aarch64: h264 chroma motion compensation NEON optimizations

Since RV40 and VC-1 use almost the same algorithm so optimizations for
those two decoders are easy to do and included.

1 files changed, 402 insertions, 0 deletions

diff --git a/libavcodec/aarch64/h264cmc_neon.S b/libavcodec/aarch64/h264cmc_neon.S
new file mode 100644
index 0000000000..50d12b4e20
--- /dev/null
+++ b/libavcodec/aarch64/h264cmc_neon.S
@@ -0,0 +1,402 @@
+/*
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ * Copyright (c) 2013 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 "libavutil/aarch64/asm.S"
+
+/* chroma_mc8(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y) */
+.macro  h264_chroma_mc8 type, codec=h264
+function ff_\type\()_\codec\()_chroma_mc8_neon, export=1
+        sxtw            x2,  w2
+  .ifc \type,avg
+        mov             x8,  x0
+  .endif
+        prfm            pldl1strm, [x1]
+        prfm            pldl1strm, [x1, x2]
+  .ifc \codec,rv40
+        movrel          x6,  rv40bias
+        lsr             w9,  w5,  #1
+        lsr             w10, w4,  #1
+        lsl             w9,  w9,  #3
+        lsl             w10, w10, #1
+        add             w9,  w9,  w10
+        add             x6,  x6,  w9, UXTW
+        ld1r            {v22.8H}, [x6]
+  .endif
+  .ifc \codec,vc1
+        movi            v22.8H,   #28
+  .endif
+        mul             w7,  w4,  w5
+        lsl             w14, w5,  #3
+        lsl             w13, w4,  #3
+        cmp             w7,  #0
+        sub             w6,  w14, w7
+        sub             w12, w13, w7
+        sub             w4,  w7,  w13
+        sub             w4,  w4,  w14
+        add             w4,  w4,  #64
+        b.eq            2f
+
+        dup             v0.8B,  w4
+        dup             v1.8B,  w12
+        ld1             {v4.8B, v5.8B}, [x1], x2
+        dup             v2.8B,  w6
+        dup             v3.8B,  w7
+        ext             v5.8B,  v4.8B,  v5.8B,  #1
+1:      ld1             {v6.8B, v7.8B}, [x1], x2
+        umull           v16.8H, v4.8B,  v0.8B
+        umlal           v16.8H, v5.8B,  v1.8B
+        ext             v7.8B,  v6.8B,  v7.8B,  #1
+        ld1             {v4.8B, v5.8B}, [x1], x2
+        umlal           v16.8H, v6.8B,  v2.8B
+        prfm            pldl1strm, [x1]
+        ext             v5.8B,  v4.8B,  v5.8B,  #1
+        umlal           v16.8H, v7.8B,  v3.8B
+        umull           v17.8H, v6.8B,  v0.8B
+        subs            w3,  w3,  #2
+        umlal           v17.8H, v7.8B, v1.8B
+        umlal           v17.8H, v4.8B, v2.8B
+        umlal           v17.8H, v5.8B, v3.8B
+        prfm            pldl1strm, [x1, x2]
+  .ifc \codec,h264
+        rshrn           v16.8B, v16.8H, #6
+        rshrn           v17.8B, v17.8H, #6
+  .else
+        add             v16.8H, v16.8H, v22.8H
+        add             v17.8H, v17.8H, v22.8H
+        shrn            v16.8B, v16.8H, #6
+        shrn            v17.8B, v17.8H, #6
+  .endif
+  .ifc \type,avg
+        ld1             {v20.8B}, [x8], x2
+        ld1             {v21.8B}, [x8], x2
+        urhadd          v16.8B, v16.8B, v20.8B
+        urhadd          v17.8B, v17.8B, v21.8B
+  .endif
+        st1             {v16.8B}, [x0], x2
+        st1             {v17.8B}, [x0], x2
+        b.gt            1b
+        ret
+
+2:      tst             w6,  w6
+        add             w12, w12, w6
