Integer legalization: fix MUL expansion

Summary:
This fixes the runtime results produces by the fallback multiplication expansion introduced in r270720.

For tests I created a fuzz tester that compares the results with Boost.Multiprecision.

Reviewers: hfinkel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D26628

llvm-svn: 286998

1 files changed, 718 insertions, 683 deletions

diff --git a/llvm/test/CodeGen/X86/mul-i512.ll b/llvm/test/CodeGen/X86/mul-i512.ll
index 7b17be51617..f9cea6d2ba3 100644
--- a/llvm/test/CodeGen/X86/mul-i512.ll
+++ b/llvm/test/CodeGen/X86/mul-i512.ll
@@ -11,193 +11,145 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %esi
 ; X32-NEXT:    andl $-8, %esp
-; X32-NEXT:    subl $632, %esp # imm = 0x278
-; X32-NEXT:    movl 12(%ebp), %ecx
-; X32-NEXT:    movl 8(%ebp), %ebx
-; X32-NEXT:    movl (%ebx), %edi
-; X32-NEXT:    movl 4(%ebx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 8(%ebx), %edx
-; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 12(%ebx), %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 8(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 12(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl (%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 4(%ecx), %ecx
+; X32-NEXT:    subl $656, %esp # imm = 0x290
+; X32-NEXT:    movl 8(%ebp), %eax
+; X32-NEXT:    movl 48(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %ecx
-; X32-NEXT:    pushl %eax
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl %edx
-; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl %eax
-; X32-NEXT:    calll __multi3
-; X32-NEXT:    addl $32, %esp
-; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl %eax
-; X32-NEXT:    calll __multi3
-; X32-NEXT:    addl $32, %esp
-; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl %eax
-; X32-NEXT:    calll __multi3
-; X32-NEXT:    addl $32, %esp
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    seto %al
-; X32-NEXT:    lahf
-; X32-NEXT:    movl %eax, %eax
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl $0, %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    adcl $0, %ecx
+; X32-NEXT:    movl 52(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 60(%ebx), %ecx
+; X32-NEXT:    movl 56(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 56(%ebx), %ecx
+; X32-NEXT:    movl 60(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 52(%ebx), %ecx
+; X32-NEXT:    movl 40(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 48(%ebx), %ecx
+; X32-NEXT:    movl 44(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 44(%ebx), %ecx
+; X32-NEXT:    movl 32(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 40(%ebx), %ecx
+; X32-NEXT:    movl 36(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 36(%ebx), %ecx
+; X32-NEXT:    movl (%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 32(%ebx), %ecx
+; X32-NEXT:    movl 4(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 28(%ebx), %ecx
+; X32-NEXT:    movl 16(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 24(%ebx), %ecx
+; X32-NEXT:    movl 20(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 20(%ebx), %ecx
+; X32-NEXT:    movl 8(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 16(%ebx), %ecx
+; X32-NEXT:    movl 12(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl %eax, %eax
-; X32-NEXT:    addb $127, %al
-; X32-NEXT:    sahf
-; X32-NEXT:    adcl $0, %ecx
+; X32-NEXT:    movl 24(%eax), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    adcl $0, %eax
+; X32-NEXT:    movl 28(%eax), %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 12(%ebp), %ecx
-; X32-NEXT:    movl 28(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 24(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 20(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 16(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 44(%ecx), %edi
-; X32-NEXT:    movl 40(%ecx), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 36(%ecx), %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 32(%ecx), %eax
+; X32-NEXT:    movl 12(%ebp), %eax
+; X32-NEXT:    movl 48(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 52(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 56(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 60(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 32(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 36(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 40(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 44(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl (%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 4(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 8(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 12(%eax), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl 16(%eax), %esi
+; X32-NEXT:    movl 20(%eax), %edi
+; X32-NEXT:    movl 24(%eax), %ebx
+; X32-NEXT:    movl 28(%eax), %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl 60(%ecx)
-; X32-NEXT:    pushl 56(%ecx)
-; X32-NEXT:    pushl 52(%ecx)
-; X32-NEXT:    pushl 48(%ecx)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl %eax
 ; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    pushl %ecx
-; X32-NEXT:    movl %ecx, %ebx
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %eax
+; X32-NEXT:    calll __multi3
+; X32-NEXT:    addl $32, %esp
+; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
+; X32-NEXT:    pushl %esi
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    movl %esi, %ebx
-; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl %esi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    pushl %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
-; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
@@ -205,7 +157,9 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    pushl %ebx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
@@ -215,133 +169,167 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl %esi
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %eax
+; X32-NEXT:    calll __multi3
+; X32-NEXT:    addl $32, %esp
+; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %esi
-; X32-NEXT:    movl %esi, %edi
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    pushl %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl %eax
+; X32-NEXT:    calll __multi3
+; X32-NEXT:    addl $32, %esp
+; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
-; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %eax
+; X32-NEXT:    calll __multi3
+; X32-NEXT:    addl $32, %esp
+; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
@@ -351,50 +339,50 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    pushl %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
-; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
@@ -402,7 +390,7 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
@@ -413,42 +401,40 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    pushl %esi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    pushl %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl %ebx
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
-; X32-NEXT:    pushl %esi
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
@@ -459,110 +445,92 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %esi
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    pushl %edi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
 ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    pushl %edi
-; X32-NEXT:    pushl $0
-; X32-NEXT:    pushl $0
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
 ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl $0
+; X32-NEXT:    pushl %edi
+; X32-NEXT:    pushl %esi
 ; X32-NEXT:    pushl %eax
 ; X32-NEXT:    calll __multi3
 ; X32-NEXT:    addl $32, %esp
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    adcl $0, %edx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    adcl $0, %esi
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl $0, %edx
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    adcl $0, %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    adcl $0, %ecx
-; X32-NEXT:    addl %edx, %eax
-; X32-NEXT:    adcl %esi, %ecx
-; X32-NEXT:    movl $0, %edx
-; X32-NEXT:    adcl $0, %edx
-; X32-NEXT:    sbbl %esi, %esi
-; X32-NEXT:    andl $1, %esi
+; X32-NEXT:    addl %esi, %eax
+; X32-NEXT:    adcl %edx, %ecx
+; X32-NEXT:    movl $0, %edi
+; X32-NEXT:    adcl $0, %edi
+; X32-NEXT:    sbbl %edx, %edx
+; X32-NEXT:    andl $1, %edx
 ; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    addl %edi, %edx
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
+; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    addl %edx, %ebx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    adcl %edx, %edi
+; X32-NEXT:    adcl %esi, %edx
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl %esi, %edx
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    adcl %esi, %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
-; X32-NEXT:    addl %eax, %esi
-; X32-NEXT:    adcl %ecx, %edi
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    addl %ecx, %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    adcl %ecx, %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl $0, %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl $0, %edi
-; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl $0, %edx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl $0, %ebx
+; X32-NEXT:    adcl %esi, %edx
+; X32-NEXT:    addl %eax, %ebx
 ; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl %ecx, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl %edi, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl %eax, %ecx
@@ -589,29 +557,88 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    addl %esi, %edx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
 ; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    adcl %edi, %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    addl %edi, %ebx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    adcl %ebx, %edi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl %edi, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl %edi, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl %edi, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    addl %eax, %ebx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    adcl %ecx, %edi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
