Skip to content

Comments

Integer dot product 4x8 packed translation#1633

Merged
svenvh merged 1 commit intoKhronosGroup:mainfrom
Quetzonarch:integer-dot-product-packed
Oct 12, 2022
Merged

Integer dot product 4x8 packed translation#1633
svenvh merged 1 commit intoKhronosGroup:mainfrom
Quetzonarch:integer-dot-product-packed

Conversation

@Quetzonarch
Copy link
Contributor

@Quetzonarch Quetzonarch commented Sep 30, 2022

This proposed bugfix updates the integer dot translation to use the correct function names from the OpenCL extension specification for packed format with scalar integer arguments (i.e. dot_4x8packed or dot_acc_sat_4x8packed instead of dot or dot_acc_sat with scalar integer arguments) to translate them into proper OpCodes. This is a fix for incorrect function names introduced in #1174
Additionally removed unused variables from visitCallDot to clean up the code.
Link to OpenCL spec:
https://registry.khronos.org/OpenCL/specs/3.0-unified/html/OpenCL_Ext.html#cl_khr_integer_dot_product

@Quetzonarch Quetzonarch force-pushed the integer-dot-product-packed branch from 5711805 to 8a393f7 Compare October 6, 2022 09:29
@MrSidims MrSidims self-requested a review October 6, 2022 09:30
Changed the integer dot translation to use the correct function names
(i.e. dot_4x8packed or dot_acc_sat_4x8packed) to translate them into
proper OpCodes. Additionally removed unused variables from visitCallDot
@Quetzonarch Quetzonarch force-pushed the integer-dot-product-packed branch from 8a393f7 to 603cd82 Compare October 6, 2022 11:28
Copy link
Contributor

@MrSidims MrSidims left a comment

Choose a reason for hiding this comment

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

LGTM
If you don't mind I added a link to OpenCL spec in the description.

@MrSidims MrSidims requested a review from svenvh October 11, 2022 13:46
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.

3 participants