![](././../../images_dir/1653356904/1.png)
## Self introduction **
PerfXPy
**
is a high-performance
**
Python
**
computing platform for data scientists, taking into account ease of use and high performance. It is mainly optimized for ARM, RISC-V and heterogeneous acceleration devices (GPU/NPU, etc.). In the future, PerfXPy will continue to integrate the accumulation of "Pengfeng Technology" in the fields of high-performance computing and heterogeneous computing. Through the underlying framework/computing library/compiler and other technologies, it will help algorithm developers to use heterogeneous computing platforms more conveniently and release hardware computing power.
## Alpha version
2022/5/20 Release V0.1, the first to support domestic chips (Feiteng).
-
System: Ubuntu/Linaro 5.5.0-12ubuntu1~16.04
-
CPU: Phytium, FT2000+/4
-
Memory: 8 GB
## "First meeting, please take care!" -
Simple installation, clean interface, and support for personalized settings
-
Support program debugging (breakpoints, variable viewing/assignment, output window, terminal console)
![](././../../images_dir/1653357221/3.png) ![](././../../images_dir/1653357255/4.png)
## "My pursuit of speed is endless!"
Note: Y-axis is time-consuming (ms), the lower the value, the faster the speed.
-
This example implements the multiplication of two higher-order matrices.
![](././../../images_dir/1653357321/5.png)
-
This example reads the handwritten digits dataset from the sklearn library, and then performs factor analysis on the dataset.
![](././../../images_dir/1653357358/6.png)
-
This example compares the MSE values of the Lw and Oas methods using Gaussian distributed data.
![](././../../images_dir/1653357391/7.png)
For more information, please see "Invitation for Internal Beta" below. (*^-^*)
## Future plans -
Support more ARM platform notebooks, desktops, servers. Such as FT2000, Kunpeng 920, Apple M1 series, AWS Graviton, and other ARM chips. At the same time, I also hope to support all kinds of domestic self-controllable operating systems as much as possible.
![](././../../images_dir/1653357478/8.png)
-
Get ready to support RISC-V 64-bit multi-core high-performance processors.
![](././../../images_dir/1653357514/9.png)
-
In order to support high-performance heterogeneous computing platforms, PerfXAPI has been prepared, which is the future and a challenge.
![](././../../images_dir/1653357542/10.png)
-
Another more important question, PerfXPy will continue to pay attention to what scientific computing toolkits, AI platform support, and data visualization components are needed by scientists and researchers in various fields.
## Internal test invitation
PerfXPy's cloud test platform (Feiteng FT2000+/4 server) has been prepared, friends who need remote experience, please add the QQ group below, and look forward to your various opinions. Thanks!
![](././../../images_dir/1653357632/11.png)
--- ![](././../../images_dir/1703074937/1.png)