ARMv6 optimised pix_norm1

Originally committed as revision 21704 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/arm/dsputil_armv6.S

@@ -564,3 +564,33 @@ function ff_sse16_armv6, export=1
 
         pop             {r4-r9, pc}
 .endfunc
+
+function ff_pix_norm1_armv6, export=1
+        push            {r4-r6, lr}
+        mov             r12, #16
+        mov             lr,  #0
+1:
+        ldm             r0,  {r2-r5}
+        uxtb16          r6,  r2
+        uxtb16          r2,  r2,  ror #8
+        smlad           lr,  r6,  r6,  lr
+        uxtb16          r6,  r3
+        smlad           lr,  r2,  r2,  lr
+        uxtb16          r3,  r3,  ror #8
+        smlad           lr,  r6,  r6,  lr
+        uxtb16          r6,  r4
+        smlad           lr,  r3,  r3,  lr
+        uxtb16          r4,  r4,  ror #8
+        smlad           lr,  r6,  r6,  lr
+        uxtb16          r6,  r5
+        smlad           lr,  r4,  r4,  lr
+        uxtb16          r5,  r5,  ror #8
+        smlad           lr,  r6,  r6,  lr
+        subs            r12, r12, #1
+        add             r0,  r0,  r1
+        smlad           lr,  r5,  r5,  lr
+        bgt             1b
+
+        mov             r0,  lr
+        pop             {r4-r6, pc}
+.endfunc

libavcodec/arm/dsputil_init_armv6.c

@@ -67,6 +67,8 @@ int ff_pix_abs8_armv6(void *s, uint8_t *blk1, uint8_t *blk2,
 int ff_sse16_armv6(void *s, uint8_t *blk1, uint8_t *blk2,
                    int line_size, int h);
 
+int ff_pix_norm1_armv6(uint8_t *pix, int line_size);
+
 void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx)
 {
     if (!avctx->lowres && (avctx->idct_algo == FF_IDCT_AUTO ||
@@ -112,4 +114,6 @@ void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx)
     c->sad[1] = ff_pix_abs8_armv6;
 
     c->sse[0] = ff_sse16_armv6;
+
+    c->pix_norm1 = ff_pix_norm1_armv6;
 }