CHANGELOG.md

0.0.8 (2016-02-21)

Features

• backend:
  • implement Backend::default() (0d4f0cc2)
  • add ability to synchronize a backend (e1bf0147)
• tensor: add resize to SharedTensor (f3bb3b44)
• feature_flag: add alternative unstable method to allocate native memory (usable via unstable_alloc feature flag) (ebaf3dac)
• export: add prelude module (import most of collenchyma via use collenchyma::prelude::*;) (4aea78df, 60c1a395)
• license: change license to dual MIT/Apache-2.0 (af5c08a0)

Bug Fixes

• cuda_context: fix repeating CUDA context creation (0bf616f8)

0.0.7 (2015-12-21)

Features

• backend_device: add device() to IBackend (f797a72a)
• tensor: add reshape to SharedTensor (9c8721ed)

Bug Fixes

• lib: various fixes concerning impl of plugins (bec27ca1)

0.0.6 (2015-12-16)

Features

• tensor: rename SharedMemory -> Tensor (acc3cbb7)

Bug Fixes

• travis: fix feature build flags (f9861a73)

0.0.5 (2015-12-09)

Bug Fixes

• use-size-types: use target dependent size types (4e4a5cd3)

Features

• bench: add benchmarks for memory synchronization (762b87ca)
• features: add native and opencl feature (3609fea1)

0.0.4 (2015-12-08)

Features

• cuda:
  • add memory synchronization (760e4c45)
  • add cuda device support (62eef6d8)
  • add memory allocation (35f7f479)
  • add basic context and memory support (b0a40d38)
• cudnn: add cudnn ffi (0bbbff83)
• library: remove last pieces of library (38dcd6a6)
• opencl: implement shared_memory for OpenCL (be47d6ba)
• perf: make error messages static strings (430c4ed6)
• plugin: move library out; replace with thin plugin mod (3bbebe9a)
• shared_memory: add dimensionality to shared_memory (13cd0905)

Performance

• blas: reduce overhead of calling blas (8b7a7aee)
• shared_memory: use linear_map for SharedMemory.copies (44ea377d)

Bug Fixes

• compilation: make cuda optional (1f933977)
• windows:
  • use size_t instead of u64 where necessary (6e9fdfbb)
  • add the link attribute to OpenCL external block (2017a10f)

0.0.3 (2015-11-30)

Features

• blas:
  • add blas native level 1 support (38273645)
  • add basic level 1 blas methods for native (62cbc4c4)
• computation: add basic design for backend-agnostic computation (a3f9534f)
• cuda:
  • add cuda structure (d42430c1)
• dot: add working dot computation (6572c010)
• extern: add backend traits for extern support (f3d50172)
• flatbox: provide slice accessors and give more allocation responsibility to shared_memory (cfbb5b13)
• memory:
  • implement MemoryType unwrappers (fbd26776)
  • add SharedMemory.latest_device() (d425fc6c)

Bug Fixes

• ci: change clippy to be optional (753dfb02)

0.0.2 (2015-11-27)

Features

• collenchyma: outline design (876ac4b0)
• computation: add basic design for backend-agnostic computation (e43f947a)
• dot: add working dot computation (f5c8fdaa)
• flatbox: provide slice accessors and give more allocation responsibility to shared_memory (a31dd493)
• lib: add foundation of the library design (f5fd0235)
• memory:
  • implement MemoryType unwrappers (a78edebb)
  • add SharedMemory.latest_device() (f9a7465d)

Bug Fixes

• ci: change clippy to be optional (db13da29)