+; X32-NEXT:    adcl %edx, %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
+; X32-NEXT:    adcl %esi, %ecx
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %ebx
 ; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
+; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl $0, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl $0, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl $0, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl $0, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl %eax, %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    adcl $0, %edi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    addl %eax, %edx
-; X32-NEXT:    adcl %ecx, %esi
+; X32-NEXT:    adcl $0, %ebx
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    adcl $0, %edx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    adcl $0, %esi
+; X32-NEXT:    addl %edi, %edx
+; X32-NEXT:    adcl %ebx, %esi
+; X32-NEXT:    movl $0, %eax
+; X32-NEXT:    adcl $0, %eax
+; X32-NEXT:    sbbl %ebx, %ebx
+; X32-NEXT:    andl $1, %ebx
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
+; X32-NEXT:    addl %ecx, %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
+; X32-NEXT:    adcl %edi, %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl %edi, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
+; X32-NEXT:    addl %edx, %eax
+; X32-NEXT:    movl %eax, %edx
+; X32-NEXT:    adcl %esi, %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
+; X32-NEXT:    adcl %ebx, %edi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
 ; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
@@ -622,18 +649,21 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    adcl %eax, {{[0-9]+}}(%esp) # 4-byte Folded Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
 ; X32-NEXT:    adcl %eax, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl %edx, %eax
+; X32-NEXT:    adcl $0, %eax
+; X32-NEXT:    adcl $0, %ecx
+; X32-NEXT:    movl %esi, %edx
 ; X32-NEXT:    adcl $0, %edx
+; X32-NEXT:    movl %edi, %esi
 ; X32-NEXT:    adcl $0, %esi
-; X32-NEXT:    adcl $0, %edi
-; X32-NEXT:    adcl $0, %ebx
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
+; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
 ; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
-; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl $0, %eax
 ; X32-NEXT:    adcl $0, %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
@@ -646,53 +676,65 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    sbbl %eax, %eax
 ; X32-NEXT:    andl $1, %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
-; X32-NEXT:    movl $0, %edi
-; X32-NEXT:    adcl $0, %edi
-; X32-NEXT:    sbbl %ebx, %ebx
-; X32-NEXT:    andl $1, %ebx
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    adcl $0, %edx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
+; X32-NEXT:    adcl $0, %esi
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl %edi, %edx
-; X32-NEXT:    adcl %ebx, %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    addl %eax, {{[0-9]+}}(%esp) # 4-byte Folded Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    adcl %eax, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    adcl $0, %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl $0, %ecx
+; X32-NEXT:    addl %edx, %eax
+; X32-NEXT:    adcl %esi, %ecx
+; X32-NEXT:    movl $0, %edx
+; X32-NEXT:    adcl $0, %edx
+; X32-NEXT:    sbbl %esi, %esi
+; X32-NEXT:    andl $1, %esi
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    adcl %edx, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    adcl %edx, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    addl %eax, %edi
+; X32-NEXT:    adcl %ecx, %ebx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
+; X32-NEXT:    adcl %esi, %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    adcl %edx, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    adcl %edx, {{[0-9]+}}(%esp) # 4-byte Folded Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
+; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
 ; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    adcl %eax, {{[0-9]+}}(%esp) # 4-byte Folded Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
-; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
@@ -750,112 +792,108 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
 ; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    adcl $0, %esi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
 ; X32-NEXT:    adcl $0, %edi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    adcl $0, %ebx
-; X32-NEXT:    addl %eax, %edi
-; X32-NEXT:    adcl %ecx, %ebx
+; X32-NEXT:    addl %eax, %esi
+; X32-NEXT:    adcl %ecx, %edi
 ; X32-NEXT:    movl $0, %eax
 ; X32-NEXT:    adcl $0, %eax
-; X32-NEXT:    sbbl %edx, %edx
-; X32-NEXT:    andl $1, %edx
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx
+; X32-NEXT:    sbbl %ecx, %ecx
