Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 4, 2020

• Clean up whitespaces

  xta0

  committed
  74208c8

  Copy full SHA for 74208c8

Commits on Feb 3, 2020

• Fix iOS build

  xta0

  committed
  8426814

  Copy full SHA for 8426814

Commits on Jan 28, 2020

• Reformat tabulation in xnnpack.h header

  Show description for 03bc407
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  03bc407

  Copy full SHA for 03bc407

• F32 Softmax operator benchmark

  Show description for 9c0db96
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  9c0db96

  Copy full SHA for 9c0db96

• Document Softmax support in README

  Show description for 49adc2d
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  49adc2d

  Copy full SHA for 49adc2d

Commits on Jan 27, 2020

• Rename SoftArgMax operator to SoftMax

  Show description for fd8e689
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  fd8e689

  Copy full SHA for fd8e689

• SoftArgMax operator

  Show description for 1edc454
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  1edc454

  Copy full SHA for 1edc454

• Remove extraneous argument from mobilenetv3 igemm benchmark.

  Show description for e104aa3
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  e104aa3

  Copy full SHA for e104aa3

Commits on Jan 25, 2020

• NEON/NEONFMA RAddStoreExpMinusMax micro-kernels

  Show description for 8137e4c
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  8137e4c

  Copy full SHA for 8137e4c

Commits on Jan 24, 2020

• SSE2/PSIMD RAddStoreExpMinusMax micro-kernels

  Show description for b39689d
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  b39689d

  Copy full SHA for b39689d

Commits on Jan 21, 2020

• Scalar RAddStoreExpMinusMax micro-kernels

  Show description for f46f675
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  f46f675

  Copy full SHA for f46f675

Commits on Jan 16, 2020

• 6x8 a53 use X8 for GPR shadow register. Eliminate GPR push/pop

  Show description for d6ebf0c
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  d6ebf0c

  Copy full SHA for d6ebf0c

• A53 GEMM / IGEMM kernel prefetches adjust by 1

  Show description for 534375d
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  534375d

  Copy full SHA for 534375d

Commits on Jan 15, 2020

• 4x12 A53 GEMM and IGEMM use X8 for temp GPR

  Show description for c03b2bd
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  c03b2bd

  Copy full SHA for c03b2bd

• Add link to the two-pass softmax paper in README

  Show description for 640b35d
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  640b35d

  Copy full SHA for 640b35d

Commits on Jan 14, 2020

• 4x8 A53 GEMM use X4 for temp GPR

  Show description for 324f2bb
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  324f2bb

  Copy full SHA for 324f2bb

• 4x8 Cortex-A53 GEMM / IGEMM use 1 GPR instead of 2.

  Show description for 7693acf
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  7693acf

  Copy full SHA for 7693acf

• 6X8 Cortex-A53 GEMM use 1 GPR instead of 2.

  Show description for f884a7b
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  f884a7b

  Copy full SHA for f884a7b

Commits on Jan 11, 2020

• Use 2 GPR registers instead of 4 for GPR loads.

  Show description for 54afb13
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  54afb13

  Copy full SHA for 54afb13

• 4x12 IGEMM use prefetch on A and B

  Show description for cbb35d0
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  cbb35d0

  Copy full SHA for cbb35d0

• 6X8 Cortex-A53 IGEMM use 1 GPR instead of 2.

  Show description for 3216758
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  3216758

  Copy full SHA for 3216758

Commits on Jan 10, 2020

• 4x12 A53 kernel use prefetches on A

  Show description for 4cd8907
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  4cd8907

  Copy full SHA for 4cd8907

Commits on Jan 9, 2020

• Fix gcc incompatibilities in NEON Sigmoid micro-kernels

  Show description for 26cda6d
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  26cda6d

  Copy full SHA for 26cda6d

• Update benchmarking results on mobile phones and Raspberry Pi boards

  Show description for 6a6a421
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  6a6a421

  Copy full SHA for 6a6a421

• A53 aarch32 pipelined. 19.2% faster 4x8 GEMM, 9.3% faster end to end

  Show description for c01d8a4
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  c01d8a4

  Copy full SHA for c01d8a4

Commits on Jan 7, 2020

• Remove prefetch of output buffer from A53 kernels.

  Show description for b177732
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  b177732

  Copy full SHA for b177732

• F32 Sigmoid micro-kernels in AVX2 implementation

  Show description for fa0a432
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  fa0a432

  Copy full SHA for fa0a432

Commits on Jan 6, 2020

• Use 1-step range reduction in NEONFMA Sigmoid micro-kernels

  Show description for 4a24a58
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  4a24a58

  Copy full SHA for 4a24a58

• NEON Sigmoid evaluation stubs with 1-step range reduction

  Show description for 77221d3
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  77221d3

  Copy full SHA for 77221d3

• AVX2 Sigmoid accuracy evaluation stubs

  Show description for 7241600
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  7241600

  Copy full SHA for 7241600

Commits on Jan 2, 2020

• Complete set of NEON F32 Sigmoid micro-kernels

  Show description for 68b3b45
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  68b3b45

  Copy full SHA for 68b3b45

• Apply patch to cpuinfo

  Show description for 1f5d9bc
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  1f5d9bc

  Copy full SHA for 1f5d9bc

• Additional Sigmoid micro-kernels and accuracy evaluation stub

  Show description for 8d3c07e
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  8d3c07e

  Copy full SHA for 8d3c07e

Commits on Dec 31, 2019

• A75 / A53 aarch32 epilogue reordered by B the same as main loop.

  Show description for 279908a
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  279908a

  Copy full SHA for 279908a

Commits on Dec 23, 2019

• Scalar F32 Sigmoid micro-kernels

  Show description for 3a77ea7
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  3a77ea7

  Copy full SHA for 3a77ea7

Pagination

Previous

Next