src/util/u_math.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2008 VMware, Inc.
   4  * All Rights Reserved.
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a
   7  * copy of this software and associated documentation files (the
   8  * "Software"), to deal in the Software without restriction, including
   9  * without limitation the rights to use, copy, modify, merge, publish,
  10  * distribute, sub license, and/or sell copies of the Software, and to
  11  * permit persons to whom the Software is furnished to do so, subject to
  12  * the following conditions:
  13  *
  14  * The above copyright notice and this permission notice (including the
  15  * next paragraph) shall be included in all copies or substantial portions
  16  * of the Software.
  17  *
  18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  19  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  20  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
  21  * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
  22  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  23  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  24  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  25  *
  26  **************************************************************************/
  27
  28
  29
  30 #include "pipe/p_config.h"
  31 #include "util/u_math.h"
  32 #include "util/u_cpu_detect.h"
  33
  34 #if defined(PIPE_ARCH_SSE)
  35 #include <xmmintrin.h>
  36 /* This is defined in pmmintrin.h, but it can only be included when -msse3 is
  37  * used, so just define it here to avoid further. */
  38 #define _MM_DENORMALS_ZERO_MASK 0x0040
  39 #endif
  40
  41
  42 /** 2^x, for x in [-1.0, 1.0) */
  43 float pow2_table[POW2_TABLE_SIZE];
  44
  45
  46 static void
  47 init_pow2_table(void)
  48 {
  49    int i;
  50    for (i = 0; i < POW2_TABLE_SIZE; i++)
  51       pow2_table[i] = exp2f((i - POW2_TABLE_OFFSET) / POW2_TABLE_SCALE);
  52 }
  53
  54
  55 /** log2(x), for x in [1.0, 2.0) */
  56 float log2_table[LOG2_TABLE_SIZE];
  57
  58
  59 static void
  60 init_log2_table(void)
  61 {
  62    unsigned i;
  63    for (i = 0; i < LOG2_TABLE_SIZE; i++)
  64       log2_table[i] = (float) log2(1.0 + i * (1.0 / LOG2_TABLE_SCALE));
  65 }
  66
  67
  68 /**
  69  * One time init for math utilities.
  70  */
  71 void
  72 util_init_math(void)
  73 {
  74    static boolean initialized = FALSE;
  75    if (!initialized) {
  76       init_pow2_table();
  77       init_log2_table();
  78       initialized = TRUE;
  79    }
  80 }
  81
  82 /**
  83  * Fetches the contents of the fpstate (mxcsr on x86) register.
  84  *
  85  * On platforms without support for it just returns 0.
  86  */
  87 unsigned
  88 util_fpstate_get(void)
  89 {
  90    unsigned mxcsr = 0;
  91
  92 #if defined(PIPE_ARCH_SSE)
  93    if (util_cpu_caps.has_sse) {
  94       mxcsr = _mm_getcsr();
  95    }
  96 #endif
  97
  98    return mxcsr;
  99 }
 100
 101 /**
 102  * Make sure that the fp treats the denormalized floating
 103  * point numbers as zero.
 104  *
 105  * This is the behavior required by D3D10. OpenGL doesn't care.
 106  */
 107 unsigned
 108 util_fpstate_set_denorms_to_zero(unsigned current_mxcsr)
 109 {
 110 #if defined(PIPE_ARCH_SSE)
 111    if (util_cpu_caps.has_sse) {
 112       /* Enable flush to zero mode */
 113       current_mxcsr |= _MM_FLUSH_ZERO_MASK;
 114       if (util_cpu_caps.has_daz) {
 115          /* Enable denormals are zero mode */
 116          current_mxcsr |= _MM_DENORMALS_ZERO_MASK;
 117       }
 118       util_fpstate_set(current_mxcsr);
 119    }
 120 #endif
 121    return current_mxcsr;
 122 }
 123
 124 /**
 125  * Set the state of the fpstate (mxcsr on x86) register.
 126  *
 127  * On platforms without support for it's a noop.
 128  */
 129 void
 130 util_fpstate_set(unsigned mxcsr)
 131 {
 132 #if defined(PIPE_ARCH_SSE)
 133    if (util_cpu_caps.has_sse) {
 134       _mm_setcsr(mxcsr);
 135    }
 136 #endif
 137 }