[GitHub] szha opened a new pull request #8642: ctypes speed improvement

[GitBox <gi...@apache.org>]

szha opened a new pull request #8642: ctypes speed improvement
URL: https://github.com/apache/incubator-mxnet/pull/8642
 
 
   ## Description ##
   Improvements on ctypes calls for creating arrays. The changes are based on the following micro benchmarks.
   ```python
   import timeit
   # ints
   setup="from array import array; import ctypes; t = [i for i in range(1000000)];"
   print('int before', timeit.timeit(stmt='(ctypes.c_uint32 * len(t))(*t)',setup=setup,number=10))
   print('int after', timeit.timeit(stmt="v = array('I',t);a = (ctypes.c_uint32 * len(v)).from_buffer(v)",setup=setup,number=10))
   
   setup="""
   import ctypes
   t = ['abc' for _ in range(100000)]
   
   import sys
   if sys.version_info[0] < 3:
       def c_str(string):
           return ctypes.c_char_p(string)
       def c_str_array(strings):
           arr = (ctypes.c_char_p * len(strings))()
           arr[:] = strings
           return arr
   else:
       def c_str(string):
           return ctypes.c_char_p(string.encode('utf-8'))
       def c_str_array(strings):
           arr = (ctypes.c_char_p * len(strings))()
           arr[:] = [s.encode('utf-8') for s in strings]
           return arr
   
   def c_array(ctype, values):
       return (ctype * len(values))(*values)
   def c_array_objs(L):
       arr = (ctypes.c_void_p * len(L))()
       arr[:] = [s.handle for s in L]
       return arr
   """
   # strings
   print('strings before', timeit.timeit(stmt='r = c_array(ctypes.c_char_p, [c_str(s) for s in t]); assert r[24601].decode("utf-8") == "abc"',setup=setup,number=10))
   print('strings after', timeit.timeit(stmt='r = c_str_array(t); assert r[24601].decode("utf-8") == "abc"',setup=setup,number=10))
   
   
   setup="""
   import ctypes
   import mxnet as mx
   t = [mx.nd.array([i]) for i in range(100000)]
   
   def c_array(ctype, values):
       return (ctype * len(values))(*values)
   def c_array_objs(L):
       arr = (ctypes.c_void_p * len(L))()
       arr[:] = [s.handle for s in L]
       return arr
   """
   # arrays
   print('handles before', timeit.timeit(stmt='r = c_array(mx.base.NDArrayHandle, [i.handle for i in t])',setup=setup,number=10))
   print('handles after', timeit.timeit(stmt='r = c_array_objs(t)',setup=setup,number=10))
   ```
   
   Results (time in seconds, the lower the better)
   
   | python version | int before | int after | strings before | strings after | handles before | handles after |
   |:----------------:|------------:|-----------:|----------------:|---------------:|----------------:|---------------:|
   | 2.7.13         | 2.50       | 0.54      | 1.52           | 0.30          | 0.46           | 0.23          |
   | 3.6.2          | 2.36       | 0.13      | 1.13           | 0.58          | 0.32           | 0.12          |
   
   ## Checklist ##
   ### Essentials ###
   - [x] Passed code style checking (`make lint`)
   - [x] Changes are complete (i.e. I finished coding on this PR)
   - [x] All changes have test coverage
   - [x] For user-facing API changes, API doc string has been updated. For new C++ functions in header files, their functionalities and arguments are well-documented. 
   - [x] To my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change
   
   ### Changes ###
   - [x] update ctypes `c_array` calls

----------------------------------------------------------------
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