blender/intern/cycles/device/device_multi.cpp
Lukas Stockner a2ebc5268f Cycles: Refactor Progress system to provide better estimates
The Progress system in Cycles had two limitations so far:
 - It just counted tiles, but ignored their size. For example, when rendering a 600x500 image with 512x512 tiles, the right 88x500 tile would count for 50% of the progress, although it only covers 15% of the image.
 - Scene update time was incorrectly counted as rendering time - therefore, the remaining time started very long and gradually decreased.

This patch fixes both problems:
First of all, the Progress now has a function to ignore time spans, and that is used to ignore scene update time.
The larger change is the tile size: Instead of counting samples per tile, so that the final value is num_samples*num_tiles, the code now counts every sample for every pixel, so that the final value is num_samples*num_pixels.

Along with that, some unused variables were removed from the Progress and Session classes.

Reviewers: brecht, sergey, #cycles

Subscribers: brecht, candreacchio, sergey

Differential Revision: https://developer.blender.org/D2214
2016-12-03 05:02:21 +01:00

363 lines
8.3 KiB
C++

/*
* Copyright 2011-2013 Blender Foundation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdlib.h>
#include <sstream>
#include "device.h"
#include "device_intern.h"
#include "device_network.h"
#include "buffers.h"
#include "util_foreach.h"
#include "util_list.h"
#include "util_logging.h"
#include "util_map.h"
#include "util_time.h"
CCL_NAMESPACE_BEGIN
class MultiDevice : public Device
{
public:
struct SubDevice {
explicit SubDevice(Device *device_)
: device(device_) {}
Device *device;
map<device_ptr, device_ptr> ptr_map;
};
list<SubDevice> devices;
device_ptr unique_ptr;
MultiDevice(DeviceInfo& info, Stats &stats, bool background_)
: Device(info, stats, background_), unique_ptr(1)
{
Device *device;
foreach(DeviceInfo& subinfo, info.multi_devices) {
device = Device::create(subinfo, sub_stats_, background);
devices.push_back(SubDevice(device));
}
#ifdef WITH_NETWORK
/* try to add network devices */
ServerDiscovery discovery(true);
time_sleep(1.0);
vector<string> servers = discovery.get_server_list();
foreach(string& server, servers) {
device = device_network_create(info, stats, server.c_str());
if(device)
devices.push_back(SubDevice(device));
}
#endif
}
~MultiDevice()
{
foreach(SubDevice& sub, devices)
delete sub.device;
}
const string& error_message()
{
foreach(SubDevice& sub, devices) {
if(sub.device->error_message() != "") {
if(error_msg == "")
error_msg = sub.device->error_message();
break;
}
}
return error_msg;
}
virtual bool show_samples() const
{
if(devices.size() > 1) {
return false;
}
return devices.front().device->show_samples();
}
bool load_kernels(const DeviceRequestedFeatures& requested_features)
{
foreach(SubDevice& sub, devices)
if(!sub.device->load_kernels(requested_features))
return false;
return true;
}
void mem_alloc(device_memory& mem, MemoryType type)
{
foreach(SubDevice& sub, devices) {
mem.device_pointer = 0;
sub.device->mem_alloc(mem, type);
sub.ptr_map[unique_ptr] = mem.device_pointer;
}
mem.device_pointer = unique_ptr++;
stats.mem_alloc(mem.device_size);
}
void mem_copy_to(device_memory& mem)
{
device_ptr tmp = mem.device_pointer;
foreach(SubDevice& sub, devices) {
mem.device_pointer = sub.ptr_map[tmp];
sub.device->mem_copy_to(mem);
}
mem.device_pointer = tmp;
}
void mem_copy_from(device_memory& mem, int y, int w, int h, int elem)
