Skip to content

vulkan: optimize coopmat2 dequant functions #10855

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Dec 21, 2024
Merged

vulkan: optimize coopmat2 dequant functions #10855

merged 1 commit into from
Dec 21, 2024

Conversation

@jeffbolznv
Copy link
Collaborator

@jeffbolznv jeffbolznv commented Dec 16, 2024

Change the code to do 16b loads when possible and extract the appropriate component late, so the code is effectively decoding a pair of elements and then selecting one. This can allow more commoning to happen in the compiler when neighboring elements are loaded.

before
| llama 3B Q5_K - Medium         |   2.16 GiB |     3.21 B | Vulkan     | 1000 |  1 |         pp512 |     5131.92  192.48 |

after
| llama 3B Q5_K - Medium         |   2.16 GiB |     3.21 B | Vulkan     | 1000 |  1 |         pp512 |     5400.29  205.68 |

@jeffbolznv
vulkan: optimize coopmat2 dequant functions
22bea1d 
Change the code to do 16b loads when possible and extract the appropriate
component late, so the code is effectively decoding a pair of elements and
then selecting one. This can allow more commoning to happen in the compiler
when neighboring elements are loaded.
@jeffbolznv jeffbolznv requested a review from 0cc4m December 16, 2024 22:28
@github-actions github-actions bot added Vulkan Issues specific to the Vulkan backend ggml changes relating to the ggml tensor library for machine learning labels Dec 16, 2024
0cc4m
0cc4m approved these changes Dec 19, 2024
View reviewed changes
Copy link
Collaborator

@0cc4m 0cc4m left a comment

Choose a reason for hiding this comment

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

I can't test this without a driver change, but the code looks fine.

@0cc4m 0cc4m merged commit a91a413 into ggml-org:master Dec 21, 2024
2 checks passed
@jeffbolznv jeffbolznv mentioned this pull request Jan 7, 2025
Merged
tinglou pushed a commit to tinglou/llama.cpp that referenced this pull request Feb 13, 2025
@jeffbolznv @tinglou
vulkan: optimize coopmat2 dequant functions (ggml-org#10855)
76220be 
Change the code to do 16b loads when possible and extract the appropriate
component late, so the code is effectively decoding a pair of elements and
then selecting one. This can allow more commoning to happen in the compiler
when neighboring elements are loaded.
arthw pushed a commit to arthw/llama.cpp that referenced this pull request Feb 26, 2025
@jeffbolznv @NeoZhangJianyu
vulkan: optimize coopmat2 dequant functions (ggml-org#10855)
e10e5f4 
Change the code to do 16b loads when possible and extract the appropriate
component late, so the code is effectively decoding a pair of elements and
then selecting one. This can allow more commoning to happen in the compiler
when neighboring elements are loaded.
mglambda pushed a commit to mglambda/llama.cpp that referenced this pull request Mar 8, 2025
@jeffbolznv @mglambda
vulkan: optimize coopmat2 dequant functions (ggml-org#10855)
c098905 
Change the code to do 16b loads when possible and extract the appropriate
component late, so the code is effectively decoding a pair of elements and
then selecting one. This can allow more commoning to happen in the compiler
when neighboring elements are loaded.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@0cc4m 0cc4m 0cc4m approved these changes

Assignees

No one assigned

Labels

ggml changes relating to the ggml tensor library for machine learning Vulkan Issues specific to the Vulkan backend

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@jeffbolznv @0cc4m