🐙 Optimise PPACK #2067
Merged
🐙 Optimise PPACK #2067
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
|
Can confirm this works on CPU SIMD, with a 5% speedup |
thorstenhater
requested review from
bcumming and
boeschf
and removed request for
bcumming
|
could you merge master? easier to see what's new that way |
|
Sure! |
boeschf
reviewed
Jan 12, 2023
Comment on lines
+268
to
+275
| "[[maybe_unused]] arb_size_type {0}index_constraints_n_contiguous = pp->index_constraints.n_contiguous;\\\n" | ||
| "[[maybe_unused]] arb_size_type {0}index_constraints_n_constant = pp->index_constraints.n_constant;\\\n" | ||
| "[[maybe_unused]] arb_size_type {0}index_constraints_n_independent = pp->index_constraints.n_independent;\\\n" | ||
| "[[maybe_unused]] arb_size_type {0}index_constraints_n_none = pp->index_constraints.n_none;\\\n" | ||
| "[[maybe_unused]] arb_index_type* __restrict__ {0}index_constraints_contiguous = pp->index_constraints.contiguous;\\\n" | ||
| "[[maybe_unused]] arb_index_type* __restrict__ {0}index_constraints_constant = pp->index_constraints.constant;\\\n" | ||
| "[[maybe_unused]] arb_index_type* __restrict__ {0}index_constraints_independent = pp->index_constraints.independent;\\\n" | ||
| "[[maybe_unused]] arb_index_type* __restrict__ {0}index_constraints_none = pp->index_constraints.none;\\\n"), |
There was a problem hiding this comment.
are these index constraints and their representation documented somewhere?
There was a problem hiding this comment.
It's an internal indicator to basically multi-version the SIMDised loops.
- contiguous: this bundle of
Nindices is{k, k+1, ..., k+N-1} - constant:
{k, k, ..., k} - independent and none: seem to be the same, we know nothing about the set
These are then used to drive the ld/st operations from/to SIMD types, so that the
actual loop can operator on full-width SIMD types.
I do not know whether this has been documented, it was already here when I came ;)
|
I like the explicit types instead of |
|
Ideally one would hope for a warning about implicit conversions as well. |
boeschf
approved these changes
Feb 1, 2023
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.
PPACK_IFACE_BLOCKwill be__restrict__ved_diis not used anymore, gone and ABI bumpedPPACK_IFACE_BLOCKNB. This builds on earlier work to test performance metrics; to be deleted and cherrypicked.