mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-06-04 00:06:52 +08:00
16 lines
356 B
Plaintext
16 lines
356 B
Plaintext
# tileMatmulAutotuner search space
|
|
#
|
|
# The autotuner tries every tile entry with every load_latency and
|
|
# store_latency value listed below. Edit this file to experiment without
|
|
# rebuilding the sample.
|
|
|
|
# tile block_m block_n block_k
|
|
tile 64 64 32
|
|
tile 128 64 32
|
|
tile 64 128 32
|
|
tile 128 128 32
|
|
|
|
# latency hint values
|
|
load_latency 2 5 8
|
|
store_latency 2 5 8
|