[X86] Enable call frame optimization ("mov to push") not only for optsize (PR26325)

[hans]

Summary

The size savings are significant, and from what I can tell, both ICC and GCC do this [1] [2].

Please let me know what you think.

[1] https://godbolt.org/g/OtrQRa
[2] https://godbolt.org/g/5We0Zx

[joerg]

test/CodeGen/X86/win32-seh-nested-finally.ll

46 ↗

(On Diff #51968)

The changes here look suspicious, doesn't the code need to restore %esp before the final popl?

test/CodeGen/X86/xmulo.ll

28 ↗

(On Diff #51968)

Each immediate push is two Bytes, materializing $0 as register is two Bytes as well, but a register push is one Byte. So for three pushes, a shorter sequence is actually:

xorl %eax, %eax
pushl %eax
pushl %eax
pushl %eax

test/CodeGen/X86/zext-fold.ll

38 ↗

(On Diff #51968)

Two register pushes are smaller than one subl?

[hans]

test/CodeGen/X86/win32-seh-nested-finally.ll

46 ↗

(On Diff #51968)

This does look weird. Reid, is there something magic about these invokes, or is mov-to-push broken here?

test/CodeGen/X86/xmulo.ll

28 ↗

(On Diff #51968)

Yes, we ca probably be more efficient here. See also PR26330 where we get this wrong the other way around.

test/CodeGen/X86/zext-fold.ll

38 ↗

(On Diff #51968)

The savings from the pushes offset the cost of the the sub.

The sub seems unnecessary though. It would be nice to not emit it.

[hans]

test/CodeGen/X86/win32-seh-nested-finally.ll

46 ↗

(On Diff #51968)

Oh wait, %esp does get restored in between, it's just not in the test expectations. Let me fix that..

[hans]

Clarify win32-seh-nested-finally.ll

[rnk]

lgtm I think this is production ready. We *mostly* build chromium with -Os, and we've fixed some bugs in this code.

lib/Target/X86/X86CallFrameOptimization.cpp

176–178 ↗

(On Diff #52139)

Huh, so we're already doing this a lot due to inalloca.

test/CodeGen/X86/win32-seh-nested-finally.ll

46 ↗

(On Diff #52139)

Great, I was halfway through checking that before I got distracted...

test/CodeGen/X86/zext-fold.ll

38–39 ↗

(On Diff #52139)

Actually, the sub probably won't be emitted on Mac and Windows, it will only appear on platforms with 16 byte stack alignment.

[joerg]

llvm/trunk/test/CodeGen/X86/win32-seh-nested-finally.ll

56

Two things here for the updated patch. If the stack alignment requirement is 32bit only OR if the pushes have realigned the stack correctly (not sure if we care about the second part), the addls can be deferred to the end of the BB.
It's also cheaper to use a pop to some scratch register if available.

62

This and the next block is missing a pop still? Also, add looks strange and out of context for the rest of the purpose of the text case.

[mkuper]

Thanks a lot, Hans!
I apologize for not making this happen myself - never got to it, and then lost the ability to benchmark...

llvm/trunk/test/CodeGen/X86/win32-seh-nested-finally.ll

56

Regarding the last part - we already have code for that in X86FrameLowering::eliminateCallFramePseudoInstr(), but it's also currently only enabled for minsize.

[hans]

llvm/trunk/test/CodeGen/X86/win32-seh-nested-finally.ll

56

Using pop for more stack restores is PR26333.
Filed PR27165 for considering using pop for stack restore beyond -Os.

62

Adding more checks in r265027.

Are you saying add instead of pop looks out of place? See Michael's comment above. Or do you mean the "addl $12, %ebp" above?

[hans]

llvm/trunk/test/CodeGen/X86/win32-seh-nested-finally.ll

56

I mean, filed PR27165 for considering not restoring the stack between calls.

llvm/trunk/lib/Target/X86/X86CallFrameOptimization.cpp

bool X86CallFrameOptimization::isProfitable(MachineFunction &MF,
   ContextVector &CallSeqVector) {
   // This transformation is always a win when we do not expect to have
   // a reserved call frame. Under other circumstances, it may be either
   // a win or a loss, and requires a heuristic.
   bool CannotReserveFrame = MF.getFrameInfo()->hasVarSizedObjects();
   if (CannotReserveFrame)
     return true;
 
-  // Don't do this when not optimizing for size.
-  if (!MF.getFunction()->optForSize())
-    return false;
-
   unsigned StackAlign = TFL->getStackAlignment();
 
   int64_t Advantage = 0;
   for (auto CC : CallSeqVector) {
     // Call sites where no parameters are passed on the stack
     // do not affect the cost, since there needs to be no
     // stack adjustment.
     if (CC.NoStackParams)

llvm/trunk/test/CodeGen/X86/2006-05-02-InstrSched1.ll

 ; REQUIRES: asserts
 ; RUN: llc < %s -march=x86 -relocation-model=static -stats 2>&1 | \
-; RUN:   grep asm-printer | grep 16
+; RUN:   grep asm-printer | grep 15
 ;
 ; It's possible to schedule this in 14 instructions by avoiding
 ; callee-save registers, but the scheduler isn't currently that
 ; conervative with registers.
 @size20 = external global i32   ; <i32*> [#uses=1]
 @in5 = external global i8*    ; <i8**> [#uses=1]
 
 define i32 @compare(i8* %a, i8* %b) nounwind {

llvm/trunk/test/CodeGen/X86/2006-11-12-CSRetCC.ll

 ; RUN: llc < %s -march=x86 | FileCheck %s
 
 target triple = "i686-pc-linux-gnu"
 @str = internal constant [9 x i8] c"%f+%f*i\0A\00"              ; <[9 x i8]*> [#uses=1]
 
 define i32 @main() {
 ; CHECK-LABEL: main:
 ; CHECK-NOT: ret
-; CHECK: subl $4, %{{.*}}
+; CHECK: subl $12, %esp
+; CHECK: pushl
+; CHECK: pushl
+; CHECK: pushl
+; CHECK: pushl
+; CHECK: pushl
+; CHECK: calll cexp
+; CHECK: addl $28, %esp
 ; CHECK: ret
 
 entry:
         %retval = alloca i32, align 4           ; <i32*> [#uses=1]
         %tmp = alloca { double, double }, align 16              ; <{ double, double }*> [#uses=4]
         %tmp1 = alloca { double, double }, align 16             ; <{ double, double }*> [#uses=4]
         %tmp2 = alloca { double, double }, align 16             ; <{ double, double }*> [#uses=3]
         %pi = alloca double, align 8            ; <double*> [#uses=2]

llvm/trunk/test/CodeGen/X86/atom-lea-sp.ll

-; RUN: llc < %s -mcpu=atom -mtriple=i686-linux  | FileCheck -check-prefix=ATOM %s
-; RUN: llc < %s -mcpu=core2 -mtriple=i686-linux | FileCheck %s
+; RUN: llc < %s -mcpu=atom -mtriple=i686-linux  -no-x86-call-frame-opt | FileCheck -check-prefix=ATOM %s
+; RUN: llc < %s -mcpu=core2 -mtriple=i686-linux -no-x86-call-frame-opt | FileCheck %s
 
 declare void @use_arr(i8*)
 declare void @many_params(i32, i32, i32, i32, i32, i32)
 
 define void @test1() nounwind {
 ; ATOM-LABEL: test1:
 ; ATOM: leal -1052(%esp), %esp
 ; ATOM-NOT: sub

llvm/trunk/test/CodeGen/X86/avx-intel-ocl.ll

 ; WIN64-LABEL: testf16_inp
 ; WIN64: vaddps  {{.*}}, {{%ymm[0-1]}}
 ; WIN64: vaddps  {{.*}}, {{%ymm[0-1]}}
 ; WIN64: leaq    {{.*}}(%rsp), %rcx
 ; WIN64: call
 ; WIN64: ret
 
 ; X32-LABEL: testf16_inp
-; X32: movl    %eax, (%esp)
 ; X32: vaddps  {{.*}}, {{%ymm[0-1]}}
 ; X32: vaddps  {{.*}}, {{%ymm[0-1]}}
+; Push is not deemed profitable if we're realigning the stack.
+; X32: {{pushl|movl}}   %eax
 ; X32: call
 ; X32: ret
 
 ; X64-LABEL: testf16_inp
 ; X64: vaddps  {{.*}}, {{%ymm[0-1]}}
 ; X64: vaddps  {{.*}}, {{%ymm[0-1]}}
 ; X64: leaq    {{.*}}(%rsp), %rdi
 ; X64: call

 define intel_ocl_bicc <16 x float> @test_prolog_epilog(<16 x float> %a, <16 x float> %b) nounwind {
    %c = call <16 x float> @func_float16(<16 x float> %a, <16 x float> %b)
    ret <16 x float> %c
 }
 
 ; test functions with integer parameters
 ; pass parameters on stack for 32-bit platform
 ; X32-LABEL: test_int
-; X32: movl {{.*}}, 4(%esp)
-; X32: movl {{.*}}, (%esp)
+; X32: pushl {{.*}}
+; X32: pushl {{.*}}
 ; X32: call
 ; X32: addl {{.*}}, %eax
 
