[x265] [PATCH] arm: Implement filterPixelToShort ARM NEON asm

dnyaneshwar at multicorewareinc.com dnyaneshwar at multicorewareinc.com
Tue Mar 1 13:16:01 CET 2016


# HG changeset patch
# User Dnyaneshwar G <dnyaneshwar at multicorewareinc.com>
# Date 1456831820 -19800
#      Tue Mar 01 17:00:20 2016 +0530
# Node ID 61e51faf9e7ee1c8056ac2f66cf51da104bfa106
# Parent  79c00b9bc2b81afef2e41526fc3c390528f3174c
arm: Implement filterPixelToShort ARM NEON asm

diff -r 79c00b9bc2b8 -r 61e51faf9e7e source/common/CMakeLists.txt
--- a/source/common/CMakeLists.txt	Tue Mar 01 12:18:18 2016 +0530
+++ b/source/common/CMakeLists.txt	Tue Mar 01 17:00:20 2016 +0530
@@ -89,7 +89,7 @@
     set(C_SRCS asm-primitives.cpp pixel.h mc.h ipfilter8.h blockcopy8.h dct8.h loopfilter.h)
 
     # add ARM assembly/intrinsic files here
-    set(A_SRCS asm.S cpu-a.S mc-a.S sad-a.S pixel-util.S ssd-a.S blockcopy8.S)
+    set(A_SRCS asm.S cpu-a.S mc-a.S sad-a.S pixel-util.S ssd-a.S blockcopy8.S ipfilter8.S)
     set(VEC_PRIMITIVES)
 
     set(ARM_ASMS "${A_SRCS}" CACHE INTERNAL "ARM Assembly Sources")
diff -r 79c00b9bc2b8 -r 61e51faf9e7e source/common/arm/asm-primitives.cpp
--- a/source/common/arm/asm-primitives.cpp	Tue Mar 01 12:18:18 2016 +0530
+++ b/source/common/arm/asm-primitives.cpp	Tue Mar 01 17:00:20 2016 +0530
@@ -33,6 +33,7 @@
 #include "blockcopy8.h"
 #include "pixel.h"
 #include "pixel-util.h"
+#include "ipfilter8.h"
 }
 
