/* $NetBSD: dtv_math.c,v 1.5 2011/08/09 01:42:24 jmcneill Exp $ */

/*-
* Copyright (c) 2011 Alan Barrett <[email protected]>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/

#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: dtv_math.c,v 1.5 2011/08/09 01:42:24 jmcneill Exp $");

#include <sys/types.h>
#include <sys/bitops.h>
#include <sys/module.h>

#include <dev/dtv/dtv_math.h>

/*
* dtv_intlog10 -- return an approximation to log10(x) * 1<<24,
* using integer arithmetic.
*
* As a special case, returns 0 when x == 0.  The mathematical
* result is -infinity.
*
* This function uses 0.5 + x/2 - 1/x as an approximation to
* log2(x) for x in the range [1.0, 2.0], and scales the input value
* to fit this range.  The resulting error is always better than
* 0.2%.
*
* Here's a table of the desired and actual results, as well
* as the absolute and relative errors, for several values of x.
*
*           x     desired      actual     err_abs err_rel
*           0           0           0          +0 +0.00000
*           1           0           0          +0 +0.00000
*           2     5050445     5050122        -323 -0.00006
*           3     8004766     7996348       -8418 -0.00105
*           4    10100890    10100887          -3 -0.00000
*           5    11726770    11741823      +15053 +0.00128
*           6    13055211    13046470       -8741 -0.00067
*           7    14178392    14158860      -19532 -0.00138
*           8    15151335    15151009        -326 -0.00002
*           9    16009532    16028061      +18529 +0.00116
*          10    16777216    16792588      +15372 +0.00092
*          11    17471670    17475454       +3784 +0.00022
*          12    18105656    18097235       -8421 -0.00047
*          13    18688868    18672077      -16791 -0.00090
*          14    19228837    19209625      -19212 -0.00100
*          15    19731537    19717595      -13942 -0.00071
*          16    20201781    20201774          -7 -0.00000
*          20    21827661    21842710      +15049 +0.00069
*          24    23156102    23147357       -8745 -0.00038
*          30    24781982    24767717      -14265 -0.00058
*          40    26878106    26893475      +15369 +0.00057
*          60    29832427    29818482      -13945 -0.00047
*         100    33554432    33540809      -13623 -0.00041
*        1000    50331648    50325038       -6610 -0.00013
*       10000    67108864    67125985      +17121 +0.00026
*      100000    83886080    83875492      -10588 -0.00013
*     1000000   100663296   100652005      -11291 -0.00011
*    10000000   117440512   117458739      +18227 +0.00016
*   100000000   134217728   134210175       -7553 -0.00006
*  1000000000   150994944   150980258      -14686 -0.00010
*  4294967295   161614248   161614192         -56 -0.00000
*/
uint32_t
dtv_intlog10(uint32_t x)
{
       uint32_t ilog2x;
       uint32_t t;
       uint32_t t1;

       if (__predict_false(x == 0))
               return 0;

       /*
        * find ilog2x = floor(log2(x)), as an integer in the range [0,31].
        */
       ilog2x = ilog2(x);

       /*
        * Set "t" to the result of shifting x left or right
        * until the most significant bit that was actually set
        * moves into the 1<<24 position.
        *
        * Now we can think of "t" as representing
        * x / 2**(floor(log2(x))),
        * as a fixed-point value with 8 integer bits and 24 fraction bits.
        *
        * This value is in the semi-closed interval [1.0, 2.0)
        * when interpreting it as a fixed-point number, or in the
        * interval [0x01000000, 0x01ffffff] when examining the
        * underlying uint32_t representation.
        */
       t = (ilog2x > 24 ? x >> (ilog2x - 24) : x << (24 - ilog2x));

       /*
        * Calculate "t1 = 1 / t" in the 8.24 fixed-point format.
        * This value is in the interval [0.5, 1.0]
        * when interpreting it as a fixed-point number, or in the
        * interval [0x00800000, 0x01000000] when examining the
        * underlying uint32_t representation.
        *
        */
       t1 = ((uint64_t)1 << 48) / t;

       /*
        * Calculate "t = ilog2x + t/2 - t1 + 0.5" in the 8.24
        * fixed-point format.
        *
        * If x is a power of 2, then t is now exactly equal to log2(x)
        * when interpreting it as a fixed-point number, or exactly
        * log2(x) << 24 when examining the underlying uint32_t
        * representation.
        *
        * If x is not a power of 2, then t is the result of
        * using the function x/2 - 1/x + 0.5 as an approximation for
        * log2(x) for x in the range [1, 2], and scaling both the
        * input and the result by the appropriate number of powers of 2.
        */
       t = (ilog2x << 24) + (t >> 1) - t1 + (1 << 23);

       /*
        * Multiply t by log10(2) to get the final result.
        *
        * log10(2) is approximately 643/2136  We divide before
        * multiplying to avoid overflow.
        */
       return t / 2136 * 643;
}

#ifdef _KERNEL
MODULE(MODULE_CLASS_MISC, dtv_math, NULL);

static int
dtv_math_modcmd(modcmd_t cmd, void *opaque)
{
       if (cmd == MODULE_CMD_INIT || cmd == MODULE_CMD_FINI)
               return 0;
       return ENOTTY;
}
#endif

#ifdef TEST_DTV_MATH
/*
* To test:
*      cc -DTEST_DTV_MATH ./dtv_math.c -lm -o ./a.out && ./a.out
*/

#include <stdio.h>
#include <inttypes.h>
#include <math.h>

int
main(void)
{
       uint32_t xlist[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,
                       14, 15, 16, 20, 24, 30, 40, 60, 100, 1000, 10000,
                       100000, 1000000, 10000000, 100000000, 1000000000,
                       0xffffffff};
       int i;

       printf("%11s %11s %11s %11s %s\n",
               "x", "desired", "actual", "err_abs", "err_rel");
       for (i = 0; i < __arraycount(xlist); i++)
       {
               uint32_t x = xlist[i];
               uint32_t desired = (uint32_t)(log10((double)x)
                                               * (double)(1<<24));
               uint32_t actual = dtv_intlog10(x);
               int32_t err_abs = actual - desired;
               double err_rel = (err_abs == 0 ? 0.0
                               : err_abs / (double)actual);

               printf("%11"PRIu32" %11"PRIu32" %11"PRIu32
                       " %+11"PRId32" %+.5f\n",
                       x, desired, actual, err_abs, err_rel);
       }
       return 0;
}

#endif /* TEST_DTV_MATH */