colin
/
CuPBoP
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2 Commits 1 Branch 0 Tags 1 MiB
C++ 89.8%
C 6.9%
CMake 2.9%
Shell 0.4%
Go to file
Ruobing Han 897af29748 add CI 2022-01-13 13:30:45 -05:00
.github/workflows add CI 2022-01-13 13:30:45 -05:00
compilation add backbone, including basic features for compilation 2022-01-11 11:01:42 -05:00
docs add CI 2022-01-13 13:30:45 -05:00
runtime add backbone, including basic features for compilation 2022-01-11 11:01:42 -05:00
.gitignore add CI 2022-01-13 13:30:45 -05:00
.pre-commit-config.yaml add CI 2022-01-13 13:30:45 -05:00
CMakeLists.txt add backbone, including basic features for compilation 2022-01-11 11:01:42 -05:00
LICENSE add backbone, including basic features for compilation 2022-01-11 11:01:42 -05:00
README.md add CI 2022-01-13 13:30:45 -05:00

README.md

CuPBoP: Cuda for Parallelized and Broad-range Processors

Introduction

CuPBoP (Cuda for parallelized and broad-range processors) is a framework aims to execute CUDA source code on non-NVIDIA devices, including CPU, GPU and other architectures.

Install

Prerequisites

  • Linux: Verified on Ubuntu 18.04
  • LLVM10.0
  • NVIDIA CUDA-toolkit
  • x86 CPU
  • pthread
  • GCC 7.5.0

Installation

  1. Clone from github

    git clone https://github.com/cupbop/CuPBoP
cd CuPBoP

  2. Build the transformer for NVVM IR to LLVM IR for X86

    mkdir build && cd build
cmake .. -DLLVM_CONFIG_PATH=`which llvm-config` # need path to llvm-config
make

Run Vecadd samples

# Generate bitcode from human-readable LLVM IR
llvm-as ../compilation/examples/vecadd/kernel-cuda-nvptx64-nvidia-cuda-sm_61.ll
# use LLVM passes to transform NVVM IR (SPMD) to LLVM IR (MPMD+SIMD).
# NOTE: we hard-code the grid size (1, 1, 1)
# and block size (1024, 1, 1) into the generated LLVM IR
./compilation/nvvm2x86 \
    ../compilation/examples/vecadd/kernel-cuda-nvptx64-nvidia-cuda-sm_61.bc \
    kernel.bc 1 1 1 32 1 1
# generate object file from LLVM IR
llc --filetype=obj kernel.bc
# link generated kernel function
# with host function and generate excutable file
g++ ../compilation/examples/vecadd/host.cpp \
    kernel.o -lpthread -o vecadd_example
# execute the executable file
./vecadd_example

Contribution

We sincerely appreciate all kinds of contributions. Please refer to CONTRIBUTING for the contributing guideline.

Author