 ; pass parameters in registers for 64-bit platform
 ; X64-LABEL: test_int
 ; X64: leal {{.*}}, %edi
 ; X64: movl {{.*}}, %esi
 ; X64: call

llvm/trunk/test/CodeGen/X86/avx512-intel-ocl.ll

 ; WIN64-LABEL: testf16_inp
 ; WIN64: vaddps  {{.*}}, {{%zmm[0-1]}}
 ; WIN64: leaq    {{.*}}(%rsp), %rcx
 ; WIN64: call
 ; WIN64: ret
 
 ; X32-LABEL: testf16_inp
 ; X32: vaddps  {{.*}}, {{%zmm[0-1]}}
-; X32: movl    %eax, (%esp)
+; Push is not deemed profitable if we're realigning the stack.
+; X32: {{pushl|movl}}   %eax
 ; X32: call
 ; X32: ret
 
 ; X64-LABEL: testf16_inp
 ; X64: vaddps  {{.*}}, {{%zmm[0-1]}}
 ; X64: leaq    {{.*}}(%rsp), %rdi
 ; X64: call
 ; X64: ret

 ; X64-LABEL: test_prolog_epilog_with_mask
 ; X64: kxorw   %k{{.*}}, %k{{.*}}, %k1
 ; X64: call
 define intel_ocl_bicc <16 x float> @test_prolog_epilog_with_mask(<16 x float> %a, <16 x i32> %x1, <16 x i32>%x2, <16 x i1> %mask) nounwind {
    %cmp_res = icmp eq <16 x i32>%x1, %x2
    %mask1 = xor <16 x i1> %cmp_res, %mask
    %c = call intel_ocl_bicc <16 x float> @func_float16_mask(<16 x float> %a, <16 x i1>%mask1)
    ret <16 x float> %c
-}
+}
 

llvm/trunk/test/CodeGen/X86/call-push.ll

-; RUN: llc < %s -mtriple=i386-apple-darwin -disable-fp-elim | FileCheck %s
+; RUN: llc < %s -mtriple=i386-apple-darwin -disable-fp-elim -no-x86-call-frame-opt | FileCheck %s
 
         %struct.decode_t = type { i8, i8, i8, i8, i16, i8, i8, %struct.range_t** }
         %struct.range_t = type { float, float, i32, i32, i32, [0 x i8] }
 
 define i32 @decode_byte(%struct.decode_t* %decode) nounwind {
 ; CHECK-LABEL: decode_byte:
 ; CHECK: pushl
 ; CHECK: popl

llvm/trunk/test/CodeGen/X86/cmpxchg-clobber-flags.ll

 define i64 @test_intervening_call(i64* %foo, i64 %bar, i64 %baz) {
 ; i386-LABEL: test_intervening_call:
 ; i386: cmpxchg8b
 ; i386-NEXT: pushl %eax
 ; i386-NEXT: seto %al
 ; i386-NEXT: lahf
 ; i386-NEXT: movl %eax, [[FLAGS:%.*]]
 ; i386-NEXT: popl %eax
-; i386-NEXT: movl %edx, 4(%esp)
-; i386-NEXT: movl %eax, (%esp)
+; i386-NEXT: subl $8, %esp
+; i386-NEXT: pushl %edx
+; i386-NEXT: pushl %eax
 ; i386-NEXT: calll bar
+; i386-NEXT: addl $16, %esp
 ; i386-NEXT: movl [[FLAGS]], %eax
 ; i386-NEXT: addb $127, %al
 ; i386-NEXT: sahf
 ; i386-NEXT: jne
 
 ; i386f-LABEL: test_intervening_call:
 ; i386f: cmpxchg8b
 ; i386f-NEXT: movl %eax, (%esp)

llvm/trunk/test/CodeGen/X86/coalescer-commute3.ll

-; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+sse2 | grep mov | count 6
+; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+sse2 -no-x86-call-frame-opt | grep mov | count 6
 
   %struct.quad_struct = type { i32, i32, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct* }
 
 define  i32 @perimeter(%struct.quad_struct* %tree, i32 %size) nounwind  {
 entry:
   switch i32 %size, label %UnifiedReturnBlock [
      i32 2, label %bb
      i32 0, label %bb50

llvm/trunk/test/CodeGen/X86/hipe-prologue.ll

define {i32, i32} @test_basic(i32 %hp, i32 %p) {
   call void @dummy_use (i32* %mem, i32 10)
   %1 = insertvalue {i32, i32} undef, i32 %hp, 0
   %2 = insertvalue {i32, i32} %1, i32 %p, 1
   ret {i32, i32} %1
 }
 
 define cc 11 {i32, i32} @test_basic_hipecc(i32 %hp, i32 %p) {
   ; X32-Linux-LABEL:       test_basic_hipecc:
-  ; X32-Linux:       leal -156(%esp), %ebx
+  ; X32-Linux:       leal -140(%esp), %ebx
   ; X32-Linux-NEXT:  cmpl 76(%ebp), %ebx
   ; X32-Linux-NEXT:  jb .LBB1_1
 
   ; X32-Linux:       ret
 
   ; X32-Linux:       .LBB1_1:
   ; X32-Linux-NEXT:  calll inc_stack_0
 

llvm/trunk/test/CodeGen/X86/i386-shrink-wrapping.ll

-; RUN: llc %s -o - -enable-shrink-wrap=true | FileCheck %s --check-prefix=CHECK --check-prefix=ENABLE
-; RUN: llc %s -o - -enable-shrink-wrap=false | FileCheck %s --check-prefix=CHECK --check-prefix=DISABLE
+; RUN: llc %s -o - -enable-shrink-wrap=true -no-x86-call-frame-opt | FileCheck %s --check-prefix=CHECK --check-prefix=ENABLE
+; RUN: llc %s -o - -enable-shrink-wrap=false -no-x86-call-frame-opt | FileCheck %s --check-prefix=CHECK --check-prefix=DISABLE
 target datalayout = "e-m:e-p:32:32-f64:32:64-f80:32-n8:16:32-S128"
 target triple = "i386-apple-macosx"
 
 @a = common global i32 0, align 4
 @d = internal unnamed_addr global i1 false
 @b = common global i32 0, align 4
 @e = common global i8 0, align 1
 @f = common global i8 0, align 1

llvm/trunk/test/CodeGen/X86/libcall-sret.ll

 ; RUN: llc -mtriple=i686-linux-gnu -o - %s | FileCheck %s
 
 @var = global i128 0
 
 ; We were trying to convert the i128 operation into a libcall, but failing to
 ; perform sret demotion when we couldn't return the result in registers. Make
 ; sure we marshal the return properly:
 
 define void @test_sret_libcall(i128 %l, i128 %r) {
 ; CHECK-LABEL: test_sret_libcall:
 
   ; Stack for call: 4(sret ptr), 16(i128 %l), 16(128 %r). So next logical
-  ; (aligned) place for the actual sret data is %esp + 40.
-; CHECK: leal 40(%esp), [[SRET_ADDR:%[a-z]+]]
-; CHECK: movl [[SRET_ADDR]], (%esp)
+  ; (aligned) place for the actual sret data is %esp + 20.
+; CHECK: leal 20(%esp), [[SRET_ADDR:%[a-z]+]]
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl 72(%esp)
+; CHECK: pushl [[SRET_ADDR]]
+
 ; CHECK: calll __multi3
-; CHECK-DAG: movl 40(%esp), [[RES0:%[a-z]+]]
-; CHECK-DAG: movl 44(%esp), [[RES1:%[a-z]+]]
-; CHECK-DAG: movl 48(%esp), [[RES2:%[a-z]+]]
-; CHECK-DAG: movl 52(%esp), [[RES3:%[a-z]+]]
+
+; CHECK: addl $44, %esp
+; CHECK-DAG: movl 8(%esp), [[RES0:%[a-z]+]]
+; CHECK-DAG: movl 12(%esp), [[RES1:%[a-z]+]]
+; CHECK-DAG: movl 16(%esp), [[RES2:%[a-z]+]]
+; CHECK-DAG: movl 20(%esp), [[RES3:%[a-z]+]]
 ; CHECK-DAG: movl [[RES0]], var
 ; CHECK-DAG: movl [[RES1]], var+4
 ; CHECK-DAG: movl [[RES2]], var+8
 ; CHECK-DAG: movl [[RES3]], var+12
   %prod = mul i128 %l, %r
   store i128 %prod, i128* @var
   ret void
 }

llvm/trunk/test/CodeGen/X86/localescape.ll

 ; X64: movl .Lalloc_func$frame_escape_1(%[[parent_fp]]), %edx
 ; X64: movq %[[str]], %rcx
 ; X64: callq printf
 ; X64: movl    $42, .Lalloc_func$frame_escape_1(%[[parent_fp]])
 ; X64: retq
 
