diff i386/dsputil_mmx.c @ 1984:ef919e9ef73e libavcodec

separate out put_signed_pixels_clamped() into its own function and implement an optimized MMX version of the function

--- a/i386/dsputil_mmx.c	Mon Apr 26 21:46:01 2004 +0000
+++ b/i386/dsputil_mmx.c	Tue Apr 27 03:58:06 2004 +0000
@@ -22,6 +22,7 @@
 
 #include "../dsputil.h"
 #include "../simple_idct.h"
+#include "mmx.h"
 
 //#undef NDEBUG
 //#include <assert.h>
@@ -293,6 +294,23 @@
 	    :"memory");
 }
 
+void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
+{
+    int i;
+    unsigned char __align8 vector128[8] =
+      { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 };
+
+    movq_m2r(*vector128, mm1);
+    for (i = 0; i < 8; i++) {
+        movq_m2r(*(block), mm0);
+        packsswb_m2r(*(block + 4), mm0);
+        block += 8;
+        paddb_r2r(mm1, mm0);
+        movq_r2m(mm0, *pixels);
+        pixels += line_size;
+    }
+}
+
 void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
 {
     const DCTELEM *p;
@@ -2160,6 +2178,7 @@
         c->diff_pixels = diff_pixels_mmx;
 #endif //CONFIG_ENCODERS
         c->put_pixels_clamped = put_pixels_clamped_mmx;
+        c->put_signed_pixels_clamped = put_signed_pixels_clamped_mmx;
         c->add_pixels_clamped = add_pixels_clamped_mmx;
         c->clear_blocks = clear_blocks_mmx;
 #ifdef CONFIG_ENCODERS