Large floating point literals

ig has asked for the wisdom of the Perl Monks concerning the following question:

I would like to better understand Perl's ability and limitations processing floating point numbers. Does anyone have suggestions for my reading list?

I believe 1.79769313486232e+308 is the largest IEEE-754 double precision floating point number. Perl (32 bit on CentOS on Intel based system) can store this value in a scalar and print it but does not accept it as a numeric literal.

use strict;
use warnings;
use Devel::Peek;

my $foo = pack("H16", "FFFFFFFFFFFFEF7F" );
Dump($foo);

my $bar = unpack("d", $foo);
Dump($bar);

my $x = $bar;
Dump($x);

my $y = 1.79769313486232e+308;
Dump($y);

__END__
SV = PV(0x99490a0) at 0x995b760
  REFCNT = 1
  FLAGS = (PADMY,POK,pPOK)
  PV = 0x9956ba8 "\377\377\377\377\377\377\357\177"\0
  CUR = 8
  LEN = 12
SV = NV(0x99832d8) at 0x995b7b0
  REFCNT = 1
  FLAGS = (PADMY,NOK,pNOK)
  NV = 1.79769313486232e+308
SV = NV(0x99832d0) at 0x995b740
  REFCNT = 1
  FLAGS = (PADMY,NOK,pNOK)
  NV = 1.79769313486232e+308
SV = NV(0x99832e0) at 0x9968360
  REFCNT = 1
  FLAGS = (PADMY,NOK,pNOK)
  NV = inf
[download]

As you can see from the above example, pack and unpack can be used to get this maximal value into a scalar and this value can be assigned from one scalar to another but using this value as a literal results in 'inf' instead of the expected value.

Any comments on why this is so?

update: The value of the largest IEEE-754 double precision floating point number, expressed as a base 10 integer is:

1797693134862315708145274237317043567980705675258449965989174768031572
+607800285387605895586327668781715404589535143824642343213268894641827
+684675467035375169860499105765512820762454900903893289440758685084551
+339423045832369032229481658085593321233482747978262041447231687381771
+80919299881250404026184124858368
[download]

The precision of an IEEE-754 double precision floating point value is approximately 16 decimal digits. Therefore, many numeric literals will be converted to this maximal value, including 1.7976931348623157e+308

When this maximal value is interpolated into a string, it is rounded to 15 digits of precision and the value it is rounded up to, when used as a numeric literal, is converted to inf rather than back to the maximum value. This value is: 1.79769313486232e+308.

Comment on Large floating point literals Select or Download Code

Replies are listed 'Best First'.
Re: Large floating point literals by BrowserUk (Patriarch) on May 20, 2009 at 21:20 UTC
You're slightly wrong. The values you are seeing displayed are subject to rounding. The biggest is: `$double = 1.797693134862314e+308;` [download] Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error. "Science is about questioning the status quo. Questioning authority". In the absence of evidence, opinion is indistinguishable from prejudice. "Too many [] have been sedated by an oppressive environment of political correctness and risk aversion."	[reply] [d/l]
Re^2: Large floating point literals by ig (Vicar) on May 20, 2009 at 22:03 UTC
Thanks!! I had, foolishly, not considered that the displayed base 10 might be rounded. Several literals slightly larger than 1.797693134862314e+308 are converted to non-infinite numbers, including two in the following test that are converted to what I believe is the largest possible value. Read more... (7 kB)	[reply] [d/l]
Re^3: Large floating point literals by BrowserUk (Patriarch) on May 20, 2009 at 23:22 UTC
Indeed. You are correct. I chose the wrong reference: `printf "%.20g\n", unpack 'd', scalar reverse pack 'NN', 0x7fefffff, 0x +ffffffff;; 1.7976931348623157e+308 printf "%.20g\n", unpack 'd', scalar reverse pack 'NN', 0x7fffffff, 0x +ffffffff;; 1.#QNAN printf "%.20g\n", unpack 'd', scalar reverse pack 'NN', 0xffefffff, 0x +ffffffff;; -1.7976931348623157e+308 printf "%.20g\n", unpack 'd', scalar reverse pack 'NN', 0xffffffff, 0x +ffffffff;; -1.#QNAN` [download] Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error. "Science is about questioning the status quo. Questioning authority". In the absence of evidence, opinion is indistinguishable from prejudice. "Too many [] have been sedated by an oppressive environment of political correctness and risk aversion."	[reply] [d/l]
Re^4: Large floating point literals by ig (Vicar) on May 21, 2009 at 00:53 UTC
Re^5: Large floating point literals by BrowserUk (Patriarch) on May 21, 2009 at 01:21 UTC
Re: Large floating point literals by zwon (Abbot) on May 20, 2009 at 21:22 UTC
The same on 64 bit Ubuntu, but the following gives me desired number: `$ perl -E'my $a=-1.797693134862314e+308; say $a;' -1.79769313486232e+308` [download] I think while converting number into binary representation perl (actually underlying c library) may loose precision, that's why you getting inf	[reply] [d/l]