Home | History | Annotate | Line # | Download | only in dtv
      1  1.5  jmcneill /* $NetBSD: dtv_math.c,v 1.5 2011/08/09 01:42:24 jmcneill Exp $ */
      2  1.1  jmcneill 
      3  1.1  jmcneill /*-
      4  1.1  jmcneill  * Copyright (c) 2011 Alan Barrett <apb (at) NetBSD.org>
      5  1.1  jmcneill  * All rights reserved.
      6  1.1  jmcneill  *
      7  1.1  jmcneill  * Redistribution and use in source and binary forms, with or without
      8  1.1  jmcneill  * modification, are permitted provided that the following conditions
      9  1.1  jmcneill  * are met:
     10  1.1  jmcneill  * 1. Redistributions of source code must retain the above copyright
     11  1.1  jmcneill  *    notice, this list of conditions and the following disclaimer.
     12  1.1  jmcneill  * 2. Redistributions in binary form must reproduce the above copyright
     13  1.1  jmcneill  *    notice, this list of conditions and the following disclaimer in the
     14  1.1  jmcneill  *    documentation and/or other materials provided with the distribution.
     15  1.1  jmcneill  *
     16  1.1  jmcneill  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     17  1.1  jmcneill  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     18  1.1  jmcneill  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     19  1.1  jmcneill  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     20  1.1  jmcneill  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     21  1.1  jmcneill  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     22  1.1  jmcneill  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     23  1.1  jmcneill  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     24  1.1  jmcneill  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     25  1.1  jmcneill  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     26  1.1  jmcneill  * POSSIBILITY OF SUCH DAMAGE.
     27  1.1  jmcneill  */
     28  1.1  jmcneill 
     29  1.1  jmcneill #include <sys/cdefs.h>
     30  1.5  jmcneill __KERNEL_RCSID(0, "$NetBSD: dtv_math.c,v 1.5 2011/08/09 01:42:24 jmcneill Exp $");
     31  1.1  jmcneill 
     32  1.2  jmcneill #include <sys/types.h>
     33  1.1  jmcneill #include <sys/bitops.h>
     34  1.5  jmcneill #include <sys/module.h>
     35  1.1  jmcneill 
     36  1.5  jmcneill #include <dev/dtv/dtv_math.h>
     37  1.4       apb 
     38  1.1  jmcneill /*
     39  1.1  jmcneill  * dtv_intlog10 -- return an approximation to log10(x) * 1<<24,
     40  1.1  jmcneill  * using integer arithmetic.
     41  1.1  jmcneill  *
     42  1.3       apb  * As a special case, returns 0 when x == 0.  The mathematical
     43  1.3       apb  * result is -infinity.
     44  1.3       apb  *
     45  1.3       apb  * This function uses 0.5 + x/2 - 1/x as an approximation to
     46  1.3       apb  * log2(x) for x in the range [1.0, 2.0], and scales the input value
     47  1.3       apb  * to fit this range.  The resulting error is always better than
     48  1.3       apb  * 0.2%.
     49  1.1  jmcneill  *
     50  1.3       apb  * Here's a table of the desired and actual results, as well
     51  1.3       apb  * as the absolute and relative errors, for several values of x.
