Merge remote-tracking branch 'qatar/master'

* qatar/master: (25 commits)
  rtpenc: Add support for G726 audio
  rtpdec: Interpret the different G726 names as bits_per_coded_sample
  rtpenc: Change rtp_send_samples to handle sample sizes other than even bytes
  rtpenc: Cast a rescaling parameter to int64_t
  h264: cap max has_b_frames at MAX_DELAYED_PIC_COUNT - 1.
  ARM: fix indentation in ff_dsputil_init_neon()
  ARM: NEON put/avg_pixels8/16 cosmetics
  ARM: add remaining NEON avg_pixels8/16 functions
  ARM: clean up NEON put/avg_pixels macros
  fate: split acodec-pcm into individual tests
  swscale: #include "libavutil/mathematics.h"
  pmpdec: don't use deprecated av_set_pts_info.
  rv34: align temporary block of "dct" coefs
  Add PlayStation Portable PMP format demuxer
  proto: Realign struct initializers
  proto: Use .priv_data_size to allocate the private context
  mmsh: Properly clean up if the second ffurl_alloc failed
  rtmp: Clean up properly if the handshake failed
  md5proto: Remove the get_file_handle function
  applehttpproto: Use the close function if the open function fails
  ...

Conflicts:
	libavcodec/vble.c
	libavformat/mmsh.c
	libavformat/pmpdec.c
	libavformat/udp.c
	tests/ref/acodec/pcm

Merged-by: Michael Niedermayer <michaelni@gmx.at>

2 files changed, 242 insertions, 152 deletions

diff --git a/libavcodec/arm/dsputil_init_neon.c b/libavcodec/arm/dsputil_init_neon.c
index ddc9d640f8..3f4dcdad5c 100644
--- a/libavcodec/arm/dsputil_init_neon.c
+++ b/libavcodec/arm/dsputil_init_neon.c
@@ -53,7 +53,19 @@ void ff_put_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
 void ff_put_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
 
 void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int);
 void ff_avg_pixels8_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
+void ff_avg_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
 
 void ff_add_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
 void ff_put_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
@@ -189,37 +201,51 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
     }
 
     if (!high_bit_depth) {
-    c->clear_block  = ff_clear_block_neon;
-    c->clear_blocks = ff_clear_blocks_neon;
-
-    c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
-    c->put_pixels_tab[0][1] = ff_put_pixels16_x2_neon;
-    c->put_pixels_tab[0][2] = ff_put_pixels16_y2_neon;
-    c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_neon;
-    c->put_pixels_tab[1][0] = ff_put_pixels8_neon;
-    c->put_pixels_tab[1][1] = ff_put_pixels8_x2_neon;
-    c->put_pixels_tab[1][2] = ff_put_pixels8_y2_neon;
-    c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_neon;
-
-    c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_neon;
-    c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_neon;
-    c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_neon;
-    c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_neon;
-    c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_neon;
-    c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_neon;
-    c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_neon;
-    c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_neon;
-
-    c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
-    c->avg_pixels_tab[1][0] = ff_avg_pixels8_neon;
+        c->clear_block  = ff_clear_block_neon;
+        c->clear_blocks = ff_clear_blocks_neon;
+
+        c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
+        c->put_pixels_tab[0][1] = ff_put_pixels16_x2_neon;
+        c->put_pixels_tab[0][2] = ff_put_pixels16_y2_neon;
+        c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_neon;
+        c->put_pixels_tab[1][0] = ff_put_pixels8_neon;
+        c->put_pixels_tab[1][1] = ff_put_pixels8_x2_neon;
+        c->put_pixels_tab[1][2] = ff_put_pixels8_y2_neon;
+        c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_neon;
+
+        c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_neon;
+        c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_neon;
+        c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_neon;
+
+        c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
+        c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_neon;
+        c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_neon;
+        c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_neon;
+        c->avg_pixels_tab[1][0] = ff_avg_pixels8_neon;
+        c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_neon;
+        c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_neon;
+        c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_neon;
+
+        c->avg_no_rnd_pixels_tab[0][0] = ff_avg_pixels16_neon;
+        c->avg_no_rnd_pixels_tab[0][1] = ff_avg_pixels16_x2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[0][2] = ff_avg_pixels16_y2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[0][3] = ff_avg_pixels16_xy2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[1][0] = ff_avg_pixels8_neon;
+        c->avg_no_rnd_pixels_tab[1][1] = ff_avg_pixels8_x2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[1][2] = ff_avg_pixels8_y2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[1][3] = ff_avg_pixels8_xy2_no_rnd_neon;
     }
 
