changeset 11532:e011e73a902b libavcodec

ARM: NEON optimised RDFT
author mru
date Tue, 23 Mar 2010 03:35:02 +0000
parents 31033caa5344
children e9640e8aeea3
files arm/Makefile arm/fft_init_arm.c arm/rdft_neon.S fft.h rdft.c
diffstat 5 files changed, 168 insertions(+), 0 deletions(-) [+]
line wrap: on
line diff
--- a/arm/Makefile	Tue Mar 23 02:17:04 2010 +0000
+++ b/arm/Makefile	Tue Mar 23 03:35:02 2010 +0000
@@ -27,6 +27,8 @@
 
 NEON-OBJS-$(CONFIG_MDCT)               += arm/mdct_neon.o               \
 
+NEON-OBJS-$(CONFIG_RDFT)               += arm/rdft_neon.o               \
+
 NEON-OBJS-$(CONFIG_H264DSP)            += arm/h264dsp_neon.o            \
                                           arm/h264idct_neon.o           \
                                           arm/h264pred_neon.o           \
--- a/arm/fft_init_arm.c	Tue Mar 23 02:17:04 2010 +0000
+++ b/arm/fft_init_arm.c	Tue Mar 23 03:35:02 2010 +0000
@@ -27,6 +27,8 @@
 void ff_imdct_half_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
 void ff_mdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
 
