Skip to content

[libspirv][ptx-nvidiacl] Change __clc__group_scratch size to 32 x i128 #18431

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

Merged
merged 5 commits into from
May 26, 2025
Merged

[libspirv][ptx-nvidiacl] Change __clc__group_scratch size to 32 x i128 #18431

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
0f5efc3
[libspirv][ptx-nvidiacl] Change __clc__group_scratch size to 32 x i128
wenju-he May 13, 2025
cf9ec4a
opaque pointer, add global variables for all sizes
wenju-he May 20, 2025
8898ee5
undef -> poison
wenju-he May 20, 2025
8c02604
remove gep
wenju-he May 22, 2025
0693f5e
i32 addrspace(3)* -> ptr addrspace(3)
wenju-he May 25, 2025
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
24 changes: 8 additions & 16 deletions libclc/libspirv/lib/amdgcn-amdhsa/group/collectives_helpers.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,49 +7,41 @@

define ptr addrspace(3) @__clc__get_group_scratch_bool() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i1], ptr addrspace(3) @__clc__group_scratch_i1, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i1
}

define ptr addrspace(3) @__clc__get_group_scratch_char() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i8], ptr addrspace(3) @__clc__group_scratch_i8, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i8
}

define ptr addrspace(3) @__clc__get_group_scratch_short() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i16], ptr addrspace(3) @__clc__group_scratch_i16, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i16
}

define i32 addrspace(3)* @__clc__get_group_scratch_int() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i32], ptr addrspace(3) @__clc__group_scratch_i32, i64 0, i64 0
ret i32 addrspace(3)* %0
ret i32 addrspace(3)* @__clc__group_scratch_i32
}

define ptr addrspace(3) @__clc__get_group_scratch_long() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i64], ptr addrspace(3) @__clc__group_scratch_i64, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i64
}

define ptr addrspace(3) @__clc__get_group_scratch_half() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i16], ptr addrspace(3) @__clc__group_scratch_i16, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i16
}

define ptr addrspace(3) @__clc__get_group_scratch_float() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i32], ptr addrspace(3) @__clc__group_scratch_i32, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i32
}

define ptr addrspace(3) @__clc__get_group_scratch_double() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i64], ptr addrspace(3) @__clc__group_scratch_i64, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i64
}

33 changes: 11 additions & 22 deletions libclc/libspirv/lib/ptx-nvidiacl/group/collectives_helpers.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,66 +8,55 @@

define ptr addrspace(3) @__clc__get_group_scratch_bool() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i1], ptr addrspace(3) @__clc__group_scratch_i1, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i1
}

define ptr addrspace(3) @__clc__get_group_scratch_char() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i8], ptr addrspace(3) @__clc__group_scratch_i8, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i8
}

define ptr addrspace(3) @__clc__get_group_scratch_short() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i16], ptr addrspace(3) @__clc__group_scratch_i16, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i16
}

define i32 addrspace(3)* @__clc__get_group_scratch_int() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i32], ptr addrspace(3) @__clc__group_scratch_i32, i64 0, i64 0
ret i32 addrspace(3)* %0
ret i32 addrspace(3)* @__clc__group_scratch_i32
}

define ptr addrspace(3) @__clc__get_group_scratch_long() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i64], ptr addrspace(3) @__clc__group_scratch_i64, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i64
}

define ptr addrspace(3) @__clc__get_group_scratch_half() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i16], ptr addrspace(3) @__clc__group_scratch_i16, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i16
}

define ptr addrspace(3) @__clc__get_group_scratch_float() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i32], ptr addrspace(3) @__clc__group_scratch_i32, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i32
}

define ptr addrspace(3) @__clc__get_group_scratch_double() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i64], ptr addrspace(3) @__clc__group_scratch_i64, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i64
}

define ptr addrspace(3) @__clc__get_group_scratch_complex_half() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i32], ptr addrspace(3) @__clc__group_scratch_i32, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i32
}

define ptr addrspace(3) @__clc__get_group_scratch_complex_float() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i64], ptr addrspace(3) @__clc__group_scratch_i64, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i64
}

define ptr addrspace(3) @__clc__get_group_scratch_complex_double() nounwind alwaysinline {
entry:
%0 = getelementptr inbounds [32 x i128], ptr addrspace(3) @__clc__group_scratch_i128, i64 0, i64 0
ret ptr addrspace(3) %0
ret ptr addrspace(3) @__clc__group_scratch_i128
}