     c->add_pixels_clamped = ff_add_pixels_clamped_neon;
     c->put_pixels_clamped = ff_put_pixels_clamped_neon;
     c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_neon;
 
-    if (CONFIG_H264_DECODER) {
-        if (!high_bit_depth) {
+    if (CONFIG_H264_DECODER && !high_bit_depth) {
         c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon;
         c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon;
         c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_neon;
@@ -295,7 +321,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
         c->avg_h264_qpel_pixels_tab[1][13] = ff_avg_h264_qpel8_mc13_neon;
         c->avg_h264_qpel_pixels_tab[1][14] = ff_avg_h264_qpel8_mc23_neon;
         c->avg_h264_qpel_pixels_tab[1][15] = ff_avg_h264_qpel8_mc33_neon;
-        }
     }
 
     if (CONFIG_VP3_DECODER) {
diff --git a/libavcodec/arm/dsputil_neon.S b/libavcodec/arm/dsputil_neon.S
index b765f23d02..c660cb0d4c 100644
--- a/libavcodec/arm/dsputil_neon.S
+++ b/libavcodec/arm/dsputil_neon.S
@@ -40,75 +40,89 @@ function ff_clear_blocks_neon, export=1
         bx              lr
 endfunc
 
-        .macro pixels16 avg=0
-.if \avg
-        mov             ip,  r0
-.endif
-1:      vld1.64         {d0, d1},  [r1], r2
-        vld1.64         {d2, d3},  [r1], r2
-        vld1.64         {d4, d5},  [r1], r2
+.macro  pixels16        rnd=1, avg=0
+  .if \avg
+        mov             r12, r0
+  .endif
+1:      vld1.64         {q0},     [r1], r2
+        vld1.64         {q1},     [r1], r2
+        vld1.64         {q2},     [r1], r2
         pld             [r1, r2, lsl #2]
-        vld1.64         {d6, d7},  [r1], r2
+        vld1.64         {q3},     [r1], r2
         pld             [r1]
         pld             [r1, r2]
         pld             [r1, r2, lsl #1]
-.if \avg
-        vld1.64         {d16,d17}, [ip,:128], r2
+  .if \avg
+        vld1.64         {q8},     [r12,:128], r2
         vrhadd.u8       q0,  q0,  q8
-        vld1.64         {d18,d19}, [ip,:128], r2
+        vld1.64         {q9},     [r12,:128], r2
         vrhadd.u8       q1,  q1,  q9
-        vld1.64         {d20,d21}, [ip,:128], r2
+        vld1.64         {q10},    [r12,:128], r2
         vrhadd.u8       q2,  q2,  q10
-        vld1.64         {d22,d23}, [ip,:128], r2
+        vld1.64         {q11},    [r12,:128], r2
         vrhadd.u8       q3,  q3,  q11
-.endif
+  .endif
         subs            r3,  r3,  #4
-        vst1.64         {d0, d1},  [r0,:128], r2
-        vst1.64         {d2, d3},  [r0,:128], r2
-        vst1.64         {d4, d5},  [r0,:128], r2
-        vst1.64         {d6, d7},  [r0,:128], r2
+        vst1.64         {q0},     [r0,:128], r2
+        vst1.64         {q1},     [r0,:128], r2
+        vst1.64         {q2},     [r0,:128], r2
+        vst1.64         {q3},     [r0,:128], r2
         bne             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels16_x2 vhadd=vrhadd.u8
-1:      vld1.64         {d0-d2},   [r1], r2
-        vld1.64         {d4-d6},   [r1], r2
+.macro  pixels16_x2     rnd=1, avg=0
+1:      vld1.64         {d0-d2},  [r1], r2
+        vld1.64         {d4-d6},  [r1], r2
         pld             [r1]
         pld             [r1, r2]
         subs            r3,  r3,  #2
         vext.8          q1,  q0,  q1,  #1
-        \vhadd          q0,  q0,  q1
+        avg             q0,  q0,  q1
         vext.8          q3,  q2,  q3,  #1
-        \vhadd          q2,  q2,  q3
-        vst1.64         {d0, d1},  [r0,:128], r2
-        vst1.64         {d4, d5},  [r0,:128], r2
+        avg             q2,  q2,  q3
+  .if \avg
+        vld1.8          {q1},     [r0,:128], r2
+        vld1.8          {q3},     [r0,:128]
+        vrhadd.u8       q0,  q0,  q1
+        vrhadd.u8       q2,  q2,  q3