     52  1.3       apb  *
     53  1.3       apb  *           x     desired      actual     err_abs err_rel
     54  1.3       apb  *           0           0           0          +0 +0.00000
     55  1.3       apb  *           1           0           0          +0 +0.00000
     56  1.3       apb  *           2     5050445     5050122        -323 -0.00006
     57  1.3       apb  *           3     8004766     7996348       -8418 -0.00105
     58  1.3       apb  *           4    10100890    10100887          -3 -0.00000
     59  1.3       apb  *           5    11726770    11741823      +15053 +0.00128
     60  1.3       apb  *           6    13055211    13046470       -8741 -0.00067
     61  1.3       apb  *           7    14178392    14158860      -19532 -0.00138
     62  1.3       apb  *           8    15151335    15151009        -326 -0.00002
     63  1.3       apb  *           9    16009532    16028061      +18529 +0.00116
     64  1.3       apb  *          10    16777216    16792588      +15372 +0.00092
     65  1.3       apb  *          11    17471670    17475454       +3784 +0.00022
     66  1.3       apb  *          12    18105656    18097235       -8421 -0.00047
     67  1.3       apb  *          13    18688868    18672077      -16791 -0.00090
     68  1.3       apb  *          14    19228837    19209625      -19212 -0.00100
     69  1.3       apb  *          15    19731537    19717595      -13942 -0.00071
     70  1.3       apb  *          16    20201781    20201774          -7 -0.00000
     71  1.3       apb  *          20    21827661    21842710      +15049 +0.00069
     72  1.3       apb  *          24    23156102    23147357       -8745 -0.00038
     73  1.3       apb  *          30    24781982    24767717      -14265 -0.00058
     74  1.3       apb  *          40    26878106    26893475      +15369 +0.00057
     75  1.3       apb  *          60    29832427    29818482      -13945 -0.00047
     76  1.3       apb  *         100    33554432    33540809      -13623 -0.00041
     77  1.3       apb  *        1000    50331648    50325038       -6610 -0.00013
     78  1.3       apb  *       10000    67108864    67125985      +17121 +0.00026
     79  1.3       apb  *      100000    83886080    83875492      -10588 -0.00013
     80  1.3       apb  *     1000000   100663296   100652005      -11291 -0.00011
     81  1.3       apb  *    10000000   117440512   117458739      +18227 +0.00016
     82  1.3       apb  *   100000000   134217728   134210175       -7553 -0.00006
     83  1.3       apb  *  1000000000   150994944   150980258      -14686 -0.00010
     84  1.3       apb  *  4294967295   161614248   161614192         -56 -0.00000
     85  1.1  jmcneill  */
     86  1.1  jmcneill uint32_t
     87  1.1  jmcneill dtv_intlog10(uint32_t x)
     88  1.1  jmcneill {
     89  1.3       apb 	uint32_t ilog2x;
     90  1.3       apb 	uint32_t t;
     91  1.3       apb 	uint32_t t1;
     92  1.3       apb 
     93  1.1  jmcneill 	if (__predict_false(x == 0))
     94  1.1  jmcneill 		return 0;
     95  1.3       apb 
     96  1.1  jmcneill 	/*
     97  1.3       apb 	 * find ilog2x = floor(log2(x)), as an integer in the range [0,31].
     98  1.1  jmcneill 	 */
     99  1.3       apb 	ilog2x = ilog2(x);
    100  1.3       apb 
    101  1.3       apb 	/*
    102  1.3       apb 	 * Set "t" to the result of shifting x left or right
    103  1.3       apb 	 * until the most significant bit that was actually set
    104  1.3       apb 	 * moves into the 1<<24 position.
    105  1.3       apb 	 *
    106  1.3       apb 	 * Now we can think of "t" as representing
    107  1.3       apb 	 * x / 2**(floor(log2(x))),
    108  1.3       apb 	 * as a fixed-point value with 8 integer bits and 24 fraction bits.
    109  1.3       apb 	 *
    110  1.3       apb 	 * This value is in the semi-closed interval [1.0, 2.0)
    111  1.3       apb 	 * when interpreting it as a fixed-point number, or in the
    112  1.3       apb 	 * interval [0x01000000, 0x01ffffff] when examining the
    113  1.3       apb 	 * underlying uint32_t representation.
    114  1.3       apb 	 */
    115  1.3       apb 	t = (ilog2x > 24 ? x >> (ilog2x - 24) : x << (24 - ilog2x));
    116  1.3       apb 
    117  1.3       apb 	/*
    118  1.3       apb 	 * Calculate "t1 = 1 / t" in the 8.24 fixed-point format.
    119  1.3       apb 	 * This value is in the interval [0.5, 1.0]
    120  1.3       apb 	 * when interpreting it as a fixed-point number, or in the
    121  1.3       apb 	 * interval [0x00800000, 0x01000000] when examining the
    122  1.3       apb 	 * underlying uint32_t representation.
    123  1.3       apb 	 *
    124  1.3       apb 	 */
    125  1.3       apb 	t1 = ((uint64_t)1 << 48) / t;
    126  1.3       apb 
    127  1.3       apb 	/*
    128  1.3       apb 	 * Calculate "t = ilog2x + t/2 - t1 + 0.5" in the 8.24
    129  1.3       apb 	 * fixed-point format.
