From 710c17a4f16a630405dd674a1e8c0f2938adf0ff Mon Sep 17 00:00:00 2001 From: villekf Date: Wed, 12 May 2021 15:36:30 +0300 Subject: [PATCH] Reformat release notes --- doc/release_notes.html | 134 ++++++++++++++++++++++++++++++++++------- release_notes.md | 22 +++---- 2 files changed, 122 insertions(+), 34 deletions(-) diff --git a/doc/release_notes.html b/doc/release_notes.html index ca2b79a..289042c 100644 --- a/doc/release_notes.html +++ b/doc/release_notes.html @@ -1,13 +1,101 @@ -Release notes

Release notes

OMEGA v1.1.1

New features
Bug fixes and enhancements

OMEGA v1.1.0

New features
Bug fixes and enhancements

OMEGA v1.0.0

New features
Bug fixes and enhancements
+.S4 { margin-left: 56px; line-height: 21px; min-height: 0px; text-align: left; white-space: pre-wrap; }

Release notes

OMEGA v1.2.0

New features

Bug fixes and enhancements

OMEGA v1.1.1

New features

Bug fixes and enhancements

OMEGA v1.1.0

New features

Bug fixes and enhancements

OMEGA v1.0.0

New features

Bug fixes and enhancements


\ No newline at end of file +--> +
\ No newline at end of file diff --git a/release_notes.md b/release_notes.md index d4f6757..1a79bbc 100644 --- a/release_notes.md +++ b/release_notes.md @@ -20,22 +20,13 @@ - This resulted in naming change for all non-MAP algorithms, regularization parameters and some relaxation parameters - Backwards compatibility is, however, maintained on all main-files except (possibly) custom prior - This has little to no effect on the normal user - -- CUDA backend should work more reliably now - - CUDA may not work on Octave - -- Fixed the use of corrections when using the forward/backward projection class - -- Improved Siddon should be about 10-20% faster with Nvidia GPUs (affects both OpenCL and CUDA) - -- Orthogonal and volume-based ray tracers should be a little faster - Added relative difference prior for implementations 1, 2 and 4 - - The current method can be memory intensive with large neighborhoods or image sizes + - The current implementation can be memory intensive with large neighborhoods or image sizes - Added support for 32-bit integer atomics (OpenCL only) - Should give about 20-30% faster computations at the expense of accuracy - - Can cause integer overflow when the number of counts is high + - Can cause integer overflow when the number of counts is high (in the range of thousands per voxel) - Numerical accuracy WILL be negatively affected - Should be used only when speed is of utmost importance @@ -46,6 +37,12 @@ ### Bug fixes and enhancements +- Fixed the use of corrections when using the forward/backward projection class + +- Improved Siddon should be about 10-20% faster with Nvidia GPUs (affects both OpenCL and CUDA) + +- Orthogonal and volume-based ray tracers should be a little faster + - Fixed raw data sampling increase when using sampling larger than 2 - Fixed custom normalization data load when the data was not in the current working directory @@ -100,6 +97,9 @@ - Fixed MetaImage load when the (raw) filename has uppercase letters - Fixed InterFile load when the (raw) file is in the working directory and is not i33 or img file + +- CUDA backend should work more reliably now + - CUDA may not work on Octave ## OMEGA v1.1.1