See $INPUT_RECORD_SEPARATOR for how to read multiline records as single chunks and tr// for a quick way to count characters.

#! perl -slw
use strict;

$/ = "\n>"; # set record separator

while( <DATA> ){
    # Split the record in two
    my( $label, $data ) =  m[ (^ [^\n]+ ) \n ( .*$ ) ]sx;
    
    # Remove extraneous '>' characters
    $label =~ s[^>][];
    $data =~ s[>$][]s;
    
    # remove newlines from the data
    $data =~ tr[\n][]d;
    
    # count the characters
    # my $count = $data =~ tr[ACGT][ACGT]; # Corrected the omission of
+ $data from this line.

    my $count = $data =~ tr[ACGT][ACGT];
    
    print "$label contains $count characters";
}

=output

P:\test>321523
gb|AE008689|:2073051-2073458, Atu4895 contains 408 characters
gb|AE008689|:c2074151-2073549, Atu4896 contains 603 characters
gb|AE008689|:c2074749-2074345, Atu4897 contains 405 characters

=cut


__DATA__
>gb|AE008689|:2073051-2073458, Atu4895
TTGGTCACATATTTTCGCTATTCGAAACCCAAATATCCCTGCGGAACACGTTTTATTGGAGCCGCCGTTT
TGCAGTTGCTCGGTGCTGGATTCTTGGCGTTCGTCTTTTGTCTGCTGGATGGGCTCACTGCAAAACCAAC
GATAATTCTCGGTCAGTTTGTATCATGCCTAGTGGGGAGCGTCGCCGGCTTTCATTTCGTGGCTTTTCGT
CGCCCAGGCACGGACGGCCAACTTTACCTTATCGCGACGTCGCTCTTGGCATTTGGGACCCATTATTGGC
TGGTGTCATATTCATTACCTGACCTTTTGCTAGCAAGATTGATTTCAGGATTTGGATCGGGTGTGGTTGT
TGCAGGAACTTTCCGACGTCGCTTTCTGGAAAATCCGGTAATTCCCTGCGTTCGATAG
>gb|AE008689|:c2074151-2073549, Atu4896
GTGTTTGGACCATATTTTCTGTGCAAACTAAACGATGACATAGGGCGATTTTTAGTGGCGGACAAATACA
GACTTCCCGAAGAGTTTTTTACCACTCGGTTTCTCGTTAGACGCATCGTACCCACAGACGCTGAAGCTAT
TTTCGAAGGGTGGAACACCGATCCCGAGGTGACGAAGTACCTGACGTGGAAACCCCACTCCGAGCTTGGC
CAGACACAGCGGGCGATTGAAGAAAATTATAGTGCGTGGAATGCAGGTACATCGTTTCCAGCTGTCATCT
GCCATCGCGAACGGCCACATGAACTAATCGGCCGTATTGATGCACGTCCGATGGGCCACAAGGTCTCTTA
CGGGTGGCTTGTCCGAAGAACCTGGTGGGGCCGGGGTGTTGCAAGCGAGGTCGTTCAACTCGCTGTAGAA
CACGCGTTATCGCATCCGCGCATCTTTCGCACCGAAGCATCCTGCGACGTTCTGAACACGGCGTCAGCAA
GAGTGATGGAAAAAGTAGGGATGACAAAGGAGGCCGTGCTTCGACGGTACCTTTTTCACCCCAATTTTTC
GAATATGCCGCGAGACGCCTTCCTGTATTCCAAGGTACGTTAA
>gb|AE008689|:c2074749-2074345, Atu4897
ATGAAACATACCATCGCAGTTCTCGGCCTGATCACCTTCTCCAGCCCGGCCTTCGCAGCATCGTGCGAGA
AAAACTTCACCGTCTCAGGCGTACCGATGGTCACGGCTGTCTCTTACAAATCCTTTCAGGAACTGCCGAA
AGCCAAAGCACCAGCTGTCCTTCAAAAGCTCGCCCAGGCCGTCGCGGCAGAAGGTTTTTCAGGTATCCAG
ATCAACAAGGCACTGTCGTCAATCGATGCCCATCAGGAAACCAGCGGAAGTGGCAGGATTCAGACGCTGC
GGGTTGTCGCCCGCCAGAAAGGCGCCGCTGTCCGGATCGATGCTGTCTTCAATATTCAGGCAGGACAGAT
CGCCGACAAAGACGTCATCCGCAAGGGCATCTGCGACATCATAAAAGGCGCGTAA
[download]