    130  1.3       apb 	 *
    131  1.3       apb 	 * If x is a power of 2, then t is now exactly equal to log2(x)
    132  1.3       apb 	 * when interpreting it as a fixed-point number, or exactly
    133  1.3       apb 	 * log2(x) << 24 when examining the underlying uint32_t
    134  1.3       apb 	 * representation.
    135  1.3       apb 	 *
    136  1.3       apb 	 * If x is not a power of 2, then t is the result of
    137  1.3       apb 	 * using the function x/2 - 1/x + 0.5 as an approximation for
    138  1.3       apb 	 * log2(x) for x in the range [1, 2], and scaling both the
    139  1.3       apb 	 * input and the result by the appropriate number of powers of 2.
    140  1.3       apb 	 */
    141  1.3       apb 	t = (ilog2x << 24) + (t >> 1) - t1 + (1 << 23);
    142  1.3       apb 
    143  1.3       apb 	/*
    144  1.3       apb 	 * Multiply t by log10(2) to get the final result.
    145  1.3       apb 	 *
    146  1.3       apb 	 * log10(2) is approximately 643/2136  We divide before
    147  1.3       apb 	 * multiplying to avoid overflow.
    148  1.3       apb 	 */
    149  1.3       apb 	return t / 2136 * 643;
    150  1.1  jmcneill }
    151  1.3       apb 
    152  1.5  jmcneill #ifdef _KERNEL
    153  1.5  jmcneill MODULE(MODULE_CLASS_MISC, dtv_math, NULL);
    154  1.5  jmcneill 
    155  1.5  jmcneill static int
    156  1.5  jmcneill dtv_math_modcmd(modcmd_t cmd, void *opaque)
    157  1.5  jmcneill {
    158  1.5  jmcneill 	if (cmd == MODULE_CMD_INIT || cmd == MODULE_CMD_FINI)
    159  1.5  jmcneill 		return 0;
    160  1.5  jmcneill 	return ENOTTY;
    161  1.5  jmcneill }
    162  1.5  jmcneill #endif
    163  1.5  jmcneill 
    164  1.3       apb #ifdef TEST_DTV_MATH
    165  1.3       apb /*
    166  1.3       apb  * To test:
    167  1.3       apb  *	cc -DTEST_DTV_MATH ./dtv_math.c -lm -o ./a.out && ./a.out
    168  1.3       apb  */
    169  1.3       apb 
    170  1.3       apb #include <stdio.h>
    171  1.3       apb #include <inttypes.h>
    172  1.3       apb #include <math.h>
    173  1.3       apb 
    174  1.3       apb int
    175  1.3       apb main(void)
    176  1.3       apb {
    177  1.3       apb 	uint32_t xlist[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,
    178  1.3       apb 			14, 15, 16, 20, 24, 30, 40, 60, 100, 1000, 10000,
    179  1.3       apb 			100000, 1000000, 10000000, 100000000, 1000000000,
    180  1.3       apb 			0xffffffff};
    181  1.3       apb 	int i;
    182  1.3       apb 
    183  1.3       apb 	printf("%11s %11s %11s %11s %s\n",
    184  1.3       apb 		"x", "desired", "actual", "err_abs", "err_rel");
    185  1.3       apb 	for (i = 0; i < __arraycount(xlist); i++)
    186  1.3       apb 	{
    187  1.3       apb 		uint32_t x = xlist[i];
    188  1.3       apb 		uint32_t desired = (uint32_t)(log10((double)x)
    189  1.3       apb 						* (double)(1<<24));
    190  1.3       apb 		uint32_t actual = dtv_intlog10(x);
    191  1.3       apb 		int32_t err_abs = actual - desired;
    192  1.3       apb 		double err_rel = (err_abs == 0 ? 0.0
    193  1.3       apb 				: err_abs / (double)actual);
    194  1.3       apb 
    195  1.3       apb 		printf("%11"PRIu32" %11"PRIu32" %11"PRIu32
    196  1.3       apb 			" %+11"PRId32" %+.5f\n",
    197  1.3       apb 			x, desired, actual, err_abs, err_rel);
    198  1.3       apb 	}
    199  1.3       apb 	return 0;
    200  1.3       apb }
    201  1.3       apb 
    202  1.3       apb #endif /* TEST_DTV_MATH */
    203