 ; X86-LABEL: print_framealloc_from_fp:
 ; X86: pushl   %esi
-; X86: subl    $8, %esp
-; X86: movl    16(%esp), %esi
-; X86: movl    Lalloc_func$frame_escape_0(%esi), %eax
-; X86: movl    %eax, 4(%esp)
-; X86: movl    $_str, (%esp)
+; X86: movl    8(%esp), %esi
+; X86: pushl   Lalloc_func$frame_escape_0(%esi)
+; X86: pushl   $_str
 ; X86: calll   _printf
-; X86: movl    Lalloc_func$frame_escape_1(%esi), %eax
-; X86: movl    %eax, 4(%esp)
-; X86: movl    $_str, (%esp)
+; X86: addl    $8, %esp
+; X86: pushl   Lalloc_func$frame_escape_1(%esi)
+; X86: pushl   $_str
 ; X86: calll   _printf
+; X86: addl    $8, %esp
 ; X86: movl    $42, Lalloc_func$frame_escape_1(%esi)
 ; X86: movl    $4, %eax
-; X86: movl    Lalloc_func$frame_escape_1(%esi,%eax), %eax
-; X86: movl    %eax, 4(%esp)
-; X86: movl    $_str, (%esp)
+; X86: pushl   Lalloc_func$frame_escape_1(%esi,%eax)
+; X86: pushl   $_str
 ; X86: calll   _printf
 ; X86: addl    $8, %esp
 ; X86: popl    %esi
 ; X86: retl
 
 define void @alloc_func(i32 %n) {
   %a = alloca i32
   %b = alloca i32, i32 2

 ; X64: movl $42, 36(%rsp)
 ; X64: movl $13, 32(%rsp)
 ; X64: xorl %ecx, %ecx
 ; X64: callq print_framealloc_from_fp
 ; X64: addq $40, %rsp
 ; X64: retq
 
 ; X86-LABEL: alloc_func_no_frameaddr:
-; X86: subl    $12, %esp
-; X86: Lalloc_func_no_frameaddr$frame_escape_0 = 8
-; X86: Lalloc_func_no_frameaddr$frame_escape_1 = 4
-; X86: movl $42, 8(%esp)
-; X86: movl $13, 4(%esp)
-; X86: movl $0, (%esp)
+; X86: subl    $8, %esp
+; X86: Lalloc_func_no_frameaddr$frame_escape_0 = 4
+; X86: Lalloc_func_no_frameaddr$frame_escape_1 = 0
+; X86: movl $42, 4(%esp)
+; X86: movl $13, (%esp)
+; X86: pushl $0
 ; X86: calll _print_framealloc_from_fp
-; X86: addl $12, %esp
+; X86: addl    $4, %esp
+; X86: addl    $8, %esp
 ; X86: retl

llvm/trunk/test/CodeGen/X86/mcu-abi.ll

 ; CHECK: movl %edx, %eax
 ; CHECK: calll __fixsfsi
 define i32 @test_lib_args(float %a, float %b) #0 {
   %ret = fptosi float %b to i32
   ret i32 %ret
 }
 
 ; CHECK-LABEL: test_fp128:
-; CHECK: movl    (%eax), %e[[CX:..]]
-; CHECK-NEXT: movl    4(%eax), %e[[DX:..]]
-; CHECK-NEXT: movl    8(%eax), %e[[SI:..]]
-; CHECK-NEXT: movl    12(%eax), %e[[AX:..]]
-; CHECK-NEXT: movl    %e[[AX]], 12(%esp)
-; CHECK-NEXT: movl    %e[[SI]], 8(%esp)
-; CHECK-NEXT: movl    %e[[DX]], 4(%esp)
-; CHECK-NEXT: movl    %e[[CX]], (%esp)
+; CHECK:      pushl   12(%eax)
+; CHECK-NEXT: pushl   8(%eax)
+; CHECK-NEXT: pushl   4(%eax)
+; CHECK-NEXT: pushl   (%eax)
 ; CHECK-NEXT: calll   __fixtfsi
 define i32 @test_fp128(fp128* %ptr) #0 {
   %v = load fp128, fp128* %ptr
   %ret = fptosi fp128 %v to i32
   ret i32 %ret
 }
 
 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i32, i1) #1

llvm/trunk/test/CodeGen/X86/memset-2.ll

 ; NOTE: Assertions have been autogenerated by update_test_checks.py
 ; RUN: llc -mtriple=i386-apple-darwin -mcpu=yonah < %s | FileCheck %s
 
 declare void @llvm.memset.i32(i8*, i8, i32, i32) nounwind
 
 define fastcc void @t1() nounwind {
 ; CHECK-LABEL: t1:
 ; CHECK:         subl $12, %esp
-; CHECK-NEXT:    movl $188, {{[0-9]+}}(%esp)
-; CHECK-NEXT:    movl $0, {{[0-9]+}}(%esp)
-; CHECK-NEXT:    movl $0, (%esp)
+; CHECK:         pushl $188
+; CHECK-NEXT:    pushl $0
+; CHECK-NEXT:    pushl $0
 ; CHECK-NEXT:    calll L_memset$stub
 ;
 entry:
   call void @llvm.memset.p0i8.i32(i8* null, i8 0, i32 188, i32 1, i1 false)
   unreachable
 }
 
 define fastcc void @t2(i8 signext %c) nounwind {

llvm/trunk/test/CodeGen/X86/mingw-alloca.ll

 declare void @bar1(i32*)
 
 define void @foo2(i32 inreg  %N) nounwind {
 entry:
 ; COFF: _foo2:
 ; COFF: andl $-16, %esp
 ; COFF: pushl %eax
 ; COFF: calll __alloca
-; COFF: movl  8028(%esp), %eax
+; COFF: movl  8012(%esp), %eax
 ; ELF: foo2:
 ; ELF: andl $-16, %esp
 ; ELF: pushl %eax
 ; ELF: calll _alloca
-; ELF: movl 8028(%esp), %eax
+; ELF: movl 8012(%esp), %eax
   %A2 = alloca [2000 x i32], align 16    ; <[2000 x i32]*> [#uses=1]
   %A2.sub = getelementptr [2000 x i32], [2000 x i32]* %A2, i32 0, i32 0    ; <i32*> [#uses=1]
   call void @bar2( i32* %A2.sub, i32 %N )
   ret void
 }
 
 declare void @bar2(i32*, i32)

llvm/trunk/test/CodeGen/X86/movtopush.ll

 ; RUN: llc < %s -mtriple=i686-windows | FileCheck %s -check-prefix=NORMAL
+; RUN: llc < %s -mtriple=i686-windows -no-x86-call-frame-opt | FileCheck %s -check-prefix=NOPUSH
 ; RUN: llc < %s -mtriple=x86_64-windows | FileCheck %s -check-prefix=X64
 ; RUN: llc < %s -mtriple=i686-windows -stackrealign -stack-alignment=32 | FileCheck %s -check-prefix=ALIGNED
 
 %class.Class = type { i32 }
 %struct.s = type { i64 }
 
 declare void @good(i32 %a, i32 %b, i32 %c, i32 %d)
 declare void @inreg(i32 %a, i32 inreg %b, i32 %c, i32 %d)
 declare x86_thiscallcc void @thiscall(%class.Class* %class, i32 %a, i32 %b, i32 %c, i32 %d)
 declare void @oneparam(i32 %a)
 declare void @eightparams(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h)
 declare void @struct(%struct.s* byval %a, i32 %b, i32 %c, i32 %d)
 
-; Here, we should have a reserved frame, so we don't expect pushes
+; We should get pushes for x86, even though there is a reserved call frame.
+; Make sure we don't touch x86-64, and that turning it off works.
 ; NORMAL-LABEL: test1:
-; NORMAL: subl    $16, %esp
-; NORMAL-NEXT: movl    $4, 12(%esp)
-; NORMAL-NEXT: movl    $3, 8(%esp)
-; NORMAL-NEXT: movl    $2, 4(%esp)
-; NORMAL-NEXT: movl    $1, (%esp)
-; NORMAL-NEXT: call
-; NORMAL-NEXT: addl $16, %esp
-define void @test1() {
-entry:
-  call void @good(i32 1, i32 2, i32 3, i32 4)
-  ret void
-}
-
-; We're optimizing for code size, so we should get pushes for x86,
-; even though there is a reserved call frame.
-; Make sure we don't touch x86-64
-; NORMAL-LABEL: test1b:
 ; NORMAL-NOT: subl {{.*}} %esp
 ; NORMAL: pushl   $4
 ; NORMAL-NEXT: pushl   $3
 ; NORMAL-NEXT: pushl   $2
 ; NORMAL-NEXT: pushl   $1
 ; NORMAL-NEXT: call
 ; NORMAL-NEXT: addl $16, %esp
-; X64-LABEL: test1b:
+; X64-LABEL: test1:
 ; X64: movl    $1, %ecx
 ; X64-NEXT: movl    $2, %edx
 ; X64-NEXT: movl    $3, %r8d
 ; X64-NEXT: movl    $4, %r9d
 ; X64-NEXT: callq   good
-define void @test1b() optsize {
-entry:
-  call void @good(i32 1, i32 2, i32 3, i32 4)
-  ret void
-}
-
-; Same as above, but for minsize
-; NORMAL-LABEL: test1c:
-; NORMAL-NOT: subl {{.*}} %esp
-; NORMAL: pushl   $4
-; NORMAL-NEXT: pushl   $3
-; NORMAL-NEXT: pushl   $2
-; NORMAL-NEXT: pushl   $1
-; NORMAL-NEXT: call
-; NORMAL-NEXT: addl $16, %esp
-define void @test1c() minsize {
+; NOPUSH-LABEL: test1:
+; NOPUSH: subl    $16, %esp
+; NOPUSH-NEXT: movl    $4, 12(%esp)
+; NOPUSH-NEXT: movl    $3, 8(%esp)
+; NOPUSH-NEXT: movl    $2, 4(%esp)
+; NOPUSH-NEXT: movl    $1, (%esp)
+; NOPUSH-NEXT: call
+; NOPUSH-NEXT: addl $16, %esp
+define void @test1() {
 entry:
   call void @good(i32 1, i32 2, i32 3, i32 4)
   ret void
 }
 
