pyq doesn't run faster parallelly

[fouvy]

Questions

• [ ] Which operating system are you using (if Linux, please provide flavour of it, i.e RedHat, CentOS or Ubuntu), is it 32-bit, or 64-bit?

  ubuntu 18.04 64-bit

• [ ] Which version of PyQ are you running? Please provide output of pyq --versions, if PyQ isn't operational, please provide Python interpreter version and PyQ version python -V; python3 -V; pip list | grep pyq:

  4.2.1

• [ ] Which version of kdb+ are you using, is it 32-bit or 64-bit?

  KDB+ 4.0 2020.03.17 64-bit

  • [ ] If on 64-bit, is your QLIC set? Please provide output env | grep QLIC on linux/macOS, or set|grep QLIC on Windows.

    yes

• [ ] Did you use virtual environment to install PyQ? If not, why?

  • no

• [ ] Where is your QHOME? Please provide output env | grep QHOME on linux/macOS, or set|grep QHOME on Windows.

  /q

• [ ] Do you use Conda? If so, what version?

  no

Steps to reproduce the issue

Expected result

Actual result

Workaround

pyq doesn't run faster parallelly with kdb+ 4.0. This is my testbench.py file

import numpy as np
from time import time

def testbench(n):
    A = np.random.random((1, n))
    # Matrix multiplication
    N = 800000
    t = time()
    for i in range(N):
        A = A*A
    delta = time() - t
    print('Dotted two %dx%d matrices in %0.2f s.' % (1, n, delta))
    return np.sum(A)

if __name__ == "__main__":
    print(testbench(40000))

Then I ran with my kdb+ like following:

q -s 8
q)\l p.q
q)p)from testbench import testbench
q)l:40000 40000 40000 40000 40000 40000 40000 40000 
q)testbench:.p.get[`testbench;<]
q)testbench peach l
Dotted two 1x40000 matrices in 35.76 s.
Dotted two 1x40000 matrices in 35.76 s.
Dotted two 1x40000 matrices in 35.81 s.
Dotted two 1x40000 matrices in 35.80 s.
Dotted two 1x40000 matrices in 35.82 s.
Dotted two 1x40000 matrices in 35.82 s.
Dotted two 1x40000 matrices in 35.83 s.
Dotted two 1x40000 matrices in 35.85 s.
0 0 0 0 0 0 0 0f
q)testbench[l 0]
Dotted two 1x40000 matrices in 10.61 s.
0f
q)\s
8i

I found that pyq cannot run faster when I testing code under kdb+ 4.0. I can run my code faster when I run peach with c library. I thought it might be kdb+ slave doesn't fork embedpy independently.

[cmccarthy1]

Hi @fouvy

The updated peach functionality introduced in kdb+ 4.0 relates to updated performance for kdb+ primitives only, this is explicitly outlined here

As such it would not be expected that the addition of distributed primitives in kdb+ 4.0 would have any impact on the speed of execution of any code on Python side as is the case in your example. These are two separate problems.

[github-actions[bot]]

Stale issue message