This is a patch update release to v2.12.1.
It has the following:

• Fix for thread safety issues
• Fix for failures in certain 2D transforms
• Fix for failures on POCL runtime
• Other minor improvements

For details on issues, please see issue trackers.

clFFT - Release Notes - version 2.12.1:

• Fixes for bugs in decomposition logic of very large problem sizes
  with environment flag CLFFT_REQUEST_LIB_NOMEMALLOC=1
• Validation of sizes that are powers of 2,3,5 and combinations
  thereof upto maximum size close to 25GB on S9170

Driver notes:

• This library version has been tested with Catalyst
  Pro driver version 15.302 on Firepro W9100 & S9170.

clFFT - Release Notes - version 2.12.0:

• Addition of radix 11 & 13 enables support for powers of 11 & 13 size transforms
• Support for 1D large size inplace transforms with no extra memory allocation
  requirement with environment flag CLFFT_REQUEST_LIB_NOMEMALLOC=1
  for complex FFTs of powers of 3,5,10 sizes. Note that support for
  power-of-2 size has been available since version 2.10.

Driver notes:

• This library version has been tested with Catalyst
  Pro driver version 15.201 on Firepro W9100.

This is a patch update release to v2.10.1.
It has the following:

• Fixes for accuracy/stability issues noted in large size real FFTs
• Directed tests to target testing of specific list of problem sizes

This is a patch update release to v2.10.0.
It has the following:

• Performance fixes for slowdown observed vs v2.8. Specifically,
  2D transforms for certain sizes showed significant slowdown.
  This issue has been fixed in this release. The fix also makes the
  performance in general better than v2.8.
• Code reorganization in transposes and performance improvements.
  There is a modest performance improvement with inplace transforms
  for large power-of-2 sizes.
• Several minor enhancements and bug fixes

This clFFT release tagged as v2.10.0 is part of AMD Compute Libraries (ACL) 1.0 GA.

clFFT - Release Notes - version 2.10.0:

• Post-callback feature that enables custom post-processing
  of output data directly by the library with user callback function
• Support for in-place transposes for power-of-2 sizes enables really
  large 1D transforms as well as supporting no additional memory
  allocation, by library, for a range of problem sizes

This clFFT release tagged as v2.8 is part of AMD Compute Libraries (ACL) 1.0 beta 2.

clFFT - Release Notes - version 2.8.0:

• Support for power-of-7 size transforms
• Pre-callback feature that enables custom pre-processing
  of input data directly by the library with user callback function
• Support for 1D large size transforms with no extra memory allocation
  requirement for certain sizes

Driver notes:

• This library version has been tested with Catalyst
  Pro driver version 14.502 on Firepro W9100.

clFFT - Release Notes - version 2.6.1:

This is a patch update release to v2.6. It has the following.

• Fix for client application crash with Linux 14.502 driver
• Performance fixes for real forward and backward transform when enabling ECC
• Performance fixes for complex transforms over a small range in power-of-2 sizes
• Improvement in API timing in client
• Several minor enhancements and bug fixes

Driver notes:

• This library version has been tested with Catalyst
  Pro driver version 14.502 on Firepro W9100.

This clFFT release tagged as v2.6 is part of AMD Compute Libraries (ACL) 1.0 beta 1.

clFFT - Release Notes - version 2.6.0:

• Significant uplift of 1D real forward transform
  performance for power-of-2 sizes
• Significant uplift of large 2D/3D real transform
  performance for power-of-2 sizes

Known issues:

• Size limit on 1D real transforms remain at 2^24
  for single precision and 2^22 for double
  precision

Driver notes:

• This library version has been tested with Catalyst
  Pro driver version 14.301 on Firepro W9100.

clFFT - Release Notes - version 2.4.0:

New features of this release:

• Significant uplift of 1D complex transform performance
• 1D large size limit relaxation for complex transforms
• 2D/3D size limit relaxation on real and complex transforms
• Binary caching feature
• Several minor fixes and improvements