fix: include cuda_bf16.h unconditionally, add --expt-relaxed-constexpr
This commit is contained in:
@@ -42,6 +42,7 @@ nvcc_cmd = [
|
||||
"--x", "cu",
|
||||
"-o", "/tmp/fmha_sm100_test.o",
|
||||
"--ptxas-options=-v",
|
||||
"--expt-relaxed-constexpr",
|
||||
]
|
||||
|
||||
print(f"nvcc command: {' '.join(nvcc_cmd)}")
|
||||
|
||||
Reference in New Issue
Block a user