Skip to content

Latest commit

 

History

History
93 lines (50 loc) · 1.19 KB

CHANGELOG.md

File metadata and controls

93 lines (50 loc) · 1.19 KB
Raw

0.1.0

First release

0.1.5

Added GPU support for static vectors and matrices in 32-bit precision

0.2.0

  • Added dynamic vectors and matrices
  • More consistent API
  • Improved type inference
  • Universal functions

0.2.1

Minor API improvement: no need to specify float32 when creating literal instances of Vector32

0.2.2

Started adding 64-bit operations on GPU

0.3.0

Support for (static) 64-bit types on the GPU

0.4.0

Support for dynamic types (32 or 64-bit) on the GPU

0.4.1

Moved term rewrites to a separate package

0.4.2

Fix for latest Nim

0.4.3

Fix for latest Nim

0.4.4

Added LAPACK solver functions

0.5.0

Fixed CUDA compilation with latest Nim

0.5.1

Added a few aggregate functions (mean, sum, ...)

0.5.2

Added more aggregate functions

0.5.3

Minor fix

0.6.0

  • Fixed various deprecations
  • Improved dependencies
  • Added Hadamard products

0.6.1

Added templates to generate vectors and matrices more efficiently

0.6.2

Minor fix

0.6.3

Minor fix

0.6.4

Fixed weird scoping with new templates to generate vectors and matrices

0.6.5

Had to restore previous version with broken scoping

0.6.5

Fixed template scoping again