What's Changed
- Fix buffer overflow: psgsrfs_ABXglobal reduced a double into float *berr by @xiaoyeli in #241
- Fix overflow: psgssvx_ABglobal broadcasts equilibration float data with MPI_DOUBLE by @xiaoyeli in #242
- Fix HIP build: missing GPU-resident solve sources and raw CUDA calls by @Heinrich-BR in #239
- Fix 'd_x' undeclared: GPU solve scratch vars require GPU_ACC && LAPACK but are used under GPU_ACC alone by @xiaoyeli in #243
- Fix CPU-path bugs in C++ 3D factorization: hard-coded MPI_DOUBLE, NULL index deref by @xiaoyeli in #245
- Bump version to 9.3.2 by @jhale in #236
Full Changelog: v9.3.1...v9.3.2