[FFmpeg-devel] [PATCH] avcodec: loongson3 optimized h264chroma put and avg with mmi

周晓勇 zhouxiaoyong at loongson.cn
Mon Jun 1 05:29:31 CEST 2015


>From 671813a26715d9d2c1b0ef9da9e4da458b709e50 Mon Sep 17 00:00:00 2001
From: ZhouXiaoyong <zhouxiaoyong at loongson.cn>
Date: Fri, 17 Apr 2015 17:09:06 +0800
Subject: [PATCH] avcodec: loongson3 optimized h264chroma put and avg with mmi


Signed-off-by: ZhouXiaoyong <zhouxiaoyong at loongson.cn>
---
 libavcodec/h264chroma.c                |   2 +
 libavcodec/h264chroma.h                |   1 +
 libavcodec/mips/Makefile               |   2 +
 libavcodec/mips/h264chroma_init_mips.c |  42 +++
 libavcodec/mips/h264chroma_mips.h      |  35 ++
 libavcodec/mips/h264chroma_mmi.c       | 582 +++++++++++++++++++++++++++++++++
 6 files changed, 664 insertions(+)


diff --git a/libavcodec/h264chroma.c b/libavcodec/h264chroma.c
index 5b3e13b..c2f1f30 100644
--- a/libavcodec/h264chroma.c
+++ b/libavcodec/h264chroma.c
@@ -54,4 +54,6 @@ av_cold void ff_h264chroma_init(H264ChromaContext *c, int bit_depth)
         ff_h264chroma_init_ppc(c, bit_depth);
     if (ARCH_X86)
         ff_h264chroma_init_x86(c, bit_depth);
+    if (ARCH_MIPS)
+        ff_h264chroma_init_mips(c, bit_depth);
 }
diff --git a/libavcodec/h264chroma.h b/libavcodec/h264chroma.h
index d4b8a0e..e0f45ad 100644
--- a/libavcodec/h264chroma.h
+++ b/libavcodec/h264chroma.h
@@ -34,5 +34,6 @@ void ff_h264chroma_init_aarch64(H264ChromaContext *c, int bit_depth);
 void ff_h264chroma_init_arm(H264ChromaContext *c, int bit_depth);
 void ff_h264chroma_init_ppc(H264ChromaContext *c, int bit_depth);
 void ff_h264chroma_init_x86(H264ChromaContext *c, int bit_depth);
+void ff_h264chroma_init_mips(H264ChromaContext *c, int bit_depth);
 
 #endif /* AVCODEC_H264CHROMA_H */
diff --git a/libavcodec/mips/Makefile b/libavcodec/mips/Makefile
index 9786175..a460f10 100644
--- a/libavcodec/mips/Makefile
+++ b/libavcodec/mips/Makefile
@@ -20,7 +20,9 @@ MIPSDSPR1-OBJS-$(CONFIG_AAC_ENCODER)      += mips/aaccoder_mips.o
 MIPSFPU-OBJS-$(CONFIG_AAC_ENCODER)        += mips/iirfilter_mips.o
 OBJS-$(CONFIG_HEVC_DECODER)               += mips/hevcdsp_init_mips.o
 OBJS-$(CONFIG_H264DSP)                    += mips/h264dsp_init_mips.o
+OBJS-$(CONFIG_H264CHROMA)                 += mips/h264chroma_init_mips.o
 MSA-OBJS-$(CONFIG_HEVC_DECODER)           += mips/hevcdsp_msa.o            \
                                              mips/hevc_mc_uni_msa.o
 MSA-OBJS-$(CONFIG_H264DSP)                += mips/h264dsp_msa.o
 LOONGSON3-OBJS-$(CONFIG_H264DSP)          += mips/h264dsp_mmi.o
+LOONGSON3-OBJS-$(CONFIG_H264CHROMA)       += mips/h264chroma_mmi.o
diff --git a/libavcodec/mips/h264chroma_init_mips.c b/libavcodec/mips/h264chroma_init_mips.c
new file mode 100644
index 0000000..588cc8a
--- /dev/null
+++ b/libavcodec/mips/h264chroma_init_mips.c
@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2015 Zhou Xiaoyong <zhouxiaoyong at loongson.cn>
+ *
+ * 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 "h264chroma_mips.h"
+
+#if HAVE_LOONGSON3
+static av_cold void h264chroma_init_mmi(H264ChromaContext *c, int bit_depth)
+{
+    int high_bit_depth = bit_depth > 8;
+
+    if (!high_bit_depth) {
+        c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_mmi;
+        c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_mmi;
+        c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_mmi;
+        c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmi;
+    }
+}
+#endif /* HAVE_LOONGSON3 */
+
+av_cold void ff_h264chroma_init_mips(H264ChromaContext *c, int bit_depth
+{
+#if HAVE_LOONGSON3
+    h264chroma_init_mmi(c, bit_depth);
+#endif /* HAVE_LOONGSON3 */
+}
diff --git a/libavcodec/mips/h264chroma_mips.h b/libavcodec/mips/h264chroma_mips.h
new file mode 100644
index 0000000..314e8a3
--- /dev/null
+++ b/libavcodec/mips/h264chroma_mips.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2015 Zhou Xiaoyong <zhouxiaoyong at loongson.cn>
+ *
+ * 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
+ */
+
+#ifndef H264_CHROMA_MIPS_H
+#define H264_CHROMA_MIPS_H
+
+#include "libavcodec/h264.h"
+
+void ff_put_h264_chroma_mc8_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y);
+void ff_avg_h264_chroma_mc8_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y);
+void ff_put_h264_chroma_mc4_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y);
+void ff_avg_h264_chroma_mc4_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y);
+
+#endif /* H264_CHROMA_MIPS_H */
diff --git a/libavcodec/mips/h264chroma_mmi.c b/libavcodec/mips/h264chroma_mmi.c
new file mode 100644
index 0000000..ef29476
--- /dev/null
+++ b/libavcodec/mips/h264chroma_mmi.c
@@ -0,0 +1,582 @@
+/*
+ * Loongson SIMD optimized h264chroma
+ *
+ * Copyright (c) 2015 Loongson Technology Corporation Limited
+ * Copyright (c) 2015 Zhou Xiaoyong <zhouxiaoyong at loongson.cn>
