kernel/

The BLAS layer over raw pointers.

Raw-pointer kernels over complex scalars: mixed real/complex products <kernel/complex.hpp>

matvec_real_complex – matvec_transpose_into_complex

Raw-pointer kernels: dense Level-2/3 BLAS, triangular solves, banded ops <kernel/dense.hpp>

combine_columns – gbmv – gemm – gemm_config – gemm_transpose_left – gemm_workspace – ger – matvec – matvec_transpose – project_columns – swap_rows – syrk_lower – transpose – trsm_lower_inplace – trsm_lower_transpose_inplace – trsm_lower_transpose_right_inplace – trsm_unit_lower_inplace – trsm_unit_lower_transpose_inplace – trsm_upper_inplace – trsm_upper_transpose_inplace – trsv_lower – trsv_lower_inplace – trsv_transpose_lower – trsv_upper – trsv_upper_inplace

Raw-pointer kernels: CSR SpMV and SpMM <kernel/sparse.hpp>

spmm – spmv – spmv_axpy

Raw-pointer kernels: BLAS-1 vector ops, Givens rotations and fused reductions <kernel/vector.hpp>

add – argmax_abs – axpby – axpbyz – axpy – axpy_norm_sq – axpy_strided – clamp – contract::alias_safe – contract::ordered – contract::throughput – copy – copy_strided – dot – dot2 – dot2_result – dot_norm_result – dot_norm_sq – fill – hadamard_div – hadamard_mul – inv – l1_norm – linear_combination_norm_sq – linf_norm – norm – norm_sq – norm_sq_strided – rot – rotg – scale – scale_copy_strided – sum – swap – swap_strided