+; X32-NEXT:    andl $1, %ecx
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
+; X32-NEXT:    pushl %eax
+; X32-NEXT:    calll __multi3
+; X32-NEXT:    addl $32, %esp
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %ebx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    addl %esi, %ebx
+; X32-NEXT:    adcl %edi, %ecx
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    addl {{[0-9]+}}(%esp), %esi
 ; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    addl %edi, %ecx
-; X32-NEXT:    adcl %ebx, %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
+; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    adcl %edx, %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
+; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
-; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    addl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
 ; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
-; X32-NEXT:    movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx # 4-byte Reload
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    addl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
-; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ebx # 4-byte Folded Reload
-; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
+; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
+; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
+; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx # 4-byte Folded Reload
 ; X32-NEXT:    adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
-; X32-NEXT:    movl 16(%ebp), %eax
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    movl %ebx, 16(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    movl %ebx, 20(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    movl %ebx, 24(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    movl %ebx, 28(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    movl %ebx, (%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
-; X32-NEXT:    movl %ebx, 4(%eax)
-; X32-NEXT:    movl %ecx, 8(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    movl %ecx, 32(%eax)
-; X32-NEXT:    movl %edx, 12(%eax)
-; X32-NEXT:    movl %edi, 36(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    movl %ecx, 40(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    movl %ecx, 44(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    movl %ecx, 48(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    movl %ecx, 52(%eax)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx # 4-byte Reload
-; X32-NEXT:    movl %ecx, 56(%eax)
-; X32-NEXT:    movl %esi, 60(%eax)
+; X32-NEXT:    adcl {{[0-9]+}}(%esp), %eax # 4-byte Folded Reload
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X32-NEXT:    movl 16(%ebp), %edi
+; X32-NEXT:    movl %esi, 4(%edi)
+; X32-NEXT:    movl 16(%ebp), %esi
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, (%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 8(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 12(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 16(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 20(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 24(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 28(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 32(%esi)
+; X32-NEXT:    movl %ebx, 36(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 40(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 44(%esi)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
+; X32-NEXT:    movl %edi, 48(%esi)
+; X32-NEXT:    movl %ecx, 52(%esi)
+; X32-NEXT:    movl %edx, 56(%esi)
+; X32-NEXT:    movl %eax, 60(%esi)
 ; X32-NEXT:    leal -12(%ebp), %esp
 ; X32-NEXT:    popl %esi
 ; X32-NEXT:    popl %edi
@@ -871,323 +909,320 @@ define void @test_512(i512* %a, i512* %b, i512* %out) nounwind {
 ; X64-NEXT:    pushq %r13
 ; X64-NEXT:    pushq %r12
 ; X64-NEXT:    pushq %rbx
-; X64-NEXT:    subq $56, %rsp
-; X64-NEXT:    movq %rdx, {{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq 24(%rdi), %rcx
-; X64-NEXT:    movq 16(%rdi), %r15
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    movq %rdx, (%rsp) # 8-byte Spill
+; X64-NEXT:    movq 24(%rdi), %rbp
+; X64-NEXT:    movq 16(%rdi), %rcx
 ; X64-NEXT:    movq %rsi, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq (%rsi), %r11
-; X64-NEXT:    movq 8(%rsi), %rbp
-; X64-NEXT:    movq %r15, %rax
-; X64-NEXT:    mulq %r11
-; X64-NEXT:    movq %rdx, %r8
-; X64-NEXT:    movq %rax, %r10
+; X64-NEXT:    movq (%rsi), %rdx
+; X64-NEXT:    movq 8(%rsi), %r8
 ; X64-NEXT:    movq %rcx, %rax
-; X64-NEXT:    movq %rcx, %r14
-; X64-NEXT:    movq %r14, {{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    mulq %r11
+; X64-NEXT:    movq %rdx, %rsi
+; X64-NEXT:    mulq %rsi
+; X64-NEXT:    movq %rdx, %r10
+; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    movq %rbp, %rax
+; X64-NEXT:    movq %rbp, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    mulq %rsi
+; X64-NEXT:    movq %rsi, %r9
 ; X64-NEXT:    movq %rdx, %rbx
 ; X64-NEXT:    movq %rax, %rsi
 ; X64-NEXT:    addq %r10, %rsi