You are including the first line when you are counting your characters so you are counting the two A's in

gb|AE008689|:2073051-2073458, Atu4895
[download]

so your counts are off by 2. If you strip off the first line you will get the right answer. You also don't need the strip off the newlines since tr[ACGT][ACGT] will only count the "ACGT" characters.

--

flounder

use strict;
use warnings;
use Data::Dumper;

my %seq_count;

local $/ = '>gb|';
while (my $seq = <DATA>) {
    $seq =~ s/>gb\|//;
    next if ! length($seq);

    # get sequence id
    my ($seq_id) = $seq =~ /\|:(.*)/;

    # remove anything other than sequence characters
    $seq =~ s/^.*\n|\n//g;

    # build hash of sequence count
    $seq_count{$seq_id} = length($seq);
}

print Dumper(\%seq_count);


__DATA__
>gb|AE008689|:2073051-2073458, Atu4895
TTGGTCACATATTTTCGCTATTCGAAACCCAAATATCCCTGCGGAACACGTTTTATTGGAGCCGCCGTTT
TGCAGTTGCTCGGTGCTGGATTCTTGGCGTTCGTCTTTTGTCTGCTGGATGGGCTCACTGCAAAACCAAC
GATAATTCTCGGTCAGTTTGTATCATGCCTAGTGGGGAGCGTCGCCGGCTTTCATTTCGTGGCTTTTCGT
CGCCCAGGCACGGACGGCCAACTTTACCTTATCGCGACGTCGCTCTTGGCATTTGGGACCCATTATTGGC
TGGTGTCATATTCATTACCTGACCTTTTGCTAGCAAGATTGATTTCAGGATTTGGATCGGGTGTGGTTGT
TGCAGGAACTTTCCGACGTCGCTTTCTGGAAAATCCGGTAATTCCCTGCGTTCGATAG
>gb|AE008689|:c2074151-2073549, Atu4896
GTGTTTGGACCATATTTTCTGTGCAAACTAAACGATGACATAGGGCGATTTTTAGTGGCGGACAAATACA
GACTTCCCGAAGAGTTTTTTACCACTCGGTTTCTCGTTAGACGCATCGTACCCACAGACGCTGAAGCTAT
TTTCGAAGGGTGGAACACCGATCCCGAGGTGACGAAGTACCTGACGTGGAAACCCCACTCCGAGCTTGGC
CAGACACAGCGGGCGATTGAAGAAAATTATAGTGCGTGGAATGCAGGTACATCGTTTCCAGCTGTCATCT
GCCATCGCGAACGGCCACATGAACTAATCGGCCGTATTGATGCACGTCCGATGGGCCACAAGGTCTCTTA
CGGGTGGCTTGTCCGAAGAACCTGGTGGGGCCGGGGTGTTGCAAGCGAGGTCGTTCAACTCGCTGTAGAA
CACGCGTTATCGCATCCGCGCATCTTTCGCACCGAAGCATCCTGCGACGTTCTGAACACGGCGTCAGCAA
GAGTGATGGAAAAAGTAGGGATGACAAAGGAGGCCGTGCTTCGACGGTACCTTTTTCACCCCAATTTTTC
GAATATGCCGCGAGACGCCTTCCTGTATTCCAAGGTACGTTAA
>gb|AE008689|:c2074749-2074345, Atu4897
ATGAAACATACCATCGCAGTTCTCGGCCTGATCACCTTCTCCAGCCCGGCCTTCGCAGCATCGTGCGAGA
AAAACTTCACCGTCTCAGGCGTACCGATGGTCACGGCTGTCTCTTACAAATCCTTTCAGGAACTGCCGAA
AGCCAAAGCACCAGCTGTCCTTCAAAAGCTCGCCCAGGCCGTCGCGGCAGAAGGTTTTTCAGGTATCCAG
ATCAACAAGGCACTGTCGTCAATCGATGCCCATCAGGAAACCAGCGGAAGTGGCAGGATTCAGACGCTGC
GGGTTGTCGCCCGCCAGAAAGGCGCCGCTGTCCGGATCGATGCTGTCTTCAATATTCAGGCAGGACAGAT
CGCCGACAAAGACGTCATCCGCAAGGGCATCTGCGACATCATAAAAGGCGCGTAA
[download]