+        dup             v0.8B, w4
+        dup             v1.8B, w12
+        b.eq            4f
+
+        ld1             {v4.8B}, [x1], x2
+3:      ld1             {v6.8B}, [x1], x2
+        umull           v16.8H, v4.8B,  v0.8B
+        umlal           v16.8H, v6.8B,  v1.8B
+        ld1             {v4.8B}, [x1], x2
+        umull           v17.8H, v6.8B,  v0.8B
+        umlal           v17.8H, v4.8B,  v1.8B
+        prfm            pldl1strm, [x1]
+  .ifc \codec,h264
+        rshrn           v16.8B, v16.8H, #6
+        rshrn           v17.8B, v17.8H, #6
+  .else
+        add             v16.8H, v16.8H, v22.8H
+        add             v17.8H, v17.8H, v22.8H
+        shrn            v16.8B, v16.8H, #6
+        shrn            v17.8B, v17.8H, #6
+  .endif
+        prfm            pldl1strm, [x1, x2]
+  .ifc \type,avg
+        ld1             {v20.8B}, [x8], x2
+        ld1             {v21.8B}, [x8], x2
+        urhadd          v16.8B, v16.8B, v20.8B
+        urhadd          v17.8B, v17.8B, v21.8B
+  .endif
+        subs            w3,  w3,  #2
+        st1             {v16.8B}, [x0], x2
+        st1             {v17.8B}, [x0], x2
+        b.gt            3b
+        ret
+
+4:      ld1             {v4.8B, v5.8B}, [x1], x2
+        ld1             {v6.8B, v7.8B}, [x1], x2
+        ext             v5.8B,  v4.8B,  v5.8B,  #1
+        ext             v7.8B,  v6.8B,  v7.8B,  #1
+        prfm            pldl1strm, [x1]
+        subs            w3,  w3,  #2
+        umull           v16.8H, v4.8B, v0.8B
+        umlal           v16.8H, v5.8B, v1.8B
+        umull           v17.8H, v6.8B, v0.8B
+        umlal           v17.8H, v7.8B, v1.8B
+        prfm            pldl1strm, [x1, x2]
+  .ifc \codec,h264
+        rshrn           v16.8B, v16.8H, #6
+        rshrn           v17.8B, v17.8H, #6
+  .else
+        add             v16.8H, v16.8H, v22.8H
+        add             v17.8H, v17.8H, v22.8H
+        shrn            v16.8B, v16.8H, #6
+        shrn            v17.8B, v17.8H, #6
+  .endif
+  .ifc \type,avg
+        ld1             {v20.8B}, [x8], x2
+        ld1             {v21.8B}, [x8], x2
+        urhadd          v16.8B, v16.8B, v20.8B
+        urhadd          v17.8B, v17.8B, v21.8B
+  .endif
+        st1             {v16.8B}, [x0], x2
+        st1             {v17.8B}, [x0], x2
+        b.gt            4b
+        ret
+endfunc
+.endm
+
+/* chroma_mc4(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y) */
+.macro  h264_chroma_mc4 type, codec=h264
+function ff_\type\()_\codec\()_chroma_mc4_neon, export=1
+        sxtw            x2,  w2
+  .ifc \type,avg
+        mov             x8,  x0
+  .endif
+        prfm            pldl1strm, [x1]
+        prfm            pldl1strm, [x1, x2]
+  .ifc \codec,rv40
+        movrel          x6,  rv40bias
+        lsr             w9,  w5,  #1
+        lsr             w10, w4,  #1
+        lsl             w9,  w9,  #3
+        lsl             w10, w10, #1
+        add             w9,  w9,  w10
+        add             x6,  x6,  w9, UXTW
+        ld1r            {v22.8H}, [x6]
+  .endif
+  .ifc \codec,vc1
+        movi            v22.8H,   #28
+  .endif
+        mul             w7,  w4,  w5
+        lsl             w14, w5,  #3
+        lsl             w13, w4,  #3
+        cmp             w7,  #0
+        sub             w6,  w14, w7
+        sub             w12, w13, w7
+        sub             w4,  w7,  w13
+        sub             w4,  w4,  w14
