Files
blis/bench
Mithun Mohan 0701a4388a Thread factorization improvements (ic ways) for BF16 LPGEMM API.
-Currently when m is small compared to n, even if MR blks (m / MR) > 1,
and total work blocks (MR blks * NR blks) < available threads, the
number of threads assigned for m dimension (ic ways) is 1. This results
in sub par performance in bandwidth bound cases. To address this, the
thread factorization is updated to increase ic ways for these cases.

AMD-Internal: [SWLCSG-3333]

Change-Id: Ife3eafc282a2b62eb212af615edb7afa40d09ae9
2025-02-06 00:51:10 -05:00
..
2021-06-04 17:45:04 +05:30
2021-05-19 14:05:01 +05:30
2021-06-08 11:54:55 +05:30
2025-01-31 06:04:16 -05:00