+        sub             r0,  r0,  r2
+  .endif
+        vst1.64         {q0},     [r0,:128], r2
+        vst1.64         {q2},     [r0,:128], r2
         bne             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels16_y2 vhadd=vrhadd.u8
-        vld1.64         {d0, d1},  [r1], r2
-        vld1.64         {d2, d3},  [r1], r2
+.macro  pixels16_y2     rnd=1, avg=0
+        vld1.64         {q0},     [r1], r2
+        vld1.64         {q1},     [r1], r2
 1:      subs            r3,  r3,  #2
-        \vhadd          q2,  q0,  q1
-        vld1.64         {d0, d1},  [r1], r2
-        \vhadd          q3,  q0,  q1
-        vld1.64         {d2, d3},  [r1], r2
+        avg             q2,  q0,  q1
+        vld1.64         {q0},     [r1], r2
+        avg             q3,  q0,  q1
+        vld1.64         {q1},     [r1], r2
         pld             [r1]
         pld             [r1, r2]
-        vst1.64         {d4, d5},  [r0,:128], r2
-        vst1.64         {d6, d7},  [r0,:128], r2
+  .if \avg
+        vld1.8          {q8},     [r0,:128], r2
+        vld1.8          {q9},     [r0,:128]
+        vrhadd.u8       q2,  q2,  q8
+        vrhadd.u8       q3,  q3,  q9
+        sub             r0,  r0,  r2
+  .endif
+        vst1.64         {q2},     [r0,:128], r2
+        vst1.64         {q3},     [r0,:128], r2
         bne             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels16_xy2 vshrn=vrshrn.u16 no_rnd=0
-        vld1.64         {d0-d2},   [r1], r2
-        vld1.64         {d4-d6},   [r1], r2
-.if \no_rnd
+.macro  pixels16_xy2    rnd=1, avg=0
+        vld1.64         {d0-d2},  [r1], r2
+        vld1.64         {d4-d6},  [r1], r2
+  .ifeq \rnd
         vmov.i16        q13, #1
-.endif
+  .endif
         pld             [r1]
         pld             [r1, r2]
         vext.8          q1,  q0,  q1,  #1
@@ -118,109 +132,129 @@ endfunc
         vaddl.u8        q9,  d4,  d6
         vaddl.u8        q11, d5,  d7
 1:      subs            r3,  r3,  #2
-        vld1.64         {d0-d2},   [r1], r2
+        vld1.64         {d0-d2},  [r1], r2
         vadd.u16        q12, q8,  q9
         pld             [r1]
-.if \no_rnd
+  .ifeq \rnd
         vadd.u16        q12, q12, q13
-.endif
+  .endif
         vext.8          q15, q0,  q1,  #1
         vadd.u16        q1 , q10, q11
-        \vshrn          d28, q12, #2
-.if \no_rnd
+        shrn            d28, q12, #2
+  .ifeq \rnd
         vadd.u16        q1,  q1,  q13
-.endif
-        \vshrn          d29, q1,  #2
+  .endif
+        shrn            d29, q1,  #2
+  .if \avg
+        vld1.8          {q8},     [r0,:128]
+        vrhadd.u8       q14, q14, q8
+  .endif
         vaddl.u8        q8,  d0,  d30
-        vld1.64         {d2-d4},   [r1], r2
+        vld1.64         {d2-d4},  [r1], r2
         vaddl.u8        q10, d1,  d31
-        vst1.64         {d28,d29}, [r0,:128], r2
+        vst1.64         {q14},    [r0,:128], r2
         vadd.u16        q12, q8,  q9
         pld             [r1, r2]
-.if \no_rnd
+  .ifeq \rnd
         vadd.u16        q12, q12, q13
-.endif
+  .endif
         vext.8          q2,  q1,  q2,  #1
         vadd.u16        q0,  q10, q11
-        \vshrn          d30, q12, #2
-.if \no_rnd
+        shrn            d30, q12, #2
+  .ifeq \rnd
         vadd.u16        q0,  q0,  q13
-.endif
-        \vshrn          d31, q0,  #2
+  .endif
+        shrn            d31, q0,  #2
+  .if \avg
+        vld1.8          {q9},     [r0,:128]
+        vrhadd.u8       q15, q15, q9
+  .endif
         vaddl.u8        q9,  d2,  d4
         vaddl.u8        q11, d3,  d5
-        vst1.64         {d30,d31}, [r0,:128], r2
+        vst1.64         {q15},    [r0,:128], r2
         bgt             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels8 avg=0
