avcodec: [loongarch] Optimize pred16x16_plane with LASX.

./ffmpeg -i ../1_h264_1080p_30fps_3Mbps.mp4 -f rawvideo -y /dev/null -an
before:295
after :296

Change-Id: I281bc739f708d45f91fc3860150944c0b8a6a5ba
Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn>
Reviewed-by: guxiwei <guxiwei-hf@loongson.cn>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

4 files changed, 204 insertions, 0 deletions

diff --git a/libavcodec/loongarch/Makefile b/libavcodec/loongarch/Makefile
index 1e1fe3fd48..30799e4e48 100644
--- a/libavcodec/loongarch/Makefile
+++ b/libavcodec/loongarch/Makefile
@@ -1,8 +1,10 @@
 OBJS-$(CONFIG_H264CHROMA)             += loongarch/h264chroma_init_loongarch.o
 OBJS-$(CONFIG_H264QPEL)               += loongarch/h264qpel_init_loongarch.o
 OBJS-$(CONFIG_H264DSP)                += loongarch/h264dsp_init_loongarch.o
+OBJS-$(CONFIG_H264PRED)               += loongarch/h264_intrapred_init_loongarch.o
 LASX-OBJS-$(CONFIG_H264CHROMA)        += loongarch/h264chroma_lasx.o
 LASX-OBJS-$(CONFIG_H264QPEL)          += loongarch/h264qpel_lasx.o
 LASX-OBJS-$(CONFIG_H264DSP)           += loongarch/h264dsp_lasx.o \
                                          loongarch/h264idct_lasx.o \
                                          loongarch/h264_deblock_lasx.o
+LASX-OBJS-$(CONFIG_H264PRED)          += loongarch/h264_intrapred_lasx.o
diff --git a/libavcodec/loongarch/h264_intrapred_init_loongarch.c b/libavcodec/loongarch/h264_intrapred_init_loongarch.c
new file mode 100644
index 0000000000..12620bd842
--- /dev/null
+++ b/libavcodec/loongarch/h264_intrapred_init_loongarch.c
@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ * Contributed by Hao Chen <chenhao@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 "libavutil/loongarch/cpu.h"
+#include "libavcodec/h264pred.h"
+#include "h264_intrapred_lasx.h"
+
+av_cold void ff_h264_pred_init_loongarch(H264PredContext *h, int codec_id,
+                                         const int bit_depth,
+                                         const int chroma_format_idc)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (bit_depth == 8) {
+        if (have_lasx(cpu_flags)) {
+            if (chroma_format_idc <= 1) {
+            }
+            if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {
+            } else {
+                if (chroma_format_idc <= 1) {
+                }
+                if (codec_id == AV_CODEC_ID_SVQ3) {
+                    h->pred16x16[PLANE_PRED8x8] = ff_h264_pred16x16_plane_svq3_8_lasx;
+                } else if (codec_id == AV_CODEC_ID_RV40) {
+                    h->pred16x16[PLANE_PRED8x8] = ff_h264_pred16x16_plane_rv40_8_lasx;
+                } else {
+                    h->pred16x16[PLANE_PRED8x8] = ff_h264_pred16x16_plane_h264_8_lasx;
+                }
+            }
+        }
+    }
+}
diff --git a/libavcodec/loongarch/h264_intrapred_lasx.c b/libavcodec/loongarch/h264_intrapred_lasx.c
new file mode 100644
index 0000000000..c38cd611b8
--- /dev/null
+++ b/libavcodec/loongarch/h264_intrapred_lasx.c
@@ -0,0 +1,121 @@
+/*
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ * Contributed by Hao Chen <chenhao@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 "libavutil/loongarch/loongson_intrinsics.h"
+#include "h264_intrapred_lasx.h"
+
+#define PRED16X16_PLANE                                                        \
+    ptrdiff_t stride_1, stride_2, stride_3, stride_4, stride_5, stride_6;      \
+    ptrdiff_t stride_8, stride_15;                                             \
+    int32_t res0, res1, res2, res3, cnt;                                       \
+    uint8_t *src0, *src1;                                                      \
+    __m256i reg0, reg1, reg2, reg3, reg4;                                      \
+    __m256i tmp0, tmp1, tmp2, tmp3;                                            \
+    __m256i shuff = {0x0B040A0509060807, 0x0F000E010D020C03, 0, 0};            \
+    __m256i mult = {0x0004000300020001, 0x0008000700060005, 0, 0};             \
+    __m256i int_mult1 = {0x0000000100000000, 0x0000000300000002,               \
+                         0x0000000500000004, 0x0000000700000006};              \
+                                                                               \
+    stride_1 = -stride;                                                        \
+    stride_2 = stride << 1;                                                    \
+    stride_3 = stride_2 + stride;                                              \
+    stride_4 = stride_2 << 1;                                                  \
+    stride_5 = stride_4 + stride;                                              \
+    stride_6 = stride_3 << 1;                                                  \
+    stride_8 = stride_4 << 1;                                                  \
+    stride_15 = (stride_8 << 1) - stride;                                      \
+    src0 = src - 1;                                                            \
+    src1 = src0 + stride_8;                                                    \
+                                                                               \