$VAR1 = {
          'c2074151-2073549, Atu4896' => 603,
          '2073051-2073458, Atu4895' => 408,
          'c2074749-2074345, Atu4897' => 405
        };
[download]

my $x=">gb|AE004895>gb|TTGGT>gb|";
my @gene= split(/>gb\|/,$x);
foreach my $tmp (@gene)
 {
   if($tmp =~ m/\w+/)
    {
     print "Length"."=".length($tmp)."\n";
   }
}
[download]

This might point you in the right direction:

use strict;
use warnings;
my ($chars, $gene);
for (<DATA>) {
    chomp; #get rid of newline
    if (/^\>/) {
        print "$gene => $chars chars\n" if $chars;
        $chars=0;
        $gene = $_;    
    }
    else {
        $chars += length;
        #$chars += tr/AGCT/AGCT/; #would work also
    }
}
print "$gene => $chars chars\n" if $chars; #print the last gene at the
+ end of the file

__DATA__
>gb|AE008689|:2073051-2073458, Atu4895
TTGGTCACATATTTTCGCTATTCGAAACCCAAATATCCCTGCGGAACACGTTTTATTGGAGCCGCCGTTT
TGCAGTTGCTCGGTGCTGGATTCTTGGCGTTCGTCTTTTGTCTGCTGGATGGGCTCACTGCAAAACCAAC
GATAATTCTCGGTCAGTTTGTATCATGCCTAGTGGGGAGCGTCGCCGGCTTTCATTTCGTGGCTTTTCGT
CGCCCAGGCACGGACGGCCAACTTTACCTTATCGCGACGTCGCTCTTGGCATTTGGGACCCATTATTGGC
TGGTGTCATATTCATTACCTGACCTTTTGCTAGCAAGATTGATTTCAGGATTTGGATCGGGTGTGGTTGT
TGCAGGAACTTTCCGACGTCGCTTTCTGGAAAATCCGGTAATTCCCTGCGTTCGATAG
>gb|AE008689|:c2074151-2073549, Atu4896
GTGTTTGGACCATATTTTCTGTGCAAACTAAACGATGACATAGGGCGATTTTTAGTGGCGGACAAATACA
GACTTCCCGAAGAGTTTTTTACCACTCGGTTTCTCGTTAGACGCATCGTACCCACAGACGCTGAAGCTAT
TTTCGAAGGGTGGAACACCGATCCCGAGGTGACGAAGTACCTGACGTGGAAACCCCACTCCGAGCTTGGC
CAGACACAGCGGGCGATTGAAGAAAATTATAGTGCGTGGAATGCAGGTACATCGTTTCCAGCTGTCATCT
GCCATCGCGAACGGCCACATGAACTAATCGGCCGTATTGATGCACGTCCGATGGGCCACAAGGTCTCTTA
CGGGTGGCTTGTCCGAAGAACCTGGTGGGGCCGGGGTGTTGCAAGCGAGGTCGTTCAACTCGCTGTAGAA
CACGCGTTATCGCATCCGCGCATCTTTCGCACCGAAGCATCCTGCGACGTTCTGAACACGGCGTCAGCAA
GAGTGATGGAAAAAGTAGGGATGACAAAGGAGGCCGTGCTTCGACGGTACCTTTTTCACCCCAATTTTTC
GAATATGCCGCGAGACGCCTTCCTGTATTCCAAGGTACGTTAA
>gb|AE008689|:c2074749-2074345, Atu4897
ATGAAACATACCATCGCAGTTCTCGGCCTGATCACCTTCTCCAGCCCGGCCTTCGCAGCATCGTGCGAGA
AAAACTTCACCGTCTCAGGCGTACCGATGGTCACGGCTGTCTCTTACAAATCCTTTCAGGAACTGCCGAA
AGCCAAAGCACCAGCTGTCCTTCAAAAGCTCGCCCAGGCCGTCGCGGCAGAAGGTTTTTCAGGTATCCAG
ATCAACAAGGCACTGTCGTCAATCGATGCCCATCAGGAAACCAGCGGAAGTGGCAGGATTCAGACGCTGC
GGGTTGTCGCCCGCCAGAAAGGCGCCGCTGTCCGGATCGATGCTGTCTTCAATATTCAGGCAGGACAGAT
CGCCGACAAAGACGTCATCCGCAAGGGCATCTGCGACATCATAAAAGGCGCGTAA
[download]

