ARM: NEON optimised h264_idct_add

Originally committed as revision 16150 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/Makefile

@@ -451,6 +451,7 @@ OBJS-$(HAVE_IWMMXT)                    += armv4l/dsputil_iwmmxt.o       \
 OBJS-$(HAVE_NEON)                      += armv4l/dsputil_neon.o         \
                                           armv4l/dsputil_neon_s.o       \
                                           armv4l/h264dsp_neon.o         \
+                                          armv4l/h264idct_neon.o        \
                                           armv4l/simple_idct_neon.o     \
 
 OBJS-$(ARCH_BFIN)                      += bfin/dsputil_bfin.o           \

libavcodec/armv4l/dsputil_neon.c

@@ -92,6 +92,8 @@ void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
 void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
                                        int beta, int8_t *tc0);
 
+void ff_h264_idct_add_neon(uint8_t *dst, DCTELEM *block, int stride);
+
 void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
 {
     c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
@@ -160,4 +162,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
     c->h264_h_loop_filter_luma = ff_h264_h_loop_filter_luma_neon;
     c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
     c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
+
+    c->h264_idct_add = ff_h264_idct_add_neon;
 }

libavcodec/armv4l/h264idct_neon.S

@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "asm.S"
+
+        .fpu neon
+
+        .text
+
+function ff_h264_idct_add_neon, export=1
+        mov             r3,  #(1<<5)
+        vmov.i16        d16, #0
+        vmov.16         d16[0],   r3
+        vld1.64         {d0-d3},  [r1,:128]
+        vadd.i16        d0,  d0,  d16
+
+        vswp            d1,  d2
+        vadd.i16        d4,  d0,  d1
+        vshr.s16        q8,  q1,  #1
+        vsub.i16        d5,  d0,  d1
+        vadd.i16        d6,  d2,  d17
+        vsub.i16        d7,  d16, d3
+        vadd.i16        q0,  q2,  q3
+        vsub.i16        q1,  q2,  q3
+
+        vtrn.16         d0,  d1
+        vtrn.16         d3,  d2
+        vtrn.32         d0,  d3
+        vtrn.32         d1,  d2
+
+        vadd.i16        d4,  d0,  d3
+        vld1.32         {d18[0]}, [r0,:32], r2
+        vswp            d1,  d3
+        vshr.s16        q8,  q1,  #1
+        vld1.32         {d19[1]}, [r0,:32], r2
+        vsub.i16        d5,  d0,  d1
+        vld1.32         {d18[1]}, [r0,:32], r2
+        vadd.i16        d6,  d16, d3
+        vld1.32         {d19[0]}, [r0,:32], r2
+        vsub.i16        d7,  d2,  d17
+        sub             r0,  r0,  r2, lsl #2
+        vadd.i16        q0,  q2,  q3
+        vsub.i16        q1,  q2,  q3
+
+        vshr.s16        q0,  q0,  #6
+        vshr.s16        q1,  q1,  #6
+
+        vaddw.u8        q0,  q0,  d18
+        vaddw.u8        q1,  q1,  d19
+
+        vqmovun.s16     d0,  q0
+        vqmovun.s16     d1,  q1
+
+        vst1.32         {d0[0]},  [r0,:32], r2
+        vst1.32         {d1[1]},  [r0,:32], r2
+        vst1.32         {d0[1]},  [r0,:32], r2
+        vst1.32         {d1[0]},  [r0,:32], r2
+
+        bx              lr
+        .endfunc