aarch64: NEON asm for missing x264_zigzag_* functions

[x264] / common / common.h

diff --git a/common/common.h b/common/common.h
index 6e63170f486c2f0adab17fb060cf194a9f2eb77f..76ae8a3c79dbfd3bb266f6547995dad2eb0c62ce 100644 (file)
--- a/common/common.h
+++ b/common/common.h
@@ -109,7 +109,7 @@ do {\
 
 #define NALU_OVERHEAD 5 // startcode + NAL type costs 5 bytes per frame
 #define FILLER_OVERHEAD (NALU_OVERHEAD+1)
-#define SEI_OVERHEAD (NALU_OVERHEAD - (h->param.b_annexb && !h->param.b_avcintra_compat && (h->out.i_nal-1)))
+#define SEI_OVERHEAD (NALU_OVERHEAD - (h->param.b_annexb && !h->param.i_avcintra_class && (h->out.i_nal-1)))
 
 /****************************************************************************
  * Includes
@@ -316,8 +316,8 @@ static ALWAYS_INLINE int x264_predictor_difference( int16_t (*mvc)[2], intptr_t
 
 static ALWAYS_INLINE uint16_t x264_cabac_mvd_sum( uint8_t *mvdleft, uint8_t *mvdtop )
 {
-    int amvd0 = abs(mvdleft[0]) + abs(mvdtop[0]);
-    int amvd1 = abs(mvdleft[1]) + abs(mvdtop[1]);
+    int amvd0 = mvdleft[0] + mvdtop[0];
+    int amvd1 = mvdleft[1] + mvdtop[1];
     amvd0 = (amvd0 > 2) + (amvd0 > 32);
     amvd1 = (amvd1 > 2) + (amvd1 > 32);
     return amvd0 + (amvd1<<8);
@@ -552,15 +552,15 @@ struct x264_t
     int             (*dequant4_mf[4])[16];   /* [4][6][16] */
     int             (*dequant8_mf[4])[64];   /* [4][6][64] */
     /* quantization matrix for trellis, [cqm][qp][coef] */
-    int             (*unquant4_mf[4])[16];   /* [4][52][16] */
-    int             (*unquant8_mf[4])[64];   /* [4][52][64] */
+    int             (*unquant4_mf[4])[16];   /* [4][QP_MAX_SPEC+1][16] */
+    int             (*unquant8_mf[4])[64];   /* [4][QP_MAX_SPEC+1][64] */
     /* quantization matrix for deadzone */
-    udctcoef        (*quant4_mf[4])[16];     /* [4][52][16] */
-    udctcoef        (*quant8_mf[4])[64];     /* [4][52][64] */
-    udctcoef        (*quant4_bias[4])[16];   /* [4][52][16] */
-    udctcoef        (*quant8_bias[4])[64];   /* [4][52][64] */
-    udctcoef        (*quant4_bias0[4])[16];  /* [4][52][16] */
-    udctcoef        (*quant8_bias0[4])[64];  /* [4][52][64] */
+    udctcoef        (*quant4_mf[4])[16];     /* [4][QP_MAX_SPEC+1][16] */
+    udctcoef        (*quant8_mf[4])[64];     /* [4][QP_MAX_SPEC+1][64] */
+    udctcoef        (*quant4_bias[4])[16];   /* [4][QP_MAX_SPEC+1][16] */
+    udctcoef        (*quant8_bias[4])[64];   /* [4][QP_MAX_SPEC+1][64] */
+    udctcoef        (*quant4_bias0[4])[16];  /* [4][QP_MAX_SPEC+1][16] */
+    udctcoef        (*quant8_bias0[4])[64];  /* [4][QP_MAX_SPEC+1][64] */
     udctcoef        (*nr_offset_emergency)[4][64];
 
     /* mv/ref cost arrays. */