>gb|AE008689|:2073051-2073458, Atu4895 => 408 chars
>gb|AE008689|:c2074151-2073549, Atu4896 => 603 chars
>gb|AE008689|:c2074749-2074345, Atu4897 => 405 chars
[download]

use Bio::SeqIO;
$in  = Bio::SeqIO->new(-file => "inputfilename",-format => 'Fasta');
while ( my $seq = $in->next_seq() ) {
   $seqstr   = $seq->seq(); # actual sequence as a string
   print $seqobj->display_id(); # human readable id of the sequence
   print length($seqstr); # length of sequence
}
[download]

#!/usr/bin/perl

use warnings;
use strict;

my $inf=shift @ARGV || or die "Need to give a input file.\n";
open (INF,shift @ARGV);

sub print_map (\$\$) {
  my $name_r=shift @_;
  my $code_r=shift @_;

  my %hash;
  print "$$name_r\n";
  while ($$code_r=~m/(\w)/g) { $hash{$1}++;}
  foreach (sort {$a cmp $b} keys %hash) {
    print "$_: $hash{$_}\n";
  }
}

my ($code,$name);
while (<INF>) {
  chomp;
  if (m/^>(.+)/) {
    print_map ($name,$code) if ($name);
    $name=$1;
  } else {
    $code.=uc($_);
  }
}
print_map($name,$code);

exit;
[download]

my $num_chars = 0;
<>;
while ( <> ) {
    do {print "$num_chars\n"; $num_chars = 0; next} if /^>/;
    $num_chars += tr/ACGT/ACGT/;
}
print "$num_chars\n";
[download]

my @num_chars = ();
my $i = -1;
while ( <> ) {
    $i++, next if /^>/;
    $num_chars[$i] += tr/ACGT/ACGT/;
}
foreach ( @num_chars ) {
    print "$_\n";
}
[download]

The first gets by with one declared scalar and one loop. The second requires more memory since it uses an array instead of a single scalar. That array is useful, though, if you want to do anything with the values other than just print them. The second also uses two loops instead of one, which could be a speed factor on large data sets. The second loop could be removed easily enough.

On the loop fixups:
Since someone (not here on PerlMonks) recently asked me why I tend to use loop fixups instead of additional tests within a loop... The first has a pre-loop fixup involving input and a post-loop fixup involving output. The second initializes the array index to -1 instead of to zero before the loop. One might wonder why I used these fixups since they may not be quite as clear to some as an additional test inside the loop. The loops are cluttered enough as it is, and sometimes visual clarity makes up for logical clarity. I find them simpler to debug when compared to something which makes an additional flow control changes, especially inside a loop. This way the loops stay more consistent and focus on the general case in which we're interested. I find this better than having the special-case code folded into the general-case code when I try to analyze a program's behavior. Also, by making these simple and easily understood actions happen outside the loop, you could see notable performance differences on large data sets. If the printing and resetting of the counter from the first example get more complex, they could be put into a single subroutine, and that subroutine could be called both from within the loop and as the last output after the loop, making the final output a single entity to debug along with the others. So it's not just optimization. I actually find it easier to do things this way. YMMV.