Skip to content

docs(repro): grouped requantization performance request - #1210

Draft
learning-chip wants to merge 21 commits into
hw-native-sys:mainfrom
learning-chip:repro/requant-pipeline
Draft

docs(repro): grouped requantization performance request#1210
learning-chip wants to merge 21 commits into
hw-native-sys:mainfrom
learning-chip:repro/requant-pipeline

Conversation

@learning-chip

Copy link
Copy Markdown
Contributor

No description provided.

vector_bool m8 = pset_b8(PAT_ALL);
vector_bool m32 = pset_b32(PAT_ALL);
vector_f8e4m3 q; ::vlds(q, src, 0, NORM);
vector_f32 qf; ::vcvt(qf, q, m8, PART_P0, MODE_ZEROING);

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

这里只处理了256 lane中的PART0,计算量天然和VMI不一样

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Pushed new fair comparisons

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants