remove a bunch of uses of long (mostly replaced by int32_t)
[ardour2.git] / libs / ardour / sse_functions_xmm.cc
blobb3e3342cb8951dc5da6bc49cef479178ef034bc0
1 /*
2 Copyright (C) 2007 Paul sDavis
3 Written by Sampo Savolainen
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2 of the License, or
8 (at your option) any later version.
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 #include <xmmintrin.h>
22 #include "ardour/types.h"
24 void
25 x86_sse_find_peaks(const ARDOUR::Sample* buf, ARDOUR::nframes_t nframes, float *min, float *max)
27 __m128 current_max, current_min, work;
29 // Load max and min values into all four slots of the XMM registers
30 current_min = _mm_set1_ps(*min);
31 current_max = _mm_set1_ps(*max);
33 // Work input until "buf" reaches 16 byte alignment
34 while ( ((intptr_t)buf) % 16 != 0 && nframes > 0) {
36 // Load the next float into the work buffer
37 work = _mm_set1_ps(*buf);
39 current_min = _mm_min_ps(current_min, work);
40 current_max = _mm_max_ps(current_max, work);
42 buf++;
43 nframes--;
46 // use 64 byte prefetch for quadruple quads
47 while (nframes >= 16) {
48 __builtin_prefetch(buf+64,0,0);
50 work = _mm_load_ps(buf);
51 current_min = _mm_min_ps(current_min, work);
52 current_max = _mm_max_ps(current_max, work);
53 buf+=4;
54 work = _mm_load_ps(buf);
55 current_min = _mm_min_ps(current_min, work);
56 current_max = _mm_max_ps(current_max, work);
57 buf+=4;
58 work = _mm_load_ps(buf);
59 current_min = _mm_min_ps(current_min, work);
60 current_max = _mm_max_ps(current_max, work);
61 buf+=4;
62 work = _mm_load_ps(buf);
63 current_min = _mm_min_ps(current_min, work);
64 current_max = _mm_max_ps(current_max, work);
65 buf+=4;
66 nframes-=16;
69 // work through aligned buffers
70 while (nframes >= 4) {
72 work = _mm_load_ps(buf);
74 current_min = _mm_min_ps(current_min, work);
75 current_max = _mm_max_ps(current_max, work);
77 buf+=4;
78 nframes-=4;
81 // work through the rest < 4 samples
82 while ( nframes > 0) {
84 // Load the next float into the work buffer
85 work = _mm_set1_ps(*buf);
87 current_min = _mm_min_ps(current_min, work);
88 current_max = _mm_max_ps(current_max, work);
90 buf++;
91 nframes--;
94 // Find min & max value in current_max through shuffle tricks
96 work = current_min;
97 work = _mm_shuffle_ps(work, work, _MM_SHUFFLE(2, 3, 0, 1));
98 work = _mm_min_ps (work, current_min);
99 current_min = work;
100 work = _mm_shuffle_ps(work, work, _MM_SHUFFLE(1, 0, 3, 2));
101 work = _mm_min_ps (work, current_min);
103 _mm_store_ss(min, work);
105 work = current_max;
106 work = _mm_shuffle_ps(work, work, _MM_SHUFFLE(2, 3, 0, 1));
107 work = _mm_max_ps (work, current_max);
108 current_max = work;
109 work = _mm_shuffle_ps(work, work, _MM_SHUFFLE(1, 0, 3, 2));
110 work = _mm_max_ps (work, current_max);
112 _mm_store_ss(max, work);