- make-blas-backend name gemm-f64 gemm-f32 gemm-strided-f64 gemm-strided-f32 gemv-f64 gemv-f32 dot-f64 dot-f32 axpy-f64 axpy-f32 conv-fwd-im2col-f32 conv-bwd-data-im2col-f32 conv-bwd-weights-im2col-f32 conv-fwd-nhwc-im2col-f32 conv-bwd-data-nhwc-im2col-f32 conv-bwd-weights-nhwc-im2col-f32 #!optional gemm-offset-f64 gemm-offset-f32 dot-offset-f64 dot-offset-f32procedure
Make a backend record from its kernels; each is a procedure or #f. The four optional kernels take matrices and vectors that start at an offset into their vectors:
gemm-offset: (M N K alpha A off-A lda transa B off-B ldb transb beta C off-C) dot-offset: (N x off-x y off-y) -> number
With them, the cells of a batched product and the rows of a framed sum are read in place. microBLAS, the system-BLAS egg and the crunch egg supply them. A backend without them still works: each cell is copied into a scratch buffer first, which gives the same values more slowly.