[x264-devel] checkasm: Try different widths for ssd_nv12

[Martin Storsjö]

x264 | branch: master | Martin Storsjö <martin at martin.st> | Fri Aug 14 00:00:57 2015 +0300| [9cbdb635a4bd78e6767e735a062c0d9a5766b849] | committer: Henrik Gramner

checkasm: Try different widths for ssd_nv12

To test all codepaths in the aarch64 neon implementation, one at
the very least needs to test with width 8, 16, 24 and 32.

> http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=9cbdb635a4bd78e6767e735a062c0d9a5766b849
---

 tools/checkasm.c |   15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/tools/checkasm.c b/tools/checkasm.c
index 183cef5..01a97c9 100644
--- a/tools/checkasm.c
+++ b/tools/checkasm.c
@@ -716,13 +716,16 @@ static int check_pixel( int cpu_ref, int cpu_new )
         used_asm = 1;
         set_func_name( "ssd_nv12" );
         uint64_t res_u_c, res_v_c, res_u_a, res_v_a;
-        pixel_c.ssd_nv12_core(   pbuf1, 368, pbuf2, 368, 360, 8, &res_u_c, &res_v_c );
-        pixel_asm.ssd_nv12_core( pbuf1, 368, pbuf2, 368, 360, 8, &res_u_a, &res_v_a );
-        if( res_u_c != res_u_a || res_v_c != res_v_a )
+        for( int w = 8; w <= 360; w += 8 )
         {
-            ok = 0;
-            fprintf( stderr, "ssd_nv12: %"PRIu64",%"PRIu64" != %"PRIu64",%"PRIu64"\n",
-                     res_u_c, res_v_c, res_u_a, res_v_a );
+            pixel_c.ssd_nv12_core(   pbuf1, 368, pbuf2, 368, w, 8, &res_u_c, &res_v_c );
+            pixel_asm.ssd_nv12_core( pbuf1, 368, pbuf2, 368, w, 8, &res_u_a, &res_v_a );
+            if( res_u_c != res_u_a || res_v_c != res_v_a )
+            {
+                ok = 0;
+                fprintf( stderr, "ssd_nv12: %"PRIu64",%"PRIu64" != %"PRIu64",%"PRIu64"\n",
+                         res_u_c, res_v_c, res_u_a, res_v_a );
+            }
         }
         call_c( pixel_c.ssd_nv12_core,   pbuf1, (intptr_t)368, pbuf2, (intptr_t)368, 360, 8, &res_u_c, &res_v_c );
         call_a( pixel_asm.ssd_nv12_core, pbuf1, (intptr_t)368, pbuf2, (intptr_t)368, 360, 8, &res_u_a, &res_v_a );