 ; If we have a reserved frame, we should have pushes
 ; NORMAL-LABEL: test2:
 ; NORMAL-NOT: subl {{.*}} %esp

llvm/trunk/test/CodeGen/X86/phys-reg-local-regalloc.ll

-; RUN: llc < %s -stack-symbol-ordering=0 -march=x86 -mtriple=i386-apple-darwin9 -mcpu=generic -regalloc=fast -optimize-regalloc=0 | FileCheck %s
-; RUN: llc -O0 < %s -stack-symbol-ordering=0 -march=x86 -mtriple=i386-apple-darwin9 -mcpu=generic -regalloc=fast | FileCheck %s
-; RUN: llc < %s -stack-symbol-ordering=0 -march=x86 -mtriple=i386-apple-darwin9 -mcpu=atom -regalloc=fast -optimize-regalloc=0 | FileCheck -check-prefix=ATOM %s
+; RUN: llc < %s -stack-symbol-ordering=0 -march=x86 -mtriple=i386-apple-darwin9 -mcpu=generic -regalloc=fast -optimize-regalloc=0 -no-x86-call-frame-opt | FileCheck %s
+; RUN: llc -O0 < %s -stack-symbol-ordering=0 -march=x86 -mtriple=i386-apple-darwin9 -mcpu=generic -regalloc=fast -no-x86-call-frame-opt | FileCheck %s
+; RUN: llc < %s -stack-symbol-ordering=0 -march=x86 -mtriple=i386-apple-darwin9 -mcpu=atom -regalloc=fast -optimize-regalloc=0 -no-x86-call-frame-opt | FileCheck -check-prefix=ATOM %s
 ; CHECKed instructions should be the same with or without -O0 except on Intel Atom due to instruction scheduling.
 
 @.str = private constant [12 x i8] c"x + y = %i\0A\00", align 1 ; <[12 x i8]*> [#uses=1]
 
 define i32 @main() nounwind {
 entry:
 ; CHECK: movl 24(%esp), %eax
 ; CHECK-NOT: movl

llvm/trunk/test/CodeGen/X86/segmented-stacks.ll

define void @test_basic() #0 {
   ret void
 
 ; X32-Linux-LABEL:       test_basic:
 
 ; X32-Linux:       cmpl %gs:48, %esp
 ; X32-Linux-NEXT:  ja      .LBB0_2
 
 ; X32-Linux:       pushl $0
-; X32-Linux-NEXT:  pushl $60
+; X32-Linux-NEXT:  pushl $44
 ; X32-Linux-NEXT:  calll __morestack
 ; X32-Linux-NEXT:  ret
 
 ; X64-Linux-LABEL:       test_basic:
 
 ; X64-Linux:       cmpq %fs:112, %rsp
 ; X64-Linux-NEXT:  ja      .LBB0_2
 

 ; X64-Darwin-NEXT: ret
 
 ; X32-MinGW-LABEL:       test_basic:
 
 ; X32-MinGW:       cmpl %fs:20, %esp
 ; X32-MinGW-NEXT:  ja      LBB0_2
 
 ; X32-MinGW:       pushl $0
-; X32-MinGW-NEXT:  pushl $48
+; X32-MinGW-NEXT:  pushl $40
 ; X32-MinGW-NEXT:  calll ___morestack
 ; X32-MinGW-NEXT:  ret
 
 ; X64-MinGW-LABEL:       test_basic:
 
 ; X64-MinGW:       cmpq %gs:40, %rsp
 ; X64-MinGW-NEXT:  ja      .LBB0_2
 

 ; X64-FreeBSD-NEXT:  ret
 
 ; X32-DFlyBSD-LABEL:       test_basic:
 
 ; X32-DFlyBSD:       cmpl %fs:16, %esp
 ; X32-DFlyBSD-NEXT:  ja      .LBB0_2
 
 ; X32-DFlyBSD:       pushl $0
-; X32-DFlyBSD-NEXT:  pushl $48
+; X32-DFlyBSD-NEXT:  pushl $40
 ; X32-DFlyBSD-NEXT:  calll __morestack
 ; X32-DFlyBSD-NEXT:  ret
 
 ; X64-DFlyBSD-LABEL:       test_basic:
 
 ; X64-DFlyBSD:       cmpq %fs:32, %rsp
 ; X64-DFlyBSD-NEXT:  ja      .LBB0_2
 
define i32 @test_nested(i32 * nest %closure, i32 %other) #0 {
        %mem = alloca i32, i32 10
        call void @dummy_use (i32* %mem, i32 10)
        ret i32 %result
 
 ; X32-Linux:       cmpl %gs:48, %esp
 ; X32-Linux-NEXT:  ja      .LBB1_2
 
 ; X32-Linux:       pushl $4
-; X32-Linux-NEXT:  pushl $60
+; X32-Linux-NEXT:  pushl $44
 ; X32-Linux-NEXT:  calll __morestack
 ; X32-Linux-NEXT:  ret
 
 ; X64-Linux:       cmpq %fs:112, %rsp
 ; X64-Linux-NEXT:  ja      .LBB1_2
 
 ; X64-Linux:       movq %r10, %rax
 ; X64-Linux-NEXT:  movabsq $56, %r10

 ; X64-Darwin-NEXT: callq ___morestack
 ; X64-Darwin-NEXT: ret
 ; X64-Darwin-NEXT: movq %rax, %r10
 
 ; X32-MinGW:       cmpl %fs:20, %esp
 ; X32-MinGW-NEXT:  ja      LBB1_2
 
 ; X32-MinGW:       pushl $4
-; X32-MinGW-NEXT:  pushl $52
+; X32-MinGW-NEXT:  pushl $44
 ; X32-MinGW-NEXT:  calll ___morestack
 ; X32-MinGW-NEXT:  ret
 
 ; X64-MinGW-LABEL: test_nested:
 ; X64-MinGW:       cmpq %gs:40, %rsp
 ; X64-MinGW-NEXT:  ja      .LBB1_2
 
 ; X64-MinGW:       movq %r10, %rax

 ; X64-FreeBSD-NEXT:  callq __morestack
 ; X64-FreeBSD-NEXT:  ret
 ; X64-FreeBSD-NEXT:  movq %rax, %r10
 
 ; X32-DFlyBSD:       cmpl %fs:16, %esp
 ; X32-DFlyBSD-NEXT:  ja      .LBB1_2
 
 ; X32-DFlyBSD:       pushl $4
-; X32-DFlyBSD-NEXT:  pushl $52
+; X32-DFlyBSD-NEXT:  pushl $44
 ; X32-DFlyBSD-NEXT:  calll __morestack
 ; X32-DFlyBSD-NEXT:  ret
 
 ; X64-DFlyBSD:       cmpq %fs:32, %rsp
 ; X64-DFlyBSD-NEXT:  ja      .LBB1_2
 
 ; X64-DFlyBSD:       movq %r10, %rax
 ; X64-DFlyBSD-NEXT:  movabsq $56, %r10

 ; X64-Darwin-NEXT: cmpq %gs:816, %r11
 ; X64-Darwin-NEXT: ja      LBB2_2
 
 ; X64-Darwin:      movabsq $40008, %r10
 ; X64-Darwin-NEXT: movabsq $0, %r11
 ; X64-Darwin-NEXT: callq ___morestack
 ; X64-Darwin-NEXT: ret
 
-; X32-MinGW:       leal -40008(%esp), %ecx
+; X32-MinGW:       leal -40000(%esp), %ecx
 ; X32-MinGW-NEXT:  cmpl %fs:20, %ecx
 ; X32-MinGW-NEXT:  ja      LBB2_2
 
 ; X32-MinGW:       pushl $0
-; X32-MinGW-NEXT:  pushl $40008
+; X32-MinGW-NEXT:  pushl $40000
 ; X32-MinGW-NEXT:  calll ___morestack
 ; X32-MinGW-NEXT:  ret
 
