lavc/aarch64: add hevc deblock chroma 8-12bit

Benched on Ampere Altra:

hevc_h_loop_filter_chroma8_c: 367.7
hevc_h_loop_filter_chroma8_neon: 31.0
hevc_h_loop_filter_chroma10_c: 396.7
hevc_h_loop_filter_chroma10_neon: 27.5
hevc_h_loop_filter_chroma12_c: 377.0
hevc_h_loop_filter_chroma12_neon: 31.7
hevc_v_loop_filter_chroma8_c: 369.0
hevc_v_loop_filter_chroma8_neon: 55.0
hevc_v_loop_filter_chroma10_c: 389.0
hevc_v_loop_filter_chroma10_neon: 54.0
hevc_v_loop_filter_chroma12_c: 389.5
hevc_v_loop_filter_chroma12_neon: 53.0

Signed-off-by: J. Dekker <jdek@itanimul.li>

3 files changed, 203 insertions, 1 deletions

diff --git a/libavcodec/aarch64/Makefile b/libavcodec/aarch64/Makefile
index 02fb51c3ab..216191640c 100644
--- a/libavcodec/aarch64/Makefile
+++ b/libavcodec/aarch64/Makefile
@@ -65,7 +65,8 @@ NEON-OBJS-$(CONFIG_VP9_DECODER)         += aarch64/vp9itxfm_16bpp_neon.o       \
                                            aarch64/vp9lpf_neon.o               \
                                            aarch64/vp9mc_16bpp_neon.o          \
                                            aarch64/vp9mc_neon.o
-NEON-OBJS-$(CONFIG_HEVC_DECODER)        += aarch64/hevcdsp_idct_neon.o         \
+NEON-OBJS-$(CONFIG_HEVC_DECODER)        += aarch64/hevcdsp_deblock_neon.o      \
+                                           aarch64/hevcdsp_idct_neon.o         \
                                            aarch64/hevcdsp_init_aarch64.o      \
                                            aarch64/hevcdsp_qpel_neon.o         \
                                            aarch64/hevcdsp_sao_neon.o
diff --git a/libavcodec/aarch64/hevcdsp_deblock_neon.S b/libavcodec/aarch64/hevcdsp_deblock_neon.S
new file mode 100644
index 0000000000..8227f65649
--- /dev/null
+++ b/libavcodec/aarch64/hevcdsp_deblock_neon.S
@@ -0,0 +1,183 @@
+/* -*-arm64-*-
+ * vim: syntax=arm64asm
+ *
+ * Copyright (c) 2014 Seppo Tomperi <seppo.tomperi@vtt.fi>
+ * Copyright (c) 2023 J. Dekker <jdek@itanimul.li>
+ *
+ * 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/aarch64/asm.S"
+#include "neon.S"
+
+.macro hevc_loop_filter_chroma_start bitdepth
+        mov             x4, x30
+        ldr             w14, [x2]
+        ldr             w15, [x2, #4]
+.if \bitdepth > 8
+        lsl             w14, w14, #(\bitdepth - 8)
+        lsl             w15, w15, #(\bitdepth - 8)
+.endif
+        adds            w2, w14, w15
+        b.eq            1f
+        dup             v16.4h, w14
+        dup             v17.4h, w15
+        trn1            v16.2d, v16.2d, v17.2d
+.if \bitdepth > 8
+        mvni            v19.8h, #((0xff << (\bitdepth - 8)) & 0xff), lsl #8
+        movi            v18.8h, #0
+.endif
+        neg             v17.8h, v16.8h
+.endm
+
+.macro hevc_loop_filter_chroma_body bitdepth
+.if \bitdepth <= 8
+        uxtl            v20.8h, v0.8b // p1
+        uxtl            v1.8h,  v1.8b // p0
+        uxtl            v2.8h,  v2.8b // q0
+        uxtl            v23.8h, v3.8b // q1
+        va              .req v20
+        vb              .req v23
+.else   // required to specify both cases as we are unable to do: v0 .req v20
+        va              .req v0
+        vb              .req v3
+.endif
+        sub             v5.8h, v2.8h, v1.8h // q0 - p0
+        sub             v6.8h, va.8h, vb.8h // p1 - q1
+        shl             v5.8h, v5.8h, #2
+        add             v5.8h, v6.8h, v5.8h
+        srshr           v5.8h, v5.8h, #3
+        clip            v17.8h, v16.8h, v5.8h
+        sqadd           v1.8h, v1.8h, v5.8h // p0 + delta
+        sqsub           v2.8h, v2.8h, v5.8h // q0 - delta
+.if \bitdepth <= 8
+        sqxtun          v1.8b, v1.8h
+        sqxtun          v2.8b, v2.8h
+.else
+        clip            v18.8h, v19.8h, v1.8h, v2.8h
+.endif
+.unreq  va
+.unreq  vb
+.endm
+
+function hevc_loop_filter_chroma_body_8_neon, export=0
+        hevc_loop_filter_chroma_body 8
+        ret
+endfunc
+
+function hevc_loop_filter_chroma_body_10_neon, export=0
