mirror of
https://github.com/tinygrad/tinygrad.git
synced 2026-01-24 22:38:16 -05:00
d878065ece
* gemm * off by factor of 5 * 50 GFLOPS * works * 91 gflops * working at 50G * works * iy * 150 GFLOPS * 150 GFLOPS * N=2048 is still fast * threading soon * multithread * pinning * throttling is sad * Align matrices to cacheline width (#361) Co-authored-by: cloud <Cloud11665@gmail.com>
29 lines
601 B
Python
Executable File
29 lines
601 B
Python
Executable File
#!/usr/bin/env python3
|
|
import os
|
|
#os.environ['OMP_NUM_THREADS'] = '1'
|
|
import time
|
|
import numpy as np
|
|
|
|
N = 2048
|
|
if __name__ == "__main__":
|
|
# N^2
|
|
A = np.random.randn(N, N).astype(np.float32)
|
|
# N^2
|
|
B = np.random.randn(N, N).astype(np.float32)
|
|
|
|
# 2N compute in N^2 output cells
|
|
flop = 2*N*N*N
|
|
#print(f"{flop / 1e9:.2f} GFLOP")
|
|
|
|
for i in range(4):
|
|
st = time.monotonic()
|
|
C = A @ B.T
|
|
et = time.monotonic()
|
|
s = et-st
|
|
print(f"{flop/s * 1e-9:.2f} GFLOP/S, {s*1e3:.2f} ms")
|
|
|
|
with open("/tmp/matmul", "wb") as f:
|
|
f.write(A.data)
|
|
f.write(B.data)
|
|
f.write(C.data)
|