https://github.com/IntelLabs/SLIDE_opt_ia

https://github.com/IntelLabs/SLIDE_opt_ia

Last synced: 7 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/IntelLabs/SLIDE_opt_ia
• Owner: IntelLabs
• License: mit
• Archived: true
• Created: 2021-03-09T03:11:25.000Z (over 3 years ago)
• Default Branch: prv_opt_ia
• Last Pushed: 2023-10-31T15:01:08.000Z (about 1 year ago)
• Last Synced: 2024-08-01T03:28:14.734Z (3 months ago)
• Language: C++
• Size: 22 MB
• Stars: 74
• Watchers: 16
• Forks: 14
• Open Issues: 3
• Metadata Files:
  • Readme: README.Intel.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Build/Run on Intel platform

## Prerequisites:

CMake >= 3.0

Intel Compiler (ICC) >= 19

## Build with ICC compiler

```

source /opt/intel/compilers_and_libraries/linux/bin/compilervars.sh -arch intel64 -platform linux

cd /path/to/slide-root

mkdir -p bin && cd bin 

# BDW (AVX2)

cmake .. -DCMAKE_CXX_COMPILER=icpc -DCMAKE_C_COMPILER=icc

# SKX/CLX (AVX512)

cmake .. -DCMAKE_CXX_COMPILER=icpc -DCMAKE_C_COMPILER=icc -DOPT_AVX512=1

# CPX (AVX512 + BF16)

cmake .. -DCMAKE_CXX_COMPILER=icpc -DCMAKE_C_COMPILER=icc -DOPT_AVX512=1 -DOPT_AVX512_BF16=1

make -j

```

## Run on Intel SKX/CLX/CPX

```

cd bin

OMP_NUM_THREADS= KMP_HW_SUBSET=s,c,t KMP_AFFINITY=compact,granularity=fine KMP_BLOCKTIME=200 ./runme ../SLIDE/Config_amz.csv

For example, on CLX8280 2Sx28c:

OMP_NUM_THREADS=112 KMP_HW_SUBSET=2s,28c,2t KMP_AFFINITY=compact,granularity=fine KMP_BLOCKTIME=200 ./runme ../SLIDE/Config_amz.csv

```

For best performance please set Batchsize=multiple-of-logic-core-number from SLIDE/Config_amz.csv.

Results can be checked from the log file under dataset:

```

tail -f dataset/log.txt

```