DeepSeek/DeepEP
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepEP synced 2025-06-26 18:28:11 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Add ENABLE_FAST_DEBUG

Browse Source
This commit is contained in:
Chenggang Zhao
2025-06-20 14:44:53 +08:00
parent 47dd77ab5f
commit 782b40a8ff
3 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
csrc/CMakeLists.txt
View File

@@ -6,6 +6,7 @@ set(CMAKE_VERBOSE_MAKEFILE ON)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O3 -fPIC") set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O3 -fPIC")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3 -fPIC") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3 -fPIC")
set(CUDA_SEPARABLE_COMPILATION ON) set(CUDA_SEPARABLE_COMPILATION ON)
list(APPEND CUDA_NVCC_FLAGS "-DENABLE_FAST_DEBUG")
list(APPEND CUDA_NVCC_FLAGS "-O3") list(APPEND CUDA_NVCC_FLAGS "-O3")
list(APPEND CUDA_NVCC_FLAGS "--ptxas-options=--verbose,--register-usage-level=10,--warn-on-local-memory-usage") list(APPEND CUDA_NVCC_FLAGS "--ptxas-options=--verbose,--register-usage-level=10,--warn-on-local-memory-usage")

7
csrc/kernels/configs.cuh
View File

@@ -7,9 +7,14 @@
#define NUM_BUFFER_ALIGNMENT_BYTES 128 #define NUM_BUFFER_ALIGNMENT_BYTES 128
#define FINISHED_SUM_TAG 1024 #define FINISHED_SUM_TAG 1024
#define NUM_WAIT_NANOSECONDS 500
#ifndef ENABLE_FAST_DEBUG
#define NUM_CPU_TIMEOUT_SECS 100 #define NUM_CPU_TIMEOUT_SECS 100
#define NUM_TIMEOUT_CYCLES 200000000000ull // 200G cycles ~= 100s #define NUM_TIMEOUT_CYCLES 200000000000ull // 200G cycles ~= 100s
#define NUM_WAIT_NANOSECONDS 500 #else
#define NUM_CPU_TIMEOUT_SECS 10
#define NUM_TIMEOUT_CYCLES 20000000000ull // 20G cycles ~= 10s
#endif
#define LOW_LATENCY_SEND_PHASE 1 #define LOW_LATENCY_SEND_PHASE 1
#define LOW_LATENCY_RECV_PHASE 2 #define LOW_LATENCY_RECV_PHASE 2

3
csrc/kernels/launch.cuh
View File

@@ -58,12 +58,9 @@ cfg.dynamicSmemBytes = smem_size;
#define SWITCH_RDMA_RANKS(case_macro) \ #define SWITCH_RDMA_RANKS(case_macro) \
switch (num_ranks / NUM_MAX_NVL_PEERS) { \ switch (num_ranks / NUM_MAX_NVL_PEERS) { \
case 2: case_macro(2); \ case 2: case_macro(2); \
case 3: case_macro(3); \
case 4: case_macro(4); \ case 4: case_macro(4); \
case 8: case_macro(8); \ case 8: case_macro(8); \
case 16: case_macro(16); \ case 16: case_macro(16); \
case 18: case_macro(18); \
case 20: case_macro(20); \
default: EP_HOST_ASSERT(false and "Unsupported RDMA ranks"); \ default: EP_HOST_ASSERT(false and "Unsupported RDMA ranks"); \
} while (false) } while (false)