+ *                    Zhang Shuangshuang <zhangshuangshuang at ict.ac.cn>
+ *
+ * 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 "h264chroma_mips.h"
+
+void ff_put_h264_chroma_mc8_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B = x * (8 - y);
+    const int C = (8 - x) * y;
+    const int D = x * y;
+    const int E = B + C;
+    int i;
+
+    av_assert2(x<8 && y<8 && x>=0 && y>=0);
+
+    if (D) {
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "ldl $4, %6                 \r\n"
+                "ldr $4, %5                 \r\n"
+                "ldl $5, %8                 \r\n"
+                "ldr $5, %7                 \r\n"
+                "daddiu $6, $0, 32          \r\n"
+                "mtc1 %9, $f6               \r\n"
+                "mtc1 %10, $f8              \r\n"
+                "mtc1 %11, $f10             \r\n"
+                "mtc1 %12, $f12             \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f16              \r\n"
+                "mtc1 $5, $f18              \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f10, $f10, $f20    \r\n"
+                "pshufh $f12, $f12, $f20    \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "punpcklbh $f16, $f16, $f20 \r\n"
+                "punpcklbh $f18, $f18, $f20 \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "dsrl32 $2, $2, 0           \r\n"
+                "dsrl32 $3, $3, 0           \r\n"
+                "dsrl32 $4, $4, 0           \r\n"
+                "dsrl32 $5, $5, 0           \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "pmullh $f16, $f10, $f16    \r\n"
+                "pmullh $f18, $f12, $f18    \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f4, $f4, $f16       \r\n"
+                "paddh $f2, $f2, $f18       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f24, $f2, $f22      \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f16              \r\n"
+                "mtc1 $5, $f18              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "punpcklbh $f16, $f16, $f20 \r\n"
+                "punpcklbh $f18, $f18, $f20 \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "pmullh $f16, $f10, $f16    \r\n"
+                "pmullh $f18, $f12, $f18    \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f4, $f4, $f16       \r\n"
+                "paddh $f2, $f2, $f18       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f24, $f2    \r\n"
+                "sdc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*src),"m"(*(src+7)),"m"(*(src+1)),"m"(*(src+8)),
+                  "m"(*(src+stride)),"m"(*(src+stride+7)),
+                  "m"(*(src+stride+1)),"m"(*(src+stride+8)),
+                  "r"(A),"r"(B),"r"(C),"r"(D)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    } else if (E) {
+        const int step = C ? stride : 1;
+
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "daddiu $6, $0, 32          \r\n"
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "mtc1 %5, $f6               \r\n"
+                "mtc1 %6, $f8               \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "dsrl32 $2, $2, 0           \r\n"
+                "dsrl32 $3, $3, 0           \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f24, $f2, $f22      \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f24, $f2    \r\n"
+                "sdc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),
+                  "m"(*(src+step)),"m"(*(src+step+7)),
+                  "r"(A),"r"(E)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    } else {
+        for (i = 0; i < h; i++) {
+            __asm__ volatile (
+                "daddiu $6, $0, 32          \r\n"
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "mtc1 %3, $f6               \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "dsrl32 $2, $2, 0           \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "psrah $f24, $f2, $f22      \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f24, $f2    \r\n"
+                "sdc1 $f2, %0               \r\n"
+                :"=m"(*dst)
+                :"m"(*src),"m"(*(src+7)),"r"(A)
+                :"$2"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    }
+}
+
+void ff_avg_h264_chroma_mc8_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B = x * (8 - y);
+    const int C = (8 - x) * y;
+    const int D = x * y;
+    const int E = B + C;
+    int i;
+
+    av_assert2(x<8 && y<8 && x>=0 && y>=0);
+
+    if (D) {
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "ldl $4, %6                 \r\n"
+                "ldr $4, %5                 \r\n"
+                "ldl $5, %8                 \r\n"
+                "ldr $5, %7                 \r\n"
+                "daddiu $6, $0, 32          \r\n"
+                "mtc1 %9, $f6               \r\n"
+                "mtc1 %10, $f8              \r\n"
+                "mtc1 %11, $f10             \r\n"
