Author: post
Date: 2012-01-31 17:23:22 +0100 (Tue, 31 Jan 2012)
New Revision: 4114
Modified:
trunk/plugins/dcp/dcp.c
Log:
Enable SSE4 DCP code after test, and DISable AVX code - seems like
miscompilation.
Modified: trunk/plugins/dcp/dcp.c
===================================================================
--- trunk/plugins/dcp/dcp.c 2012-01-31 16:22:10 UTC (rev 4113)
+++ trunk/plugins/dcp/dcp.c 2012-01-31 16:23:22 UTC (rev 4114)
@@ -471,7 +471,7 @@
pre_cache_tables(t->dcp);
if (tmp->pixelsize == 4 && (rs_detect_cpu_features() &
RS_CPU_FLAG_SSE2) && !t->dcp->read_out_curve)
{
- if ((rs_detect_cpu_features() & RS_CPU_FLAG_AVX) &&
render_AVX(t))
+ if (FALSE && (rs_detect_cpu_features() & RS_CPU_FLAG_AVX) &&
render_AVX(t))
{
/* AVX routine renders 4 pixels in parallel, but any
remaining must be */
/* calculated using C routines */
@@ -481,6 +481,16 @@
render(t);
}
}
+ else if ((rs_detect_cpu_features() & RS_CPU_FLAG_SSE4_1) &&
render_SSE4(t))
+ {
+ /* SSE4 routine renders 4 pixels in parallel, but any
remaining must be */
+ /* calculated using C routines */
+ if (tmp->w & 3)
+ {
+ t->start_x = tmp->w - (tmp->w & 3);
+ render(t);
+ }
+ }
else if (render_SSE2(t))
{
/* SSE2 routine renders 4 pixels in parallel, but any
remaining must be */
_______________________________________________
Rawstudio-commit mailing list
[email protected]
http://rawstudio.org/cgi-bin/mailman/listinfo/rawstudio-commit
