aboutsummaryrefslogtreecommitdiff
path: root/externals/gridflow/bundled/Base/GemSIMD.h
blob: 32cdb32a2d62576d3e53b53de4e72d0a472cb071 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111

/*-----------------------------------------------------------------
LOG
    GEM - Graphics Environment for Multimedia

    include file for SIMD

    Copyright (c) 1997-1999 Mark Danks. mark@danks.org
    Copyright (c) Günther Geiger. geiger@epy.co.at
    Copyright (c) 2001-2002 IOhannes m zmoelnig. forum::für::umläute. IEM. zmoelnig@iem.kug.ac.at
    For information on usage and redistribution, and for a DISCLAIMER OF ALL
    WARRANTIES, see the file, "GEM.LICENSE.TERMS" in this distribution.

-----------------------------------------------------------------*/

/* 
 * compiler-issues:
 *
 * gcc: when gcc is invoked with "-mmmx" (or "-msse2" or "-maltivec") 
 *      the defines __MMX__ (or corresponding) will be defined automatically
 *
 * vc6: you will have to install the microsoft processor-pack to use MMX/SSE2
 *      you have to have the sp5 for vc6 installed (note: do not install sp6!!)
 * vc6/vc7: (i think) you need to define __MMX__ (and friends) by hand
 */

#ifndef INCLUDE_GEMSIMD_H_
#define INCLUDE_GEMSIMD_H_

#define GEM_VECTORALIGNMENT 128

const int GEM_SIMD_NONE=0;
const int GEM_SIMD_MMX=1;
const int GEM_SIMD_SSE2=2;
const int GEM_SIMD_ALTIVEC=3;


/* include for SIMD on PC's */
#ifdef __SSE2__
#include <emmintrin.h>
// for icc this should be <dvec.h>
typedef union{
  unsigned char c[16];
  __m128i v;
} vector_128;
#elif defined __VEC__
/* for AltiVec (PowerPC) */
typedef union{
  unsigned char c[16];
  vector unsigned char v;
} vector_128;
#endif

#if defined __MMX__
# include <mmintrin.h>
// for icc this should be <ivec.h>
typedef union{
  __m64 v; unsigned char c[8];
} vector64i;

#endif


#ifdef __SSE__
#include <xmmintrin.h>

typedef union{
  __m128 m; float f[4];
} vector128f;
#endif

#include "GemExportDef.h"


/* this is a help-class to query the capabilities of the cpu
 * whenever you want to use SIMD-optimized code, you should
 * make sure you chose the code-block based on the "cpuid" value
 * of this class and NOT simply on preprocessor defines.
 *
 * this class needs only be instantiated once (and it is done in GemMan)
 * this sets the cpuid
 */
class GEM_EXTERN GemSIMD
{
 public:
   GemSIMD();
  ~GemSIMD();

  /* this gets the "cpuid" (something like GEM_SIMD_NONE) */
  static int getCPU();

  /* change the cpuid returned by getCPU()
   * you can only set the cpuid to something that is actually supported
   * by your processor
   */
  static int requestCPU(int cpuid);

  /* performs a runtime-check (if possible) to determine the capabilities
   * of the CPU
   * sets realcpuid appropriately and returns this  value
   */
  static int simd_runtime_check(void);

 private:
  /* this is the maximum capability of the CPU */
  static int realcpuid;
  /* this is the current choosen capability (normally this equals realcpuid) */
  static int cpuid;
};

#endif /* INCLUDE_GEMSIMD_H_ */