You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by GitBox <gi...@apache.org> on 2018/06/06 21:51:22 UTC
[GitHub] haojin2 commented on issue #11179: [MXNET-404] elemwise_add/sub
between rsp and rsp on GPU
haojin2 commented on issue #11179: [MXNET-404] elemwise_add/sub between rsp and rsp on GPU
URL: https://github.com/apache/incubator-mxnet/pull/11179#issuecomment-395225648
Benchmark script:
```Python
import mxnet as mx
import sys
import os
import scipy
import numpy as np
from mxnet.test_utils import rand_ndarray, assert_almost_equal
import time
def measure_cost(repeat, a, b, out=None):
# start bench
start = time.time()
results = []
for i in range(repeat):
results.append(mx.nd.elemwise_add(a, b, out=out))
for result in results:
result.wait_to_read()
end = time.time()
diff = end - start
return diff / repeat
def measure_fallback(repeat, a):
# start bench
start = time.time()
results = []
for i in range(repeat):
results.append(a.tostype('default'))
for result in results:
result.wait_to_read()
end = time.time()
diff = end - start
return diff / repeat
def main():
shape = (1000000, 512)
context = mx.gpu(0)
# context = mx.cpu()
for lhs_density in [0.01, 0.005, 0.001, 0.0005, 0.0001, 0.000]:
mx_lhs = rand_ndarray(shape, stype='row_sparse', density=lhs_density).as_in_context(context)
mx_lhs_dns = mx_lhs.tostype('default')
for rhs_density in [0.01, 0.005, 0.001, 0.0005, 0.0001, 0.000]:
mx_rhs = rand_ndarray(shape=shape, stype='row_sparse', density=rhs_density).as_in_context(context)
mx_rhs_dns = mx_rhs.tostype('default')
#warmup
sparse_cost = 0.0
dns_cost = 0.0
np_lhs = mx_lhs_dns.asnumpy()
check = mx.nd.elemwise_add(mx_lhs, mx_rhs)
np_lhs = np_lhs + mx_rhs.asnumpy()
assert_almost_equal(check.asnumpy(), np_lhs, atol=1e-5, rtol=1e-4)
mx.nd.waitall()
for i in range(100):
sparse_cost += measure_cost(1, mx_lhs, mx_rhs)
dns_cost += measure_cost(1, mx_lhs_dns, mx_rhs_dns)
print("%.2f %% %.2f %%" % (lhs_density*100, rhs_density*100), dns_cost / sparse_cost)
for rhs_density in [1.000, 0.01, 0.005, 0.001, 0.0005, 0.0001, 0.000]:
mx_lhs_dns = mx.nd.ones(shape, ctx=context)
mx_lhs = mx_lhs_dns.tostype('row_sparse')
mx_rhs = rand_ndarray(shape=shape, stype='row_sparse', density=rhs_density).as_in_context(context)
mx_rhs_dns = mx_rhs.tostype('default')
#warmup
sparse_cost = 0.0
dns_cost = 0.0
np_lhs = mx_lhs_dns.asnumpy()
mx.nd.elemwise_add(mx_lhs, mx_rhs, out=mx_lhs)
np_lhs = np_lhs + mx_rhs.asnumpy()
assert_almost_equal(mx_lhs.asnumpy(), np_lhs, atol=1e-5, rtol=1e-4)
mx.nd.waitall()
for i in range(100):
sparse_cost += measure_cost(1, mx_lhs, mx_rhs, out=mx_lhs)
dns_cost += measure_cost(1, mx_lhs_dns, mx_rhs_dns, out=mx_lhs_dns)
print("%.2f %% %.2f %%" % (1.00000*100, rhs_density*100), dns_cost / sparse_cost)
for lhs_density in [1.000, 0.01, 0.005, 0.001, 0.0005, 0.0001, 0.000]:
mx_rhs_dns = mx.nd.ones(shape, ctx=context)
mx_rhs = mx_rhs_dns.tostype('row_sparse')
mx_lhs = rand_ndarray(shape=shape, stype='row_sparse', density=lhs_density).as_in_context(context)
mx_lhs_dns = mx_lhs.tostype('default')
#warmup
sparse_cost = 0.0
dns_cost = 0.0
np_rhs = mx_rhs_dns.asnumpy()
mx.nd.elemwise_add(mx_lhs, mx_rhs, out=mx_rhs)
np_rhs = np_rhs + mx_lhs.asnumpy()
assert_almost_equal(mx_rhs.asnumpy(), np_rhs, atol=1e-5, rtol=1e-4)
mx.nd.waitall()
for i in range(100):
sparse_cost += measure_cost(1, mx_lhs, mx_rhs, out=mx_rhs)
dns_cost += measure_cost(1, mx_lhs_dns, mx_rhs_dns, out=mx_rhs_dns)
print("%.2f %% %.2f %%" % (1.00000*100, lhs_density*100), dns_cost / sparse_cost)
if __name__ == "__main__":
main()
```
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services