Enable experimental Mullapudi2016 GPU schedules for selected CI test cases

antonysigma · antonysigma · commit 3359dfd680b4 · 2025-02-19T14:26:02.000-08:00
diff --git a/apps/bilateral_grid/CMakeLists.txt b/apps/bilateral_grid/CMakeLists.txt
@@ -28,7 +28,7 @@ add_halide_library(bilateral_grid_auto_schedule FROM bilateral_grid.generator
                    AUTOSCHEDULER Halide::Mullapudi2016
                    # When target=host-cuda or host-metal, limit the GPU shared
                    # memory per block to avoid gpu kernel launch failure.
-                   PARAMS autoscheduler.last_level_cache_size=20000
+                   PARAMS autoscheduler.last_level_cache_size=20000 autoscheduler.experimental_gpu_schedule=1
                    )
 
 # Main executable
diff --git a/apps/local_laplacian/CMakeLists.txt b/apps/local_laplacian/CMakeLists.txt
@@ -23,7 +23,7 @@ add_halide_library(local_laplacian_auto_schedule FROM local_laplacian.generator
                    AUTOSCHEDULER Halide::Mullapudi2016
                    # When target=host-cuda or host-metal, limit the GPU shared
                    # memory per block to avoid gpu kernel launch failure.
-                   PARAMS autoscheduler.last_level_cache_size=30000
+                   PARAMS autoscheduler.last_level_cache_size=30000 autoscheduler.experimental_gpu_schedule=1
                    )
 
 # Main executable
diff --git a/apps/stencil_chain/CMakeLists.txt b/apps/stencil_chain/CMakeLists.txt
@@ -21,7 +21,7 @@ add_halide_library(stencil_chain_auto_schedule FROM stencil_chain.generator
                    AUTOSCHEDULER Halide::Mullapudi2016
                    # When target=host-cuda or host-metal, limit the GPU shared
                    # memory per block to avoid gpu kernel launch failure.
-                   PARAMS autoscheduler.last_level_cache_size=15000
+                   PARAMS autoscheduler.last_level_cache_size=15000 autoscheduler.experimental_gpu_schedule=1
                    )
 
 # Main executable

Original file line number	Diff line number	Diff line change
`@@ -28,7 +28,7 @@ add_halide_library(bilateral_grid_auto_schedule FROM bilateral_grid.generator`
`28`	`28`	`AUTOSCHEDULER Halide::Mullapudi2016`
`29`	`29`	`# When target=host-cuda or host-metal, limit the GPU shared`
`30`	`30`	`# memory per block to avoid gpu kernel launch failure.`
`31`		`- PARAMS autoscheduler.last_level_cache_size=20000`
	`31`	`+ PARAMS autoscheduler.last_level_cache_size=20000 autoscheduler.experimental_gpu_schedule=1`
`32`	`32`	`)`
`33`	`33`
`34`	`34`	`# Main executable`
Original file line number	Diff line number	Diff line change
`@@ -23,7 +23,7 @@ add_halide_library(local_laplacian_auto_schedule FROM local_laplacian.generator`
`23`	`23`	`AUTOSCHEDULER Halide::Mullapudi2016`
`24`	`24`	`# When target=host-cuda or host-metal, limit the GPU shared`
`25`	`25`	`# memory per block to avoid gpu kernel launch failure.`
`26`		`- PARAMS autoscheduler.last_level_cache_size=30000`
	`26`	`+ PARAMS autoscheduler.last_level_cache_size=30000 autoscheduler.experimental_gpu_schedule=1`
`27`	`27`	`)`
`28`	`28`
`29`	`29`	`# Main executable`
Original file line number	Diff line number	Diff line change
`@@ -21,7 +21,7 @@ add_halide_library(stencil_chain_auto_schedule FROM stencil_chain.generator`
`21`	`21`	`AUTOSCHEDULER Halide::Mullapudi2016`
`22`	`22`	`# When target=host-cuda or host-metal, limit the GPU shared`
`23`	`23`	`# memory per block to avoid gpu kernel launch failure.`
`24`		`- PARAMS autoscheduler.last_level_cache_size=15000`
	`24`	`+ PARAMS autoscheduler.last_level_cache_size=15000 autoscheduler.experimental_gpu_schedule=1`
`25`	`25`	`)`
`26`	`26`
`27`	`27`	`# Main executable`