/ 
Peta-Synth-0.2610180
/ Peta::Synth::Engine

NAME

Peta::Synth::Engine - the engines that render voices

VERSION

version 0.2610180

SYNOPSIS

my $engine = Peta::Synth::Engine::create('auto', rate => 48000, partials => 48);   # or 'perl', 'pdl', 'gpu'
print join ' ', Peta::Synth::Engine::available();     # gpu pdl perl
printf "%s %.1fx\n", $_->{name}, $_->{speed} for Peta::Synth::Engine::measure();

DESCRIPTION

Every engine renders the voice model of Peta::Synth::Voice and sounds the same; they differ in speed. perl is everywhere (and compiled by pperl's JIT), pdl needs PDL, gpu needs pperl with WebGPU and a graphics card. auto takes the fastest of those there are, measured, or $ENV{PETA_SYNTH_ENGINE} when set.

FUNCTIONS

create

create($name, %arg): an engine; dies if it is not available. For auto, rate and partials say what the synthesizer will have, for the measurement ("choose").

choose

choose(rate = $rate, partials => $partials)>: the name of the engine auto takes: $ENV{PETA_SYNTH_ENGINE} when set, or the fastest.

measure

measure(rate = $rate, partials => $partials)>: how fast each engine here renders a test, fastest first: [ { name => 'gpu', speed => 9.2 }, ... ], the speed in seconds of sound per second. The test is a chord of sixteen notes on four instruments and three drums, with the reverb, as the synthesizer plays them; the first quanta are not timed, so that pperl's JIT has compiled the loops. It takes a fraction of a second per engine.

Which engine is fastest depends on how much there is to render: the GPU costs a fixed time per quantum and pays from some number of partials on; the test is a moderate load.

try

The same, or undef with the reason in $@.

available, names

The engines that work here, and all of them.

perl_info

What this perl is and has: perl (pperl or perl5), version (pperl's own under pperl, not the perl it is compatible with), jit, pdl, gpu, cpus.

cpus

The number of processors.

THE ENGINE INTERFACE

new

Dies when the engine cannot run here.

name
render

render(\@voices, $n, $rate): the voices' next $n samples (a multiple of 64), each voice under its envelope and channel gains, mixed into three array references: left, right and the reverb send. The voices' phase and nstate are carried on.

AUTHOR

PetaMem s.r.o. <info@petamem.com>

COPYRIGHT

Copyright (c) 2026 PetaMem s.r.o.

LICENSE

This package is free software, dual-licensed under the Artistic License 2.0 and the BSD 2-Clause License. See the LICENSE file of the distribution.