From bc99faef1bf2e1a98e99dcf6cfba2ea58ae0a56e Mon Sep 17 00:00:00 2001 From: wernsaar Date: Sun, 31 Aug 2014 14:33:15 +0200 Subject: [PATCH] optimized sgemv_t_4.c for uneven sizes --- kernel/x86_64/sgemv_t_4.c | 28 ++++++++++++---------------- 1 file changed, 12 insertions(+), 16 deletions(-) diff --git a/kernel/x86_64/sgemv_t_4.c b/kernel/x86_64/sgemv_t_4.c index cefbaccd4..76187b57d 100644 --- a/kernel/x86_64/sgemv_t_4.c +++ b/kernel/x86_64/sgemv_t_4.c @@ -273,28 +273,24 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO } if ( m3 == 0 ) return(0); - xbuffer = buffer; + x_ptr = x; + a_ptr = a; for ( i=0; i< m3; i++ ) { - xbuffer[i] = *x_ptr; - x_ptr += inc_x; - } - j=0; - a_ptr = a; - y_ptr = y; - while ( j < n) - { - FLOAT temp = 0.0; - for( i = 0; i < m3; i++ ) + FLOAT xtemp = *x_ptr * alpha; + FLOAT *aj = a_ptr; + y_ptr = y; + for ( j=0; j