[Bug c/30013] New: Multiple flaws in decimal floating-point arithmetic conversions fixed

[Bug c/30013] New: Multiple flaws in decimal floating-point arithmetic conversions fixed

[beebe at math dot utah dot edu]

Testing of my decimal floating-point arithmetic code with gcc-4.2 and
gcc-4.3 releases turned up several bugs in gcc's type conversions that
I've had to track down and fix.

They repair several problems:

(1) Loss of trailing digits by failure to provide a suitable precision
    in conversion formats; I use the values from the formula in David
    Matula's CACM 19(3) 716--723 June 1968 paper, expressed by this
    code snippet:

        ### matula(nbits): return number of decimal digits needed to ensure
        ### correct round-trip conversion between binary and decimal
        func matula(nbits) { return ceil(nbits/log2(10) + 1) }

(2) Failure to provide prototypes of library functions that are not
    necessarily supplied by system header files: if the gcc developers
    would routinely use the gcc -Werror-implicit-function-declaration
    option in builds, this problem could have, and should have, been
    caught before release.

(3) Incorrect data types for conversion functions.

(4) Failure to match sprintf() argument types with format items.

(5) Serious loss of precision for _Decimal128 by using double strtod()
    instead of long double strtold().

The test file given below can be used to check the conversions before
and after the patch.

Here are diffs suitable for input to patch:

103 airy> diff gcc/config/dfp-bit.c.org gcc/config/dfp-bit.c
516c516
<   sprintf (buf, BFP_FMT, x);
---
>   sprintf (buf, BFP_FMT, (double)x);

104 airy> diff gcc/config/dfp-bit.h.org gcc/config/dfp-bit.h
241c241
< #define BFP_FMT "%e"
---
> #define BFP_FMT "%.9e"
246c246
< #define BFP_FMT "%e"
---
> #define BFP_FMT "%.17e"
253,255c253,257
< #define BFP_FMT "%e"
< #define BFP_VIA_TYPE double
< #define STR_TO_BFP strtod
---
> #define BFP_FMT "%.36Le"
> #define BFP_VIA_TYPE long double
> /* strtold is declared in <stdlib.h> only for C99.  */
> extern long double strtold (const char *, char **);
> #define STR_TO_BFP strtold
422c424
< typedef float DFtype __attribute__ ((mode (DF)));
---
> typedef double DFtype __attribute__ ((mode (DF)));
424c426
< typedef float XFtype __attribute__ ((mode (XF)));
---
> typedef long double XFtype __attribute__ ((mode (XF)));

Here is the test file:

% cat dec-cast-bug.c
/***********************************************************************
[23-Nov-2006]
***********************************************************************/

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

void
test(long double x)
{
    float f;
    double d;
    long double y;
    _Decimal32 d32;
    _Decimal64 d64;
    _Decimal128 d128;

    d32 = (_Decimal32)x;
    (void)printf(" (_Decimal32)(%.21Lg) = 0x%08x\n", x,
                 ((unsigned int *)(&d32))[0]);

    d64 = (_Decimal64)x;
    (void)printf(" (_Decimal64)(%.21Lg) = 0x%08x_%08x\n", x,
                 ((unsigned int *)(&d64))[1],
                 ((unsigned int *)(&d64))[0]);

    d128 = (_Decimal128)x;
    (void)printf("(_Decimal128)(%.21Lg) = 0x%08x_%08x_%08x_%08x\n", x,
                 ((unsigned int *)(&d128))[3],
                 ((unsigned int *)(&d128))[2],
                 ((unsigned int *)(&d128))[1],
                 ((unsigned int *)(&d128))[0]);

    (void)printf("\n");

    f = (float)x;
    (void)printf("      (float)(%.21Lg) = %.9g             = %a\n", x,
(double)f, (double)f);

    d = (double)x;
    (void)printf("     (double)(%.21Lg) = %.17g     = %a\n", x, (double)d,
(double)d);

    y = (long double)x;
    (void)printf("(long double)(%.21Lg) = %.21Lg = %La\n", x, (long double)y,
(long double)y);

    (void)printf("\n");

    y = (long double)(float)x;
    (void)printf("      (long double)(float)(%.21Lg) = %.21Lg = %La\n", x, y,
y);

    y = (long double)(double)x;
    (void)printf("     (long double)(double)(%.21Lg) = %.21Lg = %La\n", x, y,
y);

    y = (long double)(long double)x;
    (void)printf("(long double)(long double)(%.21Lg) = %.21Lg = %La\n", x, y,
y);

