What's Changed
- Deprecate COMBFLEN also for GPU by @samhatfield in #123
- Qualify all USE statements in GPU version by @samhatfield in #124
- Fix incorrect abort message in inv_trans.F90 by @marsdeno in #129
- Remove eckit and fckit from CI build testing by @marsdeno in #130
- Add missing USE statements by @samhatfield in #127
- [GPU] Add missing ASYNC statements by @samhatfield in #133
- Allow NOPT_MEMORY_TR to be controlled by driver command-line option. by @marsdeno in #136
- Printing bug in setup_trans0 by @marsdeno in #137
- Enable GPU feature without MPI requirement by @wdeconinck in #140
- [GPU] Fix potential NaNs when using cutlass by @lukasm91 in #132
- Clean up hicfft by @lukasm91 in #143
- No pinned flags in setup_trans by @lukasm91 in #142
- Fix corner cases with many ranks by @lukasm91 in #145
- Transform all GPU symbols to become unique by @wdeconinck in #141
- Fix calculation of zstepavg by @samhatfield in #147
- Tidy GEMMs by @samhatfield in #146
- Configurable warmup iterations for ectrans-benchmark by @wdeconinck in #149
- Fix typo in TRGTOL by @samhatfield in #154
- Add FFT implementation without graphs by @lukasm91 in #150
- CMake fixes for test linker and OpenACC search by @reuterbal in #156
- Update paths to HIP includes to support ROCm >= 6 by @samhatfield in #152
- Move more TPM modules and setup routines to common by @wdeconinck in #148
- Remove default values for TPM_GEN variables by @samhatfield in #159
- Pin buffers in benchmark by @lukasm91 in #158
- Make in-place FFT optional by @samhatfield in #155
- Add support for AMD GPUs by @samhatfield in #157
New Contributors
Full Changelog: 1.4.0...1.5.0