+    reg0 = __lasx_xvldx(src0, -stride);                                        \
+    reg1 = __lasx_xvldx(src, (8 - stride));                                    \
+    reg0 = __lasx_xvilvl_d(reg1, reg0);                                        \
+    reg0 = __lasx_xvshuf_b(reg0, reg0, shuff);                                 \
+    reg0 = __lasx_xvhsubw_hu_bu(reg0, reg0);                                   \
+    reg0 = __lasx_xvmul_h(reg0, mult);                                         \
+    res1 = (src1[0] - src0[stride_6]) +                                        \
+        2 * (src1[stride] - src0[stride_5]) +                                  \
+        3 * (src1[stride_2] - src0[stride_4]) +                                \
+        4 * (src1[stride_3] - src0[stride_3]) +                                \
+        5 * (src1[stride_4] - src0[stride_2]) +                                \
+        6 * (src1[stride_5] - src0[stride]) +                                  \
+        7 * (src1[stride_6] - src0[0]) +                                       \
+        8 * (src0[stride_15] - src0[stride_1]);                                \
+    reg0 = __lasx_xvhaddw_w_h(reg0, reg0);                                     \
+    reg0 = __lasx_xvhaddw_d_w(reg0, reg0);                                     \
+    reg0 = __lasx_xvhaddw_q_d(reg0, reg0);                                     \
+    res0 = __lasx_xvpickve2gr_w(reg0, 0);                                      \
+
+#define PRED16X16_PLANE_END                                                    \
+    res2 = (src0[stride_15] + src[15 - stride] + 1) << 4;                      \
+    res3 = 7 * (res0 + res1);                                                  \
+    res2 -= res3;                                                              \
+    reg0 = __lasx_xvreplgr2vr_w(res0);                                         \
+    reg1 = __lasx_xvreplgr2vr_w(res1);                                         \
+    reg2 = __lasx_xvreplgr2vr_w(res2);                                         \
+    reg3 = __lasx_xvmul_w(reg0, int_mult1);                                    \
+    reg4 = __lasx_xvslli_w(reg0, 3);                                           \
+    reg4 = __lasx_xvadd_w(reg4, reg3);                                         \
+    for (cnt = 8; cnt--;) {                                                    \
+        tmp0 = __lasx_xvadd_w(reg2, reg3);                                     \
+        tmp1 = __lasx_xvadd_w(reg2, reg4);                                     \
+        tmp0 = __lasx_xvssrani_hu_w(tmp1, tmp0, 5);                            \
+        tmp0 = __lasx_xvpermi_d(tmp0, 0xD8);                                   \
+        reg2 = __lasx_xvadd_w(reg2, reg1);                                     \
+        tmp2 = __lasx_xvadd_w(reg2, reg3);                                     \
+        tmp3 = __lasx_xvadd_w(reg2, reg4);                                     \
+        tmp1 = __lasx_xvssrani_hu_w(tmp3, tmp2, 5);                            \
+        tmp1 = __lasx_xvpermi_d(tmp1, 0xD8);                                   \
+        tmp0 = __lasx_xvssrani_bu_h(tmp1, tmp0, 0);                            \
+        reg2 = __lasx_xvadd_w(reg2, reg1);                                     \
+        __lasx_xvstelm_d(tmp0, src, 0, 0);                                     \
+        __lasx_xvstelm_d(tmp0, src, 8, 2);                                     \
+        src += stride;                                                         \
+        __lasx_xvstelm_d(tmp0, src, 0, 1);                                     \
+        __lasx_xvstelm_d(tmp0, src, 8, 3);                                     \
+        src += stride;                                                         \
+    }
+
+
+void ff_h264_pred16x16_plane_h264_8_lasx(uint8_t *src, ptrdiff_t stride)
+{
+    PRED16X16_PLANE
+    res0 = (5 * res0 + 32) >> 6;
+    res1 = (5 * res1 + 32) >> 6;
+    PRED16X16_PLANE_END
+}
+
+void ff_h264_pred16x16_plane_rv40_8_lasx(uint8_t *src, ptrdiff_t stride)
+{
+    PRED16X16_PLANE
+    res0 = (res0 + (res0 >> 2)) >> 4;
+    res1 = (res1 + (res1 >> 2)) >> 4;
+    PRED16X16_PLANE_END
+}
+
+void ff_h264_pred16x16_plane_svq3_8_lasx(uint8_t *src, ptrdiff_t stride)
+{
+    PRED16X16_PLANE
+    cnt  = (5 * (res0/4)) / 16;
+    res0 = (5 * (res1/4)) / 16;
+    res1 = cnt;
+    PRED16X16_PLANE_END
+}
diff --git a/libavcodec/loongarch/h264_intrapred_lasx.h b/libavcodec/loongarch/h264_intrapred_lasx.h
new file mode 100644
index 0000000000..0c2653300c
--- /dev/null
+++ b/libavcodec/loongarch/h264_intrapred_lasx.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ * Contributed by Hao Chen <chenhao@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 AVCODEC_LOONGARCH_H264_INTRAPRED_LASX_H
+#define AVCODEC_LOONGARCH_H264_INTRAPRED_LASX_H
+
+#include "libavcodec/avcodec.h"
+
+void ff_h264_pred16x16_plane_h264_8_lasx(uint8_t *src, ptrdiff_t stride);
+void ff_h264_pred16x16_plane_rv40_8_lasx(uint8_t *src, ptrdiff_t stride);
+void ff_h264_pred16x16_plane_svq3_8_lasx(uint8_t *src, ptrdiff_t stride);
+
+#endif  // #ifndef AVCODEC_LOONGARCH_H264_INTRAPRED_LASX_H