v9.3.2

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