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
  
        
    
Uh oh!
There was an error while loading. Please reload this page.
-
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