While testing performance improvements elsewhere, I noticed some
micro-optimizations could give a small ~2-3% speedup in my test
using the git async API to parse a large inbox.
The `read' perlfunc already has read-in-full behavior (unless
git is killed unexpectedly), so there's no point in using a
loop. SearchIdxShard in the parallel v2 indexing code path
never looped on `read', either.
Furthermore, we can avoid method dispatch overhead on ->getline
and ->print by using `readline' and `print' as ops which can be
resolved during the Perl compilation phase.
Finally, avoid passing the IO handle around as a parameter,
since avoiding hash lookups with a local variable has its own
costs in stack and refcount bumping.
Best off all, there's less code :>
-sub read_cat_in_full ($$$) {
- my ($self, $in, $left) = @_;
- my $offset = 0;
- my $buf = '';
- while ($left > 0) {
- my $r = read($in, $buf, $left, $offset);
- defined($r) or fail($self, "read failed: $!");
- $r == 0 and fail($self, 'exited unexpectedly');
- $left -= $r;
- $offset += $r;
- }
- my $r = read($in, my $lf, 1);
- defined($r) or fail($self, "read failed: $!");
- fail($self, 'newline missing after blob') if ($r != 1 || $lf ne "\n");
+sub read_cat_in_full ($$) {
+ my ($self, $len) = @_;
+ ++$len; # for final "\n" added by git
+ read($self->{in}, my $buf, $len) == $len or fail($self, 'short read');
+ chop($buf) eq "\n" or fail($self, 'newline missing after blob');
-sub _cat_async_step ($$$) {
- my ($self, $inflight, $in) = @_;
+sub _cat_async_step ($$) {
+ my ($self, $inflight) = @_;
my $pair = shift @$inflight or die 'BUG: inflight empty';
my ($cb, $arg) = @$pair;
local $/ = "\n";
my $pair = shift @$inflight or die 'BUG: inflight empty';
my ($cb, $arg) = @$pair;
local $/ = "\n";
- my $head = $in->getline;
+ my $head = readline($self->{in});
$head =~ / missing$/ and return
eval { $cb->(undef, undef, undef, undef, $arg) };
$head =~ /^([0-9a-f]{40}) (\S+) ([0-9]+)$/ or
fail($self, "Unexpected result from async git cat-file: $head");
my ($oid_hex, $type, $size) = ($1, $2, $3 + 0);
$head =~ / missing$/ and return
eval { $cb->(undef, undef, undef, undef, $arg) };
$head =~ /^([0-9a-f]{40}) (\S+) ([0-9]+)$/ or
fail($self, "Unexpected result from async git cat-file: $head");
my ($oid_hex, $type, $size) = ($1, $2, $3 + 0);
- my $bref = read_cat_in_full($self, $in, $size);
+ my $bref = read_cat_in_full($self, $size);
eval { $cb->($bref, $oid_hex, $type, $size, $arg) };
}
sub cat_async_wait ($) {
my ($self) = @_;
my $inflight = delete $self->{inflight} or return;
eval { $cb->($bref, $oid_hex, $type, $size, $arg) };
}
sub cat_async_wait ($) {
my ($self) = @_;
my $inflight = delete $self->{inflight} or return;
while (scalar(@$inflight)) {
while (scalar(@$inflight)) {
- _cat_async_step($self, $inflight, $in);
+ _cat_async_step($self, $inflight);
}
}
sub cat_file {
my ($self, $obj, $ref) = @_;
}
}
sub cat_file {
my ($self, $obj, $ref) = @_;
- my ($retried, $in, $head);
cat_async_wait($self);
again:
batch_prepare($self);
cat_async_wait($self);
again:
batch_prepare($self);
- $self->{out}->print($obj, "\n") or fail($self, "write error: $!");
+ print { $self->{out} } $obj, "\n" or fail($self, "write error: $!");
+ $head = readline($self->{in});
if ($head =~ / missing$/) {
if (!$retried && alternates_changed($self)) {
$retried = 1;
if ($head =~ / missing$/) {
if (!$retried && alternates_changed($self)) {
$retried = 1;
my $size = $1;
$$ref = $size if $ref;
my $size = $1;
$$ref = $size if $ref;
- read_cat_in_full($self, $in, $size);
+ read_cat_in_full($self, $size);
}
sub batch_prepare ($) { _bidi_pipe($_[0], qw(--batch in out pid)) }
}
sub batch_prepare ($) { _bidi_pipe($_[0], qw(--batch in out pid)) }
sub check {
my ($self, $obj) = @_;
_bidi_pipe($self, qw(--batch-check in_c out_c pid_c err_c));
sub check {
my ($self, $obj) = @_;
_bidi_pipe($self, qw(--batch-check in_c out_c pid_c err_c));
- $self->{out_c}->print($obj, "\n") or fail($self, "write error: $!");
+ print { $self->{out_c} } $obj, "\n" or fail($self, "write error: $!");
- chomp(my $line = $self->{in_c}->getline);
+ chomp(my $line = readline($self->{in_c}));
my ($hex, $type, $size) = split(' ', $line);
# Future versions of git.git may show 'ambiguous', but for now,
my ($hex, $type, $size) = split(' ', $line);
# Future versions of git.git may show 'ambiguous', but for now,
my ($self, $oid, $cb, $arg) = @_;
my $inflight = $self->{inflight} or die 'BUG: not in async';
if (scalar(@$inflight) >= MAX_INFLIGHT) {
my ($self, $oid, $cb, $arg) = @_;
my $inflight = $self->{inflight} or die 'BUG: not in async';
if (scalar(@$inflight) >= MAX_INFLIGHT) {
- _cat_async_step($self, $inflight, $self->{in});
+ _cat_async_step($self, $inflight);
- $self->{out}->print($oid, "\n") or fail($self, "write error: $!");
+ print { $self->{out} } $oid, "\n" or fail($self, "write error: $!");
push(@$inflight, [ $cb, $arg ]);
}
push(@$inflight, [ $cb, $arg ]);
}