    (void)printf("\n");

    f = (float)(_Decimal32)x;
    (void)printf(" (float)(_Decimal32)(%.21Lg) = %.9g = %a\n", x, (double)f,
(double)f);

    f = (float)(_Decimal64)x;
    (void)printf(" (float)(_Decimal64)(%.21Lg) = %.9g = %a\n", x, (double)f,
(double)f);

    f = (float)(_Decimal128)x;
    (void)printf("(float)(_Decimal128)(%.21Lg) = %.9g = %a\n", x, (double)f,
(double)f);

    (void)printf("\n");

    d = (double)(_Decimal32)x;
    (void)printf(" (double)(_Decimal32)(%.21Lg) = %.17g = %a\n", x, (double)d,
(double)d);

    d = (double)(_Decimal64)x;
    (void)printf(" (double)(_Decimal64)(%.21Lg) = %.17g = %a\n", x, (double)d,
(double)d);

    d = (double)(_Decimal128)x;
    (void)printf("(double)(_Decimal128)(%.21Lg) = %.17g = %a\n", x, (double)d,
(double)d);

    (void)printf("\n");

    y = (long double)(_Decimal32)x;
    (void)printf(" (long double)(_Decimal32)(%.21Lg) = %.21Lg = %La\n", x,
(long double)y, (long double)y);

    y = (long double)(_Decimal64)x;
    (void)printf(" (long double)(_Decimal64)(%.21Lg) = %.21Lg = %La\n", x,
(long double)y, (long double)y);

    y = (long double)(_Decimal128)x;
    (void)printf("(long double)(_Decimal128)(%.21Lg) = %.21Lg = %La\n", x,
(long double)y, (long double)y);

    (void)printf("\n");

    y = (long double)(float)(_Decimal32)x;
    (void)printf(" (long double)(float)(_Decimal32)(%.21Lg) = %.21Lg = %La\n",
x, y, y);

    y = (long double)(float)(_Decimal64)x;
    (void)printf(" (long double)(float)(_Decimal64)(%.21Lg) = %.21Lg = %La\n",
x, y, y);

    y = (long double)(float)(_Decimal128)x;
    (void)printf("(long double)(float)(_Decimal128)(%.21Lg) = %.21Lg = %La\n",
x, y, y);

    (void)printf("\n");

    y = (long double)(double)(_Decimal32)x;
    (void)printf(" (long double)(double)(_Decimal32)(%.21Lg) = %.21Lg = %La\n",
x, y, y);

    y = (long double)(double)(_Decimal64)x;
    (void)printf(" (long double)(double)(_Decimal64)(%.21Lg) = %.21Lg = %La\n",
x, y, y);

    y = (long double)(double)(_Decimal128)x;
    (void)printf("(long double)(double)(_Decimal128)(%.21Lg) = %.21Lg = %La\n",
x, y, y);

    (void)printf("\n");

    y = (long double)(_Decimal32)x;
    (void)printf(" (long double)(_Decimal32)(%.21Lg) = %.21Lg = %La\n", x, y,
y);

    y = (long double)(_Decimal64)x;
    (void)printf(" (long double)(_Decimal64)(%.21Lg) = %.21Lg = %La\n", x, y,
y);

    y = (long double)(_Decimal128)x;
    (void)printf("(long double)(_Decimal128)(%.21Lg) = %.21Lg = %La\n", x, y,
y);
}

int
main(int argc, char* argv[])
{
    test(3.141592653589793238462643383279502884L);

    return (EXIT_SUCCESS);
}


-- 
           Summary: Multiple flaws in decimal floating-point arithmetic
                    conversions fixed
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: beebe at math dot utah dot edu
 GCC build triplet: x86_64-unknown-linux-gnu
  GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: x86_64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-11-29 02:13 -------
422c424
< typedef float DFtype __attribute__ ((mode (DF)));
---
> typedef double DFtype __attribute__ ((mode (DF)));
424c426
< typedef float XFtype __attribute__ ((mode (XF)));
---
> typedef long double XFtype __attribute__ ((mode (XF)));



This part really should not matter as mode tells the compiler what the real
type is.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2006-11-29 02:13 -------
> /* strtold is declared in <stdlib.h> only for C99.  */
> extern long double strtold (const char *, char **);
> #define STR_TO_BFP strtold

Well use -std=gnu99 then.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[bje at gcc dot gnu dot org]

-- 

bje at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |bje at gcc dot gnu dot org
                   |dot org                     |
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2006-11-29 02:45:38
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2006-11-29 02:57 -------
PS the testcase violates C aliasing rules.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[janis at gcc dot gnu dot org]

------- Comment #4 from janis at gcc dot gnu dot org  2007-08-17 00:40 -------
Ben just mentioned this PR to me, I hadn't seen it before.