-1:      vld1.64         {d0}, [r1], r2
-        vld1.64         {d1}, [r1], r2
-        vld1.64         {d2}, [r1], r2
+.macro  pixels8         rnd=1, avg=0
+1:      vld1.64         {d0},     [r1], r2
+        vld1.64         {d1},     [r1], r2
+        vld1.64         {d2},     [r1], r2
         pld             [r1, r2, lsl #2]
-        vld1.64         {d3}, [r1], r2
+        vld1.64         {d3},     [r1], r2
         pld             [r1]
         pld             [r1, r2]
         pld             [r1, r2, lsl #1]
-.if \avg
-        vld1.64         {d4}, [r0,:64], r2
+  .if \avg
+        vld1.64         {d4},     [r0,:64], r2
         vrhadd.u8       d0,  d0,  d4
-        vld1.64         {d5}, [r0,:64], r2
+        vld1.64         {d5},     [r0,:64], r2
         vrhadd.u8       d1,  d1,  d5
-        vld1.64         {d6}, [r0,:64], r2
+        vld1.64         {d6},     [r0,:64], r2
         vrhadd.u8       d2,  d2,  d6
-        vld1.64         {d7}, [r0,:64], r2
+        vld1.64         {d7},     [r0,:64], r2
         vrhadd.u8       d3,  d3,  d7
         sub             r0,  r0,  r2,  lsl #2
-.endif
+  .endif
         subs            r3,  r3,  #4
-        vst1.64         {d0}, [r0,:64], r2
-        vst1.64         {d1}, [r0,:64], r2
-        vst1.64         {d2}, [r0,:64], r2
-        vst1.64         {d3}, [r0,:64], r2
+        vst1.64         {d0},     [r0,:64], r2
+        vst1.64         {d1},     [r0,:64], r2
+        vst1.64         {d2},     [r0,:64], r2
+        vst1.64         {d3},     [r0,:64], r2
         bne             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels8_x2 vhadd=vrhadd.u8
-1:      vld1.64         {d0, d1},  [r1], r2
+.macro  pixels8_x2      rnd=1, avg=0
+1:      vld1.64         {q0},     [r1], r2
         vext.8          d1,  d0,  d1,  #1
-        vld1.64         {d2, d3},  [r1], r2
+        vld1.64         {q1},     [r1], r2
         vext.8          d3,  d2,  d3,  #1
         pld             [r1]
         pld             [r1, r2]
         subs            r3,  r3,  #2
         vswp            d1,  d2
-        \vhadd          q0,  q0,  q1
-        vst1.64         {d0},      [r0,:64], r2
-        vst1.64         {d1},      [r0,:64], r2
+        avg             q0,  q0,  q1
+  .if \avg
+        vld1.8          {d4},     [r0,:64], r2
+        vld1.8          {d5},     [r0,:64]
+        vrhadd.u8       q0,  q0,  q2
+        sub             r0,  r0,  r2
+  .endif
+        vst1.64         {d0},     [r0,:64], r2
+        vst1.64         {d1},     [r0,:64], r2
         bne             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels8_y2 vhadd=vrhadd.u8
-        vld1.64         {d0},      [r1], r2
-        vld1.64         {d1},      [r1], r2
+.macro  pixels8_y2      rnd=1, avg=0
+        vld1.64         {d0},     [r1], r2
+        vld1.64         {d1},     [r1], r2
 1:      subs            r3,  r3,  #2
-        \vhadd          d4,  d0,  d1
-        vld1.64         {d0},      [r1], r2
-        \vhadd          d5,  d0,  d1
-        vld1.64         {d1},      [r1], r2
+        avg             d4,  d0,  d1
+        vld1.64         {d0},     [r1], r2
+        avg             d5,  d0,  d1
+        vld1.64         {d1},     [r1], r2
         pld             [r1]
         pld             [r1, r2]
-        vst1.64         {d4},      [r0,:64], r2
-        vst1.64         {d5},      [r0,:64], r2
+  .if \avg
+        vld1.8          {d2},     [r0,:64], r2
+        vld1.8          {d3},     [r0,:64]
+        vrhadd.u8       q2,  q2,  q1
+        sub             r0,  r0,  r2
+  .endif
+        vst1.64         {d4},     [r0,:64], r2
+        vst1.64         {d5},     [r0,:64], r2
         bne             1b
         bx              lr
-        .endm
+.endm
 