 namespace X265_NS {
@@ -42,6 +43,33 @@
 {
     if (cpuMask & X265_CPU_NEON)
     {
+        // filterPixelToShort
+        p.pu[LUMA_4x4].convert_p2s   = PFX(filterPixelToShort_4x4_neon);
+        p.pu[LUMA_4x8].convert_p2s   = PFX(filterPixelToShort_4x8_neon);
+        p.pu[LUMA_4x16].convert_p2s  = PFX(filterPixelToShort_4x16_neon);
+        p.pu[LUMA_8x4].convert_p2s   = PFX(filterPixelToShort_8x4_neon);
+        p.pu[LUMA_8x8].convert_p2s   = PFX(filterPixelToShort_8x8_neon);
+        p.pu[LUMA_8x16].convert_p2s  = PFX(filterPixelToShort_8x16_neon);
+        p.pu[LUMA_8x32].convert_p2s  = PFX(filterPixelToShort_8x32_neon);
+        p.pu[LUMA_12x16].convert_p2s = PFX(filterPixelToShort_12x16_neon);
+        p.pu[LUMA_16x4].convert_p2s  = PFX(filterPixelToShort_16x4_neon);
+        p.pu[LUMA_16x8].convert_p2s  = PFX(filterPixelToShort_16x8_neon);
+        p.pu[LUMA_16x12].convert_p2s = PFX(filterPixelToShort_16x12_neon);
+        p.pu[LUMA_16x16].convert_p2s = PFX(filterPixelToShort_16x16_neon);
+        p.pu[LUMA_16x32].convert_p2s = PFX(filterPixelToShort_16x32_neon);
+        p.pu[LUMA_16x64].convert_p2s = PFX(filterPixelToShort_16x64_neon);
+        p.pu[LUMA_24x32].convert_p2s = PFX(filterPixelToShort_24x32_neon);
+        p.pu[LUMA_32x8].convert_p2s  = PFX(filterPixelToShort_32x8_neon);
+        p.pu[LUMA_32x16].convert_p2s = PFX(filterPixelToShort_32x16_neon);
+        p.pu[LUMA_32x24].convert_p2s = PFX(filterPixelToShort_32x24_neon);
+        p.pu[LUMA_32x32].convert_p2s = PFX(filterPixelToShort_32x32_neon);
+        p.pu[LUMA_32x64].convert_p2s = PFX(filterPixelToShort_32x64_neon);
+        p.pu[LUMA_48x64].convert_p2s = PFX(filterPixelToShort_48x64_neon);
+        p.pu[LUMA_64x16].convert_p2s = PFX(filterPixelToShort_64x16_neon);
+        p.pu[LUMA_64x32].convert_p2s = PFX(filterPixelToShort_64x32_neon);
+        p.pu[LUMA_64x48].convert_p2s = PFX(filterPixelToShort_64x48_neon);
+        p.pu[LUMA_64x64].convert_p2s = PFX(filterPixelToShort_64x64_neon);
+
         // Block_fill
         p.cu[BLOCK_4x4].blockfill_s   = PFX(blockfill_s_4x4_neon);
         p.cu[BLOCK_8x8].blockfill_s   = PFX(blockfill_s_8x8_neon);
diff -r 79c00b9bc2b8 -r 61e51faf9e7e source/common/arm/ipfilter8.S
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/source/common/arm/ipfilter8.S	Tue Mar 01 17:00:20 2016 +0530
@@ -0,0 +1,694 @@
+/*****************************************************************************
+ * Copyright (C) 2016 x265 project
+ *
+ * Authors: Dnyaneshwar G <dnyaneshwar at multicorewareinc.com>
+ * 
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+ *
+ * This program is also available under a commercial proprietary license.
+ * For more information, contact us at license @ x265.com.
+ *****************************************************************************/
+
+#include "asm.S"
+
+.section .rodata
+
+.align 4
+
+.text
+
+// filterPixelToShort(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride)
+function x265_filterPixelToShort_4x4_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 2
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {d4}, [r2], r3
+    vst1.16     {d6}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_4x8_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 4
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {d4}, [r2], r3
+    vst1.16     {d6}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_4x16_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 8
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {d4}, [r2], r3
+    vst1.16     {d6}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_8x4_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 2
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2}, [r2], r3
+    vst1.16     {q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_8x8_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 4
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2}, [r2], r3
+    vst1.16     {q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_8x16_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 8
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2}, [r2], r3
+    vst1.16     {q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_8x32_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 16
+    vld1.u8     {d0}, [r0], r1
+    vld1.u8     {d2}, [r0], r1
+    vmovl.u8    q0, d0
+    vmovl.u8    q1, d2
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2}, [r2], r3
+    vst1.16     {q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_12x16_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 16
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {d4, d5, d6}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_16x4_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 4
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_16x8_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 8
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_16x12_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 12
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_16x16_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 16
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_16x32_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 32
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_16x64_neon
+    add         r3, r3
+    vmov.u16    q8, #64
+    vmov.u16    q9, #8192
+    vneg.s16    q9, q9
+.rept 64
+    vld1.u8     {d2-d3}, [r0], r1
+    vmovl.u8    q0, d2
+    vmovl.u8    q1, d3
+    vmov        q2, q9
+    vmov        q3, q9
+    vmla.s16    q2, q0, q8
+    vmla.s16    q3, q1, q8
+    vst1.16     {q2-q3}, [r2], r3
+.endr   
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_24x32_neon
+    add         r3, r3
+    sub         r3, #32
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+.rept 32
+    vld1.u8     {d18, d19, d20}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmla.s16    q2, q11, q0
+    vst1.16     {q2}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_32x8_neon
+    add         r3, r3
+    sub         r3, #32
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+.rept 8
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_32x16_neon
+    add         r3, r3
+    sub         r3, #32
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #8
+.loop_filterP2S_32x16:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_32x16
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_32x24_neon
+    add         r3, r3
+    sub         r3, #32
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #12
+.loop_filterP2S_32x24:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_32x24
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_32x32_neon
+    add         r3, r3
+    sub         r3, #32
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #16
+.loop_filterP2S_32x32:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_32x32
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_32x64_neon
+    add         r3, r3
+    sub         r3, #32
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #32
+.loop_filterP2S_32x64:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_32x64
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_64x16_neon
+    add         r3, r3
+    sub         r1, #32
+    sub         r3, #96
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #8
+.loop_filterP2S_64x16:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0]!
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2]!
+
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_64x16
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_64x32_neon
+    add         r3, r3
+    sub         r1, #32
+    sub         r3, #96
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #16
+.loop_filterP2S_64x32:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0]!
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2]!
+
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_64x32
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_64x48_neon
+    add         r3, r3
+    sub         r1, #32
+    sub         r3, #96
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #24
+.loop_filterP2S_64x48:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0]!
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2]!
+
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_64x48
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_64x64_neon
+    add         r3, r3
+    sub         r1, #32
+    sub         r3, #96
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #32
+.loop_filterP2S_64x64:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0]!
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2]!
+
+    vld1.u8     {q9-q10}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_64x64
+    bx          lr
+endfunc
+
+function x265_filterPixelToShort_48x64_neon
+    add         r3, r3
+    sub         r1, #32
+    sub         r3, #64
+    vmov.u16    q0, #64
+    vmov.u16    q1, #8192
+    vneg.s16    q1, q1
+    mov         r12, #32
+.loop_filterP2S_48x64:
+    subs        r12, #1
+.rept 2
+    vld1.u8     {q9-q10}, [r0]!
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmovl.u8    q11, d20
+    vmovl.u8    q10, d21
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2]!
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q11, q0
+    vmla.s16    q3, q10, q0
+    vst1.16     {q2-q3}, [r2]!
+
+    vld1.u8     {q9}, [r0], r1
+    vmovl.u8    q8, d18
+    vmovl.u8    q9, d19
+    vmov        q2, q1
+    vmov        q3, q1
+    vmla.s16    q2, q8, q0
+    vmla.s16    q3, q9, q0
+    vst1.16     {q2-q3}, [r2], r3
+.endr
+    bgt         .loop_filterP2S_48x64
+    bx          lr
+endfunc
diff -r 79c00b9bc2b8 -r 61e51faf9e7e source/common/arm/ipfilter8.h
--- a/source/common/arm/ipfilter8.h	Tue Mar 01 12:18:18 2016 +0530
+++ b/source/common/arm/ipfilter8.h	Tue Mar 01 17:00:20 2016 +0530
@@ -25,4 +25,30 @@
 #ifndef X265_IPFILTER8_ARM_H
 #define X265_IPFILTER8_ARM_H
 
+void x265_filterPixelToShort_4x4_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_4x8_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_4x16_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x4_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x8_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x16_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x32_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_12x16_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x4_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x8_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x12_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x16_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x32_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x64_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_24x32_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x8_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x16_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x24_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x32_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x64_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_48x64_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x16_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x32_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x48_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x64_neon(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+
 #endif // ifndef X265_IPFILTER8_ARM_H


More information about the x265-devel mailing list