[TKW] Work on buffer ops #492
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
Hardcode84
commented
Feb 11, 2025
•
Hardcode84
commented
- Split read/write ops indexing on thread-dependent and thread-independent.
- Use symbolic vals for strides instead of extracting them from memref.
- For now only replace gather/scatter ops with buffer ops.
Hardcode84
force-pushed
the
buffer-ops-indexing
branch
from
6355a62 to
eac4567
Compare
harsh-nod
requested changes
Feb 13, 2025
| @@ -75,15 +76,29 @@ def _get_start_indices( | |||
| return start_indices | |||
|
|
|||
|
|
|||
| def _split_index(src: IndexExpr) -> tuple[IndexExpr, IndexExpr]: | |||
There was a problem hiding this comment.
Could you add a comment explaining why it is not sufficient to just substitute in subs_th , why we need to compute the difference and the need for safe_subs as well?
| IndexingContext.current(), symbolic_shape, allow_mixed_shapes=True | ||
| ) | ||
| if ( | ||
| emitter.params.get("use_buffer_load_ops", False) |
There was a problem hiding this comment.
Can you refactor this into a variable/function so something like
use_buffer_ops = emitter.params.get("use_buffer_load_ops", False)
has_integer_strides = all(isinstance(s, int) for s in strides)
can_emit_buffer_ops = use_buffer_ops and has_integer_strides and is_gather
if can_emit_buffer_ops:
...
harsh-nod
reviewed
Feb 13, 2025
| elements_per_thread: int, | ||
| mask: Optional[Value], | ||
| offsets_vec: Optional[Value], | ||
| ) -> Value: | ||
| if mask is None and offsets_vec is None: | ||
| return vector_d.load(vector_type, mem, start_indices) | ||
|
|
||
| is_gather = offsets_vec is not None |
There was a problem hiding this comment.
Don't we also need to check the address space since we can have gathers from shared memory? In this case, we cannot use buffer ops right? (Also same for scatters from shared memory)
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
This reverts commit cfef902. Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Signed-off-by: Ivan Butygin <[email protected]>
Hardcode84
force-pushed
the
buffer-ops-indexing
branch
from
670e275 to
f0cacae
Compare
Signed-off-by: Ivan Butygin <[email protected]>
xintin
pushed a commit
to xintin/iree-turbine
that referenced
this pull request
Feb 14, 2025
* Split read/write ops indexing on thread-dependent and thread-independent. * Use symbolic vals for strides instead of extracting them from memref. * For now only replace gather/scatter ops with buffer ops. --------- Signed-off-by: Ivan Butygin <[email protected]> Signed-off-by: xintin <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.