Skip to content

Conversation

cyyever
Copy link

@cyyever cyyever commented Aug 24, 2025

No description provided.

@@ -103,7 +103,7 @@ enum xnn_status xnn_setup_batch_matrix_multiply_nc_qp8_f32_qc8w(
xnn_operator_t batch_matrix_multiply_op, //
void* workspace, //
const int8_t* input_a, //
const int8_t* input_b, //
const float* input_b, //
Copy link
Contributor

Choose a reason for hiding this comment

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

this doesnt look right. the input_b is qc8w which is int8_t

Copy link
Author

@cyyever cyyever Aug 26, 2025

Choose a reason for hiding this comment

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

But the function definition in src/operators/batch-matrix-multiply-nc.c uses const float* input_b.
PS: it looks like I should change the definition.

Copy link
Contributor

@fbarchard fbarchard left a comment

Choose a reason for hiding this comment

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

on 2nd look, i think you got it right
qp8_f32_qc8w

a is 8 bit
b is 8 bit
output is float

Are you seeing a build error/warning that this fixes?

@cyyever
Copy link
Author

cyyever commented Sep 12, 2025

I tried lto and GCC or Clang complained it. It still can't build with lto after these fixes.

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