Light

wu-kan / hpl-ai Goto Github PK

View Code? Open in Web Editor NEW

25.0 25.0 5.0 1.01 MB

An implementation of HPL-AI Mixed-Precision Benchmark based on hpl-2.3

Home Page: https://wu-kan.cn/2021/03/14/HPL-AI/

License: MIT License

Makefile 0.24% M4 0.11% C 60.99% C++ 38.66%

benchmarking gmres hpc hpl hpl-ai linpack mixed-precision mpi performance

hpl-ai's Introduction

星合の空

更严肃的自我介绍/简历

随笔

hpl-ai's People

Contributors

Stargazers

Watchers

Forkers

schuangs atestgroup233 kuangllbnu sysuyy duzhuqi

hpl-ai's Issues

感谢作者放出可用的 HPL-AI

工作中需要在FPGA上实现一个支持FP64-FP16混合精度的加速器，基于您基础上我只需要少量修改就可以完成交付了

HPL failed residual checks on linux-centos7-aarch64

$ spack debug report
* **Spack:** 0.16.1
* **Python:** 3.7.5
* **Platform:** linux-centos7-aarch64
$ spack find -v --loaded
==> 26 installed packages
-- linux-centos7-aarch64 / [email protected] ----------------------------
[email protected]~binutils~bootstrap~nvptx~piclibs~strip languages=c,c++,fortran patches=b620e9257ec6b9845f961931b0aa92c35b37e72b15d88ee392c7b67620ebaa2f,dc1ca240b7fb70112ae6cc47cd86925adf78d29ed9d0c26b0c51d52e40ceca0e
[email protected]
[email protected]
[email protected]
[email protected] patches=7a6dd71bcda4803d6b89612706a17b8816e1acd5dd9bf1bec29cf748f3b60008
[email protected]+optimize+pic+shared

-- linux-centos7-aarch64 / [email protected] ----------------------------
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]~cuda~ipo+openmp+shared build_type=RelWithDebInfo cuda_arch=none
[email protected]
[email protected]~cairo~cuda~gl~libudev+libxml2~netloc~nvml+pci+shared
[email protected]
[email protected]
[email protected]
[email protected]~python
[email protected]+sigsegv patches=3877ab548f88597ab2327a2230ee048d2d07ace1062efe81fc92e91b7f39cd00,fc9b61654a3ba1a8d6cd78ce087e7c96366c290bc8d2c299f09828d793b853c8
[email protected]~symlinks+termlib
[email protected] patches=4e1d78cbbb85de625bad28705e748856033eaafab92a66dffd383a3d7e00cc94
[email protected]~consistent_fpcsr~ilp64+pic+shared threads=openmp
[email protected]~atomics~cuda~cxx~cxx_exceptions+gpfs~java~legacylaunchers~lustre~memchecker~pmi~singularity~sqlite3+static~thread_multiple+vt+wrapper-rpath fabrics=none schedulers=none
[email protected]+cpanm+shared+threads
[email protected]
[email protected]~pic
[email protected]+optimize+pic+shared

对hpl-ai+ascend的运行结果有点困惑

我在昇腾设备上运行了hpl-ai+ascend代码，成功获得了PASS的结果。我知道大概步骤是先用半精度获得一个近似的结果，然后用数值迭代的方法获得满足精度要求的最终结果。我找出了代码中数值迭代的代码HPL_pir，我让程序直接跳过HPL_pir，我本以为会获得一个近似但不准确的结果，但输出是这样的：

================================================================================
HPL-AI Mixed-Precision Benchmark v2.3d  --  April 23, 2021
Written by Junkang Huang and Kan Wu,  Sun Yat-sen University
================================================================================

An explanation of the input/output parameters follows:
T/V    : Wall time / encoded variant.
N      : The order of the coefficient matrix A.
NB     : The partitioning blocking factor.
P      : The number of process rows.
Q      : The number of process columns.
Time   : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.

The following parameter values will be used:

N      :   32768 
NB     :    4096     4096 
PMAP   : Row-major process mapping
P      :       2 
Q      :       4 
PFACT  :   Right 
NBMIN  :       2 
NDIV   :       2 
RFACT  :   Right 
BCAST  :   1ring 
DEPTH  :       1 
SWAP   : Spread-roll (long)
L1     : no-transposed form
U      : transposed form
EQUIL  : yes
ALIGN  : 16 HPLAI_T_AFLOAT precision words

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

- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
      ||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be               1.110223e-16
- Computational tests pass if scaled residuals are less than                16.0

================================================================================
T/V                N    NB     P     Q               Time                 Gflops
--------------------------------------------------------------------------------
WR10R2R2       32768  4096     2     4              39.43             5.9493e+02
HPLAI_pdgesv() start time Thu Aug 26 04:52:00 2021

HPLAI_pdgesv() end time   Thu Aug 26 04:52:40 2021

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=   2.74877907e+11 ...... FAILED
||Ax-b||_oo  . . . . . . . . . . . . . . . . . =           0.499998
||A||_oo . . . . . . . . . . . . . . . . . . . =      114629.501221
||A||_1  . . . . . . . . . . . . . . . . . . . =       74376.728001
||x||_oo . . . . . . . . . . . . . . . . . . . =           0.000000
||x||_1  . . . . . . . . . . . . . . . . . . . =           0.000000
||b||_oo . . . . . . . . . . . . . . . . . . . =           0.499998
================================================================================
T/V                N    NB     P     Q               Time                 Gflops
--------------------------------------------------------------------------------
WR10R2R2       32768  4096     2     4              33.78             6.9446e+02
HPLAI_pdgesv() start time Thu Aug 26 04:52:42 2021

HPLAI_pdgesv() end time   Thu Aug 26 04:53:16 2021

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=   2.74877907e+11 ...... FAILED
||Ax-b||_oo  . . . . . . . . . . . . . . . . . =           0.499998
||A||_oo . . . . . . . . . . . . . . . . . . . =      114629.501221
||A||_1  . . . . . . . . . . . . . . . . . . . =       74376.728001
||x||_oo . . . . . . . . . . . . . . . . . . . =           0.000000
||x||_1  . . . . . . . . . . . . . . . . . . . =           0.000000
||b||_oo . . . . . . . . . . . . . . . . . . . =           0.499998
================================================================================

Finished      2 tests with the following results:
              0 tests completed and passed residual checks,
              2 tests completed and failed residual checks,
              0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.
================================================================================

x的范数是0，似乎x并没有计算出来。
目前我不确定是我编译运行的操作问题，还是我对代码的理解有问题，还是代码存在bug？
我的操作步骤是这样的：

spack unload -a
spack load [email protected]
spack load [email protected]%[email protected]
spack load [email protected]%[email protected]
spack load [email protected]%[email protected]
spack load [email protected]%[email protected]
spack load [email protected]%[email protected] arch=linux-ubuntu18.04-aarch64

autoreconf -ivf
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest/arm64-linux
./configure LIBS="-L$DDK_PATH/fwkacllib/lib64 -lascendcl -lacl_op_compiler" CPPFLAGS="-I$DDK_PATH/fwkacllib/include -DHPLAI_ACL_BLASPP_GEMM"
make -j
OMP_NUM_THREADS=24 $(which mpirun) -n 8 testing/xhpl_ai

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.