 ; X64-NEXT:    adcq $0, %rbx
-; X64-NEXT:    movq %r15, %rax
-; X64-NEXT:    mulq %rbp
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    movq %rcx, %r11
+; X64-NEXT:    movq %r11, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    mulq %r8
 ; X64-NEXT:    movq %rdx, %rcx
-; X64-NEXT:    movq %rax, %r9
-; X64-NEXT:    addq %rsi, %r9
+; X64-NEXT:    movq %rax, %r12
+; X64-NEXT:    addq %rsi, %r12
 ; X64-NEXT:    adcq $0, %rcx
 ; X64-NEXT:    addq %rbx, %rcx
-; X64-NEXT:    sbbq %rax, %rax
-; X64-NEXT:    andl $1, %eax
-; X64-NEXT:    movq %r15, %rbx
+; X64-NEXT:    sbbq %rbx, %rbx
+; X64-NEXT:    andl $1, %ebx
+; X64-NEXT:    movq %rbp, %rax
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rsi
+; X64-NEXT:    movq %rax, %rbp
+; X64-NEXT:    addq %rcx, %rbp
+; X64-NEXT:    adcq %rbx, %rsi
+; X64-NEXT:    xorl %ecx, %ecx
+; X64-NEXT:    movq %r9, %rbx
 ; X64-NEXT:    movq %rbx, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %rbx, %rdx
-; X64-NEXT:    imulq %rbp, %rdx
-; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %rbp, %r12
-; X64-NEXT:    movq %r12, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    leaq (%r8,%rdx), %rbp
-; X64-NEXT:    imulq %r11, %r14
-; X64-NEXT:    movq %r14, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    addq %r14, %rbp
-; X64-NEXT:    addq %r10, %rcx
-; X64-NEXT:    adcq %rax, %rbp
-; X64-NEXT:    xorl %r14d, %r14d
-; X64-NEXT:    movq %r11, %rax
-; X64-NEXT:    mulq %r14
-; X64-NEXT:    movq %rdx, %r10
-; X64-NEXT:    movq %r10, {{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %rax, %rsi
-; X64-NEXT:    movq %rsi, {{[0-9]+}}(%rsp) # 8-byte Spill
 ; X64-NEXT:    movq %rbx, %rax
-; X64-NEXT:    mulq %r14
-; X64-NEXT:    movq %rdx, %r14
-; X64-NEXT:    movq %rax, %rbx
-; X64-NEXT:    addq %rsi, %rbx
-; X64-NEXT:    adcq %r10, %r14
-; X64-NEXT:    addq %rcx, %rbx
-; X64-NEXT:    adcq %rbp, %r14
+; X64-NEXT:    mulq %rcx
+; X64-NEXT:    movq %rdx, %r13
+; X64-NEXT:    movq %rax, %r10
+; X64-NEXT:    movq %r11, %rax
+; X64-NEXT:    mulq %rcx
+; X64-NEXT:    movq %rdx, %r9
+; X64-NEXT:    movq %r9, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    movq %rax, %r15
+; X64-NEXT:    movq %r15, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    addq %r10, %r15
+; X64-NEXT:    adcq %r13, %r9
+; X64-NEXT:    addq %rbp, %r15
+; X64-NEXT:    adcq %rsi, %r9
 ; X64-NEXT:    movq %rdi, -{{[0-9]+}}(%rsp) # 8-byte Spill
 ; X64-NEXT:    movq (%rdi), %rax
-; X64-NEXT:    movq %rax, %r13
-; X64-NEXT:    mulq %r11
+; X64-NEXT:    movq %rax, %r14
+; X64-NEXT:    mulq %rbx
+; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    movq %rdx, %r11
+; X64-NEXT:    movq 8(%rdi), %rax
 ; X64-NEXT:    movq %rax, %rcx
 ; X64-NEXT:    movq %rcx, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    mulq %rbx
+; X64-NEXT:    movq %rdx, %rbp
+; X64-NEXT:    movq %rax, %rsi
+; X64-NEXT:    addq %r11, %rsi
+; X64-NEXT:    adcq $0, %rbp
+; X64-NEXT:    movq %r14, %rax
+; X64-NEXT:    movq %r8, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rbx
+; X64-NEXT:    addq %rsi, %rax
+; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    adcq $0, %rbx
+; X64-NEXT:    addq %rbp, %rbx
+; X64-NEXT:    sbbq %rdi, %rdi
+; X64-NEXT:    andl $1, %edi
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    mulq %r8
 ; X64-NEXT:    movq %rdx, %rsi
-; X64-NEXT:    movq %rsi, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq 8(%rdi), %rax
-; X64-NEXT:    movq %rax, %rdi
-; X64-NEXT:    mulq %r11
 ; X64-NEXT:    movq %rax, %rbp
-; X64-NEXT:    addq %rcx, %rbp
-; X64-NEXT:    adcq $0, %rdx
-; X64-NEXT:    movq %rdx, (%rsp) # 8-byte Spill
-; X64-NEXT:    movq %r13, %rax
-; X64-NEXT:    mulq %r12
+; X64-NEXT:    addq %rbx, %rbp
+; X64-NEXT:    adcq %rdi, %rsi
+; X64-NEXT:    movq %r14, %rcx
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    mulq %rdx
+; X64-NEXT:    movq %rdx, %r14
+; X64-NEXT:    movq %rax, %r11
+; X64-NEXT:    addq %r11, %r10
+; X64-NEXT:    adcq %r14, %r13
+; X64-NEXT:    addq %rbp, %r10
+; X64-NEXT:    adcq %rsi, %r13
+; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %r10 # 8-byte Folded Reload
+; X64-NEXT:    adcq %r12, %r13
+; X64-NEXT:    adcq $0, %r15
+; X64-NEXT:    adcq $0, %r9
+; X64-NEXT:    movq %r9, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rsi # 8-byte Reload
+; X64-NEXT:    movq 16(%rsi), %r8
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    movq %rcx, %r9
+; X64-NEXT:    movq %r9, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rdi
+; X64-NEXT:    movq %rax, %r12
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rbx
+; X64-NEXT:    movq %rax, %rbp
+; X64-NEXT:    addq %rdi, %rbp
+; X64-NEXT:    adcq $0, %rbx
+; X64-NEXT:    movq 24(%rsi), %rdi
+; X64-NEXT:    movq %r9, %rax
+; X64-NEXT:    mulq %rdi
+; X64-NEXT:    movq %rdx, %rsi
 ; X64-NEXT:    addq %rbp, %rax
 ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq $0, %rdx
+; X64-NEXT:    adcq $0, %rsi
+; X64-NEXT:    addq %rbx, %rsi
+; X64-NEXT:    sbbq %rbp, %rbp
+; X64-NEXT:    andl $1, %ebp
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    mulq %rdi
