Function arm_nn_mat_mult_nt_t_f16

Function Documentation

arm_cmsis_nn_status arm_nn_mat_mult_nt_t_f16(const float16_t *lhs, const float16_t *rhs, const float16_t *bias, float16_t *dst, int32_t lhs_rows, int32_t rhs_rows, int32_t rhs_cols, int32_t row_address_offset, float16_t activation_min, float16_t activation_max)

Matrix multiply with non-transposed lhs and transposed rhs rows (float32).

Note

Accumulation width per leg. MVE legs accumulate in float16: per-k on the gather path (rhs_cols below the contiguous-K threshold), lane-partial sums then one reduction on the contiguous-K path, and float16 lanes then one reduction on the remainder rows. Error grows with rhs_cols there; the K=1024 tester cases carry measured tolerance overrides for this reason. Float16 accumulation is the chosen MVE trade-off (throughput over the last ulps); see AmbiqAI/ns-cmsis-nn#417. The scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE) accumulates bias and every product in float32 and rounds to float16 once before the clamp (AmbiqAI/ns-cmsis-nn#449, #457).

Parameters:

  • lhs – [in] Left-hand matrix stored row-major.

  • rhs – [in] Right-hand matrix stored row-major, one row per output channel.

  • bias – [in] Optional bias vector.

  • dst – [out] Output matrix.

  • lhs_rows – [in] Number of rows in lhs.

  • rhs_rows – [in] Number of rows in rhs.

  • rhs_cols – [in] Number of columns in rhs.

  • row_address_offset – [in] Output row stride, expressed in elements.

  • activation_min – [in] Lower clamp bound.

  • activation_max – [in] Upper clamp bound.

Returns:

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.