rv40: NEON optimised chroma MC

Signed-off-by: Mans Rullgard <mans@mansr.com>

5 files changed, 118 insertions, 5 deletions

diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index c125a59078..a948e6db3f 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -68,6 +68,8 @@ NEON-OBJS-$(CONFIG_RV30_DECODER)       += arm/rv34dsp_init_neon.o       \
 
 NEON-OBJS-$(CONFIG_RV40_DECODER)       += arm/rv34dsp_init_neon.o       \
                                           arm/rv34dsp_neon.o            \
+                                          arm/rv40dsp_init_neon.o       \
+                                          arm/h264cmc_neon.o            \
 
 NEON-OBJS-$(CONFIG_VP3_DECODER)        += arm/vp3dsp_neon.o
 
diff --git a/libavcodec/arm/h264cmc_neon.S b/libavcodec/arm/h264cmc_neon.S
index e10adaca10..a6feadd189 100644
--- a/libavcodec/arm/h264cmc_neon.S
+++ b/libavcodec/arm/h264cmc_neon.S
@@ -21,8 +21,8 @@
 #include "asm.S"
 
 /* chroma_mc8(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y) */
-.macro  h264_chroma_mc8 type
-function ff_\type\()_h264_chroma_mc8_neon, export=1
+.macro  h264_chroma_mc8 type, codec=h264
+function ff_\type\()_\codec\()_chroma_mc8_neon, export=1
         push            {r4-r7, lr}
         ldrd            r4,  [sp, #20]
   .ifc \type,avg
@@ -31,6 +31,15 @@ function ff_\type\()_h264_chroma_mc8_neon, export=1
         pld             [r1]
         pld             [r1, r2]
 
+  .ifc \codec,rv40
+        movrel          r6,  rv40bias
+        lsr             r7,  r5,  #1
+        add             r6,  r6,  r7,  lsl #3
+        lsr             r7,  r4,  #1
+        add             r6,  r6,  r7,  lsl #1
+        vld1.16         {d22[],d23[]}, [r6,:16]
+  .endif
+
 A       muls            r7,  r4,  r5
 T       mul             r7,  r4,  r5
 T       cmp             r7,  #0
@@ -67,10 +76,17 @@ T       cmp             r7,  #0
         vmlal.u8        q9,  d7,  d1
         vmlal.u8        q9,  d4,  d2
         vmlal.u8        q9,  d5,  d3
-        vrshrn.u16      d16, q8,  #6
         vld1.8          {d6, d7}, [r5], r4
         pld             [r1]
+  .ifc \codec,h264
+        vrshrn.u16      d16, q8,  #6
         vrshrn.u16      d17, q9,  #6
+  .else
+        vadd.u16        q8,  q8,  q11
+        vadd.u16        q9,  q9,  q11
+        vshrn.u16       d16, q8,  #6
+        vshrn.u16       d17, q9,  #6
+  .endif
   .ifc \type,avg
         vld1.8          {d20}, [lr,:64], r2
         vld1.8          {d21}, [lr,:64], r2
@@ -102,8 +118,15 @@ T       cmp             r7,  #0
         vmull.u8        q9,  d6,  d0
         vmlal.u8        q9,  d4,  d1
         vld1.8          {d6}, [r5], r4
+  .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
         vrshrn.u16      d17, q9,  #6
+  .else
+        vadd.u16        q8,  q8,  q11
+        vadd.u16        q9,  q9,  q11
+        vshrn.u16       d16, q8,  #6
+        vshrn.u16       d17, q9,  #6
+  .endif
   .ifc \type,avg
         vld1.8          {d20}, [lr,:64], r2
         vld1.8          {d21}, [lr,:64], r2
@@ -131,8 +154,15 @@ T       cmp             r7,  #0
         vmlal.u8        q9,  d7,  d1
         pld             [r1]
         vext.8          d5,  d4,  d5,  #1
+  .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
         vrshrn.u16      d17, q9,  #6
+  .else
+        vadd.u16        q8,  q8,  q11
+        vadd.u16        q9,  q9,  q11
+        vshrn.u16       d16, q8,  #6
+        vshrn.u16       d17, q9,  #6
+  .endif
   .ifc \type,avg
         vld1.8          {d20}, [lr,:64], r2
         vld1.8          {d21}, [lr,:64], r2
@@ -149,8 +179,8 @@ endfunc
 .endm
 
 /* chroma_mc4(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y) */
-.macro  h264_chroma_mc4 type
-function ff_\type\()_h264_chroma_mc4_neon, export=1
+.macro  h264_chroma_mc4 type, codec=h264
+function ff_\type\()_\codec\()_chroma_mc4_neon, export=1
         push            {r4-r7, lr}
         ldrd            r4,  [sp, #20]
   .ifc \type,avg
@@ -159,6 +189,15 @@ function ff_\type\()_h264_chroma_mc4_neon, export=1
         pld             [r1]
         pld             [r1, r2]
 