+; X64-NEXT:    movq %rdx, %rbx
+; X64-NEXT:    movq %rax, %r9
+; X64-NEXT:    addq %rsi, %r9
+; X64-NEXT:    adcq %rbp, %rbx
+; X64-NEXT:    movq %r8, %rax
+; X64-NEXT:    xorl %ecx, %ecx
+; X64-NEXT:    mulq %rcx
 ; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    addq %rsi, %r8
-; X64-NEXT:    movq %r8, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq %rax, %r9
-; X64-NEXT:    movq %r9, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq $0, %rbx
+; X64-NEXT:    movq %rax, %rbp
+; X64-NEXT:    addq %rbp, %r11
+; X64-NEXT:    adcq %rdx, %r14
+; X64-NEXT:    addq %r9, %r11
+; X64-NEXT:    adcq %rbx, %r14
+; X64-NEXT:    addq %r10, %r12
+; X64-NEXT:    movq %r12, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    adcq %r13, -{{[0-9]+}}(%rsp) # 8-byte Folded Spill
+; X64-NEXT:    adcq $0, %r11
 ; X64-NEXT:    adcq $0, %r14
+; X64-NEXT:    addq %r15, %r11
+; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %r14 # 8-byte Folded Reload
+; X64-NEXT:    adcq $0, %rcx
+; X64-NEXT:    movq %rcx, %r13
+; X64-NEXT:    sbbq %r9, %r9
+; X64-NEXT:    andl $1, %r9d
 ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
-; X64-NEXT:    movq 16(%rcx), %r10
-; X64-NEXT:    movq %r13, %rax
-; X64-NEXT:    mulq %r10
-; X64-NEXT:    movq %rdx, %r8
-; X64-NEXT:    movq %rax, %r15
-; X64-NEXT:    movq %rdi, %rax
-; X64-NEXT:    movq %rdi, %r12
-; X64-NEXT:    movq %r12, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    mulq %r10
-; X64-NEXT:    movq %rdx, %rbp
-; X64-NEXT:    movq %rax, %rdi
-; X64-NEXT:    addq %r15, %rdi
-; X64-NEXT:    adcq $0, %rbp
-; X64-NEXT:    movq 24(%rcx), %rsi
-; X64-NEXT:    movq %r13, %rcx
 ; X64-NEXT:    movq %rcx, %rax
-; X64-NEXT:    mulq %rsi
-; X64-NEXT:    movq %rsi, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %rdx, %r13
-; X64-NEXT:    movq %rax, %r9
-; X64-NEXT:    addq %rdi, %r9
-; X64-NEXT:    adcq $0, %r13
-; X64-NEXT:    addq %rbp, %r13
-; X64-NEXT:    sbbq %rax, %rax
-; X64-NEXT:    andl $1, %eax
-; X64-NEXT:    movq %rcx, %rdx
-; X64-NEXT:    movq %rcx, %rdi
-; X64-NEXT:    imulq %rsi, %rdx
-; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %r12, %rbp
-; X64-NEXT:    imulq %r10, %rbp
-; X64-NEXT:    leaq (%r8,%rdx), %rcx
-; X64-NEXT:    addq %rbp, %rcx
-; X64-NEXT:    addq %r15, %r13
-; X64-NEXT:    adcq %rax, %rcx
-; X64-NEXT:    movq %r10, %rax
-; X64-NEXT:    xorl %esi, %esi
-; X64-NEXT:    mulq %rsi
+; X64-NEXT:    mulq %r8
 ; X64-NEXT:    movq %rdx, %r15
 ; X64-NEXT:    movq %rax, %r12
-; X64-NEXT:    movq %rdi, %rax
-; X64-NEXT:    mulq %rsi
-; X64-NEXT:    xorl %esi, %esi
-; X64-NEXT:    movq %rdx, {{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    addq %rax, %r12
-; X64-NEXT:    adcq %rdx, %r15
-; X64-NEXT:    addq %r13, %r12
-; X64-NEXT:    adcq %rcx, %r15
-; X64-NEXT:    addq %r8, -{{[0-9]+}}(%rsp) # 8-byte Folded Spill
-; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %r9 # 8-byte Folded Reload
-; X64-NEXT:    movq %r9, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq $0, %r12
-; X64-NEXT:    adcq $0, %r15
-; X64-NEXT:    addq %rbx, %r12
-; X64-NEXT:    adcq %r14, %r15
-; X64-NEXT:    movq %r15, %r8
-; X64-NEXT:    adcq $0, %rsi
-; X64-NEXT:    movq %rsi, %r13
-; X64-NEXT:    sbbq %r15, %r15
-; X64-NEXT:    movq %r11, %rax
-; X64-NEXT:    mulq -{{[0-9]+}}(%rsp) # 8-byte Folded Reload
-; X64-NEXT:    movq %rax, %rsi
-; X64-NEXT:    movq %rdx, %rbx
-; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rbx # 8-byte Folded Reload
-; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rbx # 8-byte Folded Reload
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r10 # 8-byte Reload
 ; X64-NEXT:    movq %r10, %rax
-; X64-NEXT:    movq %rdi, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rsi
+; X64-NEXT:    movq %rax, %rbx
+; X64-NEXT:    addq %r15, %rbx
+; X64-NEXT:    adcq $0, %rsi
+; X64-NEXT:    movq %rcx, %rax
 ; X64-NEXT:    mulq %rdi
-; X64-NEXT:    movq %rdx, %r14
-; X64-NEXT:    addq %rbp, %r14
-; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %r14 # 8-byte Folded Reload
-; X64-NEXT:    andl $1, %r15d
-; X64-NEXT:    addq %rsi, %rax
-; X64-NEXT:    adcq %rbx, %r14
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbx # 8-byte Reload
-; X64-NEXT:    addq (%rsp), %rbx # 8-byte Folded Reload
+; X64-NEXT:    movq %rdx, %rcx
+; X64-NEXT:    addq %rbx, %rax
+; X64-NEXT:    movq %rax, %rbx
+; X64-NEXT:    adcq $0, %rcx
+; X64-NEXT:    addq %rsi, %rcx
 ; X64-NEXT:    sbbq %rsi, %rsi
-; X64-NEXT:    movq %rdi, %rdx
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rdi # 8-byte Reload
-; X64-NEXT:    imulq %rdi, %rdx
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
-; X64-NEXT:    addq %rcx, %rdx
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
-; X64-NEXT:    imulq %r11, %rbp
-; X64-NEXT:    addq %rdx, %rbp
 ; X64-NEXT:    andl $1, %esi
-; X64-NEXT:    movq %rbx, %rdx
-; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rdx # 8-byte Folded Reload
-; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq %rsi, %rbp
-; X64-NEXT:    movq %rbp, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    addq %rdx, %rax
