Tile size limitation #9701
Unanswered
chiachia36
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
After checking some files about tilng (LLVMCPUTileFuseAndVectorizeLinalgTensorOps.cpp and TileAndDistributeToWorkgroupsPass.cpp)
I guess that lowering config info (e.g. lowering_config = <tile_sizes = [[32, 32], [8, 8, 0], [0, 0, 8]]>) means workgroup size, L1 block size and vector size.
I tried to use lowering config to make a proper tile for resnet 50.
First I add compilation info
on the first convolution.
However when I tried to compile it, I got error messages.
I tried other compilation info before, and I could compile it successfully.
I guess there are some limitations about tiling size, and it causes the error I mention above.
Could you give me some advices and help?
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions