EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Fix the style of sgl kernel (#10398)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-09-12 22:20:21 -07:00
committed by GitHub
parent 99757cc3e6
commit c9ec4cae5b
5 changed files with 19 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
sgl-kernel/CMakeLists.txt
View File

@@ -157,6 +157,7 @@ set(SGL_KERNEL_CUDA_FLAGS
"-DCUTLASS_DEBUG_TRACE_LEVEL=0"
"--expt-relaxed-constexpr"
"--expt-extended-lambda"
# The following flag leads to the CMAKE_BUILD_PARALLEL_LEVEL breaking,
# it triggers OOM with low memory host. Extract the threads number to
# option named SGL_KERNEL_COMPILE_THREADS, default value 32.
@@ -169,7 +170,8 @@ set(SGL_KERNEL_CUDA_FLAGS
"-Xcompiler=-Wno-terminate"
"-Xcompiler=-Wfatal-errors"
"-Xcompiler=-ftemplate-backtrace-limit=1"
"-Xcudafe=--diag_suppress=177" # variable was declared but never referenced
"-Xcudafe=--diag_suppress=177" # variable was declared but never referenced
"-Xcudafe=--diag_suppress=2361" # invalid narrowing conversion from "char" to "signed char"
# uncomment to debug
# "--ptxas-options=-v"
@@ -299,11 +301,12 @@ set(SOURCES
"csrc/grammar/apply_token_bitmask_inplace_cuda.cu"
"csrc/mamba/causal_conv1d.cu"
"csrc/moe/cutlass_moe/w4a8/scaled_mm_entry.cu"
"csrc/moe/cutlass_moe/w4a8/w4a8_moe_data.cu"
"csrc/moe/cutlass_moe/w4a8/w4a8_grouped_mm_c3x.cu"
"csrc/moe/marlin_moe_wna16/ops.cu"
"csrc/mamba/causal_conv1d.cu"
"csrc/moe/moe_align_kernel.cu"
"csrc/moe/moe_fused_gate.cu"
"csrc/moe/moe_topk_softmax_kernels.cu"