tinygrad/test/external/speed_compare_cuda_ptx.py

61 lines
2.2 KiB
Python
Raw Normal View History

2024-03-09 02:17:49 +08:00
import itertools
from tinygrad import Device
from tinygrad.engine.realize import CompiledRunner
from tinygrad.helpers import getenv, colorize_float
2024-03-09 02:17:49 +08:00
from extra.optimization.helpers import load_worlds, ast_str_to_lin
2024-05-15 14:12:59 +08:00
from tinygrad.engine.search import bufs_from_lin
from tinygrad.runtime.ops_cuda import PTXCompiler, PTXRenderer, CUDACompiler
2024-03-09 02:17:49 +08:00
if __name__ == "__main__":
ast_strs = load_worlds(filter_reduce=False, filter_novariable=True)
# no bfloat16 for ptx at the moment
ast_strs = [x for x in ast_strs if "dtypes.bfloat16" not in x]
2024-03-09 02:17:49 +08:00
dev = Device["CUDA"]
ptx = PTXRenderer(dev.arch)
2024-03-09 02:17:49 +08:00
# NUM=112 python3 test/external/speed_compare_cuda_ptx.py
single = getenv("NUM", -1)
if single != -1: ast_strs = ast_strs[single:single+1]
average_tm_cuda, average_tm_ptx = 0, 0
for num,ast in enumerate(ast_strs):
# cuda compile
dev.compiler = CUDACompiler(dev.arch)
lin = ast_str_to_lin(ast, opts=dev.renderer)
2024-03-09 02:17:49 +08:00
lin.hand_coded_optimizations()
cuda_prg = CompiledRunner(lin.to_program())
2024-03-09 02:17:49 +08:00
bufs = bufs_from_lin(lin)
# ptx compile
dev.compiler = PTXCompiler(dev.arch)
lin = ast_str_to_lin(ast, opts=ptx)
2024-03-09 02:17:49 +08:00
lin.hand_coded_optimizations()
lin.linearize()
ptx_prg = CompiledRunner(lin.to_program())
# warmup
2024-03-09 02:17:49 +08:00
try:
cuda_prg(bufs, {}, wait=True)
2024-03-09 02:17:49 +08:00
except RuntimeError:
print("cuda failed ast:", num)
2024-03-09 02:17:49 +08:00
continue
ptx_prg(bufs, {}, wait=True)
tm_cuda, tm_ptx = [], []
for i in range(5):
tm_cuda.append(cuda_prg(bufs, {}, wait=True))
tm_ptx.append(ptx_prg(bufs, {}, wait=True))
average_tm_cuda += min(tm_cuda)
average_tm_ptx += min(tm_ptx)
ratio = min(tm_ptx)/min(tm_cuda)
print(f"{average_tm_ptx/average_tm_cuda:5.2f}x -- {num:4d} {colorize_float(ratio)} {min(tm_ptx)*1e6:7.2f} us", lin.name)
if ratio > 1.5:
def fix(x): return x.replace('\t', ' ').strip()
ll1, ll2 = cuda_prg.lib.decode().split('\n'), ptx_prg.lib.decode().split('\n')
2024-03-09 02:17:49 +08:00
if single != -1:
for ln, (l1, l2) in enumerate(itertools.zip_longest(ll1, ll2, fillvalue='')):
print(f"{ln:5d} | {fix(l1):80s} | {fix(l2):80s}")
print(len(ll1), len(ll2), "RATIO", ratio, "us", min(tm_ptx)*1e6)