 ; X64-MinGW-LABEL: test_large:
 ; X64-MinGW:       leaq -40040(%rsp), %r11
 ; X64-MinGW-NEXT:  cmpq %gs:40, %r11
 ; X64-MinGW-NEXT:  ja      .LBB2_2
 
 ; X64-MinGW:       movabsq $40040, %r10
 ; X64-MinGW-NEXT:  movabsq $32, %r11
 ; X64-MinGW-NEXT:  callq __morestack
 ; X64-MinGW-NEXT:  retq
 
 ; X64-FreeBSD:       leaq -40008(%rsp), %r11
 ; X64-FreeBSD-NEXT:  cmpq %fs:24, %r11
 ; X64-FreeBSD-NEXT:  ja      .LBB2_2
 
 ; X64-FreeBSD:       movabsq $40008, %r10
 ; X64-FreeBSD-NEXT:  movabsq $0, %r11
 ; X64-FreeBSD-NEXT:  callq __morestack
 ; X64-FreeBSD-NEXT:  ret
 
-; X32-DFlyBSD:       leal -40008(%esp), %ecx
+; X32-DFlyBSD:       leal -40000(%esp), %ecx
 ; X32-DFlyBSD-NEXT:  cmpl %fs:16, %ecx
 ; X32-DFlyBSD-NEXT:  ja      .LBB2_2
 
 ; X32-DFlyBSD:       pushl $0
-; X32-DFlyBSD-NEXT:  pushl $40008
+; X32-DFlyBSD-NEXT:  pushl $40000
 ; X32-DFlyBSD-NEXT:  calll __morestack
 ; X32-DFlyBSD-NEXT:  ret
 
 ; X64-DFlyBSD:       leaq -40008(%rsp), %r11
 ; X64-DFlyBSD-NEXT:  cmpq %fs:32, %r11
 ; X64-DFlyBSD-NEXT:  ja      .LBB2_2
 
 ; X64-DFlyBSD:       movabsq $40008, %r10
define fastcc void @test_fastcc() #0 {
         ret void
 
 ; X32-Linux-LABEL:       test_fastcc:
 
 ; X32-Linux:       cmpl %gs:48, %esp
 ; X32-Linux-NEXT:  ja      .LBB3_2
 
 ; X32-Linux:       pushl $0
-; X32-Linux-NEXT:  pushl $60
+; X32-Linux-NEXT:  pushl $44
 ; X32-Linux-NEXT:  calll __morestack
 ; X32-Linux-NEXT:  ret
 
 ; X64-Linux-LABEL:       test_fastcc:
 
 ; X64-Linux:       cmpq %fs:112, %rsp
 ; X64-Linux-NEXT:  ja      .LBB3_2
 

 ; X64-Darwin-NEXT: ret
 
 ; X32-MinGW-LABEL:       test_fastcc:
 
 ; X32-MinGW:       cmpl %fs:20, %esp
 ; X32-MinGW-NEXT:  ja      LBB3_2
 
 ; X32-MinGW:       pushl $0
-; X32-MinGW-NEXT:  pushl $48
+; X32-MinGW-NEXT:  pushl $40
 ; X32-MinGW-NEXT:  calll ___morestack
 ; X32-MinGW-NEXT:  ret
 
 ; X64-MinGW-LABEL:       test_fastcc:
 
 ; X64-MinGW:       cmpq %gs:40, %rsp
 ; X64-MinGW-NEXT:  ja      .LBB3_2
 

 ; X64-FreeBSD-NEXT:  ret
 
 ; X32-DFlyBSD-LABEL:       test_fastcc:
 
 ; X32-DFlyBSD:       cmpl %fs:16, %esp
 ; X32-DFlyBSD-NEXT:  ja      .LBB3_2
 
 ; X32-DFlyBSD:       pushl $0
-; X32-DFlyBSD-NEXT:  pushl $48
+; X32-DFlyBSD-NEXT:  pushl $40
 ; X32-DFlyBSD-NEXT:  calll __morestack
 ; X32-DFlyBSD-NEXT:  ret
 
 ; X64-DFlyBSD-LABEL:       test_fastcc:
 
 ; X64-DFlyBSD:       cmpq %fs:32, %rsp
 ; X64-DFlyBSD-NEXT:  ja      .LBB3_2
 

 
 ; X64-Darwin:      movabsq $40008, %r10
 ; X64-Darwin-NEXT: movabsq $0, %r11
 ; X64-Darwin-NEXT: callq ___morestack
 ; X64-Darwin-NEXT: ret
 
 ; X32-MinGW-LABEL:       test_fastcc_large:
 
-; X32-MinGW:       leal -40008(%esp), %eax
+; X32-MinGW:       leal -40000(%esp), %eax
 ; X32-MinGW-NEXT:  cmpl %fs:20, %eax
 ; X32-MinGW-NEXT:  ja      LBB4_2
 
 ; X32-MinGW:       pushl $0
-; X32-MinGW-NEXT:  pushl $40008
+; X32-MinGW-NEXT:  pushl $40000
 ; X32-MinGW-NEXT:  calll ___morestack
 ; X32-MinGW-NEXT:  ret
 
 ; X64-MinGW-LABEL:       test_fastcc_large:
 
 ; X64-MinGW:       leaq -40040(%rsp), %r11
 ; X64-MinGW-NEXT:  cmpq %gs:40, %r11
 ; X64-MinGW-NEXT:  ja      .LBB4_2

 
 ; X64-FreeBSD:       movabsq $40008, %r10
 ; X64-FreeBSD-NEXT:  movabsq $0, %r11
 ; X64-FreeBSD-NEXT:  callq __morestack
 ; X64-FreeBSD-NEXT:  ret
 
 ; X32-DFlyBSD-LABEL:       test_fastcc_large:
 
-; X32-DFlyBSD:       leal -40008(%esp), %eax
+; X32-DFlyBSD:       leal -40000(%esp), %eax
 ; X32-DFlyBSD-NEXT:  cmpl %fs:16, %eax
 ; X32-DFlyBSD-NEXT:  ja      .LBB4_2
 
 ; X32-DFlyBSD:       pushl $0
-; X32-DFlyBSD-NEXT:  pushl $40008
+; X32-DFlyBSD-NEXT:  pushl $40000
 ; X32-DFlyBSD-NEXT:  calll __morestack
 ; X32-DFlyBSD-NEXT:  ret
 
 ; X64-DFlyBSD-LABEL:       test_fastcc_large:
 
 ; X64-DFlyBSD:       leaq -40008(%rsp), %r11
 ; X64-DFlyBSD-NEXT:  cmpq %fs:32, %r11
 ; X64-DFlyBSD-NEXT:  ja      .LBB4_2

llvm/trunk/test/CodeGen/X86/seh-catch-all-win32.ll

 ; CHECK: popl %edi
 ; CHECK: popl %ebx
 ; CHECK: retl
 ; CHECK: LBB0_[[lpbb:[0-9]+]]: # %__except{{$}}
 ;       stackrestore
 ; CHECK: movl -24(%ebp), %esp
 ;       EH state -1
 ; CHECK: movl [[code_offs]](%ebp), %[[code:[a-z]+]]
-; CHECK-DAG: movl %[[code]], 4(%esp)
-; CHECK-DAG: movl $_str, (%esp)
+; CHECK: pushl %[[code]]
+; CHECK: pushl $_str
 ; CHECK: calll _printf
 
 ; CHECK: .section .xdata,"dr"
 ; CHECK: Lmain$parent_frame_offset = [[reg_offs]]
 ; CHECK: .p2align 2
 ; CHECK: L__ehtable$main
 ; CHECK-NEXT: .long -1
 ; CHECK-NEXT: .long _filt$main

llvm/trunk/test/CodeGen/X86/seh-stack-realign.ll

 
 ; Check that we can get the exception code from eax to the printf.
 