+hevc_loop_filter_chroma_body_12_neon:
+        hevc_loop_filter_chroma_body 10
+        ret
+endfunc
+
+// void ff_hevc_h_loop_filter_chroma_8_neon(uint8_t *_pix, ptrdiff_t _stride, int *_tc, uint8_t *_no_p, uint8_t *_no_q);
+
+.macro hevc_h_loop_filter_chroma bitdepth
+function ff_hevc_h_loop_filter_chroma_\bitdepth\()_neon, export=1
+        hevc_loop_filter_chroma_start \bitdepth
+        sub             x0, x0, x1, lsl #1
+.if \bitdepth > 8
+        ld1             {v0.8h}, [x0], x1
+        ld1             {v1.8h}, [x0], x1
+        ld1             {v2.8h}, [x0], x1
+        ld1             {v3.8h}, [x0]
+.else
+        ld1             {v0.8b}, [x0], x1
+        ld1             {v1.8b}, [x0], x1
+        ld1             {v2.8b}, [x0], x1
+        ld1             {v3.8b}, [x0]
+.endif
+        sub             x0, x0, x1, lsl #1
+        bl              hevc_loop_filter_chroma_body_\bitdepth\()_neon
+.if \bitdepth > 8
+        st1             {v1.8h}, [x0], x1
+        st1             {v2.8h}, [x0]
+.else
+        st1             {v1.8b}, [x0], x1
+        st1             {v2.8b}, [x0]
+.endif
+1:      ret             x4
+endfunc
+.endm
+
+.macro hevc_v_loop_filter_chroma bitdepth
+function ff_hevc_v_loop_filter_chroma_\bitdepth\()_neon, export=1
+        hevc_loop_filter_chroma_start \bitdepth
+.if \bitdepth > 8
+        sub             x0, x0, #4
+        add             x3, x0, x1
+        lsl             x1, x1, #1
+        ld1             {v0.d}[0], [x0], x1
+        ld1             {v1.d}[0], [x3], x1
+        ld1             {v2.d}[0], [x0], x1
+        ld1             {v3.d}[0], [x3], x1
+        ld1             {v0.d}[1], [x0], x1
+        ld1             {v1.d}[1], [x3], x1
+        ld1             {v2.d}[1], [x0], x1
+        ld1             {v3.d}[1], [x3], x1
+        transpose_4x8H  v0, v1, v2, v3, v28, v29, v30, v31
+.else
+        sub             x0, x0, #2
+        add             x3, x0, x1
+        lsl             x1, x1, #1
+        ld1             {v0.s}[0], [x0], x1
+        ld1             {v1.s}[0], [x3], x1
+        ld1             {v2.s}[0], [x0], x1
+        ld1             {v3.s}[0], [x3], x1
+        ld1             {v0.s}[1], [x0], x1
+        ld1             {v1.s}[1], [x3], x1
+        ld1             {v2.s}[1], [x0], x1
+        ld1             {v3.s}[1], [x3], x1
+        transpose_4x8B  v0, v1, v2, v3, v28, v29, v30, v31
+.endif
+        sub             x0, x0, x1, lsl #2
+        sub             x3, x3, x1, lsl #2
+        bl              hevc_loop_filter_chroma_body_\bitdepth\()_neon
+.if \bitdepth > 8
+        transpose_4x8H  v0, v1, v2, v3, v28, v29, v30, v31
+        st1             {v0.d}[0], [x0], x1
+        st1             {v1.d}[0], [x3], x1
+        st1             {v2.d}[0], [x0], x1
+        st1             {v3.d}[0], [x3], x1
+        st1             {v0.d}[1], [x0], x1
+        st1             {v1.d}[1], [x3], x1
+        st1             {v2.d}[1], [x0], x1
+        st1             {v3.d}[1], [x3]
+.else
+        transpose_4x8B  v0, v1, v2, v3, v28, v29, v30, v31
+        st1             {v0.s}[0], [x0], x1
+        st1             {v1.s}[0], [x3], x1
+        st1             {v2.s}[0], [x0], x1
+        st1             {v3.s}[0], [x3], x1
+        st1             {v0.s}[1], [x0], x1
+        st1             {v1.s}[1], [x3], x1
+        st1             {v2.s}[1], [x0], x1
+        st1             {v3.s}[1], [x3]
+.endif
+1:      ret             x4
+endfunc
+.endm
+
+hevc_h_loop_filter_chroma 8
+hevc_h_loop_filter_chroma 10
+hevc_h_loop_filter_chroma 12
+
+hevc_v_loop_filter_chroma 8
+hevc_v_loop_filter_chroma 10
+hevc_v_loop_filter_chroma 12
diff --git a/libavcodec/aarch64/hevcdsp_init_aarch64.c b/libavcodec/aarch64/hevcdsp_init_aarch64.c
index 1deefca0a2..a923bae35c 100644
--- a/libavcodec/aarch64/hevcdsp_init_aarch64.c
+++ b/libavcodec/aarch64/hevcdsp_init_aarch64.c
@@ -25,6 +25,18 @@
 #include "libavutil/aarch64/cpu.h"
 #include "libavcodec/hevcdsp.h"
 
