ggml: aarch64: implement SVE kernels for q2_k_q8_k vector dot #12064
Conversation
This file contains hidden or 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
github-actions
bot
added
the
ggml
ggerganov
reviewed
Feb 25, 2025
ggerganov
reviewed
Feb 27, 2025
ggml/src/ggml-cpu/ggml-cpu-quants.c
Outdated
|
|
||
| svint32_t sumi1 = svdup_n_s32(0); | ||
|
|
||
| for (int j = 0; j < QK_K/256; ++j) { |
There was a problem hiding this comment.
These loops seem redundant - are they needed? Can you simplify by knowing that this will always be a single iteration?
There was a problem hiding this comment.
Yes. This loop always run for one time. The code is simplified considering the for loop will run for 1 time only. I can remove the loop.
ggerganov
approved these changes
Feb 28, 2025
mglambda
pushed a commit
to mglambda/llama.cpp
that referenced
this pull request
Mar 8, 2025
…rg#12064) * Added SVE Support for Q2_K Quantized Models * Use 4-space indentation in the switch cases * removed comments lines * Remove the loop Retain the curly bracess for better understanding of code * Remove the comment like added for q3_k_q8_k kernel --------- Co-authored-by: vithulep <[email protected]>
arthw
pushed a commit
to arthw/llama.cpp
that referenced
this pull request
Mar 19, 2025
…rg#12064) * Added SVE Support for Q2_K Quantized Models * Use 4-space indentation in the switch cases * removed comments lines * Remove the loop Retain the curly bracess for better understanding of code * Remove the comment like added for q3_k_q8_k kernel --------- Co-authored-by: vithulep <[email protected]>
mostlyuseful
pushed a commit
to mostlyuseful/llama.cpp
that referenced
this pull request
May 12, 2025
…rg#12064) * Added SVE Support for Q2_K Quantized Models * Use 4-space indentation in the switch cases * removed comments lines * Remove the loop Retain the curly bracess for better understanding of code * Remove the comment like added for q3_k_q8_k kernel --------- Co-authored-by: vithulep <[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.
This PR introduces support for SVE (Scalable Vector Extensions) kernels for the q2_K_q8_K vector dot on the Arm architecture. A similar proposal for SVE support is made in PR 7433 and 11227.
This PR contains the SVE implementation of the vector dot used to compute the Q2_K quantization.
By running a Q2_K quantized model of mistral-7b-v01, on Graviton 3E (Perf 21 XL), Accuracy and Performance are measured.
Performance
The performance enhancement with this PR (SVE) is ~ x1.03 to x1.09 faster than the NEON implementation.
The command used to measure the performance is
Perplexity
I have ran perplexity with the NEON(Original) and SVE (This PR) Implementation.
And below is the summary.
This correction does not appear to have any impact on accuracy.