256x256x64 bf16 spill; 256x256x128 fp8 not spill

example/01_gemm/CMakeLists.txt

@@ -27,10 +27,12 @@ add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_streamk_v3)
 add_example_executable(example_gemm_xdl_fp16_v3 gemm_xdl_fp16_v3.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_v3)
 add_example_executable(example_gemm_xdl_fp8_v3 gemm_xdl_fp8_v3.cpp)
+target_compile_options(example_gemm_xdl_fp8_v3 PRIVATE -mllvm -greedy-reverse-local-assignment=1 -save-temps=$PWD -Wno-gnu-line-marker)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp8_v3)
 add_example_executable(example_gemm_xdl_fp16_fp8_v3 gemm_xdl_fp16_fp8_v3.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_fp8_v3)
 add_example_executable(example_gemm_xdl_bf16_v3 gemm_xdl_bf16_v3.cpp)
+target_compile_options(example_gemm_xdl_bf16_v3 PRIVATE -mllvm -greedy-reverse-local-assignment=1 -save-temps=$PWD -Wno-gnu-line-marker)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_bf16_v3)
 
 add_example_executable(example_gemm_xdl_wavelet_fp16 gemm_xdl_wavelet_fp16.cpp)

example/01_gemm/gemm_xdl_bf16_v3.cpp

@@ -28,10 +28,10 @@ using DeviceGemmV2Instance =
         ADataType,   BDataType,  CDataType,  AccDataType,  CShuffleDataType, 
         PassThrough, PassThrough, PassThrough, GemmDefault, 
         256,
-        128, 128, 
+        256, 256, 
         64, 8, 8,
         16,   16,
-        4,    4,
+        8,    8,
         S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>, 
         2, 8, 8, 0,
         S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>, 

example/01_gemm/gemm_xdl_fp8_v3.cpp

@@ -28,10 +28,10 @@ using DeviceGemmV2Instance =
         ADataType,   BDataType,  CDataType,  AccDataType,  CShuffleDataType, 
         PassThrough, PassThrough, PassThrough, GemmDefault, 
         256,
-        224, 256, 
+        256, 256, 
         128, 16, 16,
         16,   16,
-        7,    8, 
+        8,    8, 
         S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>, 
         2, 16, 16, 0,
         S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>,