misc: update build setup (#2306)

.gitignore

@@ -218,3 +218,5 @@ work_dirs/
 *.exe
 *.out
 *.app
+
+compile_commands.json

sgl-kernel/Makefile

@@ -0,0 +1,19 @@
+.PHONY: tree ln install build clean test
+
+tree:
+	@tree --prune -I "__pycache__|*.egg-info|*.so|build"
+
+ln:
+	@rm -rf build && cmake . -DCMAKE_EXPORT_COMPILE_COMMANDS=1 -DCMAKE_CUDA_COMPILER=nvcc -B build && rm -rf compile_commands.json && ln -s build/compile_commands.json compile_commands.json
+
+install:
+	@pip install -e .
+
+build:
+	@python3 setup.py bdist_wheel
+
+clean:
+	@rm -rf build dist *.egg-info
+
+test:
+	@pytest tests/

sgl-kernel/build.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+set -ex
+
+docker run --rm -it \
+    -v "$(pwd)":/sgl-kernel \
+    pytorch/manylinux-builder:cuda12.1 \
+    bash -c "
+    pip install --no-cache-dir torch==2.4.0 --index-url https://download.pytorch.org/whl/cu121 && \
+    export TORCH_CUDA_ARCH_LIST='7.5 8.0 8.9 9.0+PTX' && \
+    cd /sgl-kernel && \
+    python setup.py bdist_wheel
+    "

sgl-kernel/setup.py

@@ -13,6 +13,18 @@
                 "src/sgl-kernel/csrc/warp_reduce.cc",
                 "src/sgl-kernel/csrc/warp_reduce_kernel.cu",
             ],
+            extra_compile_args={
+                "nvcc": [
+                    "-O3",
+                    "-Xcompiler",
+                    "-fPIC",
+                    "-gencode=arch=compute_75,code=sm_75",
+                    "-gencode=arch=compute_80,code=sm_80",
+                    "-gencode=arch=compute_89,code=sm_89",
+                    "-gencode=arch=compute_90,code=sm_90",
+                ],
+                "cxx": ["-O3"],
+            },
         )
     ],
     cmdclass={"build_ext": BuildExtension},

sgl-kernel/src/sgl-kernel/csrc/warp_reduce.cc

@@ -1,5 +1,4 @@
 #include <torch/extension.h>
-#include <vector>
 
 torch::Tensor warp_reduce_cuda(torch::Tensor input);