 ; CHECK-LABEL: _main:
 ; CHECK: Lmain$frame_escape_0 = [[code_offs:[-0-9]+]]
 ; CHECK: movl %esp, [[reg_offs:[-0-9]+]](%esi)
 ; CHECK: movl $L__ehtable$main,
 ;       EH state 0
-; CHECK: movl $0, 40(%esi)
+; CHECK: movl $0, 32(%esi)
 ; CHECK: calll _crash
 ; CHECK: retl
 ; CHECK: LBB0_[[lpbb:[0-9]+]]: # %__except
 ;       Restore ESP
 ; CHECK: movl -24(%ebp), %esp
 ;       Restore ESI
-; CHECK: leal -44(%ebp), %esi
+; CHECK: leal -36(%ebp), %esi
 ;       Restore EBP
-; CHECK: movl 12(%esi), %ebp
+; CHECK: movl 4(%esi), %ebp
 ; CHECK: movl [[code_offs]](%esi), %[[code:[a-z]+]]
-; CHECK-DAG: movl %[[code]], 4(%esp)
-; CHECK-DAG: movl $_str, (%esp)
+; CHECK: pushl %[[code]]
+; CHECK: pushl $_str
 ; CHECK: calll _printf
 
 ; CHECK: .section .xdata,"dr"
 ; CHECK: Lmain$parent_frame_offset = [[reg_offs]]
 ; CHECK: L__ehtable$main
 ; CHECK-NEXT: .long -1
 ; CHECK-NEXT: .long _filt$main
 ; CHECK-NEXT: .long LBB0_[[lpbb]]

llvm/trunk/test/CodeGen/X86/shrink-wrap-chkstk.ll

false:
   ret i32 %tmp.0
 }
 
 ; CHECK-LABEL: @use_eax_before_prologue@8: # @use_eax_before_prologue
 ; CHECK: movl %ecx, %eax
 ; CHECK: cmpl %edx, %eax
 ; CHECK: jge LBB1_2
 ; CHECK: pushl %eax
-; CHECK: movl $4100, %eax
+; CHECK: movl $4092, %eax
 ; CHECK: calll __chkstk
-; CHECK: movl 4100(%esp), %eax
+; CHECK: movl 4092(%esp), %eax
 ; CHECK: calll _doSomething
 ; CHECK: LBB1_2:
 ; CHECK: retl

llvm/trunk/test/CodeGen/X86/sse-intel-ocl.ll

 ; RUN: llc < %s -mtriple=i386-pc-win32 -mcpu=nehalem | FileCheck -check-prefix=WIN32 %s
 ; RUN: llc < %s -mtriple=x86_64-win32 -mcpu=nehalem | FileCheck -check-prefix=WIN64 %s
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=nehalem | FileCheck -check-prefix=NOT_WIN %s
 
 declare <16 x float> @func_float16_ptr(<16 x float>, <16 x float> *)
 declare <16 x float> @func_float16(<16 x float>, <16 x float>)
 ; WIN64: testf16_inp
 ; WIN64: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN64: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN64: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN64: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN64: leaq    {{.*}}(%rsp), %rcx
 ; WIN64: call
 ; WIN64: ret
 
 ; WIN32: testf16_inp
-; WIN32: movl    %eax, (%esp)
+; WIN32: pushl   %eax
 ; WIN32: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN32: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN32: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN32: addps  {{.*}}, {{%xmm[0-3]}}
 ; WIN32: call
 ; WIN32: ret
 
 ; NOT_WIN: testf16_inp

llvm/trunk/test/CodeGen/X86/tailcall-stackalign.ll

-; RUN: llc < %s  -mtriple=i686-unknown-linux  -tailcallopt | FileCheck %s
+; RUN: llc < %s  -mtriple=i686-unknown-linux  -tailcallopt -no-x86-call-frame-opt | FileCheck %s
 ; Linux has 8 byte alignment so the params cause stack size 20 when tailcallopt
 ; is enabled, ensure that a normal fastcc call has matching stack size
 
 
 define fastcc i32 @tailcallee(i32 %a1, i32 %a2, i32 %a3, i32 %a4) {
        ret i32 %a3
 }
 

llvm/trunk/test/CodeGen/X86/twoaddr-coalesce.ll

-; RUN: llc < %s -march=x86 | grep mov | count 4
+; RUN: llc < %s -march=x86 | grep mov | count 2
 ; rdar://6523745
 
 @"\01LC" = internal constant [4 x i8] c"%d\0A\00"    ; <[4 x i8]*> [#uses=1]
 
 define i32 @foo() nounwind {
 bb1.thread:
   br label %bb1
 

llvm/trunk/test/CodeGen/X86/vararg-callee-cleanup.ll

-; RUN: llc -mtriple=i686-pc-windows < %s | FileCheck %s
+; RUN: llc -mtriple=i686-pc-windows -no-x86-call-frame-opt < %s | FileCheck %s
 
 target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
 
 declare x86_thiscallcc void @thiscall_thunk(i8* %this, ...)
 define i32 @call_varargs_thiscall_thunk(i8* %a, i32 %b, i32 %c, i32 %d) {
   call x86_thiscallcc void (i8*, ...) @thiscall_thunk(i8* %a, i32 1, i32 2)
   call x86_thiscallcc void (i8*, ...) @thiscall_thunk(i8* %a, i32 1, i32 2)
   %t1 = add i32 %b, %c

llvm/trunk/test/CodeGen/X86/win-catchpad-csrs.ll

 ; X86: pushl %esi
 ; X86: subl ${{[0-9]+}}, %esp
 ; X86: calll _getint
 ; X86: calll _getint
 ; X86: calll _getint
 ; X86: calll _getint
 ; X86: calll _useints
 ; X86: movl $0, -{{[0-9]+}}(%ebp)
-; X86: movl $1, (%esp)
+; X86: pushl $1
 ; X86: calll _f
 ; X86: [[contbb:LBB0_[0-9]+]]: # %try.cont
 ; X86: popl %esi
 ; X86: popl %edi
 ; X86: popl %ebx
 ; X86: popl %ebp
 ; X86: retl
 
 ; X86: [[restorebb:LBB0_[0-9]+]]:
 ; X86: addl $12, %ebp
 ; X86: jmp [[contbb]]
 
 ; X86: "?catch$[[catch1bb:[0-9]+]]@?0?try_catch_catch@4HA":
 ; X86: LBB0_[[catch1bb]]: # %handler1{{$}}
 ; X86: pushl %ebp
 ; X86-NOT: pushl
 ; X86: subl $16, %esp
 ; X86: addl $12, %ebp
 ; X86: movl $1, -{{[0-9]+}}(%ebp)
-; X86: movl $2, (%esp)
+; X86: pushl $2
 ; X86: calll _f
 ; X86: movl $[[restorebb]], %eax
 ; X86-NEXT: addl $16, %esp
 ; X86-NEXT: popl %ebp
 ; X86-NEXT: retl
 
 ; X86: L__ehtable$try_catch_catch:
 ; X86: $handlerMap$0$try_catch_catch:

llvm/trunk/test/CodeGen/X86/win-catchpad.ll

 try.cont:
   ret i32 0
 }
 
 ; X86-LABEL: _try_catch_catch:
 ; X86: movl %esp, -[[sp_offset:[0-9]+]](%ebp)
 ; X86: movl $0, -{{[0-9]+}}(%ebp)
 ; X86: leal -[[local_offs:[0-9]+]](%ebp), %[[addr_reg:[a-z]+]]
-; X86-DAG: movl %[[addr_reg]], 4(%esp)
-; X86-DAG: movl $1, (%esp)
+; X86-DAG: pushl %[[addr_reg]]
+; X86-DAG: pushl $1
 ; X86: calll _f
 ; X86: [[contbb:LBB0_[0-9]+]]: # %try.cont
 ; X86: retl
 
 ; X86: [[restorebb1:LBB0_[0-9]+]]: # Block address taken
 ; X86-NEXT:                        # %handler1
 ; X86-NEXT: addl $12, %ebp
 ; X86: jmp [[contbb]]
 
 ; FIXME: These should be de-duplicated.
 ; X86: [[restorebb2:LBB0_[0-9]+]]: # Block address taken
 ; X86-NEXT:                        # %handler2
 ; X86-NEXT: addl $12, %ebp
 ; X86: jmp [[contbb]]
 
 ; X86: "?catch$[[catch1bb:[0-9]+]]@?0?try_catch_catch@4HA":
 ; X86: LBB0_[[catch1bb]]: # %handler1{{$}}
 ; X86: pushl %ebp
 ; X86: subl $8, %esp
 ; X86: addl $12, %ebp
 ; X86: movl %esp, -[[sp_offset]](%ebp)
 ; X86-DAG: movl -32(%ebp), %[[e_reg:[a-z]+]]
 ; X86-DAG: leal -[[local_offs]](%ebp), %[[addr_reg:[a-z]+]]
 ; X86-DAG: movl $1, -{{[0-9]+}}(%ebp)
-; X86-DAG: movl %[[addr_reg]], 4(%esp)
-; X86-DAG: movl %[[e_reg]], (%esp)
+; X86: pushl %[[addr_reg]]
+; X86: pushl %[[e_reg]]
 ; X86: calll _f
-; X86-NEXT: movl $[[restorebb1]], %eax
-; X86-NEXT: addl $8, %esp
-; X86-NEXT: popl %ebp
-; X86-NEXT: retl
+; X86: addl $8, %esp
+; X86: movl $[[restorebb1]], %eax
+; X86: addl $8, %esp
+; X86: popl %ebp
+; X86: retl
 
 ; X86: "?catch$[[catch2bb:[0-9]+]]@?0?try_catch_catch@4HA":
 ; X86: LBB0_[[catch2bb]]: # %handler2{{$}}
 ; X86: pushl %ebp
 ; X86: subl $8, %esp
 ; X86: addl $12, %ebp
 ; X86: movl %esp, -[[sp_offset]](%ebp)
 ; X86-DAG: leal -[[local_offs]](%ebp), %[[addr_reg:[a-z]+]]
 ; X86-DAG: movl $1, -{{[0-9]+}}(%ebp)
-; X86-DAG: movl %[[addr_reg]], 4(%esp)
-; X86-DAG: movl $3, (%esp)
+; X86: pushl %[[addr_reg]]
+; X86: pushl $3
 ; X86: calll _f
-; X86-NEXT: movl $[[restorebb2]], %eax
-; X86-NEXT: addl $8, %esp
-; X86-NEXT: popl %ebp
-; X86-NEXT: retl
+; X86: addl $8, %esp
+; X86: movl $[[restorebb2]], %eax
+; X86: addl $8, %esp
+; X86: popl %ebp
+; X86: retl
 