I'm working on a patch to support TFmode for powerpc*-linux, and I'll talk to
HJ about proper support for XFmode.  Initially we didn't support long double
because  we were testing on powerpc64-linux where there was no C library
support for 128-bit long double.  Now that there is, we can support it.

Thanks for the test, I'll modify it so it doesn't violate aliasing rules and
add it to the testsuite.  Please report other problems you find.


-- 

janis at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |janis at gcc dot gnu dot org
         AssignedTo|bje at gcc dot gnu dot org  |janis at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[janis at gcc dot gnu dot org]

------- Comment #5 from janis at gcc dot gnu dot org  2007-08-17 20:44 -------
Why use "%.9e", "%.17e", and "%.36Le" to write the binary float values to a
string, instead of using lengths of FLT_DIG, DBL_DIG, and LDBL_DIG?  For
i686-linux those are 6, 15, and 18.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[janis at gcc dot gnu dot org]

------- Comment #6 from janis at gcc dot gnu dot org  2007-09-11 01:11 -------
Subject: Bug 30013

Author: janis
Date: Tue Sep 11 01:11:16 2007
New Revision: 128361

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=128361
Log:
gcc/
        PR c/30013
        * config/dfp-bit.c: Don't skip TFmode conversions; move strto*
        declarations to top.
        (DFP_TO_BFP): Use for either XFmode or TFmode.
        (BFP_TO_DFP): Use for either XFmode or TFmode; always use cast
        of BFP_VIA_TYPE.
        * config/dfp-bit.h: Include float.h.
        (LONG_DOUBLE_HAS_XF_MODE, LONG_DOUBLE_HAS_TF_MODE): Define if long
        double is one of these modes, rather than using LIBGCC_HAS_*F_MODE
        which doesn't mean the same thing.
        (BFP_KIND): Use 4 to mean TFmode.
        (BFP_FMT): Specify the number of decimal digits based on the
        number of mantissa digits.
        (BFP_VIA_TYPE): Binary float type to use as cast for sprintf.
        (BFP_TO_DFP, DFP_TO_BFP): Define names for TFmode variants.
        (STR_TO_BFP): Use strtold for XFmode or TFmode.
        (TFtype): Define if TFmode is supported.
        * doc/libgcc.texi (Decimal float library routines): Document
        TF conversion functions.

gcc/testsuite/
        * gcc.dg/dfp/convert-bfp.c: Replace SKIP_LONG_DOUBLE with runtime
        checks for size of long double.
        * gcc.dg/dfp/convert.h: New file.
        * gcc.dg/dfp/convert-bfp-2.c: New test.
        * gcc.dg/dfp/convert-bfp-3.c: Ditto.
        * gcc.dg/dfp/convert-bfp-4.c: Ditto.
        * gcc.dg/dfp/convert-bfp-5.c: Ditto.
        * gcc.dg/dfp/convert-bfp-6.c: Ditto.
        * gcc.dg/dfp/convert-bfp-7.c: Ditto.
        * gcc.dg/dfp/convert-bfp-8.c: Ditto.
        * gcc.dg/dfp/convert-bfp-9.c: Ditto.
        * gcc.dg/dfp/convert-bfp-10.c: Ditto.
        * gcc.dg/dfp/convert-bfp-11.c: Ditto.

Added:
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-10.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-11.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-2.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-3.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-4.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-5.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-6.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-7.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-8.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp-9.c
    trunk/gcc/testsuite/gcc.dg/dfp/convert.h
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/dfp-bit.c
    trunk/gcc/config/dfp-bit.h
    trunk/gcc/doc/libgcc.texi
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/gcc.dg/dfp/convert-bfp.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013

[Bug c/30013] Multiple flaws in decimal floating-point arithmetic conversions fixed

[janis at gcc dot gnu dot org]

------- Comment #7 from janis at gcc dot gnu dot org  2007-10-23 16:50 -------
Fixed in what will be 4.3.0.


-- 

janis at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30013