vm_gemm_batch_ctx struct
#include <src/features/gemm.c>
Context for a batched GEMM parallel-for range.
Public variables
-
vm_
float_ t*const * C - Output panels.
- int ldc
- Leading dimension of each C.
-
const vm_
float_ t*const * A - Input A panels (or NULL).
- int lda
- Leading dimension of each A.
-
const vm_
float_ t*const * B - Per-batch B panels (or NULL).
-
const vm_
float_ t* B_shared - Shared B when
sharedis set. - int ldb
- Leading dimension of B.
- int M
- int N
- int K
- Panel sizes.
-
vm_
float_ t alpha -
vm_
float_ t beta - GEMM scales.
- bool transA
- bool transB
- Transpose flags.
-
vm_
layout_ t layout - Memory layout.
- int shared
- Non-zero when B is shared.