diff options
Diffstat (limited to 'test/CodeGen/X86/win32_sret.ll')
-rw-r--r-- | test/CodeGen/X86/win32_sret.ll | 83 |
1 files changed, 83 insertions, 0 deletions
diff --git a/test/CodeGen/X86/win32_sret.ll b/test/CodeGen/X86/win32_sret.ll index d8ecd44..8728712 100644 --- a/test/CodeGen/X86/win32_sret.ll +++ b/test/CodeGen/X86/win32_sret.ll @@ -2,9 +2,11 @@ ; some setups (e.g., Atom) from affecting the output. ; RUN: llc < %s -mcpu=core2 -mtriple=i686-pc-win32 | FileCheck %s -check-prefix=WIN32 ; RUN: llc < %s -mcpu=core2 -mtriple=i686-pc-mingw32 | FileCheck %s -check-prefix=MINGW_X86 +; RUN: llc < %s -mcpu=core2 -mtriple=i686-pc-cygwin | FileCheck %s -check-prefix=CYGWIN ; RUN: llc < %s -mcpu=core2 -mtriple=i386-pc-linux | FileCheck %s -check-prefix=LINUX ; RUN: llc < %s -mcpu=core2 -O0 -mtriple=i686-pc-win32 | FileCheck %s -check-prefix=WIN32 ; RUN: llc < %s -mcpu=core2 -O0 -mtriple=i686-pc-mingw32 | FileCheck %s -check-prefix=MINGW_X86 +; RUN: llc < %s -mcpu=core2 -O0 -mtriple=i686-pc-cygwin | FileCheck %s -check-prefix=CYGWIN ; RUN: llc < %s -mcpu=core2 -O0 -mtriple=i386-pc-linux | FileCheck %s -check-prefix=LINUX ; The SysV ABI used by most Unixes and Mingw on x86 specifies that an sret pointer @@ -21,6 +23,9 @@ entry: ; MINGW_X86-LABEL: _sret1: ; MINGW_X86: {{retl$}} +; CYGWIN-LABEL: _sret1: +; CYGWIN: retl $4 + ; LINUX-LABEL: sret1: ; LINUX: retl $4 @@ -38,6 +43,9 @@ entry: ; MINGW_X86-LABEL: _sret2: ; MINGW_X86: {{retl$}} +; CYGWIN-LABEL: _sret2: +; CYGWIN: retl $4 + ; LINUX-LABEL: sret2: ; LINUX: retl $4 @@ -56,6 +64,9 @@ entry: ; MINGW_X86-LABEL: _sret3: ; MINGW_X86: {{retl$}} +; CYGWIN-LABEL: _sret3: +; CYGWIN: retl $4 + ; LINUX-LABEL: sret3: ; LINUX: retl $4 @@ -77,6 +88,9 @@ entry: ; MINGW_X86-LABEL: _sret4: ; MINGW_X86: {{retl$}} +; CYGWIN-LABEL: _sret4: +; CYGWIN: retl $4 + ; LINUX-LABEL: sret4: ; LINUX: retl $4 @@ -98,6 +112,7 @@ entry: ret void ; WIN32-LABEL: {{^}}"?foo@C5@@QAE?AUS5@@XZ": ; MINGW_X86-LABEL: {{^}}"?foo@C5@@QAE?AUS5@@XZ": +; CYGWIN-LABEL: {{^}}"?foo@C5@@QAE?AUS5@@XZ": ; LINUX-LABEL: {{^}}"?foo@C5@@QAE?AUS5@@XZ": ; The address of the return structure is passed as an implicit parameter. @@ -115,6 +130,7 @@ 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: @@ -135,6 +151,7 @@ entry: 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. @@ -145,6 +162,9 @@ define void @test6_f(%struct.test6* %x) nounwind { ; MINGW_X86: movl 8(%ebp), %eax ; MINGW_X86: movl %eax, (%e{{([a-d]x)|(sp)}}) +; CYGWIN: movl 8(%ebp), %eax +; CYGWIN: movl %eax, (%e{{([a-d]x)|(sp)}}) + ; The sret pointer is (%esp) ; WIN32: leal 8(%esp), %[[REG:e[a-d]x]] ; WIN32-NEXT: movl %[[REG]], (%e{{([a-d]x)|(sp)}}) @@ -153,8 +173,71 @@ define void @test6_f(%struct.test6* %x) nounwind { ; MINGW_X86-NEXT: leal 8(%esp), %ecx ; MINGW_X86-NEXT: calll _test6_g +; CYGWIN-NEXT: leal 8(%esp), %ecx +; 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 8(%ebp), %ecx +; MINGW_X86: movl 8(%ebp), %ecx +; CYGWIN: movl 8(%ebp), %ecx + +; The sret pointer is (%esp) +; WIN32: leal 8(%esp), %[[REG:e[a-d]x]] +; WIN32-NEXT: movl %[[REG]], (%e{{([a-d]x)|(sp)}}) +; MINGW_X86: leal 8(%esp), %[[REG:e[a-d]x]] +; MINGW_X86-NEXT: movl %[[REG]], (%e{{([a-d]x)|(sp)}}) +; CYGWIN: leal 8(%esp), %[[REG:e[a-d]x]] +; CYGWIN-NEXT: movl %[[REG]], (%e{{([a-d]x)|(sp)}}) + + %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* %in, i32 0, i32 0 + %d = getelementptr %struct.test7* %out, i32 0, i32 0 + %v = load i32* %s + store i32 %v, i32* %d + call void @clobber_eax() + ret void + +; Make sure we return the second parameter in %eax. +; WIN32-LABEL: _test7_g: +; WIN32: calll _clobber_eax +; WIN32: movl {{.*}}, %eax +; WIN32: retl +} + +declare void @clobber_eax() + +; Test what happens if the first parameter has to be split by codegen. +; Realistically, no frontend will generate code like this, but here it is for +; completeness. +define void @test8_f(i64 inreg %a, i64* sret %out) { + store i64 %a, i64* %out + call void @clobber_eax() + ret void + +; WIN32-LABEL: _test8_f: +; WIN32: movl {{[0-9]+}}(%esp), %[[out:[a-z]+]] +; WIN32-DAG: movl %edx, 4(%[[out]]) +; WIN32-DAG: movl %eax, (%[[out]]) +; WIN32: calll _clobber_eax +; WIN32: movl {{.*}}, %eax +; WIN32: retl +} |