[FFmpeg-devel] [FFmpeg-Devel] [GSoC] [PATCH 3/4] Added asm implmentations of some deinterlace functions

[Tucker DiNapoli]

From: Tucker DiNapoli <T.DiNapoli42 at gmail.com>

I also added a makefile which assembles this file into libpostproc. I
haven't yet modified the c code to use these functions yet.

diff --git a/libpostproc/x86/Makefile b/libpostproc/x86/Makefile
new file mode 100644
index 0000000..06838ca
--- /dev/null
+++ b/libpostproc/x86/Makefile
@@ -0,0 +1 @@
+YASM-OBJS-$(CONFIG_POSTPROC) += x86/deinterlace.o
diff --git a/libpostproc/x86/deinterlace.asm b/libpostproc/x86/deinterlace.asm
new file mode 100644
index 0000000..6e669bb
--- /dev/null
+++ b/libpostproc/x86/deinterlace.asm
@@ -0,0 +1,167 @@
+;*
+;* DeInterlacing filters written using SIMD extensions
+;* Copyright (C) 2015 Tucker DiNapoli (T.Dinapoli at gmail.com)
+;*
+;* Adapted from inline assembly:
+;* Copyright (C) 2001-2002 Michael Niedermayer (michaelni at gmx.at)
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg is free software; you can redistribute it and/or modify
+;* it under the terms of the GNU General Public License as published by
+;* the Free Software Foundation; either version 2 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 General Public License for more details.
+;*
+;* You should have received a copy of the GNU 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 "PPUtil.asm"
+;; All deinterlace functions operate on N 8x8 blocks at a time, where N
+;; is the size (in bytes) of the simd registers being used divided
+;; by 8, so 2 for xmm, and 4 for ymm.
+
+;; NOTE: The function names are camel case for compatibility with existing 
+;; postprocessing code, eventually the names of postprocessing functions
+;; will be changed to use lowercase and underscores
+
+;; Deinterlace blocks using linear interpolation
+;; Set each line 2n+1 to (line 2n + line 2n+2)/2
+%macro gen_deinterlace_interpolate_linear 0
+cglobal deInterlaceInterpolateLinear, 2, 4, 2;, src, stride
+    lea r1, [r1 + r2 * 4]
+    lea r3, [r1 + r2]
+    lea r4, [r3 + r2 * 4]
+    mova m0, [r1]
+    mova m1, [r3 + r2]
+    pavgb m0,m1
+    mova [r3], m0
+    mova m0, [r1 + r2 * 4]
+    pavgb m1,m0
+    mova [r3 + r2 * 2],m1
+    mova m1, [r4 + r2]
+    pavgb m0,m1
+    mova [r4], m0
+    mova m1, [r1 + r2 * 8]
+    pavgb m0,m1
+    mova [r4 + r2 * 2], m0
+    RET
+%endmacro
+;; Deinterlace blocks using cubic interpolation
+;; Line 2n+1 = (9(2n) + 9(2n+2) - (2n-2) - (2n+4))/16
+%macro gen_deinterlace_interpolate_cubic 0
+cglobal deInterlaceInterpolateCubic, 2, 5, 5;, src, stride
+    lea r3, [r2 + r2 * 2]
+    add r1,r3
+    lea r3, [r1 + r2]
+    lea r4, [r3 + r2 * 4]
+    lea r5, [r4 + r2 * 4]
+    pxor m4,m4
+%ifnmacro deint_cubic
+;; given 5 lines a,b,c,d,e: a = c-3, b = c-1, d = c+1, e = c + 2
+;; set c = (9b + 9d - a - b)/16
+%macro deint_cubic 5
+    mova m0,%1
+    mova m1,%2
+    mova m2,%4
+    mova m3,%5
+    pavgb m1,m2 ;(b+d)/2
+    pavgb m0,m3 ;(a+e)/2
+    ;; convert each byte into a word
+    mova m2,m1
+    punpcklbw m1, m4
+    punpckhbw m2, m4
+    mova m0,m3
+    punpcklbw m0, m4
+    punpckhbw m3, m4
+    ;; calculate the pixel values
+    psubw m0, m1 ;L(a+e - (b+d))/2
+    psubw m3, m2 ;H(a+e - (b+d))/2
+    psraw m0, 3
+    psraw m3, 3
+    psubw m1, m0 ;L(9(b+d) - (a+e))/16
+    psubw m3, m2 ;H(9(b+d) - (a+e))/16
+    ;; convert the words back into bytes using unsigned saturation
+    packuswb m1, m3
+    mova %3, m1
+%endmacro
+%endif
+    deint_cubic [r1], [r3 + r2], [r3 + r2 *2],\
+                [r1 + r2 *4], [r4 + r2]
+    deint_cubic [r3 + r2], [r1 + r2 * 4], [r4],\
+                [r4 + r2], [r1 + r2 * 8]
+    deint_cubic [r1 + r2 * 4], [r4 + r2], [r4 + r2 * 2],\
+                [r1 + r2 * 8], [r5]
+    deint_cubic [r4 + r2], [r1 + r2 * 8], [r4 + r2 * 4],\
+                [r5], [r5 + r2 * 2]
+    RET
+%endmacro
+
+;; deinterlace blocks by seting every line n to (n-1 + 2n + n+1)/4
+%macro gen_deinterlace_blend_linear 0
+cglobal deInterlaceBlendLinear, 3, 5, 2 ;src, stride, tmp
+    lea r1, [r1 + r2 * 4]
+    lea r4, [r1 + r2]
+    lea r5, [r4 + r2 * 4]
+    mova m0, [r3] ;L0
+    mova m1, [r4] ;L2
+    mova m2, [r1] ;L1
+    pavgb m0, m1 ;L0+L2
+    pavgb m0, m2 ;L0 + 2L1 + L2 / 4
+    mova [r1], m0
+    mova m0, [r4 + r2 * 2] ;L3
+    pavgb m2, m0
+    pavgb m2, m1
+    mova [r4], m2 ;L4
+    mova m2, [r4 + r2 * 2]
+    pavgb m1, m2
+    pavgb m1, m0
+    mova [r4+r2], m1 ;L5
+    mova m1, [r1 + r2 * 4]
+    pavgb m0, m1
+    pavgb m0, m2
+    mova [r4 + r2 * 2], m0 ;L6
+    mova m0, [r5]
+    pavgb m2, m0
+    pavgb m2, m1
+    mova [r1 + r2 * 4], m2 ;L7
+    mova m2, [r5 + r2]
+    pavgb m1, m2
+    pavgb m1, m0
+    mova [r5], m1
+    mova m1, [r5 + r2 * 2]
+    pavgb m0, m1
+    pavgb m0, m2
+    mova [r5 + r2], m0
+    mova m0, [r1 + r2 * 8]
+    pavgb m2, m0
+    pavgb m2, m1
+    mova [r5 + r2 * 2], m2
+    mova [r3], m1
+    RET
+%endmacro
+;; I'm not exactly sure how to insure the following only get built if
+;; the specified instruction set is available.
+;; If the INIT_XXX macros do that then great, otherwise I'll correct it
+SECTION_TEXT
+
+INIT_MMX mmx2
+gen_deinterlace_interpolate_linear
+gen_deinterlace_interpolate_cubic
+gen_deinterlace_blend_linear
+
+INIT_XMM sse2
+gen_deinterlace_interpolate_linear
+gen_deinterlace_interpolate_cubic
+gen_deinterlace_blend_linear
+
+INIT_YMM avx2
+gen_deinterlace_interpolate_linear
+gen_deinterlace_interpolate_cubic
+gen_deinterlace_blend_linear
-- 
2.3.3