+                "mtc1 %12, $f12             \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f16              \r\n"
+                "mtc1 $5, $f18              \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f10, $f10, $f20    \r\n"
+                "pshufh $f12, $f12, $f20    \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "punpcklbh $f16, $f16, $f20 \r\n"
+                "punpcklbh $f18, $f18, $f20 \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "dsrl32 $2, $2, 0           \r\n"
+                "dsrl32 $3, $3, 0           \r\n"
+                "dsrl32 $4, $4, 0           \r\n"
+                "dsrl32 $5, $5, 0           \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "pmullh $f16, $f10, $f16    \r\n"
+                "pmullh $f18, $f12, $f18    \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f4, $f4, $f16       \r\n"
+                "paddh $f2, $f2, $f18       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f24, $f2, $f22      \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f16              \r\n"
+                "mtc1 $5, $f18              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "punpcklbh $f16, $f16, $f20 \r\n"
+                "punpcklbh $f18, $f18, $f20 \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "pmullh $f16, $f10, $f16    \r\n"
+                "pmullh $f18, $f12, $f18    \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f4, $f4, $f16       \r\n"
+                "paddh $f2, $f2, $f18       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f24, $f2    \r\n"
+                "ldc1 $f4, %0               \r\n"
+                "pavgb $f2, $f2, $f4        \r\n"
+                "sdc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),"m"(*(src+1)),"m"(*(src+8)),
+                  "m"(*(src+stride)),"m"(*(src+stride+7)),
+                  "m"(*(src+stride+1)),"m"(*(src+stride+8)),
+                  "r"(A),"r"(B),"r"(C),"r"(D)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    } else {
+        const int step = C ? stride : 1;
+
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "daddiu $6, $0, 32          \r\n"
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "mtc1 %5, $f6               \r\n"
+                "mtc1 %6, $f8               \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "dsrl32 $2, $2, 0           \r\n"
+                "dsrl32 $3, $3, 0           \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f24, $f2, $f22      \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f24, $f2    \r\n"
+                "ldc1 $f4, %0               \r\n"
+                "pavgb $f2, $f2, $f4        \r\n"
+                "sdc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),
+                  "m"(*(src+step)),"m"(*(src+step+7)),"r"(A),"r"(E)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    }
+}
+
+void ff_put_h264_chroma_mc4_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B = x * (8 - y);
+    const int C = (8 - x) *  y;
+    const int D = x *  y;
+    const int E = B + C;
+    int i;
+
+    av_assert2(x<8 && y<8 && x>=0 && y>=0);
+
+    if (D) {
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "ldl $4, %6                 \r\n"
+                "ldr $4, %5                 \r\n"
+                "ldl $5, %8                 \r\n"
+                "ldr $5, %7                 \r\n"
+                "daddiu $6, $0, 32          \r\n"
+                "mtc1 %9, $f6               \r\n"
+                "mtc1 %10, $f8              \r\n"
+                "mtc1 %11, $f10             \r\n"
+                "mtc1 %12, $f12             \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f16              \r\n"
+                "mtc1 $5, $f18              \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f10, $f10, $f20    \r\n"
+                "pshufh $f12, $f12, $f20    \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "punpcklbh $f16, $f16, $f20 \r\n"
+                "punpcklbh $f18, $f18, $f20 \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "pmullh $f16, $f10, $f16    \r\n"
+                "pmullh $f18, $f12, $f18    \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f4, $f4, $f16       \r\n"
+                "paddh $f2, $f2, $f18       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f2, $f2     \r\n"
+                "swc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),"m"(*(src+1)),"m"(*(src+8)),
+                  "m"(*(src+stride)),"m"(*(src+stride+7)),
+                  "m"(*(src+stride+1)),"m"(*(src+stride+8)),
+                  "r"(A),"r"(B),"r"(C),"r"(D)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    } else if (E) {
+        const int step = C ? stride : 1;
+
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "daddiu $4, $0, 32          \r\n"
+                "mtc1 %5, $f6               \r\n"
