Pull requests

• cpu: evaluate arrays past 2^31 elements and sum with compensation

  #3731·melonakos opened Sep 12, 2026Member·

  ·

• api: convert images to float before saving them

  #3730·melonakos opened Sep 12, 2026Member·

  ·

• api: materialise a strided view before assigning into it

  #3729·melonakos opened Sep 12, 2026Member·

  ·

• opencl: size the Harris responses grid by columns then rows

  #3728·melonakos opened Sep 11, 2026Member·

  ·

• opencl: keep half-precision matrix-vector products on the GEMM path

  #3727·melonakos opened Sep 11, 2026Member·

  ·

• opencl: reduce moments in local memory without float atomics

  #3726·melonakos opened Sep 11, 2026Member·

  ·

• cuda: serialise constant-memory uploads with their kernel launches

  #3725·melonakos opened Sep 11, 2026Member·

  ·

• build: add AF_WITH_GRAPHICS to build without Forge, OpenGL and glad

  #3724·melonakos opened Sep 10, 2026Member·

  ·

• api: let host() copy sparse arrays as dense

  #3723·melonakos opened Sep 10, 2026Member·

  ·
  1

• build: compile against fmt 10 and 11

  #3722·melonakos opened Sep 10, 2026Member·

  ·
  1

• cpu: fix the FAST non-max bounds check

  #3720·melonakos opened Sep 10, 2026Member·

  ·

• windows: check FormatMessage result in getErrorMessage

  #3719·melonakos opened Sep 10, 2026Member·

  ·
  1

• unified: do not call a null symbol in af_get_last_error

  #3718·melonakos opened Sep 10, 2026Member·

  ·
  1

• cuda: honour the documented 64-byte d_name buffer in af_device_info

  #3717·melonakos opened Sep 10, 2026Member·

  ·
  1

• cuda: build with CUDA 13 and target Blackwell

  #3715·melonakos opened Sep 10, 2026Member·

  ·
  1

• Fix ambiguous std::abs call in confidenceCC

  #3710·daeho-ro opened Jul 12, 2026·5 tasks·

  ·

• cuda: guard removed cuFFT error enums

  #3706·LwhJesse opened May 13, 2026·

  ·

• build: add missing direct includes

  #3705·LwhJesse opened May 13, 2026·

  ·
  1

• cuda: use strided batched GEMM for linear batch layouts

  #3704·LwhJesse opened May 12, 2026·

  ·

• Add support for CUDA 13.1 and Blackwell GPUS

  #3698·christophe-murphy opened Mar 7, 2026Contributor·

  ·

• Revert cl2hpp dependency version.

  #3697·christophe-murphy opened Mar 7, 2026Contributor·

  ·
  1

• support CUDA 13.1, SM 12.0

  #3696·ktdq opened Dec 7, 2025Contributor·

  ·
  1

• Fixes to mean and better precision on sum (cpu)

  #3687·willyborn opened Aug 4, 2025Contributor·3 of 5 tasks·

  ·
  4

• Fixes sub array (opencl, cuda, cpu, oneapi) for orb

  #3670·willyborn opened Jun 22, 2025Contributor·3 of 5 tasks·

  ·
  4

• if not Win check added

  #3653·ved1beta opened Apr 15, 2025·3 of 5 tasks·

  ·
  1

  1