-        .macro pixels8_xy2 vshrn=vrshrn.u16 no_rnd=0
-        vld1.64         {d0, d1},  [r1], r2
-        vld1.64         {d2, d3},  [r1], r2
-.if \no_rnd
+.macro  pixels8_xy2     rnd=1, avg=0
+        vld1.64         {q0},     [r1], r2
+        vld1.64         {q1},     [r1], r2
+  .ifeq \rnd
         vmov.i16        q11, #1
-.endif
+  .endif
         pld             [r1]
         pld             [r1, r2]
         vext.8          d4,  d0,  d1,  #1
@@ -228,70 +262,101 @@ endfunc
         vaddl.u8        q8,  d0,  d4
         vaddl.u8        q9,  d2,  d6
 1:      subs            r3,  r3,  #2
-        vld1.64         {d0, d1},  [r1], r2
+        vld1.64         {q0},     [r1], r2
         pld             [r1]
         vadd.u16        q10, q8,  q9
         vext.8          d4,  d0,  d1,  #1
-.if \no_rnd
+  .ifeq \rnd
         vadd.u16        q10, q10, q11
-.endif
+  .endif
         vaddl.u8        q8,  d0,  d4
-        \vshrn          d5,  q10, #2
-        vld1.64         {d2, d3},  [r1], r2
+        shrn            d5,  q10, #2
+        vld1.64         {q1},     [r1], r2
         vadd.u16        q10, q8,  q9
         pld             [r1, r2]
-.if \no_rnd
+  .if \avg
+        vld1.8          {d7},     [r0,:64]
+        vrhadd.u8       d5,  d5,  d7
+  .endif
+  .ifeq \rnd
         vadd.u16        q10, q10, q11
-.endif
-        vst1.64         {d5},      [r0,:64], r2
-        \vshrn          d7,  q10, #2
+  .endif
+        vst1.64         {d5},     [r0,:64], r2
+        shrn            d7,  q10, #2
+  .if \avg
+        vld1.8          {d5},     [r0,:64]
+        vrhadd.u8       d7,  d7,  d5
+  .endif
         vext.8          d6,  d2,  d3,  #1
         vaddl.u8        q9,  d2,  d6
-        vst1.64         {d7},      [r0,:64], r2
+        vst1.64         {d7},     [r0,:64], r2
         bgt             1b
         bx              lr
-        .endm
-
-        .macro pixfunc pfx name suf rnd_op args:vararg
+.endm
+
+.macro  pixfunc         pfx, name, suf, rnd=1, avg=0
+  .if \rnd
+    .macro avg  rd, rn, rm
+        vrhadd.u8       \rd, \rn, \rm
+    .endm
+    .macro shrn rd, rn, rm
+        vrshrn.u16      \rd, \rn, \rm
+    .endm
+  .else
+    .macro avg  rd, rn, rm
+        vhadd.u8        \rd, \rn, \rm
+    .endm
+    .macro shrn rd, rn, rm
+        vshrn.u16       \rd, \rn, \rm
+    .endm
+  .endif
 function ff_\pfx\name\suf\()_neon, export=1
-        \name \rnd_op \args
+        \name           \rnd, \avg
 endfunc
-        .endm
+        .purgem         avg
+        .purgem         shrn
+.endm
 
