ggml : use F16C conversion when available #1506
Closed
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
|
I tried this a while ago, but IIRC converting a single value from FP16 to FP32 with a lookup table is faster. If we used SIMD to convert 8 values at a time, it is likely that F16C would be faster that way. |
|
Agree with @slaren that the table look-up is faster. This PR makes Q5_1 inference about 7% slower for me. The upcoming AVX-NE-CONVERT instructions may change that, though (namely |
ggerganov
added
the
demo
|
I did some tests as well and I agree - it does not help |
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Outdated suggestions cannot be applied.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.
Somehow this has been replace via lookup table even if F16C intrinsics are available