 ; X86: L__ehtable$try_catch_catch:
 ; X86: $handlerMap$0$try_catch_catch:
 ; X86-NEXT:   .long   0
 ; X86-NEXT:   .long   "??_R0H@8"
 ; X86-NEXT:   .long   -20
 ; X86-NEXT:   .long   "?catch$[[catch1bb]]@?0?try_catch_catch@4HA"
 ; X86-NEXT:   .long   64

llvm/trunk/test/CodeGen/X86/win-cleanuppad.ll

 
 cleanup.outer:                                      ; preds = %invoke.cont.1, %cleanup.inner, %entry
   %1 = cleanuppad within none []
   call x86_thiscallcc void @"\01??1Dtor@@QAE@XZ"(%struct.Dtor* %o1) #2 [ "funclet"(token %1) ]
   cleanupret from %1 unwind to caller
 }
 
 ; X86-LABEL: _nested_cleanup:
-; X86: movl    $1, (%esp)
+; X86: pushl   $1
 ; X86: calll   _f
-; X86: movl    $2, (%esp)
+; X86: pushl   $2
 ; X86: calll   _f
-; X86: movl    $3, (%esp)
+; X86: pushl   $3
 ; X86: calll   _f
 
 ; X86: "?dtor$[[cleanup_inner:[0-9]+]]@?0?nested_cleanup@4HA":
 ; X86: LBB1_[[cleanup_inner]]: # %cleanup.inner{{$}}
 ; X86: pushl %ebp
 ; X86: leal    {{.*}}(%ebp), %ecx
 ; X86: calll   "??1Dtor@@QAE@XZ"
 ; X86: popl %ebp

llvm/trunk/test/CodeGen/X86/win32-eh-states.ll

catch.7:
   catchret from %p2 to label %try.cont.9
 }
 
 ; X86-LABEL: _f:
 ; X86: movl $-1, [[state:[-0-9]+]](%ebp)
 ; X86: movl $___ehhandler$f, {{.*}}
 ;
 ; X86: movl $0, [[state]](%ebp)
-; X86: movl $1, (%esp)
+; X86: pushl $1
 ; X86: calll _may_throw
 ;
 ; X86: movl $1, [[state]](%ebp)
-; X86: movl $2, (%esp)
+; X86: pushl $2
 ; X86: calll _may_throw
 ;
 ; X86: movl $2, [[state]](%ebp)
-; X86: movl $3, (%esp)
+; X86: pushl $3
 ; X86: calll _may_throw
 ;
 ; X86: movl $3, [[state]](%ebp)
-; X86: movl $4, (%esp)
+; X86: pushl $4
 ; X86: calll _may_throw
 
 
 ; X64-LABEL: f:
 ; X64-LABEL: $ip2state$f:
 ; X64-NEXT:   .long .Lfunc_begin0@IMGREL
 ; X64-NEXT:   .long -1
 ; X64-NEXT:   .long .Ltmp{{.*}}@IMGREL+1
unreachable:                                      ; preds = %entry
   unreachable
 }
 
 ; X86-LABEL: _g:
 ; X86: movl $-1, [[state:[-0-9]+]](%ebp)
 ; X86: movl $___ehhandler$g, {{.*}}
 ;
 ; X86: movl $1, [[state]](%ebp)
-; X86: movl $-1, (%esp)
+; X86: pushl $-1
 ; X86: calll _may_throw
 ;
 ; X86: movl $2, [[state]](%ebp)
-; X86: movl $0, (%esp)
+; X86: pushl $0
 ; X86: calll _may_throw
 ;
 ; X86: movl $3, [[state]](%ebp)
-; X86: movl $1, (%esp)
+; X86: pushl $1
 ; X86: calll _may_throw
 ;
 ; X86: movl $2, [[state]](%ebp)
-; X86: movl $2, (%esp)
+; X86: pushl $2
 ; X86: calll _may_throw
 
 ; X64-LABEL: g:
 ; X64-LABEL: $ip2state$g:
 ; X64-NEXT:   .long .Lfunc_begin1@IMGREL
 ; X64-NEXT:   .long -1
 ; X64-NEXT:   .long .Ltmp{{.*}}@IMGREL+1
 ; X64-NEXT:   .long 1

llvm/trunk/test/CodeGen/X86/win32-seh-catchpad.ll

 
 invoke.cont:                                      ; preds = %entry
   br label %__try.cont
 }
 
 ; CHECK-LABEL: _try_except:
 ;     Store state #0
 ; CHECK: movl $0, -[[state:[0-9]+]](%ebp)
-; CHECK: movl $1, (%esp)
+; CHECK: pushl $1
 ; CHECK: calll _f
 ; CHECK: movl $-1, -[[state]](%ebp)
-; CHECK: movl $3, (%esp)
+; CHECK: pushl $3
 ; CHECK: calll _f
 ; CHECK: retl
 
 ;   __except
 ; CHECK: movl $-1, -[[state]](%ebp)
-; CHECK: movl $2, (%esp)
+; CHECK: pushl $2
 ; CHECK: calll _f
 
 ; CHECK: .section        .xdata,"dr"
 ; CHECK: L__ehtable$try_except:
 ; CHECK:         .long   -1                          # ToState
 ; CHECK:         .long   _try_except_filter_catchall # Filter
 ; CHECK:         .long   LBB0_1
 
__except:
   ret void
 }
 
 ; CHECK-LABEL: _code_in_catchpad:
 ; CHECK: # %__except.ret
 ; CHECK-NEXT:         movl    -24(%ebp), %esp
 ; CHECK-NEXT:         addl    $12, %ebp
 ; CHECK-NEXT:         movl    $-1, -16(%ebp)
-; CHECK-NEXT:         movl    $2, (%esp)
+; CHECK-NEXT:         pushl   $2
 ; CHECK-NEXT:         calll   _f
 
 
 ; Function Attrs: nounwind readnone
 declare i8* @llvm.frameaddress(i32) #1
 
 ; Function Attrs: nounwind readnone
 declare i8* @llvm.x86.seh.recoverfp(i8*, i8*) #1

llvm/trunk/test/CodeGen/X86/win32-seh-nested-finally.ll

 attributes #1 = { noinline nounwind "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
 attributes #2 = { nounwind readnone }
 attributes #3 = { noinline }
 
 ; CHECK: _nested_finally:
 ; CHECK: movl $-1, -[[state:[0-9]+]](%ebp)
 ; CHECK: movl {{.*}}, %fs:0
 ; CHECK: movl $1, -[[state]](%ebp)
-; CHECK: movl $1, (%esp)
+; CHECK: pushl $1
 ; CHECK: calll _f
+; CHECK: addl $4, %esp
 ; CHECK: movl $0, -[[state]](%ebp)
-; CHECK: movl $2, (%esp)
+; CHECK: pushl $2
 ; CHECK: calll _f
+; CHECK: addl $4, %esp
 ; CHECK: movl $-1, -[[state]](%ebp)
-; CHECK: movl $3, (%esp)
+; CHECK: pushl $3
 ; CHECK: calll _f
+; CHECK: addl $4, %esp
 ; CHECK: retl
 
 ; CHECK: LBB0_[[inner:[0-9]+]]: # %ehcleanup
 ; CHECK: pushl %ebp
 ; CHECK: addl $12, %ebp
-; CHECK: movl $2, (%esp)
+; CHECK: pushl $2
 ; CHECK: calll _f
 ; CHECK: popl %ebp
 ; CHECK: retl
 
 ; CHECK: LBB0_[[outer:[0-9]+]]: # %ehcleanup.3
 ; CHECK: pushl %ebp
 ; CHECK: addl $12, %ebp
-; CHECK: movl $3, (%esp)
+; CHECK: pushl $3
 ; CHECK: calll _f
 ; CHECK: popl %ebp
 ; CHECK: retl
 
 ; CHECK: L__ehtable$nested_finally:
 ; CHECK:        .long   -1 # ToState
 ; CHECK:        .long   0  # Null
 ; CHECK:        .long   "?dtor$[[outer]]@?0?nested_finally@4HA" # FinallyFunclet
 ; CHECK:        .long   0  # ToState
 ; CHECK:        .long   0  # Null
 ; CHECK:        .long   "?dtor$[[inner]]@?0?nested_finally@4HA" # FinallyFunclet

llvm/trunk/test/CodeGen/X86/win32_sret.ll

entry:
   call x86_thiscallcc void @"\01?foo@C5@@QAE?AUS5@@XZ"(%struct.S5* sret %s, %class.C5* %c)
 ; WIN32-LABEL:      {{^}}_call_foo5:
 ; MINGW_X86-LABEL:  {{^}}_call_foo5:
 ; CYGWIN-LABEL:     {{^}}_call_foo5:
 ; LINUX-LABEL:      {{^}}call_foo5:
 