+        add             w4,  w4,  #64
+        b.eq            2f
+
+        dup             v24.8B,  w4
+        dup             v25.8B,  w12
+        ld1             {v4.8B}, [x1], x2
+        dup             v26.8B,  w6
+        dup             v27.8B,  w7
+        ext             v5.8B,  v4.8B,  v5.8B, #1
+        trn1            v0.2S,  v24.2S, v25.2S
+        trn1            v2.2S,  v26.2S, v27.2S
+        trn1            v4.2S,  v4.2S,  v5.2S
+1:      ld1             {v6.8B}, [x1], x2
+        ext             v7.8B,  v6.8B,  v7.8B, #1
+        trn1            v6.2S,  v6.2S,  v7.2S
+        umull           v18.8H, v4.8B,  v0.8B
+        umlal           v18.8H, v6.8B,  v2.8B
+        ld1             {v4.8B}, [x1], x2
+        ext             v5.8B,  v4.8B,  v5.8B, #1
+        trn1            v4.2S,  v4.2S,  v5.2S
+        prfm            pldl1strm, [x1]
+        umull           v19.8H, v6.8B,  v0.8B
+        umlal           v19.8H, v4.8B,  v2.8B
+        trn1            v30.2D, v18.2D, v19.2D
+        trn2            v31.2D, v18.2D, v19.2D
+        add             v18.8H, v30.8H, v31.8H
+  .ifc \codec,h264
+        rshrn           v16.8B, v18.8H, #6
+  .else
+        add             v18.8H, v18.8H, v22.8H
+        shrn            v16.8B, v18.8H, #6
+  .endif
+        subs            w3,  w3,  #2
+        prfm            pldl1strm, [x1, x2]
+  .ifc \type,avg
+        ld1             {v20.S}[0], [x8], x2
+        ld1             {v20.S}[1], [x8], x2
+        urhadd          v16.8B, v16.8B, v20.8B
+  .endif
+        st1             {v16.S}[0], [x0], x2
+        st1             {v16.S}[1], [x0], x2
+        b.gt            1b
+        ret
+
+2:      tst             w6,  w6
+        add             w12, w12, w6
+        dup             v30.8B, w4
+        dup             v31.8B, w12
+        trn1            v0.2S,  v30.2S, v31.2S
+        trn2            v1.2S,  v30.2S, v31.2S
+        b.eq            4f
+
+        ext             v1.8B,  v0.8B,  v1.8B, #4
+        ld1             {v4.S}[0], [x1], x2
+3:      ld1             {v4.S}[1], [x1], x2
+        umull           v18.8H, v4.8B,  v0.8B
+        ld1             {v4.S}[0], [x1], x2
+        umull           v19.8H, v4.8B,  v1.8B
+        trn1            v30.2D, v18.2D, v19.2D
+        trn2            v31.2D, v18.2D, v19.2D
+        add             v18.8H, v30.8H, v31.8H
+        prfm            pldl1strm, [x1]
+  .ifc \codec,h264
+        rshrn           v16.8B, v18.8H, #6
+  .else
+        add             v18.8H, v18.8H, v22.8H
+        shrn            v16.8B, v18.8H, #6
+  .endif
+  .ifc \type,avg
+        ld1             {v20.S}[0], [x8], x2
+        ld1             {v20.S}[1], [x8], x2
+        urhadd          v16.8B, v16.8B, v20.8B
+  .endif
+        subs            w3,  w3,  #2
+        prfm            pldl1strm, [x1, x2]
+        st1             {v16.S}[0], [x0], x2
+        st1             {v16.S}[1], [x0], x2
+        b.gt            3b
+        ret
+
+4:      ld1             {v4.8B}, [x1], x2
+        ld1             {v6.8B}, [x1], x2
+        ext             v5.8B,  v4.8B,  v5.8B, #1
+        ext             v7.8B,  v6.8B,  v7.8B, #1
+        trn1            v4.2S,  v4.2S,  v5.2S
+        trn1            v6.2S,  v6.2S,  v7.2S
+        umull           v18.8H, v4.8B,  v0.8B
+        umull           v19.8H, v6.8B,  v0.8B
+        subs            w3,  w3,  #2
