Skip to content

Update _i32_matmul_unreduced_CGGI to use single bfloat16 (instead of 4) for int32 multiplcations. #90

Merged
copybara-service[bot] merged 1 commit intomainfrom
test_854282648
Jan 13, 2026
Merged

Update _i32_matmul_unreduced_CGGI to use single bfloat16 (instead of 4) for int32 multiplcations. #90
copybara-service[bot] merged 1 commit intomainfrom
test_854282648

Conversation

@copybara-service
Copy link

@copybara-service copybara-service bot commented Jan 10, 2026

Update _i32_matmul_unreduced_CGGI to use single bfloat16 (instead of 4) for int32 multiplcations.

This takes into consideration the fact that the rlwe ciphertexts in blind_rotate are decomposed using a log_base < 8 i.e. all values are < 255. Currently the log_base = 4 in bool_params.py.

@copybara-service copybara-service bot force-pushed the test_854282648 branch 7 times, most recently from f4676ed to 4f761fb Compare January 13, 2026 04:30
…4) for int32 multiplcations.

This takes into consideration the fact that the rlwe ciphertexts in blind_rotate are decomposed using a log_base < 8 i.e. all values are < 255. Currently the log_base = 4 in bool_params.py.

PiperOrigin-RevId: 855504683
@copybara-service copybara-service bot merged commit f4add75 into main Jan 13, 2026
@copybara-service copybara-service bot deleted the test_854282648 branch January 13, 2026 04:50
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.

1 participant