HN Explorer

Marat_Dukhan

↗ HN profile [ 87.7 ms ] full profile
Karma
625
Created
November 3, 2012 (13y ago)
Submissions
0
XNNPACK TLM @ Google, previously QNNPACK lead @ Facebook, Ph.D. student @ Georgia Tech, and the author of NNPACK library.

My Ph.D. projects:

www.github.com/Maratyszcza/NNPACK

www.github.com/Maratyszcza/PeachPy

www.github.com/Maratyszcza/Opcodes

Cool demos:

www.github.com/Maratyszcza/blis-bench

www.github.com/Maratyszcza/laff-demos

  1. Faster Quantized Neural Network Inference with XNNPack (blog.tensorflow.org)
    18 points 16
  2. MediaPipe Pose (google.github.io)
    1 points 0
  3. M1 Mac Mini at MacStadium (macstadium.com)
    2 points 0
  4. Background Features in Google Meet, Powered by Web ML (ai.googleblog.com)
    427 points 275
  5. Supercharging TensorFlow.js with SIMD and multi-threading (blog.tensorflow.org)
    63 points 19
  6. Introducing the WebAssembly backend for TensorFlow.js (blog.tensorflow.org)
    4 points 0
  7. Fast Sparse ConvNets (arxiv.org)
    4 points 1
  8. Show HN: detecting cache latency inside a Web browser (maratyszcza.github.io)
    107 points 49
  9. mInference demo: ImageNet convnets inside a Web browser (WAsm/Asm.js/PNaCl) (minference.com)
    1 points 0
  10. NNPACK – acceleration package for neural networks on multi-core CPUs (github.com)
    8 points 1
  11. Show HN: PeachPy – x86-64 assembler in Python, with Go support (github.com)
    4 points 0
  12. In-Browser Memory Latency Benchmark (PNaCl/Asm.js) (maratyszcza.github.io)
    2 points 0
  13. Benchmark of matrix-matrix multiplication in Web browsers (github.com)
    2 points 0
  14. BLIS for the Web: High-Performance Computing in a Web Browser (cs.utexas.edu)
    2 points 0
  15. Spec: JavaScript Shared Memory, Atomics, and Locks (docs.google.com)
    6 points 0
  16. Yeppp! a trivial test (nosubstance.me)
    1 points 0
  17. Show HN: Open-source SIMD-accelerated math library for C/C++/C#/Java/FORTRAN (yeppp.info)
    9 points 3
  18. Intel AVX-512, SHA, and MPX (Memory Protection) instruction extensions (software.intel.com)
    1 points 0
  19. Yeppp! performance numbers (yeppp.info)
    1 points 0
  20. GCC 4.8 release (gcc.gnu.org)
    110 points 40