Repository navigation
-P flag appears to be broken in example README command #18
Description
Summary:
It seems as though the -P flag and it's other synonyms break the fft_bench.py script for a particular command given in the README
python fft_bench.py -P -r -t 1 -d float32 -o 5 100000000
Output:
(fft-bench) harlowjo@harlowjo-mobl:~/repositories/IntelPython/fft_benchmark$ python fft_bench.py -P -r -t 1 -d float32 -o 5 100000000
/home/harlowjo/miniconda3/envs/fft-bench/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.23.5
warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
TAG: CONDA_DEFAULT_ENV = fft-bench
TAG: numpy.__mkl_version__ = 2023.0.1
prefix,module,function,threads,dtype,size,place,time
Traceback (most recent call last):
File "/home/harlowjo/repositories/IntelPython/fft_benchmark/fft_bench.py", line 169, in <module>
perf_times = perf.time_func(func, arr, kwargs, **time_kwargs)
File "/home/harlowjo/repositories/IntelPython/fft_benchmark/perf.py", line 161, in time_func
res = func(buf, **kwargs)
File "/home/harlowjo/miniconda3/envs/fft-bench/lib/python3.10/site-packages/mkl_fft/_scipy_fft_backend.py", line 113, in __ua_function__
return fn(*args, **kwargs)
TypeError: rfft() got an unexpected keyword argument 'overwrite_x'
Steps to reproduce:
I followed the README instructions nearly identically
conda create -n fft-bench -c intel numpy scipyconda activate fft-benchpython fft_bench.py -P -r -t 1 -d float32 -o 5 100000000
I noticed I can run the same command without -P and it works.
Running it with --overwrite-x and --in-place caused the same error as above.
Please let me know if I did something wrong with my setup or if you need more info
More info:
scipy-1.7.3 | py310h77d8492_8 22.9 MB intel
numpy-1.23.5 | py310h52df89b_7 4 KB intel
mkl-2023.1.0 | intel_46342 197.2 MB intel
mkl-service-2.4.0 | py310h75d02e3_15 58 KB intel
mkl_fft-1.3.5 | py310h28f0b46_11 260 KB intel
mkl_random-1.2.2 | py310h0b06908_51 375 KB intel
mkl_umath-0.1.1 | py310h450dca2_61 296 KB intel
Activity
oleksandr-pavlyk commentedon May 9, 2023
The
rfftdoes not provide in-place operation mode. Please do not use-Pwhen compute FFT of real-value floating-point type datasets.