32-LABEL: test2:
+; X32: # %bb.0: # %entry
+; X32-NEXT: pushl %ebp
+; X32-NEXT: .cfi_def_cfa_offset 8
+; X32-NEXT: .cfi_offset %ebp, -8
+; X32-NEXT: movl %esp, %ebp
+; X32-NEXT: .cfi_def_cfa_register %ebp
+; X32-NEXT: andl $-8, %esp
+; X32-NEXT: subl $48, %esp
+; X32-NEXT: movl 12(%ebp), %ecx
+; X32-NEXT: movl 8(%ebp), %eax
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X32-NEXT: paddd %xmm0, %xmm1
+; X32-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; X32-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm0, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: paddsw (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: paddusw (%ecx), %mm0
+; X32-NEXT: movq %mm0, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X32-NEXT: psubd %xmm1, %xmm0
+; X32-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; X32-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm0, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: psubsw (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: psubusw (%ecx), %mm0
+; X32-NEXT: movq %mm0, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X32-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; X32-NEXT: pmuludq %xmm1, %xmm0
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X32-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; X32-NEXT: pmuludq %xmm2, %xmm1
+; X32-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; X32-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; X32-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; X32-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm0, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: pmulhw (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: pmaddwd (%ecx), %mm0
+; X32-NEXT: movq %mm0, (%esp)
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X32-NEXT: movq %mm0, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X32-NEXT: pand %xmm0, %xmm1
+; X32-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; X32-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X32-NEXT: por %xmm1, %xmm0
+; X32-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,2,2,3,4,5,6,7]
+; X32-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; X32-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; X32-NEXT: movq %xmm1, (%eax)
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X32-NEXT: pxor %xmm0, %xmm1
+; X32-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; X32-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X32-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X32-NEXT: movq %xmm0, (%eax)
+; X32-NEXT: emms
+; X32-NEXT: movl %ebp, %esp
+; X32-NEXT: popl %ebp
+; X32-NEXT: retl
+;
+; X64-LABEL: test2:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-NEXT: paddd %xmm0, %xmm1
+; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq -{{[0-9]+}}(%rsp), %mm0
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: paddsw (%rsi), %mm0
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: paddusw (%rsi), %mm0
+; X64-NEXT: movq %mm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-NEXT: psubd %xmm1, %xmm0
+; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq -{{[0-9]+}}(%rsp), %mm0
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: psubsw (%rsi), %mm0
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: psubusw (%rsi), %mm0
+; X64-NEXT: movq %mm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; X64-NEXT: pmuludq %xmm1, %xmm0
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; X64-NEXT: pmuludq %xmm2, %xmm1
+; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; X64-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq -{{[0-9]+}}(%rsp), %mm0
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: pmulhw (%rsi), %mm0
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: pmaddwd (%rsi), %mm0
+; X64-NEXT: movq %mm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X64-NEXT: movq %mm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-NEXT: pand %xmm0, %xmm1
+; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X64-NEXT: por %xmm1, %xmm0
+; X64-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,2,2,3,4,5,6,7]
+; X64-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; X64-NEXT: movq %xmm1, (%rdi)
+; X64-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X64-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-NEXT: pxor %xmm0, %xmm1
+; X64-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; X64-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-NEXT: movq %xmm0, (%rdi)
+; X64-NEXT: emms
+; X64-NEXT: retq
entry:
%tmp1 = load x86_mmx, x86_mmx* %A
%tmp3 = load x86_mmx, x86_mmx* %B
@@ -167,15 +559,78 @@ entry:
ret void
}
-; X32-LABEL: test3
define <1 x i64> @test3(<1 x i64>* %a, <1 x i64>* %b, i32 %count) nounwind {
+; X32-LABEL: test3:
+; X32: # %bb.0: # %entry
+; X32-NEXT: pushl %ebp
+; X32-NEXT: movl %esp, %ebp
+; X32-NEXT: pushl %ebx
+; X32-NEXT: pushl %edi
+; X32-NEXT: pushl %esi
+; X32-NEXT: andl $-8, %esp
+; X32-NEXT: subl $16, %esp
+; X32-NEXT: cmpl $0, 16(%ebp)
+; X32-NEXT: je .LBB3_1
+; X32-NEXT: # %bb.2: # %bb26.preheader
+; X32-NEXT: xorl %ebx, %ebx
+; X32-NEXT: xorl %eax, %eax
+; X32-NEXT: xorl %edx, %edx
+; X32-NEXT: .p2align 4, 0x90
+; X32-NEXT: .LBB3_3: # %bb26
+; X32-NEXT: # =>This Inner Loop Header: Depth=1
+; X32-NEXT: movl 8(%ebp), %ecx
+; X32-NEXT: movl %ecx, %esi
+; X32-NEXT: movl (%esi,%ebx,8), %ecx
+; X32-NEXT: movl 4(%esi,%ebx,8), %esi
+; X32-NEXT: movl 12(%ebp), %edi
+; X32-NEXT: addl (%edi,%ebx,8), %ecx
+; X32-NEXT: adcl 4(%edi,%ebx,8), %esi
+; X32-NEXT: addl %eax, %ecx
+; X32-NEXT: movl %ecx, (%esp)
+; X32-NEXT: adcl %edx, %esi
+; X32-NEXT: movl %esi, {{[0-9]+}}(%esp)
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: movd %xmm0, %eax
+; X32-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,0,1]
+; X32-NEXT: movd %xmm0, %edx
+; X32-NEXT: incl %ebx
+; X32-NEXT: cmpl 16(%ebp), %ebx
+; X32-NEXT: jb .LBB3_3
+; X32-NEXT: jmp .LBB3_4
+; X32-NEXT: .LBB3_1:
+; X32-NEXT: xorl %eax, %eax
+; X32-NEXT: xorl %edx, %edx
+; X32-NEXT: .LBB3_4: # %bb31
+; X32-NEXT: leal -12(%ebp), %esp
+; X32-NEXT: popl %esi
+; X32-NEXT: popl %edi
+; X32-NEXT: popl %ebx
+; X32-NEXT: popl %ebp
+; X32-NEXT: retl
+;
+; X64-LABEL: test3:
+; X64: # %bb.0: # %entry
+; X64-NEXT: xorl %r8d, %r8d
+; X64-NEXT: xorl %eax, %eax
+; X64-NEXT: testl %edx, %edx
+; X64-NEXT: je .LBB3_2
+; X64-NEXT: .p2align 4, 0x90
+; X64-NEXT: .LBB3_1: # %bb26
+; X64-NEXT: # =>This Inner Loop Header: Depth=1
+; X64-NEXT: movslq %r8d, %r8
+; X64-NEXT: movq (%rdi,%r8,8), %rcx
+; X64-NEXT: addq (%rsi,%r8,8), %rcx
+; X64-NEXT: addq %rcx, %rax
+; X64-NEXT: incl %r8d
+; X64-NEXT: cmpl %edx, %r8d
+; X64-NEXT: jb .LBB3_1
+; X64-NEXT: .LBB3_2: # %bb31
+; X64-NEXT: retq
entry:
%tmp2942 = icmp eq i32 %count, 0
br i1 %tmp2942, label %bb31, label %bb26
bb26:
-; X32: addl
-; X32: adcl
%i.037.0 = phi i32 [ 0, %entry ], [ %tmp25, %bb26 ]
%sum.035.0 = phi <1 x i64> [ zeroinitializer, %entry ], [ %tmp22, %bb26 ]
%tmp13 = getelementptr <1 x i64>, <1 x i64>* %b, i32 %i.037.0
@@ -194,94 +649,186 @@ bb31:
}
; There are no MMX operations here, so we use XMM or i64.
-; X64-LABEL: ti8
define void @ti8(double %a, double %b) nounwind {
+; X32-LABEL: ti8:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: paddb %xmm0, %xmm1
+; X32-NEXT: movq %xmm1, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti8:
+; X64: # %bb.0: # %entry
+; X64-NEXT: paddb %xmm1, %xmm0
+; X64-NEXT: movq %xmm0, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to <8 x i8>
%tmp2 = bitcast double %b to <8 x i8>
%tmp3 = add <8 x i8> %tmp1, %tmp2
-; X64: paddb
store <8 x i8> %tmp3, <8 x i8>* null
ret void
}
-; X64-LABEL: ti16
define void @ti16(double %a, double %b) nounwind {
+; X32-LABEL: ti16:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: paddw %xmm0, %xmm1
+; X32-NEXT: movq %xmm1, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti16:
+; X64: # %bb.0: # %entry
+; X64-NEXT: paddw %xmm1, %xmm0
+; X64-NEXT: movq %xmm0, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to <4 x i16>
%tmp2 = bitcast double %b to <4 x i16>
%tmp3 = add <4 x i16> %tmp1, %tmp2
-; X64: paddw
store <4 x i16> %tmp3, <4 x i16>* null
ret void
}
-; X64-LABEL: ti32
define void @ti32(double %a, double %b) nounwind {
+; X32-LABEL: ti32:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT: paddd %xmm0, %xmm1
+; X32-NEXT: movq %xmm1, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti32:
+; X64: # %bb.0: # %entry
+; X64-NEXT: paddd %xmm1, %xmm0
+; X64-NEXT: movq %xmm0, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to <2 x i32>
%tmp2 = bitcast double %b to <2 x i32>
%tmp3 = add <2 x i32> %tmp1, %tmp2
-; X64: paddd
store <2 x i32> %tmp3, <2 x i32>* null
ret void
}
-; X64-LABEL: ti64
define void @ti64(double %a, double %b) nounwind {
+; X32-LABEL: ti64:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT: addl {{[0-9]+}}(%esp), %eax
+; X32-NEXT: adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT: movl %eax, 0
+; X32-NEXT: movl %ecx, 4
+; X32-NEXT: retl
+;
+; X64-LABEL: ti64:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movq %xmm0, %rax
+; X64-NEXT: movq %xmm1, %rcx
+; X64-NEXT: addq %rax, %rcx
+; X64-NEXT: movq %rcx, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to <1 x i64>
%tmp2 = bitcast double %b to <1 x i64>
%tmp3 = add <1 x i64> %tmp1, %tmp2
-; X64: addq
store <1 x i64> %tmp3, <1 x i64>* null
ret void
}
; MMX intrinsics calls get us MMX instructions.
-; X64-LABEL: ti8a
define void @ti8a(double %a, double %b) nounwind {
+; X32-LABEL: ti8a:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: paddb {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %mm0, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti8a:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movdq2q %xmm0, %mm0
+; X64-NEXT: movdq2q %xmm1, %mm1
+; X64-NEXT: paddb %mm0, %mm1
+; X64-NEXT: movq %mm1, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to x86_mmx
-; X64: movdq2q
%tmp2 = bitcast double %b to x86_mmx
-; X64: movdq2q
%tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.b(x86_mmx %tmp1, x86_mmx %tmp2)
store x86_mmx %tmp3, x86_mmx* null
ret void
}
-; X64-LABEL: ti16a
define void @ti16a(double %a, double %b) nounwind {
+; X32-LABEL: ti16a:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: paddw {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %mm0, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti16a:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movdq2q %xmm0, %mm0
+; X64-NEXT: movdq2q %xmm1, %mm1
+; X64-NEXT: paddw %mm0, %mm1
+; X64-NEXT: movq %mm1, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to x86_mmx
-; X64: movdq2q
%tmp2 = bitcast double %b to x86_mmx
-; X64: movdq2q
%tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.w(x86_mmx %tmp1, x86_mmx %tmp2)
store x86_mmx %tmp3, x86_mmx* null
ret void
}
-; X64-LABEL: ti32a
define void @ti32a(double %a, double %b) nounwind {
+; X32-LABEL: ti32a:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: paddd {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %mm0, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti32a:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movdq2q %xmm0, %mm0
+; X64-NEXT: movdq2q %xmm1, %mm1
+; X64-NEXT: paddd %mm0, %mm1
+; X64-NEXT: movq %mm1, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to x86_mmx
-; X64: movdq2q
%tmp2 = bitcast double %b to x86_mmx
-; X64: movdq2q
%tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.d(x86_mmx %tmp1, x86_mmx %tmp2)
store x86_mmx %tmp3, x86_mmx* null
ret void
}
-; X64-LABEL: ti64a
define void @ti64a(double %a, double %b) nounwind {
+; X32-LABEL: ti64a:
+; X32: # %bb.0: # %entry
+; X32-NEXT: movq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: paddq {{[0-9]+}}(%esp), %mm0
+; X32-NEXT: movq %mm0, 0
+; X32-NEXT: retl
+;
+; X64-LABEL: ti64a:
+; X64: # %bb.0: # %entry
+; X64-NEXT: movdq2q %xmm0, %mm0
+; X64-NEXT: movdq2q %xmm1, %mm1
+; X64-NEXT: paddq %mm0, %mm1
+; X64-NEXT: movq %mm1, 0
+; X64-NEXT: retq
entry:
%tmp1 = bitcast double %a to x86_mmx
-; X64: movdq2q
%tmp2 = bitcast double %b to x86_mmx
-; X64: movdq2q
%tmp3 = tail call x86_mmx @llvm.x86.mmx.padd.q(x86_mmx %tmp1, x86_mmx %tmp2)
store x86_mmx %tmp3, x86_mmx* null
ret void
More information about the llvm-commits
mailing list