 
 ; Load the address of the result and put it onto stack
-; (through %ecx in the -O0 build).
-; WIN32:      leal {{[0-9]+}}(%esp), %e{{[a-d]}}x
-; WIN32:      movl %e{{[a-d]}}x, (%e{{([a-d]x)|(sp)}})
-
 ; The this pointer goes to ECX.
-; WIN32-NEXT: leal {{[0-9]+}}(%esp), %ecx
+; (through %ecx in the -O0 build).
+; WIN32:      leal {{[0-9]*}}(%esp), %e{{[a-d]}}x
+; WIN32:      leal {{[0-9]*}}(%esp), %ecx
+; WIN32:      pushl %e{{[a-d]}}x
 ; WIN32-NEXT: calll "?foo@C5@@QAE?AUS5@@XZ"
 ; WIN32:      retl
   ret void
 }
 
 
 %struct.test6 = type { i32, i32, i32 }
 define void @test6_f(%struct.test6* %x) nounwind {
 ; WIN32-LABEL: _test6_f:
 ; MINGW_X86-LABEL: _test6_f:
 ; CYGWIN-LABEL: _test6_f:
 ; LINUX-LABEL: test6_f:
 
 ; The %x argument is moved to %ecx. It will be the this pointer.
-; WIN32: movl    20(%esp), %ecx
-
-; The %x argument is moved to (%esp). It will be the this pointer. With -O0
-; we copy esp to ecx and use (ecx) instead of (esp).
-; MINGW_X86: movl    20(%esp), %eax
-; MINGW_X86: movl    %eax, (%e{{([a-d]x)|(sp)}})
+; WIN32: movl    16(%esp), %ecx
 
-; CYGWIN: movl    20(%esp), %eax
-; CYGWIN: movl    %eax, (%e{{([a-d]x)|(sp)}})
 
 ; The sret pointer is (%esp)
-; WIN32:          leal    4(%esp), %[[REG:e[a-d]x]]
-; WIN32-NEXT:     movl    %[[REG]], (%e{{([a-d]x)|(sp)}})
+; WIN32:          leal    (%esp), %[[REG:e[a-d]x]]
+; WIN32-NEXT:     pushl   %[[REG]]
 
 ; The sret pointer is %ecx
-; MINGW_X86-NEXT: leal    4(%esp), %ecx
+; The %x argument is moved to (%esp). It will be the this pointer.
+; MINGW_X86:      leal    (%esp), %ecx
+; MINGW_X86-NEXT: pushl   16(%esp)
 ; MINGW_X86-NEXT: calll   _test6_g
 
-; CYGWIN-NEXT: leal    4(%esp), %ecx
+; CYGWIN:      leal    (%esp), %ecx
+; CYGWIN-NEXT: pushl   16(%esp)
 ; CYGWIN-NEXT: calll   _test6_g
 
   %tmp = alloca %struct.test6, align 4
   call x86_thiscallcc void @test6_g(%struct.test6* sret %tmp, %struct.test6* %x)
   ret void
 }
 declare x86_thiscallcc void @test6_g(%struct.test6* sret, %struct.test6*)
 
 ; Flipping the parameters at the IR level generates the same code.
 %struct.test7 = type { i32, i32, i32 }
 define void @test7_f(%struct.test7* %x) nounwind {
 ; WIN32-LABEL: _test7_f:
 ; MINGW_X86-LABEL: _test7_f:
 ; CYGWIN-LABEL: _test7_f:
 ; LINUX-LABEL: test7_f:
 
 ; The %x argument is moved to %ecx on all OSs. It will be the this pointer.
-; WIN32:      movl    20(%esp), %ecx
-; MINGW_X86:  movl    20(%esp), %ecx
-; CYGWIN:     movl    20(%esp), %ecx
+; WIN32:      movl    16(%esp), %ecx
+; MINGW_X86:  movl    16(%esp), %ecx
+; CYGWIN:     movl    16(%esp), %ecx
 
 ; The sret pointer is (%esp)
-; WIN32:          leal    4(%esp), %[[REG:e[a-d]x]]
-; WIN32-NEXT:     movl    %[[REG]], (%e{{([a-d]x)|(sp)}})
-; MINGW_X86:      leal    4(%esp), %[[REG:e[a-d]x]]
-; MINGW_X86-NEXT: movl    %[[REG]], (%e{{([a-d]x)|(sp)}})
-; CYGWIN:         leal    4(%esp), %[[REG:e[a-d]x]]
-; CYGWIN-NEXT:    movl    %[[REG]], (%e{{([a-d]x)|(sp)}})
+; WIN32:          leal    (%esp), %[[REG:e[a-d]x]]
+; WIN32-NEXT:     pushl   %[[REG]]
+; MINGW_X86:      leal    (%esp), %[[REG:e[a-d]x]]
+; MINGW_X86-NEXT: pushl   %[[REG]]
+; CYGWIN:         leal    (%esp), %[[REG:e[a-d]x]]
+; CYGWIN-NEXT:    pushl   %[[REG]]
 
   %tmp = alloca %struct.test7, align 4
   call x86_thiscallcc void @test7_g(%struct.test7* %x, %struct.test7* sret %tmp)
   ret void
 }
 
 define x86_thiscallcc void @test7_g(%struct.test7* %in, %struct.test7* sret %out) {
   %s = getelementptr %struct.test7, %struct.test7* %in, i32 0, i32 0

llvm/trunk/test/CodeGen/X86/xmulo.ll

 ; RUN: llc %s -o - | FileCheck %s
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128"
 target triple = "i386-apple-macosx10.8.0"
 
 declare {i64, i1} @llvm.umul.with.overflow.i64(i64, i64) nounwind readnone
 declare i32 @printf(i8*, ...)
 
 @.str = private unnamed_addr constant [10 x i8] c"%llx, %d\0A\00", align 1
 
 define i32 @t1() nounwind {
 ; CHECK-LABEL: t1:
-; CHECK:  movl $0, 12(%esp)
-; CHECK:  movl $0, 8(%esp)
-; CHECK:  movl $72, 4(%esp)
+; CHECK:  pushl $0
+; CHECK:  pushl $0
+; CHECK:  pushl $72
 
     %1 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 9, i64 8)
     %2 = extractvalue {i64, i1} %1, 0
     %3 = extractvalue {i64, i1} %1, 1
     %4 = zext i1 %3 to i32
     %5 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([10 x i8], [10 x i8]* @.str, i32 0, i32 0), i64 %2, i32 %4)
     ret i32 0
 }
 
 define i32 @t2() nounwind {
 ; CHECK-LABEL: t2:
-; CHECK:  movl $0, 12(%esp)
-; CHECK:  movl $0, 8(%esp)
-; CHECK:  movl $0, 4(%esp)
+; CHECK:  pushl $0
+; CHECK:  pushl $0
+; CHECK:  pushl $0
 
     %1 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 9, i64 0)
     %2 = extractvalue {i64, i1} %1, 0
     %3 = extractvalue {i64, i1} %1, 1
     %4 = zext i1 %3 to i32
     %5 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([10 x i8], [10 x i8]* @.str, i32 0, i32 0), i64 %2, i32 %4)
     ret i32 0
 }
 
 define i32 @t3() nounwind {
 ; CHECK-LABEL: t3:
-; CHECK:  movl $1, 12(%esp)
-; CHECK:  movl $-1, 8(%esp)
-; CHECK:  movl $-9, 4(%esp)
+; CHECK:  pushl $1
+; CHECK:  pushl $-1
+; CHECK:  pushl $-9
 
     %1 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 9, i64 -1)
     %2 = extractvalue {i64, i1} %1, 0
     %3 = extractvalue {i64, i1} %1, 1
     %4 = zext i1 %3 to i32
     %5 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([10 x i8], [10 x i8]* @.str, i32 0, i32 0), i64 %2, i32 %4)
     ret i32 0
 }

llvm/trunk/test/CodeGen/X86/zext-fold.ll

 define void @test3(i8 %x) nounwind readnone {
   %A = and i8 %x, -32
   %B = zext i8 %A to i32
   call void @use(i32 %B, i8 %x)
   ret void
 }
 ; CHECK: test3
 ; CHECK: movzbl {{[0-9]+}}(%esp), [[REGISTER:%e[a-z]{2}]]
-; CHECK-NEXT: movl [[REGISTER]], 4(%esp)
+; CHECK: subl $8, %esp
+; CHECK-NEXT: pushl [[REGISTER]]
 ; CHECK-NEXT: andl $224, [[REGISTER]]
-; CHECK-NEXT: movl [[REGISTER]], (%esp)
+; CHECK-NEXT: pushl [[REGISTER]]
 ; CHECK-NEXT: call{{.*}}use