-        .macro pixfunc2 pfx name args:vararg
-        pixfunc \pfx \name
-        pixfunc \pfx \name \args
-        .endm
+.macro  pixfunc2        pfx, name, avg=0
+        pixfunc         \pfx, \name,          rnd=1, avg=\avg
+        pixfunc         \pfx, \name, _no_rnd, rnd=0, avg=\avg
+.endm
 
 function ff_put_h264_qpel16_mc00_neon, export=1
         mov             r3,  #16
 endfunc
 
-        pixfunc  put_ pixels16
-        pixfunc2 put_ pixels16_x2,  _no_rnd, vhadd.u8
-        pixfunc2 put_ pixels16_y2,  _no_rnd, vhadd.u8
-        pixfunc2 put_ pixels16_xy2, _no_rnd, vshrn.u16, 1
+        pixfunc         put_, pixels16,     avg=0
+        pixfunc2        put_, pixels16_x2,  avg=0
+        pixfunc2        put_, pixels16_y2,  avg=0
+        pixfunc2        put_, pixels16_xy2, avg=0
 
 function ff_avg_h264_qpel16_mc00_neon, export=1
         mov             r3,  #16
 endfunc
 
-        pixfunc  avg_ pixels16,, 1
+        pixfunc         avg_, pixels16,     avg=1
+        pixfunc2        avg_, pixels16_x2,  avg=1
+        pixfunc2        avg_, pixels16_y2,  avg=1
+        pixfunc2        avg_, pixels16_xy2, avg=1
 
 function ff_put_h264_qpel8_mc00_neon, export=1
         mov             r3,  #8
 endfunc
 
-        pixfunc  put_ pixels8
-        pixfunc2 put_ pixels8_x2,   _no_rnd, vhadd.u8
-        pixfunc2 put_ pixels8_y2,   _no_rnd, vhadd.u8
-        pixfunc2 put_ pixels8_xy2,  _no_rnd, vshrn.u16, 1
+        pixfunc         put_, pixels8,     avg=0
+        pixfunc2        put_, pixels8_x2,  avg=0
+        pixfunc2        put_, pixels8_y2,  avg=0
+        pixfunc2        put_, pixels8_xy2, avg=0
 
 function ff_avg_h264_qpel8_mc00_neon, export=1
         mov             r3,  #8
 endfunc
 
-        pixfunc  avg_ pixels8,, 1
+        pixfunc         avg_, pixels8,     avg=1
+        pixfunc2        avg_, pixels8_x2,  avg=1
+        pixfunc2        avg_, pixels8_y2,  avg=1
+        pixfunc2        avg_, pixels8_xy2, avg=1
 
 function ff_put_pixels_clamped_neon, export=1
         vld1.64         {d16-d19}, [r0,:128]!