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.
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
Numerical overflow happens on GEMMLowpOutputStage #1012
Comments
Some updates: I modified the
we will get:
The expected result should be The codes:
|
@morgolock Please help to look at this question, thanks you:) |
Some updates:
just run:
the codes:
|
I am currently unable to judge whether my
GEMMLowpOutputStageInfo
configuration is wrong or there are some bugs in the codes. TheGEMMLowpOutputStageInfo
I configured is listed below:The phenomenon is that when convert s32 to s8, there might be a numerical overflow. For example:
Given:
The output of the
GEMMLowpMatrixMultiplyCore
returns -56 which might be overflow (100 * 2 - 127 - 128 -1=-56)the expected target s8 value should be 127 (clamp(100*2, -128, 127)=127).
The text was updated successfully, but these errors were encountered: