Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Apr 14, 2020

• Remove -ffast-math compile option.

  

  Danliran

  authored and

  Maratyszcza

  committedApr 14, 2020
  57616b9

Commits on Apr 12, 2020

• Update README

  Maratyszcza

  committedApr 12, 2020
  fb78b9b

• Migrate off of legacy pthreadpool API

  Maratyszcza

  committedApr 12, 2020
  a10d5ae

Commits on Mar 12, 2020

• enables NEON backend for armv7s

  

  dipidoo

  authored and

  Maratyszcza

  committedMar 12, 2020
  bda381b

Commits on Jan 2, 2020

• recognizes arm64(e) architectures for iOS

  

  dipidoo

  authored and

  Maratyszcza

  committedJan 2, 2020
  f345fc2

Commits on Dec 21, 2019

• avoids NEON backend on iphonesimulator

  

  dipidoo

  authored and

  Maratyszcza

  committedDec 21, 2019
  d062380

• adapts CMAKE_SYSTEM_NAME filter for CMake 3.14

  

  dipidoo

  authored and

  Maratyszcza

  committedDec 21, 2019
  7c4dac1

Commits on Nov 13, 2019

• Fix CMake Android toolchain path in README

  Maratyszcza

  committedNov 13, 2019
  a344d23

Commits on Oct 8, 2019

• Fix buffer overflow in zero'ing block (Maratyszcza#171)

  

  ajtulloch

  authored and

  Maratyszcza

  committedOct 8, 2019
  24b5530

Commits on Mar 23, 2019

• Adapt based on review feedback

  

  t-vi

  authored and

  Maratyszcza

  committedMar 23, 2019
  c039579

• Fix x86 32 bit detection

  

  t-vi

  authored and

  Maratyszcza

  committedMar 23, 2019
  293a700

Commits on Mar 20, 2019

• fix typo

  

  zenogantner

  authored and

  Maratyszcza

  committedMar 20, 2019
  da5c94b

Commits on Sep 3, 2018

• Remove invalid restrict attribute in softmax ukernel

  

  Marat Dukhan

  committedSep 3, 2018
  1e005b0

Commits on Jun 22, 2018

• Fix bug in validation of pooling parameters

  

  Marat Dukhan

  committedJun 22, 2018
  af40ea7

Commits on Jun 21, 2018

• Fix project name in cmake/DownloadCpuinfo.cmake

  

  Marat Dukhan

  committedJun 21, 2018
  d204380

Commits on Jun 14, 2018

• Fix missing clog dependency in confu build

  

  Marat Dukhan

  committedJun 14, 2018
  f28b0ac

Commits on Jun 13, 2018

• Rewrite input/output Winograd transform on NEON

  

  Marat Dukhan

  committedJun 13, 2018
  48cd8e0

• Micro-benchmark for Winograd transform on NEON

  

  Marat Dukhan

  committedJun 13, 2018
  e98dbd7

• Generalized vmuladd[q]_lane(i)_f32 functions for ARM NEON

  

  Marat Dukhan

  committedJun 13, 2018
  7bc75d6

Commits on Jun 12, 2018

• Fix ISA check for NEON FP16 compute

  

  Marat Dukhan

  committedJun 12, 2018
  864dd14

• Assembly implementation of NEON H4GEMM full ukernel for AArch32

  

  Marat Dukhan

  committedJun 12, 2018
  f2024bf

• Proper ISA check for NEON-FP16-ARITH version of H4GEMM unit test

  

  Marat Dukhan

  committedJun 12, 2018
  fd79cb2

Commits on Jun 11, 2018

• Use VLDM instructions instead of VLD1 in H4GEMM micro-kernel

  

  Marat Dukhan

  committedJun 11, 2018
  26c32d0

• H4GEMM micro-benchmark

  

  Marat Dukhan

  committedJun 11, 2018
  34942f8

• Fix compatibility with recent cpuinfo changes

  

  Marat Dukhan

  committedJun 11, 2018
  bfd8374

• Rewrite full H4GEMM NEON ukernel

  

  Marat Dukhan

  committedJun 11, 2018
  a1a7f76

• Unit tests for HXGEMM ukernels

  

  Marat Dukhan

  committedJun 11, 2018
  9c88dc8

• Rename TupleGEMM test to SXGEMM test

  

  Marat Dukhan

  committedJun 11, 2018
  52b29e9

• Test both accumulate-C and overwrite-C modes in SGEMM and SXGEMM ukernels

  

  Marat Dukhan

  committedJun 11, 2018
  633894a

Commits on May 22, 2018

• Specify non-executable stack for assembly files

  

  Marat Dukhan

  committedMay 22, 2018
  3eb0d45

Commits on May 11, 2018

• Optimize nnp_s4gemm_only_3x3__aarch32_neon2 for better perf on Samsung cores

  

  Marat Dukhan

  committedMay 11, 2018
  42d9355

Commits on Apr 30, 2018

• Optimized S4GEMM ukernel for Meerkat

  

  Marat Dukhan

  committedApr 30, 2018
  5233bc8

Commits on Apr 25, 2018

• AArch32 assembly versions of fast H4GEMM

  

  Marat Dukhan

  committedApr 25, 2018
  1a99ef6

• Fix 1x1 direct convolution benchmark

  

  Marat Dukhan

  committedApr 25, 2018
  085547d

• Relax accuracy checks in WT8x8 with FP16

  

  Marat Dukhan

  committedApr 25, 2018
  8d4498d

Pagination

PreviousNext