+void ff_rdft_calc_neon(struct RDFTContext *s, FFTSample *z);
+
 av_cold void ff_fft_init_arm(FFTContext *s)
 {
     if (HAVE_NEON) {
@@ -38,3 +40,11 @@
         s->permutation  = FF_MDCT_PERM_INTERLEAVE;
     }
 }
+
+#if CONFIG_RDFT
+av_cold void ff_rdft_init_arm(RDFTContext *s)
+{
+    if (HAVE_NEON)
+        s->rdft_calc    = ff_rdft_calc_neon;
+}
+#endif
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/arm/rdft_neon.S	Tue Mar 23 03:35:02 2010 +0000
@@ -0,0 +1,151 @@
+/*
+ * ARM NEON optimised RDFT
+ * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
+ *
+ * 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 "asm.S"
+
+        preserve8
+
+function ff_rdft_calc_neon, export=1
+        push            {r4-r8,lr}
+
+        ldr             r6,  [r0, #4]           @ inverse
+        mov             r4,  r0
+        mov             r5,  r1
+
+        lsls            r6,  r6,  #31
+        bne             1f
+        add             r0,  r4,  #20
+        bl              ff_fft_permute_neon
+        add             r0,  r4,  #20
+        mov             r1,  r5
+        bl              ff_fft_calc_neon
+1:
+        ldr             r12, [r4, #0]           @ nbits
+        mov             r2,  #1
+        lsl             r12, r2,  r12
+        add             r0,  r5,  #8
+        add             r1,  r5,  r12, lsl #2
+        lsr             r12, r12, #2
+        ldr             r2,  [r4, #12]          @ tcos
+        sub             r12, r12, #2
+        ldr             r3,  [r4, #16]          @ tsin
+        mov             r7,  r0
+        sub             r1,  r1,  #8
+        mov             lr,  r1
+        mov             r8,  #-8
+        vld1.32         {d0},     [r0,:64]!     @ d1[0,1]
+        vld1.32         {d1},     [r1,:64], r8  @ d2[0,1]
+        vld1.32         {d4},     [r2,:64]!     @ tcos[i]
+        vld1.32         {d5},     [r3,:64]!     @ tsin[i]
+        vmov.f32        d18, #0.5               @ k1
+        vdup.32         d19, r6
+        pld             [r0, #32]
+        veor            d19, d18, d19           @ k2
+        vmov.i32        d16, #0
+        vmov.i32        d17, #1<<31
+        pld             [r1, #-32]
+        vtrn.32         d16, d17
+        pld             [r2, #32]
+        vrev64.32       d16, d16                @ d16=1,0 d17=0,1
+        pld             [r3, #32]
+2:
+        veor            q1,  q0,  q8            @ -d1[0],d1[1], d2[0],-d2[1]
+        vld1.32         {d24},    [r0,:64]!     @  d1[0,1]
+        vadd.f32        d0,  d0,  d3            @  d1[0]+d2[0], d1[1]-d2[1]
+        vld1.32         {d25},    [r1,:64], r8  @  d2[0,1]
+        vadd.f32        d1,  d2,  d1            @ -d1[0]+d2[0], d1[1]+d2[1]
+        veor            q3,  q12, q8            @ -d1[0],d1[1], d2[0],-d2[1]
+        pld             [r0, #32]
+        vmul.f32        q10, q0,  q9            @  ev.re, ev.im, od.im, od.re
+        pld             [r1, #-32]
+        vadd.f32        d0,  d24, d7            @  d1[0]+d2[0], d1[1]-d2[1]
+        vadd.f32        d1,  d6,  d25           @ -d1[0]+d2[0], d1[1]+d2[1]
+        vmul.f32        q11, q0,  q9            @  ev.re, ev.im, od.im, od.re
+        veor            d7,  d21, d16           @ -od.im, od.re
+        vrev64.32       d3,  d21                @  od.re, od.im
+        veor            d6,  d20, d17           @  ev.re,-ev.im
+        veor            d2,  d3,  d16           @ -od.re, od.im
+        vmla.f32        d20, d3,  d4[1]
+        vmla.f32        d20, d7,  d5[1]
+        vmla.f32        d6,  d2,  d4[1]
+        vmla.f32        d6,  d21, d5[1]
+        vld1.32         {d4},     [r2,:64]!     @  tcos[i]
+        veor            d7,  d23, d16           @ -od.im, od.re
+        vld1.32         {d5},     [r3,:64]!     @  tsin[i]
+        veor            d24, d22, d17           @  ev.re,-ev.im
+        vrev64.32       d3,  d23                @  od.re, od.im
+        pld             [r2, #32]
+        veor            d2,  d3,  d16           @ -od.re, od.im
+        pld             [r3, #32]
+        vmla.f32        d22, d3,  d4[0]
+        vmla.f32        d22, d7,  d5[0]
+        vmla.f32        d24, d2,  d4[0]
+        vmla.f32        d24, d23, d5[0]
+        vld1.32         {d0},     [r0,:64]!     @  d1[0,1]
+        vld1.32         {d1},     [r1,:64], r8  @  d2[0,1]
+        vst1.32         {d20},    [r7,:64]!
+        vst1.32         {d6},     [lr,:64], r8
+        vst1.32         {d22},    [r7,:64]!
+        vst1.32         {d24},    [lr,:64], r8
+        subs            r12, r12, #2
+        bgt             2b
+
+        veor            q1,  q0,  q8            @ -d1[0],d1[1], d2[0],-d2[1]
+        vadd.f32        d0,  d0,  d3            @  d1[0]+d2[0], d1[1]-d2[1]
+        vadd.f32        d1,  d2,  d1            @ -d1[0]+d2[0], d1[1]+d2[1]
+        ldr             r2,  [r4, #8]           @  sign_convention
+        vmul.f32        q10, q0,  q9            @  ev.re, ev.im, od.im, od.re
+        add             r0,  r0,  #4
+        bfc             r2,  #0,  #31
+        vld1.32         {d0[0]},  [r0,:32]
+        veor            d7,  d21, d16           @ -od.im, od.re
+        vrev64.32       d3,  d21                @  od.re, od.im
+        veor            d6,  d20, d17           @  ev.re,-ev.im
+        vld1.32         {d22},    [r5,:64]
+        vdup.32         d1,  r2
+        vmov            d23, d22
+        veor            d2,  d3,  d16           @ -od.re, od.im
+        vtrn.32         d22, d23
+        veor            d0,  d0,  d1
+        veor            d23, d23, d17
+        vmla.f32        d20, d3,  d4[1]
+        vmla.f32        d20, d7,  d5[1]
+        vmla.f32        d6,  d2,  d4[1]
+        vmla.f32        d6,  d21, d5[1]
+        vadd.f32        d22, d22, d23
+        vst1.32         {d20},    [r7,:64]
+        vst1.32         {d6},     [lr,:64]
+        vst1.32         {d0[0]},  [r0,:32]
+        vst1.32         {d22},    [r5,:64]
+
+        cmp             r6,  #0
+        popeq           {r4-r8,pc}
+
+        vmul.f32        d22, d22, d18
+        vst1.32         {d22},    [r5,:64]
+        add             r0,  r4,  #20
+        mov             r1,  r5
+        bl              ff_fft_permute_neon
+        add             r0,  r4,  #20
+        mov             r1,  r5
+        pop             {r4-r8,lr}
+        b               ff_fft_calc_neon
+endfunc
--- a/fft.h	Tue Mar 23 02:17:04 2010 +0000
+++ b/fft.h	Tue Mar 23 03:35:02 2010 +0000
@@ -207,6 +207,8 @@
 int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans);
 void ff_rdft_end(RDFTContext *s);
 
+void ff_rdft_init_arm(RDFTContext *s);
+
 static av_always_inline void ff_rdft_calc(RDFTContext *s, FFTSample *data)
 {
     s->rdft_calc(s, data);
--- a/rdft.c	Tue Mar 23 02:17:04 2010 +0000
+++ b/rdft.c	Tue Mar 23 03:35:02 2010 +0000
@@ -121,6 +121,9 @@
     }
 #endif
     s->rdft_calc   = ff_rdft_calc_c;
+
+    if (ARCH_ARM) ff_rdft_init_arm(s);
+
     return 0;
 }