+        trn1            v30.2D, v18.2D, v19.2D
+        trn2            v31.2D, v18.2D, v19.2D
+        add             v18.8H, v30.8H, v31.8H
+        prfm            pldl1strm, [x1]
+  .ifc \codec,h264
+        rshrn           v16.8B, v18.8H, #6
+  .else
+        add             v18.8H, v18.8H, v22.8H
+        shrn            v16.8B, v18.8H, #6
+  .endif
+  .ifc \type,avg
+        ld1             {v20.S}[0], [x8], x2
+        ld1             {v20.S}[1], [x8], x2
+        urhadd          v16.8B, v16.8B, v20.8B
+  .endif
+        prfm            pldl1strm, [x1]
+        st1             {v16.S}[0], [x0], x2
+        st1             {v16.S}[1], [x0], x2
+        b.gt            4b
+        ret
+endfunc
+.endm
+
+.macro  h264_chroma_mc2 type
+function ff_\type\()_h264_chroma_mc2_neon, export=1
+        sxtw            x2,  w2
+        prfm            pldl1strm, [x1]
+        prfm            pldl1strm, [x1, x2]
+        orr             w7,  w4,  w5
+        cbz             w7,  2f
+
+        mul             w7,  w4,  w5
+        lsl             w14, w5,  #3
+        lsl             w13, w4,  #3
+        sub             w6,  w14, w7
+        sub             w12, w13, w7
+        sub             w4,  w7,  w13
+        sub             w4,  w4,  w14
+        add             w4,  w4,  #64
+        dup             v0.8B,  w4
+        dup             v2.8B,  w12
+        dup             v1.8B,  w6
+        dup             v3.8B,  w7
+        trn1            v0.4H,  v0.4H,  v2.4H
+        trn1            v1.4H,  v1.4H,  v3.4H
+1:
+        ld1             {v4.S}[0],  [x1], x2
+        ld1             {v4.S}[1],  [x1], x2
+        rev64           v5.2S,  v4.2S
+        ld1             {v5.S}[1],  [x1]
+        ext             v6.8B,  v4.8B,  v5.8B,  #1
+        ext             v7.8B,  v5.8B,  v4.8B,  #1
+        trn1            v4.4H,  v4.4H,  v6.4H
+        trn1            v5.4H,  v5.4H,  v7.4H
+        umull           v16.8H, v4.8B,  v0.8B
+        umlal           v16.8H, v5.8B,  v1.8B
+  .ifc \type,avg
+        ld1             {v18.H}[0], [x0], x2
+        ld1             {v18.H}[2], [x0]
+        sub             x0,  x0,  x2
+  .endif
+        rev64           v17.4S, v16.4S
+        add             v16.8H, v16.8H, v17.8H
+        rshrn           v16.8B, v16.8H, #6
+  .ifc \type,avg
+        urhadd          v16.8B, v16.8B, v18.8B
+  .endif
+        st1             {v16.H}[0], [x0], x2
+        st1             {v16.H}[2], [x0], x2
+        subs            w3,  w3,  #2
+        b.gt            1b
+        ret
+
+2:
+        ld1             {v16.H}[0], [x1], x2
+        ld1             {v16.H}[1], [x1], x2
+  .ifc \type,avg
+        ld1             {v18.H}[0], [x0], x2
+        ld1             {v18.H}[1], [x0]
+        sub             x0,  x0,  x2
+        urhadd          v16.8B, v16.8B, v18.8B
+  .endif
+        st1             {v16.H}[0], [x0], x2
+        st1             {v16.H}[1], [x0], x2
+        subs            w3,  w3,  #2
+        b.gt            2b
+        ret
+endfunc
+.endm
+
+        h264_chroma_mc8 put
+        h264_chroma_mc8 avg
+        h264_chroma_mc4 put
+        h264_chroma_mc4 avg
+        h264_chroma_mc2 put
+        h264_chroma_mc2 avg
+
+#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
+
+#if CONFIG_VC1_DECODER
+        h264_chroma_mc8 put, vc1
+        h264_chroma_mc8 avg, vc1
+        h264_chroma_mc4 put, vc1
+        h264_chroma_mc4 avg, vc1
+#endif