Increase decoding chunk size for much better performance.

[pulseview.git] / pv / data / segment.cpp

/*
 * This file is part of the PulseView project.
 *
 * Copyright (C) 2017 Soeren Apel <soeren@apelpie.net>
 * Copyright (C) 2012 Joel Holdsworth <joel@airwebreathe.org.uk>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, see <http://www.gnu.org/licenses/>.
 */

#include "segment.hpp"

#include <assert.h>
#include <stdlib.h>
#include <string.h>

#include <vector>

using std::lock_guard;
using std::recursive_mutex;
using std::vector;

namespace pv {
namespace data {

const uint64_t Segment::MaxChunkSize = 10*1024*1024;  /* 10MiB */

Segment::Segment(uint64_t samplerate, unsigned int unit_size) :
	sample_count_(0),
	start_time_(0),
	samplerate_(samplerate),
	unit_size_(unit_size),
	iterator_count_(0),
	mem_optimization_requested_(false)
{
	lock_guard<recursive_mutex> lock(mutex_);
	assert(unit_size_ > 0);

	// Determine the number of samples we can fit in one chunk
	// without exceeding MaxChunkSize
	chunk_size_ = std::min(MaxChunkSize,
		(MaxChunkSize / unit_size_) * unit_size_);

	// Create the initial chunk
	current_chunk_ = new uint8_t[chunk_size_];
	data_chunks_.push_back(current_chunk_);
	used_samples_ = 0;
	unused_samples_ = chunk_size_ / unit_size_;
}

Segment::~Segment()
{
	lock_guard<recursive_mutex> lock(mutex_);

	for (uint8_t* chunk : data_chunks_)
		delete[] chunk;
}

uint64_t Segment::get_sample_count() const
{
	lock_guard<recursive_mutex> lock(mutex_);
	return sample_count_;
}

const pv::util::Timestamp& Segment::start_time() const
{
	return start_time_;
}

double Segment::samplerate() const
{
	return samplerate_;
}

void Segment::set_samplerate(double samplerate)
{
	samplerate_ = samplerate;
}

unsigned int Segment::unit_size() const
{
	return unit_size_;
}

void Segment::free_unused_memory()
{
	lock_guard<recursive_mutex> lock(mutex_);

	// Do not mess with the data chunks if we have iterators pointing at them
	if (iterator_count_ > 0) {
		mem_optimization_requested_ = true;
		return;
	}

	// No more data will come in, so re-create the last chunk accordingly
	uint8_t* resized_chunk = new uint8_t[used_samples_ * unit_size_];
	memcpy(resized_chunk, current_chunk_, used_samples_ * unit_size_);

	delete[] current_chunk_;
	current_chunk_ = resized_chunk;

	data_chunks_.pop_back();
	data_chunks_.push_back(resized_chunk);
}

void Segment::append_single_sample(void *data)
{
	lock_guard<recursive_mutex> lock(mutex_);

	// There will always be space for at least one sample in
	// the current chunk, so we do not need to test for space

	memcpy(current_chunk_ + (used_samples_ * unit_size_),
		data, unit_size_);
	used_samples_++;
	unused_samples_--;

	if (unused_samples_ == 0) {
		current_chunk_ = new uint8_t[chunk_size_];
		data_chunks_.push_back(current_chunk_);
		used_samples_ = 0;
		unused_samples_ = chunk_size_ / unit_size_;
	}

	sample_count_++;
}

void Segment::append_samples(void* data, uint64_t samples)
{
	lock_guard<recursive_mutex> lock(mutex_);

	if (unused_samples_ >= samples) {
		// All samples fit into the current chunk
		memcpy(current_chunk_ + (used_samples_ * unit_size_),
			data, (samples * unit_size_));
		used_samples_ += samples;
		unused_samples_ -= samples;
	} else {
		// Only a part of the samples fit, split data up between chunks
		memcpy(current_chunk_ + (used_samples_ * unit_size_),
			data, (unused_samples_ * unit_size_));
		const uint64_t remaining_samples = samples - unused_samples_;

		// If we're out of memory, this will throw std::bad_alloc
		current_chunk_ = new uint8_t[chunk_size_];
		data_chunks_.push_back(current_chunk_);
		memcpy(current_chunk_, (uint8_t*)data + (unused_samples_ * unit_size_),
			(remaining_samples * unit_size_));

		used_samples_ = remaining_samples;
		unused_samples_ = (chunk_size_ / unit_size_) - remaining_samples;
	}

	if (unused_samples_ == 0) {
		// If we're out of memory, this will throw std::bad_alloc
		current_chunk_ = new uint8_t[chunk_size_];
		data_chunks_.push_back(current_chunk_);
		used_samples_ = 0;
		unused_samples_ = chunk_size_ / unit_size_;
	}

	sample_count_ += samples;
}

uint8_t* Segment::get_raw_samples(uint64_t start, uint64_t count) const
{
	assert(start < sample_count_);
	assert(start + count <= sample_count_);
	assert(count > 0);

	lock_guard<recursive_mutex> lock(mutex_);

	uint8_t* dest = new uint8_t[count * unit_size_];
	uint8_t* dest_ptr = dest;

	uint64_t chunk_num = (start * unit_size_) / chunk_size_;
	uint64_t chunk_offs = (start * unit_size_) % chunk_size_;

	while (count > 0) {
		const uint8_t* chunk = data_chunks_[chunk_num];

		uint64_t copy_size = std::min(count * unit_size_,
			chunk_size_ - chunk_offs);

		memcpy(dest_ptr, chunk + chunk_offs, copy_size);

		dest_ptr += copy_size;
		count -= (copy_size / unit_size_);

		chunk_num++;
		chunk_offs = 0;
	}

	return dest;
}

SegmentRawDataIterator* Segment::begin_raw_sample_iteration(uint64_t start)
{
	SegmentRawDataIterator* it = new SegmentRawDataIterator;

	assert(start < sample_count_);

	iterator_count_++;

	it->sample_index = start;
	it->chunk_num = (start * unit_size_) / chunk_size_;
	it->chunk_offs = (start * unit_size_) % chunk_size_;
	it->chunk = data_chunks_[it->chunk_num];
	it->value = it->chunk + it->chunk_offs;

	return it;
}

void Segment::continue_raw_sample_iteration(SegmentRawDataIterator* it, uint64_t increase)
{
	lock_guard<recursive_mutex> lock(mutex_);

	if (it->sample_index > sample_count_)
	{
		// Fail gracefully if we are asked to deliver data we don't have
		return;
	} else {
		it->sample_index += increase;
		it->chunk_offs += (increase * unit_size_);
	}

	if (it->chunk_offs > (chunk_size_ - 1)) {
		it->chunk_num++;
		it->chunk_offs -= chunk_size_;
		it->chunk = data_chunks_[it->chunk_num];
	}

	it->value = it->chunk + it->chunk_offs;
}

void Segment::end_raw_sample_iteration(SegmentRawDataIterator* it)
{
	delete it;

	iterator_count_--;

	if ((iterator_count_ == 0) && mem_optimization_requested_) {
		mem_optimization_requested_ = false;
		free_unused_memory();
	}
}


} // namespace data
} // namespace pv