Files
project_6/upstream_ref/sgemm_edtallison/Makefile
dylan 284804ac53 data: cat 3 SGEMM repos — siboehm, wangzyon, edtallison (full clone, no --depth)
Sources:
  siboehm/SGEMM_CUDA        → upstream_ref/sgemm_siboehm/     (25 files)
  wangzyon/NVIDIA_SGEMM_PRACTICE → upstream_ref/nvidia_sgemm_practice/ (23 files, filled gaps)
  edtallison/sgemm-cuda      → upstream_ref/sgemm_edtallison/  (41 files)

All files cat'd one by one from git clone (no --depth).
These are the 3 public SGEMM repos that can compile on CUDA 10.2 + CoreX ivcore10.

Key files for BI-V100 porting:
  kernel 10 (warp tiling) — already proven on device with WARPSIZE=64
  kernel 11/12 (double buffering) — next optimization target
  sgemm.cu + runner.cu — complete build+benchmark harness
  CMakeLists.txt — build system reference
2026-08-15 06:58:07 +00:00

35 lines
933 B
Makefile

.PHONY: all build debug clean profile bench cuobjdump
CMAKE := cmake
BUILD_DIR := build
BENCHMARK_DIR := benchmark_results
all: build
build:
@mkdir -p $(BUILD_DIR)
@cd $(BUILD_DIR) && $(CMAKE) -DCMAKE_BUILD_TYPE=Release ..
@$(MAKE) -C $(BUILD_DIR)
debug:
@mkdir -p $(BUILD_DIR)
@cd $(BUILD_DIR) && $(CMAKE) -DCMAKE_BUILD_TYPE=Debug ..
@$(MAKE) -C $(BUILD_DIR)
clean:
@rm -rf $(BUILD_DIR)
FUNCTION := $$(cuobjdump -symbols build/sgemm | grep -i Warptiling | awk '{print $$NF}')
cuobjdump: build
@cuobjdump -arch sm_86 -sass -fun $(FUNCTION) build/sgemm | c++filt > build/cuobjdump.sass
@cuobjdump -arch sm_86 -ptx -fun $(FUNCTION) build/sgemm | c++filt > build/cuobjdump.ptx
# Usage: make profile KERNEL=<integer> PREFIX=<optional string>
profile: build
@ncu --set full --export $(BENCHMARK_DIR)/$(PREFIX)kernel_$(KERNEL) --force-overwrite $(BUILD_DIR)/sgemm $(KERNEL)
bench: build
@bash gen_benchmark_results.sh