github/ROCm
1
1
Fork 0
mirror of https://github.com/ROCm/ROCm.git synced 2026-04-05 03:01:17 -04:00
Code Issues Packages Projects Releases Wiki Activity

[BACKEND] Add back dot.wait when generating async_dot (#2478)

Browse Source 
Based on discussion this is needed to make sure there is no race
condition when reading shared memory.
This commit is contained in:
Thomas Raoux
2023-10-10 21:45:28 -07:00
committed by GitHub
parent 4749072fbd
commit 6f46c93b9e
3 changed files with 16 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
test/TritonGPU/wspipeline.mlir
View File

@@ -22,9 +22,10 @@
// CHECK: triton_gpu.extract_slice
// CHECK: triton_gpu.extract_slice
// CHECK: triton_nvidia_gpu.dot_async
// CHECK: triton_nvidia_gpu.dot_wait {{.*}} pendings = 1
// CHECK: triton_nvidia_gpu.consumer_release
// CHECK: scf.yield
// CHECK: triton_nvidia_gpu.dot_wait
// CHECK: triton_nvidia_gpu.dot_wait {{.*}} pendings = 0
// CHECK: async_agent = dense<1> : vector<1xi32>
#blocked = #triton_gpu.blocked<{sizePerThread = [8, 1], threadsPerWarp = [4, 8], warpsPerCTA = [1, 4], order = [0, 1], CTAsPerCGA = [1, 1], CTASplitNum = [1, 1], CTAOrder = [0, 1]}>