ATS-GPU Change Log

This file contains all important changes to ATS-GPU

[24.1.0] - 2024-1-15

Fixed

• ATS-GPU-OCT and ATS-GPU-NUFFT API Reference documentation.
• Issue in ATS-GPU-OCT and -NUFFT CUDA sample programs where transfer length was incorrectly modified in NPT mode.

Added

• Support for buffer averaging in ATS_GPU_SetProcessingPipeline() and related C sample program.
• MATLAB sample programs in ATS-GPU-BASE for record averaging, boxcar averaging and buffer averaging with processing pipeline.
• Python sample program in ATS-GPU-BASE for buffer averaging.

Changed

• Return code errors thrown inside function files in MATLAB ATS-GPU-OCT and ATS-GPU-NUFFT code samples.
• ATS-GPU-Setup checks for valid software license.
• Logging capability.

[23.1.1] - 2023-10-3

Added

• Offline C sample program in ATS-GPU-NUFFT to post-process data with NUFFT algorithm.
• Link to documentation home in html user manuals.
• Option to output resampled data in ATS-GPU-NUFFT.

Changed

[23.1.0] - 2023-6-27

Fixed

• ATS_GPU_NUFFT_Setup() returning incorrect buffer size.
• Managed buffer scheme not working for ATS-GPU-OCT and -NUFFT.
• De-interleaving for very large buffers.
• Fix issue where ApiBufferOverflow errors are detected as ApiWaitTimeout errors.
• Fix 12-bit unpacking for record and channel interleaving.

Changed

• ATS_GPU_GetBuffer() returns immediately when buffer overflows occur.
• Outputs of ATS-GPU-OCT appear as output interleaved.
• Change ALAZAR_PACKING to C enum.
• Remove non-portable Visual Studio projects and rely on CMake to generate them.
• Load NVML library at runtime.
• Separate packages install directories under Linux.
• Place ATS-GPU dynamic link libraries in %WINDIR%/System32 for Windows users and in /usr/lib/ for Linux users.
• Change acquisition implementation to support very large buffers.

Added

• Support for ARM64.
• ATS_GPU_NUFFT_GetLinearizationFromKclock API.
• ATS-CUDA API to convert data to volts.
• ATS-CUDA API to multiply the records by a reference record.
• C sample program to perform boxcar averaging (ATS-GPU-BASE).
• Log failed loadings of libraries.
• Sample program for ATS-GPU-BASE without CUDA dependency.
• Support for boxcar averaging in ATS_GPU_SetProcessingPipeline() and related C sample program.
• Python sample program for boxcar averaging with processing pipeline.
• Add dll information in property page under Windows.

[4.2.2] - 2022-4-28

• No user-visible change

[4.2.1] - 2022-4-27

Added

• Put libcufft in linux installer.
• Raw data output.
• ATS_GPU_AllocBuffer() and ATS_GPU_GetBuffer() return pointer to cudaStream_t.
• Square root of amplitude output.
• RPM package.
• Places dlls in System32.
• Modular processing pipeline.
• Record averaging API.
• API to get GPU serial number.

Changed

• ATS_GPU_OCT_Setup and ATS_GPU_NUFFT_Setup initiate default window function.

Fixed

• Uninstallers for different packages do not overwrite each other.

[4.1.0] - 2020-3-4

Added

• API to query compute capability.
• API call to return libraries version number.
• ATS-CUDA-OCT requires an AlazarTech board in order to be used.
• ATS-CUDA-NUFFT low-level library to perform non-uniform FFTs.
• ATS-GPU-NUFFT high-level library to perform non-uniform FFTs using various programming languages.

Fixed

• ats-gpu-oct-sample.vi generates window function.
• MATLAB and Python code samples support complex window functions.
• Prevent ATS_GPU_OCT_GetBuffer() from doing a stream synchronize by default.

Changed

• Renamed input data types enum.

[4.0.1] - 2019-4-23

Fixed

• ATS_GPU_OCT result buffers take number of outputs into account
• AcqToGPU_OCT_CUDA sample code uses bytesPerDmaBuffer for buffer posting and allocation.
• NVCC options supports all GPU architectures.

Changed

• NVCC '-arch' flag 75.

[4.0.0] - 2019-1-22

Added

• Separate GPU processing stages, and make them available through new ATS-CUDA and ATS-CUDA-OCT interfaces.
• Support for unpacking and de-interleaving.
• Support GPUs with compute capabilities 7.5

Changed

• ATS-GPU library built with CUDA toolkit v10.0

Deprecated

• GPUs with compute capability <3.0 due to CUDA toolkit 10.0 update.

[3.7.0] - 2018-8-22

Changed

• Modification in ATS-GPU-BASE unpacking and de-interleaving kernel to correct for inaccurate number of blocks.

Added

• Lock-in amplification library.

[3.6.0] - 2018-4-23

Changed

• Modification in Acquisition thread to have the ability to pass non-infinite values of records per acquisition to `ATS_GPU_Setup()`

Added

• Support for deinterleaving and unpacking digitizer data on the GPU prior to delivering data buffers to users.

[3.5.0] - 2017-6-29

Changed

• Completely reworked ATS-GPU with a new API

Added

• OCT processing library

Deprecated

• ATS-GPU 3.0.0 and earlier interface

[3.0.0] - 2017-4-13

Added

• FFT processing library `ATS_GPU_FFT_CUDA`.
• Averaging option in `ATS_GPU_DLL_CUDA`.
• Example codes in C++, MATLAB, Python and LabVIEW.

Deprecated

• OpenCL-based ATS-GPU.

Changed

• Directory architecture since OpenCL is deprecated.

[2.0.0] - 2016-11-4

Added

• Support for CUDA with a sample inverting kernel and example program.
• Support for Linux with CMake for the CUDA library and example program.

Changed

• Directory architecture to improve clarity between OpenCL and CUDA implementations.

[1.1.0] - 2016-9-13

Added

• support for amplitude, phase, real and imaginary values output.

Changed

• kernel code changed for easier readability.