Skip to content

[NVPTX] Add intrinsic range to nvvm_read_ptx_sreg_laneid #153099

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 1 commit into from
Closed

[NVPTX] Add intrinsic range to nvvm_read_ptx_sreg_laneid #153099

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
4 changes: 4 additions & 0 deletions llvm/lib/Target/NVPTX/NVVMIntrRange.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,6 +130,10 @@ static bool runNVVMIntrRange(Function &F) {
if (OverallClusterRank)
return addRangeAttr(1, FunctionClusterRank + 1, II);
break;

// Lane ID
case Intrinsic::nvvm_read_ptx_sreg_laneid:
return addRangeAttr(0, 32, II);
Comment on lines +134 to +136
Copy link
Member

Choose a reason for hiding this comment

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

Looks like we already attach this range info to the declaration of the intrinsic via TableGen attrs. Is there any reason we need to attach it to each call as well?

llvm-project/llvm/include/llvm/IR/IntrinsicsNVVM.td

Lines 1759 to 1760 in ef50227

def int_nvvm_read_ptx_sreg_laneid
: PTXReadSRegIntrinsic_r32<[Range<RetIndex, 0, WARP_SIZE>]>;

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I added this for completeness of the pass. Won't the range info through TableGen attrs be available only while InstructionSelection? and not before.

Copy link
Member

Choose a reason for hiding this comment

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

As far as I know these attributes will be part of the IR and basically indistinguishable from if they were added to the call itself. Here is a quick demo:

https://cuda.godbolt.org/z/b6fEsKddc

Unless we find a use-case for adding to the call as well, I don't think it makes sense to include it here for completeness.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

that makes sense!

default:
return false;
}
Expand Down
10 changes: 10 additions & 0 deletions llvm/test/CodeGen/NVPTX/intr-range.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -135,6 +135,16 @@ define ptx_kernel i32 @test_cluster_dim() "nvvm.cluster_dim"="4,4,1" {
ret i32 %11
}

define ptx_kernel i32 @test_laneid() "nvvm.cluster_dim"="4,4,1" {
; CHECK-LABEL: define ptx_kernel i32 @test_laneid(
; CHECK-SAME: ) #[[ATTR4]] {
; CHECK-NEXT: [[TMP1:%.*]] = call range(i32 0, 32) i32 @llvm.nvvm.read.ptx.sreg.laneid()
; CHECK-NEXT: ret i32 [[TMP1]]
;
%1 = call i32 @llvm.nvvm.read.ptx.sreg.laneid()
ret i32 %1
}


; DEFAULT-DAG: declare noundef range(i32 0, 1024) i32 @llvm.nvvm.read.ptx.sreg.tid.x()
; DEFAULT-DAG: declare noundef range(i32 0, 1024) i32 @llvm.nvvm.read.ptx.sreg.tid.y()
Expand Down