shithub: opus

Info  •  Files  •  Log  •  Branches

Neon: Make gcc actually generate VMLA instructions for sparse mul

Otherwise it was splitting the mla into a mul and an add

--- a/dnn/vec_neon.h

+++ b/dnn/vec_neon.h

@@ -187,13 +187,13 @@

 	for (j=0;j<cols;j++)

 	{

-	    float xj= x[*idx++];

+	    float32x4_t xj= vld1q_dup_f32(&x[*idx++]);

 	    float32x4_t wvec;

-	    wvec = vld1q_f32(&w[0]); y0_3 = vmlaq_n_f32(y0_3, wvec, xj);

-	    wvec = vld1q_f32(&w[4]); y4_7 = vmlaq_n_f32(y4_7, wvec, xj);

-	    wvec = vld1q_f32(&w[8]); y8_11 = vmlaq_n_f32(y8_11, wvec, xj);

-	    wvec = vld1q_f32(&w[12]); y12_15 = vmlaq_n_f32(y12_15, wvec, xj);

+	    wvec = vld1q_f32(&w[0]); y0_3 = vmlaq_f32(y0_3, wvec, xj);

+	    wvec = vld1q_f32(&w[4]); y4_7 = vmlaq_f32(y4_7, wvec, xj);

+	    wvec = vld1q_f32(&w[8]); y8_11 = vmlaq_f32(y8_11, wvec, xj);

+	    wvec = vld1q_f32(&w[12]); y12_15 = vmlaq_f32(y12_15, wvec, xj);

 	    w += 16;

 	}

-- 

⑨