HIP: bump requirement to rocm 6.1 (llama/15296)

HIP: add cmake option to enable compiler output of kernel resource usage metrics (llama/15103)

HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (llama/14930)

HIP: disable rocwmma on gfx12 by default until rocm 7.0 (llama/14202)

CUDA/HIP: Share the same unified memory allocation logic. (llama/12934)

HIP: implement FlashAttention via rocWMMA for CDNA and RDNA3+ (llama/12032)

CUDA: app option to compile without FlashAttention (llama/12025)

HIP: force max threads per block to be 1024 (llama/11621)

CUDA: use mma PTX instructions for FlashAttention (llama/11583)

HIP: require at least HIP 5.5

Hip: disable VMM on hip as it seams that it dosent work in some configurations (llama/11420)

hip : Add hipGraph and VMM support to ROCM (llama/11362)

ggml : do not define GGML_USE_CUDA when building with GGML_BACKEND_DL (llama/11211)

ggml : add support for dynamic loading of backends (llama/10469)

CUDA: remove DMMV, consolidate F16 mult mat vec (llama/10318)

ggml : build backends as libraries (llama/10256)