bench_sparse_cuda.py

import os
import subprocess

from src.codegen import vbr_spmm_cuda_codegen, vbr_spmv_cuda_codegen

BENCHMARK_FREQ = 5

def bench_spmv(bad: bool):
    if bad:
        gen_dir = "Generated_SpMV_cuda_Sparse_Bad"
        benchfile_name = "benchmarks_spmv_1_rel_bad_cuda.csv"
        dense_blocks_only = True
    else:
        gen_dir = "Generated_SpMV_cuda_Sparse"
        benchfile_name = "benchmarks_spmv_1_rel_cuda.csv"
        dense_blocks_only = False
    vbr_files = os.listdir("Generated_VBR_Sparse")
    print("Benchmarking executor")
    with open(os.path.join("results", benchfile_name), "w") as fMy:
        for filename in vbr_files:
            fname = filename[:-len(".vbr")]
            spmv_file = fname + ".cu"
            print(filename, flush=True)
            # compile the generated code for SpMV operation
            vbr_spmv_cuda_codegen(fname, dense_blocks_only=dense_blocks_only, dir_name=gen_dir, vbr_dir="Generated_VBR_Sparse")
            subprocess.run(["nvcc", "-O3", "-o", fname, spmv_file], cwd=gen_dir)
            output = subprocess.run(["./"+fname], capture_output=True, cwd=gen_dir)
            execution_times = []
            for i in range(BENCHMARK_FREQ):
                print(f"Benchmarking executor iteration", i, flush=True)
                output = subprocess.run(["./"+fname], capture_output=True, cwd=gen_dir)
                execution_time = output.stdout.decode("utf-8").split("\n")[0].split(" = ")[1]
                execution_times.append(execution_time)
            # save execution times to file
            p = f"{fname},{','.join(execution_times)}\n"
            print(p, flush=True)
            fMy.write(p)

def bench_spmm(bad: bool):
    if bad:
        gen_dir = "Generated_SpMM_cuda_Sparse_Bad"
        benchfile_name = "benchmarks_spmm_1_rel_bad_cuda.csv"
        dense_blocks_only = True
    else:
        gen_dir = "Generated_SpMM_cuda_Sparse"
        benchfile_name = "benchmarks_spmm_1_rel_cuda.csv"
        dense_blocks_only = False
    vbr_files = os.listdir("Generated_VBR_Sparse")
    print("Benchmarking executor")
    with open(os.path.join("results", benchfile_name), "w") as fMy:
        for filename in vbr_files:
            fname = filename[:-len(".vbr")]
            spmm_file = fname + ".cu"
            print(filename, flush=True)
            # compile the generated code for SpMV operation
            vbr_spmm_cuda_codegen(fname, dense_blocks_only=dense_blocks_only, dir_name=gen_dir, vbr_dir="Generated_VBR_Sparse")
            subprocess.run(["nvcc", "-O3", "-o", fname, spmm_file], cwd=gen_dir)
            output = subprocess.run(["./"+fname], capture_output=True, cwd=gen_dir)
            execution_times = []
            for i in range(BENCHMARK_FREQ):
                print(f"Benchmarking executor iteration", i, flush=True)
                output = subprocess.run(["./"+fname], capture_output=True, cwd=gen_dir)
                execution_time = output.stdout.decode("utf-8").split("\n")[0].split(" = ")[1]
                execution_times.append(execution_time)
            # save execution times to file
            p = f"{fname},{','.join(execution_times)}\n"
            print(p, flush=True)
            fMy.write(p)

if __name__ == "__main__":
    bench_spmv(bad=True)
    bench_spmv(bad=False)
    bench_spmm(bad=True)
    bench_spmm(bad=False)