// blas implementation
//
-// uses _sdot on the padded input to compute each output elements at once
+// uses gemv on the padded input to compute each output elements at once
//
// TODO
-// - switch to sgemv to factorise over activations->shape[j]
// - avoid copy when padding_start == 0
// - optimize copying using tensor helpers
aubio_tensor_t *activations)
{
uint_t i, j;
- smpl_t bias, acc;
uint_t sdot_size = c->kernel->shape[0] * c->kernel->shape[1];
- uint_t input_stride = c->stride_shape * c->padded_input->shape[1]
+ uint_t input_stride = c->stride_shape * c->padded_input->shape[1];
AUBIO_ASSERT(c && input_tensor && activations);
if (aubio_conv1d_check_output_shape(c, input_tensor, activations))