+  .ifc \codec,rv40
+        movrel          r6,  rv40bias
+        lsr             r7,  r5,  #1
+        add             r6,  r6,  r7,  lsl #3
+        lsr             r7,  r4,  #1
+        add             r6,  r6,  r7,  lsl #1
+        vld1.16         {d22[],d23[]}, [r6,:16]
+  .endif
+
 A       muls            r7,  r4,  r5
 T       mul             r7,  r4,  r5
 T       cmp             r7,  #0
@@ -199,7 +238,12 @@ T       cmp             r7,  #0
         vld1.8          {d6},     [r5], r4
         vadd.i16        d16, d16, d17
         vadd.i16        d17, d18, d19
+  .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
+  .else
+        vadd.u16        q8,  q8,  q11
+        vshrn.u16       d16, q8,  #6
+  .endif
         subs            r3,  r3,  #2
         pld             [r1]
   .ifc \type,avg
@@ -236,7 +280,12 @@ T       cmp             r7,  #0
         vld1.32         {d4[1]},  [r5], r4
         vadd.i16        d16, d16, d17
         vadd.i16        d17, d18, d19
+  .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
+  .else
+        vadd.u16        q8,  q8,  q11
+        vshrn.u16       d16, q8,  #6
+  .endif
   .ifc \type,avg
         vld1.32         {d20[0]}, [lr,:32], r2
         vld1.32         {d20[1]}, [lr,:32], r2
@@ -266,7 +315,12 @@ T       cmp             r7,  #0
         vadd.i16        d16, d16, d17
         vadd.i16        d17, d18, d19
         pld             [r1]
+  .ifc \codec,h264
         vrshrn.u16      d16, q8,  #6
+  .else
+        vadd.u16        q8,  q8,  q11
+        vshrn.u16       d16, q8,  #6
+  .endif
   .ifc \type,avg
         vld1.32         {d20[0]}, [lr,:32], r2
         vld1.32         {d20[1]}, [lr,:32], r2
@@ -352,9 +406,25 @@ function ff_\type\()_h264_chroma_mc2_neon, export=1
 endfunc
 .endm
 
+#if CONFIG_H264_DECODER
         h264_chroma_mc8 put
         h264_chroma_mc8 avg
         h264_chroma_mc4 put
         h264_chroma_mc4 avg
         h264_chroma_mc2 put
         h264_chroma_mc2 avg
+#endif
+
+#if CONFIG_RV40_DECODER
+const   rv40bias
+        .short           0, 16, 32, 16
+        .short          32, 28, 32, 28
+        .short           0, 32, 16, 32
+        .short          32, 28, 32, 28
+endconst
+
+        h264_chroma_mc8 put, rv40
+        h264_chroma_mc8 avg, rv40
+        h264_chroma_mc4 put, rv40
+        h264_chroma_mc4 avg, rv40
+#endif
diff --git a/libavcodec/arm/rv40dsp_init_neon.c b/libavcodec/arm/rv40dsp_init_neon.c
new file mode 100644
index 0000000000..aa4a88da1a
--- /dev/null
+++ b/libavcodec/arm/rv40dsp_init_neon.c
@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2011 Janne Grunau <janne-libav@jannau.net>
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavcodec/avcodec.h"
+#include "libavcodec/rv34dsp.h"
+
+void ff_put_rv40_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
+void ff_put_rv40_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
+
+void ff_avg_rv40_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
+void ff_avg_rv40_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
+
+void ff_rv40dsp_init_neon(RV34DSPContext *c, DSPContext* dsp)
+{
+    c->put_chroma_pixels_tab[0] = ff_put_rv40_chroma_mc8_neon;
+    c->put_chroma_pixels_tab[1] = ff_put_rv40_chroma_mc4_neon;
+    c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_neon;
+    c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_neon;
+}
diff --git a/libavcodec/rv34dsp.h b/libavcodec/rv34dsp.h
index a1636e6eb5..695af06970 100644
--- a/libavcodec/rv34dsp.h
+++ b/libavcodec/rv34dsp.h
@@ -59,5 +59,6 @@ void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp);
 void ff_rv34dsp_init_neon(RV34DSPContext *c, DSPContext *dsp);
 
 void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp);
+void ff_rv40dsp_init_neon(RV34DSPContext *c, DSPContext *dsp);
 
 #endif /* AVCODEC_RV34DSP_H */
diff --git a/libavcodec/rv40dsp.c b/libavcodec/rv40dsp.c
index f193b6050d..06bdf18c42 100644
--- a/libavcodec/rv40dsp.c
+++ b/libavcodec/rv40dsp.c
@@ -534,4 +534,6 @@ av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
 
     if (HAVE_MMX)
         ff_rv40dsp_init_x86(c, dsp);
+    if (HAVE_NEON)
+        ff_rv40dsp_init_neon(c, dsp);
 }