Skip to content

Latest commit

 

History

105 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

tc-benchmark

Code for Tensor Core Micro Benchmarking for PIC2 MEEC

Current Tasks

  • Half Precision
    • FMA benchmark
    • Compute Roof -> FLOPs
  • Alternate Formats
    • Bf16
    • Tf32
  • Single Precision
  • Double Precision
  • Integer

Benchmarking Results

Inputs Accumulators Shape FMA/clk/SM FLOPs
FP16 FP16 m16n8k8 511.438
FP16 FP16 m16n8k16 511.820
FP16 FP16 m8n8k4ª 1.0093
FP16 FP32 m16n8k8 255.675
FP16 FP32 m16n8k16 255.942
FP16 FP32 m8n8k4ª 1.0441

Notes: ª - "mma.sync.m8n8k4 is optimized for target architecture sm_70 and may have substantially reduced performance on other target architectures." 1

About

Code for Tensor Core Micro Benchmarking, for PIC2 MEEC

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages