Skip to content

Conversation

@victor-eds
Copy link
Contributor

Add support for layout conversion shuffles in which rows managed by a single thread are contiguous in the output matrix.

Step 2/2 to #2749

Add support for layout conversion shuffles in which rows managed by a single thread are contiguous in the output matrix.

Signed-off-by: victor-eds <[email protected]>
@victor-eds victor-eds requested review from a team, etiotto and whitneywhtsang November 22, 2024 11:27
@victor-eds victor-eds self-assigned this Nov 22, 2024
// CHECK: %[[VAL_1:.*]] = llvm.extractvalue %[[VAL_0]][0] : !llvm.struct<(f16, f16)>
// CHECK: %[[VAL_2:.*]] = llvm.extractvalue %[[VAL_0]][1] : !llvm.struct<(f16, f16)>
// COM: Check the shuffles are "coalesced"
// CHECK: llvm.call spir_funccc @_Z17sub_group_shuffleDhj(%[[VAL_1]]
Copy link
Contributor

Choose a reason for hiding this comment

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

ah would be nice if CHECK-COUNT could work with more than one line (so we could check that a pattern involving more than one line repeats a specified number of times). AFAIK this is not possible though.

int32_t laneSize) {
std::vector<std::vector<int32_t>> bases;
std::vector<int32_t> curr(2);
int i = 1;
Copy link
Contributor

Choose a reason for hiding this comment

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

[NIT]: int -> int32_t (consistency with surrounding code)

@etiotto etiotto merged commit 2a72ba2 into intel:main Nov 26, 2024
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants