maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/kernarg-stack-alignment.ll

45 lines
1.2 KiB
LLVM
Raw Normal View History

AMDGPU: Fix kernel argument alignment impacting stack size Don't use AllocateStack because kernel arguments have nothing to do with the stack. The ensureMaxAlignment call was still changing the stack alignment. llvm-svn: 273080
2016-06-18 13:15:53 +08:00
; RUN: llc -O0 -march=amdgcn -verify-machineinstrs < %s | FileCheck %s
; Test that the alignment of kernel arguments does not impact the
; alignment of the stack
; CHECK-LABEL: {{^}}no_args:
AMDGPU: Always allocate emergency stack slot at offset 0 This allows us to ensure that 0 is never a valid pointer to a user object, and ensures that the offset is always legal without needing a register to access it. This comes at the cost of usable offsets and wasted stack space. llvm-svn: 295877
2017-02-23 05:05:25 +08:00
; CHECK: ScratchSize: 5{{$}}
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @no_args() {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i8, addrspace(5)
store volatile i8 0, i8 addrspace(5)* %alloca
AMDGPU: Fix kernel argument alignment impacting stack size Don't use AllocateStack because kernel arguments have nothing to do with the stack. The ensureMaxAlignment call was still changing the stack alignment. llvm-svn: 273080
2016-06-18 13:15:53 +08:00
ret void
}
; CHECK-LABEL: {{^}}force_align32:
AMDGPU: Always allocate emergency stack slot at offset 0 This allows us to ensure that 0 is never a valid pointer to a user object, and ensures that the offset is always legal without needing a register to access it. This comes at the cost of usable offsets and wasted stack space. llvm-svn: 295877
2017-02-23 05:05:25 +08:00
; CHECK: ScratchSize: 5{{$}}
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @force_align32(<8 x i32>) {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i8, addrspace(5)
store volatile i8 0, i8 addrspace(5)* %alloca
AMDGPU: Fix kernel argument alignment impacting stack size Don't use AllocateStack because kernel arguments have nothing to do with the stack. The ensureMaxAlignment call was still changing the stack alignment. llvm-svn: 273080
2016-06-18 13:15:53 +08:00
ret void
}
; CHECK-LABEL: {{^}}force_align64:
AMDGPU: Always allocate emergency stack slot at offset 0 This allows us to ensure that 0 is never a valid pointer to a user object, and ensures that the offset is always legal without needing a register to access it. This comes at the cost of usable offsets and wasted stack space. llvm-svn: 295877
2017-02-23 05:05:25 +08:00
; CHECK: ScratchSize: 5{{$}}
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @force_align64(<16 x i32>) {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i8, addrspace(5)
store volatile i8 0, i8 addrspace(5)* %alloca
AMDGPU: Fix kernel argument alignment impacting stack size Don't use AllocateStack because kernel arguments have nothing to do with the stack. The ensureMaxAlignment call was still changing the stack alignment. llvm-svn: 273080
2016-06-18 13:15:53 +08:00
ret void
}
; CHECK-LABEL: {{^}}force_align128:
AMDGPU: Always allocate emergency stack slot at offset 0 This allows us to ensure that 0 is never a valid pointer to a user object, and ensures that the offset is always legal without needing a register to access it. This comes at the cost of usable offsets and wasted stack space. llvm-svn: 295877
2017-02-23 05:05:25 +08:00
; CHECK: ScratchSize: 5{{$}}
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @force_align128(<32 x i32>) {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i8, addrspace(5)
store volatile i8 0, i8 addrspace(5)* %alloca
AMDGPU: Fix kernel argument alignment impacting stack size Don't use AllocateStack because kernel arguments have nothing to do with the stack. The ensureMaxAlignment call was still changing the stack alignment. llvm-svn: 273080
2016-06-18 13:15:53 +08:00
ret void
}
; CHECK-LABEL: {{^}}force_align256:
AMDGPU: Always allocate emergency stack slot at offset 0 This allows us to ensure that 0 is never a valid pointer to a user object, and ensures that the offset is always legal without needing a register to access it. This comes at the cost of usable offsets and wasted stack space. llvm-svn: 295877
2017-02-23 05:05:25 +08:00
; CHECK: ScratchSize: 5{{$}}
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @force_align256(<64 x i32>) {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i8, addrspace(5)
store volatile i8 0, i8 addrspace(5)* %alloca
AMDGPU: Fix kernel argument alignment impacting stack size Don't use AllocateStack because kernel arguments have nothing to do with the stack. The ensureMaxAlignment call was still changing the stack alignment. llvm-svn: 273080
2016-06-18 13:15:53 +08:00
ret void
}