+                "mtc1 %6, $f8               \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f10              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f10, $f10, $f20    \r\n"
+                "daddiu $4, $0, 6           \r\n"
+                "mtc1 $4, $f22              \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "paddh $f2, $f2, $f10       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f2, $f20    \r\n"
+                "swc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),"m"(*(src+step)),
+                  "m"(*(src+step+7)),"r"(A),"r"(E)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    } else {
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "lwl $2, %2                 \r\n"
+                "lwr $2, %1                 \r\n"
+                "sw $2, %0                  \r\n"
+                : "=m"(*dst)
+                : "m"(*src),"m"(*(src+3))
+                : "$2"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    }
+}
+
+void ff_avg_h264_chroma_mc4_mmi(uint8_t *dst, uint8_t *src, int stride,
+        int h, int x, int y)
+{
+    const int A = (8 - x) *(8 - y);
+    const int B = x * (8 - y);
+    const int C = (8 - x) * y;
+    const int D = x * y;
+    int i;
+
+    av_assert2(x<8 && y<8 && x>=0 && y>=0);
+
+    if (D) {
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "ldl $4, %6                 \r\n"
+                "ldr $4, %5                 \r\n"
+                "ldl $5, %8                 \r\n"
+                "ldr $5, %7                 \r\n"
+                "daddiu $6, $0, 32          \r\n"
+                "mtc1 %9, $f6               \r\n"
+                "mtc1 %10, $f8              \r\n"
+                "mtc1 %11, $f10             \r\n"
+                "mtc1 %12, $f12             \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f16              \r\n"
+                "mtc1 $5, $f18              \r\n"
+                "mtc1 $6, $f14              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f10, $f10, $f20    \r\n"
+                "pshufh $f12, $f12, $f20    \r\n"
+                "pshufh $f14, $f14, $f20    \r\n"
+                "punpcklbh $f16, $f16, $f20 \r\n"
+                "punpcklbh $f18, $f18, $f20 \r\n"
+                "daddiu $6, $0, 6           \r\n"
+                "mtc1 $6, $f22              \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "pmullh $f16, $f10, $f16    \r\n"
+                "pmullh $f18, $f12, $f18    \r\n"
+                "paddh $f2, $f2, $f14       \r\n"
+                "paddh $f4, $f4, $f16       \r\n"
+                "paddh $f2, $f2, $f18       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f2, $f2     \r\n"
+                "lwc1 $f4, %0               \r\n"
+                "pavgb $f2, $f2, $f4        \r\n"
+                "swc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),"m"(*(src+1)),"m"(*(src+8)),
+                  "m"(*(src+stride)),"m"(*(src+stride+7)),
+                  "m"(*(src+stride+1)),"m"(*(src+stride+8)),
+                  "r"(A),"r"(B),"r"(C),"r"(D)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    } else {
+        const int E = B + C;
+        const int step = C ? stride : 1;
+
+        for (i=0; i<h; i++) {
+            __asm__ volatile (
+                "ldl $2, %2                 \r\n"
+                "ldr $2, %1                 \r\n"
+                "ldl $3, %4                 \r\n"
+                "ldr $3, %3                 \r\n"
+                "daddiu $4, $0, 32          \r\n"
+                "mtc1 %5, $f6               \r\n"
+                "mtc1 %6, $f8               \r\n"
+                "mtc1 $0, $f20              \r\n"
+                "mtc1 $2, $f2               \r\n"
+                "mtc1 $3, $f4               \r\n"
+                "mtc1 $4, $f10              \r\n"
+                "punpcklbh $f2, $f2, $f20   \r\n"
+                "punpcklbh $f4, $f4, $f20   \r\n"
+                "pshufh $f6, $f6, $f20      \r\n"
+                "pshufh $f8, $f8, $f20      \r\n"
+                "pshufh $f10, $f10, $f20    \r\n"
+                "daddiu $4, $0, 6           \r\n"
+                "mtc1 $4, $f22              \r\n"
+                "pmullh $f2, $f2, $f6       \r\n"
+                "pmullh $f4, $f4, $f8       \r\n"
+                "paddh $f2, $f2, $f10       \r\n"
+                "paddh $f2, $f2, $f4        \r\n"
+                "psrah $f2, $f2, $f22       \r\n"
+                "packushb $f2, $f2, $f20    \r\n"
+                "lwc1 $f4, %0               \r\n"
+                "pavgb $f2, $f2, $f4        \r\n"
+                "swc1 $f2, %0               \r\n"
+                : "=m"(*dst)
+                : "m"(*(src)),"m"(*(src+7)),"m"(*(src+step)),
+                  "m"(*(src+step+7)),"r"(A),"r"(E)
+                : "$2","$3","$4","$5","$6"
+            );
+
+            dst += stride;
+            src += stride;
+        }
+    }
+}
-- 
2.1.4



More information about the ffmpeg-devel mailing list