PR #8106 [SYCL] Fix the sub group size of Intel

[SYCL] Fix the sub group size of Intel #8106

airMeng merged 11 commits into ggml-org:master from luoyu-intel:intel_sgsize

github-actions added build

github-actions added SYCL

mofosyne added Review Complexity : Low

airMeng requested a review from

airMeng 1 year ago

airMeng requested a review from

AidanBeltonS 1 year ago

airMeng commented on 2024-06-26

airMeng approved these changes on 2024-06-26

luoyu-intel force pushed 1 year ago

github-actions added ggml

luoyu-intel marked this pull request as draft 1 year ago

luoyu-intel force pushed 1 year ago

airMeng removed Review Complexity : Low

airMeng added Review Complexity : High

joeatodd requested changes on 2024-06-28

luoyu-intel marked this pull request as ready for review 1 year ago

joeatodd approved these changes on 2024-07-01

airMeng commented on 2024-07-01

use warp_size macro for all sycl kernels

e18bf85d

fix mask of permute_sub_group_by_xor

a85ad06e

fix rms_norm with correct warp number

9a48a453

fix rms_norm_f32

33df09d9

fix group_norm_f32

90e03280

move norm to norm.cpp file

a2936f40

sync master

4cd48c7c

fix quantize bug

61f0cd58

fix mmvq's batch size

eb0d1325

add warp_size macro

e4f1516e

fix compile issue

8117e6d4

luoyu-intel force pushed to 8117e6d4 1 year ago

airMeng merged d08c20ed into master 1 year ago

luoyu-intel commented on 2024-07-02

luoyu-intel deleted the intel_sgsize branch 1 year ago

Reviewers

airMeng

joeatodd

qnixsynapse

AidanBeltonS

Assignees

No one assigned

Labels

build Review Complexity : High ggml SYCL

Milestone

No milestone

llama.cpp [SYCL] Fix the sub group size of Intel #8106 Merged

[SYCL] Fix the sub group size of Intel #8106

llama.cpp
[SYCL] Fix the sub group size of Intel
#8106

Merged