+void ff_hevc_v_loop_filter_chroma_8_neon(uint8_t *_pix, ptrdiff_t _stride,
+                                         const int *_tc, const uint8_t *_no_p, const uint8_t *_no_q);
+void ff_hevc_v_loop_filter_chroma_10_neon(uint8_t *_pix, ptrdiff_t _stride,
+                                          const int *_tc, const uint8_t *_no_p, const uint8_t *_no_q);
+void ff_hevc_v_loop_filter_chroma_12_neon(uint8_t *_pix, ptrdiff_t _stride,
+                                          const int *_tc, const uint8_t *_no_p, const uint8_t *_no_q);
+void ff_hevc_h_loop_filter_chroma_8_neon(uint8_t *_pix, ptrdiff_t _stride,
+                                         const int *_tc, const uint8_t *_no_p, const uint8_t *_no_q);
+void ff_hevc_h_loop_filter_chroma_10_neon(uint8_t *_pix, ptrdiff_t _stride,
+                                          const int *_tc, const uint8_t *_no_p, const uint8_t *_no_q);
+void ff_hevc_h_loop_filter_chroma_12_neon(uint8_t *_pix, ptrdiff_t _stride,
+                                          const int *_tc, const uint8_t *_no_p, const uint8_t *_no_q);
 void ff_hevc_add_residual_4x4_8_neon(uint8_t *_dst, const int16_t *coeffs,
                                      ptrdiff_t stride);
 void ff_hevc_add_residual_4x4_10_neon(uint8_t *_dst, const int16_t *coeffs,
@@ -117,6 +129,8 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
     if (!have_neon(av_get_cpu_flags())) return;
 
     if (bit_depth == 8) {
+        c->hevc_h_loop_filter_chroma   = ff_hevc_h_loop_filter_chroma_8_neon;
+        c->hevc_v_loop_filter_chroma   = ff_hevc_v_loop_filter_chroma_8_neon;
         c->add_residual[0]             = ff_hevc_add_residual_4x4_8_neon;
         c->add_residual[1]             = ff_hevc_add_residual_8x8_8_neon;
         c->add_residual[2]             = ff_hevc_add_residual_16x16_8_neon;
@@ -167,6 +181,8 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
         c->put_hevc_qpel_bi[9][0][1]   = ff_hevc_put_hevc_qpel_bi_h16_8_neon;
     }
     if (bit_depth == 10) {
+        c->hevc_h_loop_filter_chroma   = ff_hevc_h_loop_filter_chroma_10_neon;
+        c->hevc_v_loop_filter_chroma   = ff_hevc_v_loop_filter_chroma_10_neon;
         c->add_residual[0]             = ff_hevc_add_residual_4x4_10_neon;
         c->add_residual[1]             = ff_hevc_add_residual_8x8_10_neon;
         c->add_residual[2]             = ff_hevc_add_residual_16x16_10_neon;
@@ -180,6 +196,8 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
         c->idct_dc[3]                  = ff_hevc_idct_32x32_dc_10_neon;
     }
     if (bit_depth == 12) {
+        c->hevc_h_loop_filter_chroma   = ff_hevc_h_loop_filter_chroma_12_neon;
+        c->hevc_v_loop_filter_chroma   = ff_hevc_v_loop_filter_chroma_12_neon;
         c->add_residual[0]             = ff_hevc_add_residual_4x4_12_neon;
         c->add_residual[1]             = ff_hevc_add_residual_8x8_12_neon;
         c->add_residual[2]             = ff_hevc_add_residual_16x16_12_neon;