-; X64-NEXT:    adcq %rbp, %r14
-; X64-NEXT:    addq %rcx, %r12
-; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %r8 # 8-byte Folded Reload
-; X64-NEXT:    movq %r8, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq %r13, %rax
-; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq %r15, %r14
+; X64-NEXT:    movq %r10, %rax
+; X64-NEXT:    mulq %rdi
+; X64-NEXT:    addq %rcx, %rax
+; X64-NEXT:    adcq %rsi, %rdx
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rsi # 8-byte Reload
+; X64-NEXT:    addq %rbp, %rsi
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
+; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rcx # 8-byte Folded Reload
+; X64-NEXT:    addq %rax, %rsi
+; X64-NEXT:    adcq %rdx, %rcx
+; X64-NEXT:    addq %r11, %r12
+; X64-NEXT:    movq %r12, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    adcq %r14, %rbx
+; X64-NEXT:    movq %rbx, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    adcq %r13, %rsi
+; X64-NEXT:    movq %rsi, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    adcq %r9, %rcx
+; X64-NEXT:    movq %rcx, -{{[0-9]+}}(%rsp) # 8-byte Spill
 ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
 ; X64-NEXT:    movq 32(%rcx), %rsi
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
-; X64-NEXT:    imulq %rsi, %rbp
+; X64-NEXT:    imulq %rsi, %rdi
 ; X64-NEXT:    movq %rsi, %rax
-; X64-NEXT:    mulq %r10
-; X64-NEXT:    movq %rax, %rbx
-; X64-NEXT:    addq %rbp, %rdx
-; X64-NEXT:    movq 40(%rcx), %r8
-; X64-NEXT:    imulq %r8, %r10
-; X64-NEXT:    addq %rdx, %r10
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rax, %r11
+; X64-NEXT:    addq %rdi, %rdx
+; X64-NEXT:    movq 40(%rcx), %r9
+; X64-NEXT:    imulq %r9, %r8
+; X64-NEXT:    addq %rdx, %r8
 ; X64-NEXT:    movq 48(%rcx), %rax
-; X64-NEXT:    movq %rcx, %rbp
-; X64-NEXT:    movq %rax, %rcx
-; X64-NEXT:    imulq %rdi, %rcx
-; X64-NEXT:    movq %rdi, %r9
-; X64-NEXT:    mulq %r11
-; X64-NEXT:    movq %rax, %r15
-; X64-NEXT:    addq %rcx, %rdx
-; X64-NEXT:    movq 56(%rbp), %rdi
-; X64-NEXT:    imulq %r11, %rdi
-; X64-NEXT:    addq %rdx, %rdi
-; X64-NEXT:    addq %rbx, %r15
-; X64-NEXT:    adcq %r10, %rdi
-; X64-NEXT:    movq %r11, %rax
+; X64-NEXT:    movq %rcx, %rbx
+; X64-NEXT:    movq %rax, %rdi
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
+; X64-NEXT:    imulq %rcx, %rdi
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
+; X64-NEXT:    mulq %rbp
+; X64-NEXT:    movq %rax, %r12
+; X64-NEXT:    addq %rdi, %rdx
+; X64-NEXT:    movq 56(%rbx), %rbx
+; X64-NEXT:    imulq %rbp, %rbx
+; X64-NEXT:    addq %rdx, %rbx
+; X64-NEXT:    addq %r11, %r12
+; X64-NEXT:    adcq %r8, %rbx
+; X64-NEXT:    movq %rbp, %rax
+; X64-NEXT:    movq %rbp, %r8
 ; X64-NEXT:    mulq %rsi
-; X64-NEXT:    movq %rdx, %r10
-; X64-NEXT:    movq %r10, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq %rax, %r13
-; X64-NEXT:    movq %r9, %rax
+; X64-NEXT:    movq %rdx, %rdi
+; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
+; X64-NEXT:    movq %rcx, %rax
 ; X64-NEXT:    mulq %rsi
-; X64-NEXT:    movq %rdx, %rbp
-; X64-NEXT:    movq %rax, %rcx
-; X64-NEXT:    addq %r13, %rcx
-; X64-NEXT:    adcq $0, %rbp
-; X64-NEXT:    movq %r11, %rax
-; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rsi
+; X64-NEXT:    movq %rax, %rbp
+; X64-NEXT:    addq %rdi, %rbp
+; X64-NEXT:    adcq $0, %rsi
+; X64-NEXT:    movq %r8, %rax
+; X64-NEXT:    mulq %r9
+; X64-NEXT:    movq %rdx, %rdi
+; X64-NEXT:    movq %rax, %r11
+; X64-NEXT:    addq %rbp, %r11
+; X64-NEXT:    adcq $0, %rdi
+; X64-NEXT:    addq %rsi, %rdi
+; X64-NEXT:    sbbq %rsi, %rsi
+; X64-NEXT:    andl $1, %esi
+; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    mulq %r9
+; X64-NEXT:    movq %rdx, %r14
+; X64-NEXT:    movq %rax, %r15
+; X64-NEXT:    addq %rdi, %r15
+; X64-NEXT:    adcq %rsi, %r14
+; X64-NEXT:    addq %r12, %r15
+; X64-NEXT:    adcq %rbx, %r14
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rdx # 8-byte Reload
+; X64-NEXT:    movq 56(%rdx), %rcx
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r8 # 8-byte Reload
+; X64-NEXT:    imulq %r8, %rcx
+; X64-NEXT:    movq 48(%rdx), %rbp
 ; X64-NEXT:    movq %rdx, %rbx
-; X64-NEXT:    addq %rcx, %rax
-; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    adcq $0, %rbx
-; X64-NEXT:    addq %rbp, %rbx
-; X64-NEXT:    sbbq %rax, %rax
-; X64-NEXT:    imulq %r8, %r11
-; X64-NEXT:    movq %r9, %rcx
-; X64-NEXT:    imulq %rsi, %rcx
-; X64-NEXT:    addq %r10, %r11
-; X64-NEXT:    addq %r11, %rcx
-; X64-NEXT:    andl $1, %eax
-; X64-NEXT:    addq %r13, %rbx
-; X64-NEXT:    adcq %rax, %rcx
-; X64-NEXT:    addq %r15, %rbx
-; X64-NEXT:    adcq %rdi, %rcx
-; X64-NEXT:    movq %rcx, -{{[0-9]+}}(%rsp) # 8-byte Spill
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
-; X64-NEXT:    movq 56(%rcx), %rsi
+; X64-NEXT:    movq %r8, %rax
+; X64-NEXT:    mulq %rbp
+; X64-NEXT:    movq %rax, %rsi
+; X64-NEXT:    addq %rcx, %rdx
 ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r10 # 8-byte Reload
-; X64-NEXT:    imulq %r10, %rsi
-; X64-NEXT:    movq 48(%rcx), %rdi
-; X64-NEXT:    movq %r10, %rax
