Skip to content

[NVPTX] Annotate CUDA kernel pointer arguments with .ptr .space .align attributes. #79646

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 14 commits into from
Closed

[NVPTX] Annotate CUDA kernel pointer arguments with .ptr .space .align attributes. #79646

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
d5bd021
Enable .ptr .global .align attributes for kernel attributes for CUDA
Vandana2896 Jan 26, 2024
a952cc1
Rearrange code, add comment
Vandana2896 Feb 2, 2024
f5e7276
Fixed clang formatting
Vandana2896 Feb 2, 2024
761d8a0
Update NVPTXAsmPrinter.cpp
Vandana2896 Feb 20, 2024
3d49f30
Update NVPTXAsmPrinter.cpp
Vandana2896 Feb 20, 2024
424667b
Update .global and .align
Vandana2896 Mar 11, 2024
637da98
Merge branch 'main' of https://github.com/Vandana2896/llvm-project in…
Vandana2896 Mar 11, 2024
3912211
Fix comment
Vandana2896 Mar 11, 2024
ce29dc1
add addrspace
Vandana2896 Apr 25, 2024
fbedea5
Fix typo
Vandana2896 Apr 25, 2024
75cce1b
CHECK-LABEL
Vandana2896 Apr 25, 2024
9fc5d6c
Update testcase
Vandana2896 Apr 27, 2024
9171d70
Merge branch 'main' into main
Vandana2896 Apr 30, 2024
14291a9
upadte test
Vandana2896 May 20, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 0 additions & 1 deletion llvm/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1597,7 +1597,6 @@ void NVPTXAsmPrinter::emitFunctionParamList(const Function *F, raw_ostream &O) {
// CUDA kernels assume that pointers are in global address space
// See:
// https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#parameter-state-space
assert(addrSpace == 0 && "Invalid address space");
Copy link
Member

@Artem-B Artem-B May 20, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this assertion is still (partially?) valid. The only case when we want to emit .global is if the pointer is generic or explicitly in the global AS.

O << ".ptr .global ";
if (I->getParamAlign().valueOrOne() != 1) {
Align ParamAlign = I->getParamAlign().value();
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/NVPTX/i1-param.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ target triple = "nvptx-nvidia-cuda"

; CHECK: .entry foo
; CHECK: .param .u8 foo_param_0
; CHECK: .param .u64 foo_param_1
; CHECK: .param .u64 .ptr .global foo_param_1
define void @foo(i1 %p, ptr %out) {
%val = zext i1 %p to i32
store i32 %val, ptr %out
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/NVPTX/kernel-param-align.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@
; CHECK-LABEL: func_align
; CHECK: .param .u64 .ptr .global .align 16 func_align_param_0,
; CHECK: .param .u64 .ptr .global func_align_param_1,
; CHECK: .param .u64 .ptr .global func_align_param_2
; CHECK: .param .u32 .ptr .global func_align_param_2
Copy link
Member

@Artem-B Artem-B May 20, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is still wrong. addrspace(3) is not a global pointer. It's a shared AS pointer.

Speaking of which, please also add a test for explicitly annotated const and global AS pointers.

define void @func_align(ptr nocapture readonly align 16 %input, ptr nocapture %out, ptr addrspace(3) %n) {
entry:
%0 = addrspacecast ptr %out to ptr addrspace(1)
Expand Down