-; X64-NEXT:    mulq %rdi
-; X64-NEXT:    movq %rax, %r9
-; X64-NEXT:    addq %rsi, %rdx
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r8 # 8-byte Reload
-; X64-NEXT:    imulq %r8, %rdi
-; X64-NEXT:    addq %rdx, %rdi
-; X64-NEXT:    movq 32(%rcx), %rsi
-; X64-NEXT:    movq 40(%rcx), %rbp
+; X64-NEXT:    imulq %r10, %rbp
+; X64-NEXT:    addq %rdx, %rbp
+; X64-NEXT:    movq 32(%rbx), %rdi
+; X64-NEXT:    movq 40(%rbx), %r12
 ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rax # 8-byte Reload
 ; X64-NEXT:    movq %rax, %rcx
-; X64-NEXT:    imulq %rbp, %rcx
-; X64-NEXT:    mulq %rsi
-; X64-NEXT:    movq %rax, %r15
+; X64-NEXT:    imulq %r12, %rcx
+; X64-NEXT:    mulq %rdi
+; X64-NEXT:    movq %rax, %r9
 ; X64-NEXT:    addq %rcx, %rdx
-; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rax # 8-byte Reload
-; X64-NEXT:    imulq %rsi, %rax
-; X64-NEXT:    addq %rdx, %rax
-; X64-NEXT:    addq %r9, %r15
-; X64-NEXT:    adcq %rdi, %rax
-; X64-NEXT:    movq %rax, %r11
-; X64-NEXT:    movq %rsi, %rax
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r13 # 8-byte Reload
+; X64-NEXT:    imulq %rdi, %r13
+; X64-NEXT:    addq %rdx, %r13
+; X64-NEXT:    addq %rsi, %r9
+; X64-NEXT:    adcq %rbp, %r13
+; X64-NEXT:    movq %rdi, %rax
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rcx
+; X64-NEXT:    movq %rax, %rsi
+; X64-NEXT:    movq %r12, %rax
+; X64-NEXT:    mulq %r8
+; X64-NEXT:    movq %rdx, %rbx
+; X64-NEXT:    movq %rax, %rbp
+; X64-NEXT:    addq %rcx, %rbp
+; X64-NEXT:    adcq $0, %rbx
+; X64-NEXT:    movq %rdi, %rax
 ; X64-NEXT:    mulq %r10
-; X64-NEXT:    movq %rdx, %r13
-; X64-NEXT:    movq %rax, %r9
-; X64-NEXT:    movq %rbp, %rax
+; X64-NEXT:    movq %rdx, %rcx
+; X64-NEXT:    movq %rax, %rdi
+; X64-NEXT:    addq %rbp, %rdi
+; X64-NEXT:    adcq $0, %rcx
+; X64-NEXT:    addq %rbx, %rcx
+; X64-NEXT:    sbbq %rbp, %rbp
+; X64-NEXT:    andl $1, %ebp
+; X64-NEXT:    movq %r12, %rax
 ; X64-NEXT:    mulq %r10
-; X64-NEXT:    movq %rdx, %rdi
-; X64-NEXT:    movq %rax, %rcx
-; X64-NEXT:    addq %r9, %rcx
-; X64-NEXT:    adcq $0, %rdi
-; X64-NEXT:    movq %rsi, %rax
-; X64-NEXT:    mulq %r8
 ; X64-NEXT:    addq %rcx, %rax
-; X64-NEXT:    adcq $0, %rdx
-; X64-NEXT:    addq %rdi, %rdx
-; X64-NEXT:    sbbq %rcx, %rcx
-; X64-NEXT:    imulq %r8, %rsi
-; X64-NEXT:    imulq %r10, %rbp
-; X64-NEXT:    addq %r13, %rsi
-; X64-NEXT:    addq %rsi, %rbp
-; X64-NEXT:    andl $1, %ecx
-; X64-NEXT:    addq %r9, %rdx
-; X64-NEXT:    adcq %rcx, %rbp
-; X64-NEXT:    addq %r15, %rdx
-; X64-NEXT:    adcq %r11, %rbp
-; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %r13 # 8-byte Folded Reload
-; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rax # 8-byte Folded Reload
-; X64-NEXT:    adcq %rbx, %rdx
-; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rbp # 8-byte Folded Reload
-; X64-NEXT:    addq %r12, %r13
+; X64-NEXT:    adcq %rbp, %rdx
+; X64-NEXT:    addq %r9, %rax
+; X64-NEXT:    adcq %r13, %rdx
+; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rsi # 8-byte Folded Reload
+; X64-NEXT:    adcq %r11, %rdi
+; X64-NEXT:    adcq %r15, %rax
+; X64-NEXT:    adcq %r14, %rdx
+; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rsi # 8-byte Folded Reload
+; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rdi # 8-byte Folded Reload
 ; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rax # 8-byte Folded Reload
 ; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rdx # 8-byte Folded Reload
-; X64-NEXT:    adcq %r14, %rbp
-; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rsi # 8-byte Reload
-; X64-NEXT:    addq {{[0-9]+}}(%rsp), %rsi # 8-byte Folded Reload
-; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
-; X64-NEXT:    adcq {{[0-9]+}}(%rsp), %rcx # 8-byte Folded Reload
-; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rsi # 8-byte Folded Reload
-; X64-NEXT:    movq %rsi, %rdi
-; X64-NEXT:    adcq -{{[0-9]+}}(%rsp), %rcx # 8-byte Folded Reload
-; X64-NEXT:    movq %rcx, %rsi
-; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbx # 8-byte Reload
-; X64-NEXT:    movq %rbx, 16(%rcx)
-; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbx # 8-byte Reload
-; X64-NEXT:    movq %rbx, 24(%rcx)
-; X64-NEXT:    movq %rdi, (%rcx)
-; X64-NEXT:    movq %r13, 32(%rcx)
-; X64-NEXT:    movq %rsi, 8(%rcx)
-; X64-NEXT:    movq %rax, 40(%rcx)
-; X64-NEXT:    movq %rdx, 48(%rcx)
-; X64-NEXT:    movq %rbp, 56(%rcx)
-; X64-NEXT:    addq $56, %rsp
+; X64-NEXT:    movq (%rsp), %rcx # 8-byte Reload
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
+; X64-NEXT:    movq %rbp, (%rcx)
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
+; X64-NEXT:    movq %rbp, 8(%rcx)
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
+; X64-NEXT:    movq %rbp, 16(%rcx)
+; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
+; X64-NEXT:    movq %rbp, 24(%rcx)
+; X64-NEXT:    movq %rsi, 32(%rcx)
+; X64-NEXT:    movq %rdi, 40(%rcx)
+; X64-NEXT:    movq %rax, 48(%rcx)
+; X64-NEXT:    movq %rdx, 56(%rcx)
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:    popq %rbx
 ; X64-NEXT:    popq %r12
 ; X64-NEXT:    popq %r13