{
device_ptr tmp = mem.device_pointer;
int i = 0, sub_h = h/devices.size();
foreach(SubDevice& sub, devices) {
int sy = y + i*sub_h;
int sh = (i == (int)devices.size() - 1)? h - sub_h*i: sub_h;
mem.device_pointer = sub.ptr_map[tmp];
sub.device->mem_copy_from(mem, sy, w, sh, elem);
i++;
}
mem.device_pointer = tmp;
}
void mem_zero(device_memory& mem)
{
device_ptr tmp = mem.device_pointer;
foreach(SubDevice& sub, devices) {
mem.device_pointer = sub.ptr_map[tmp];
sub.device->mem_zero(mem);
}
mem.device_pointer = tmp;
}
void mem_free(device_memory& mem)
{
device_ptr tmp = mem.device_pointer;
foreach(SubDevice& sub, devices) {
mem.device_pointer = sub.ptr_map[tmp];
sub.device->mem_free(mem);
sub.ptr_map.erase(sub.ptr_map.find(tmp));
}
mem.device_pointer = 0;
stats.mem_free(mem.device_size);
}
void const_copy_to(const char *name, void *host, size_t size)
{
foreach(SubDevice& sub, devices)
sub.device->const_copy_to(name, host, size);
}
void tex_alloc(const char *name,
device_memory& mem,
InterpolationType
interpolation,
ExtensionType extension)
{
VLOG(1) << "Texture allocate: " << name << ", "
<< string_human_readable_number(mem.memory_size()) << " bytes. ("
<< string_human_readable_size(mem.memory_size()) << ")";
foreach(SubDevice& sub, devices) {
mem.device_pointer = 0;
sub.device->tex_alloc(name, mem, interpolation, extension);
sub.ptr_map[unique_ptr] = mem.device_pointer;
}
mem.device_pointer = unique_ptr++;
stats.mem_alloc(mem.device_size);
}
void tex_free(device_memory& mem)
{
device_ptr tmp = mem.device_pointer;
foreach(SubDevice& sub, devices) {
mem.device_pointer = sub.ptr_map[tmp];
sub.device->tex_free(mem);
sub.ptr_map.erase(sub.ptr_map.find(tmp));
}
mem.device_pointer = 0;
stats.mem_free(mem.device_size);
}
void pixels_alloc(device_memory& mem)
{
foreach(SubDevice& sub, devices) {
mem.device_pointer = 0;
sub.device->pixels_alloc(mem);
sub.ptr_map[unique_ptr] = mem.device_pointer;
}
mem.device_pointer = unique_ptr++;
}
void pixels_free(device_memory& mem)
{
device_ptr tmp = mem.device_pointer;
foreach(SubDevice& sub, devices) {
mem.device_pointer = sub.ptr_map[tmp];
sub.device->pixels_free(mem);
sub.ptr_map.erase(sub.ptr_map.find(tmp));
}
mem.device_pointer = 0;
}
void pixels_copy_from(device_memory& mem, int y, int w, int h)
{
device_ptr tmp = mem.device_pointer;
int i = 0, sub_h = h/devices.size();
foreach(SubDevice& sub, devices) {
int sy = y + i*sub_h;
int sh = (i == (int)devices.size() - 1)? h - sub_h*i: sub_h;
mem.device_pointer = sub.ptr_map[tmp];
sub.device->pixels_copy_from(mem, sy, w, sh);
i++;
}
mem.device_pointer = tmp;
}
void draw_pixels(device_memory& rgba, int y, int w, int h, int dx, int dy, int width, int height, bool transparent,
const DeviceDrawParams &draw_params)
{
device_ptr tmp = rgba.device_pointer;
int i = 0, sub_h = h/devices.size();
int sub_height = height/devices.size();
foreach(SubDevice& sub, devices) {
int sy = y + i*sub_h;
int sh = (i == (int)devices.size() - 1)? h - sub_h*i: sub_h;
int sheight = (i == (int)devices.size() - 1)? height - sub_height*i: sub_height;
int sdy = dy + i*sub_height;
/* adjust math for w/width */
rgba.device_pointer = sub.ptr_map[tmp];
sub.device->draw_pixels(rgba, sy, w, sh, dx, sdy, width, sheight, transparent, draw_params);
i++;
}
rgba.device_pointer = tmp;
}
void map_tile(Device *sub_device, RenderTile& tile)
{
foreach(SubDevice& sub, devices) {
if(sub.device == sub_device) {
if(tile.buffer) tile.buffer = sub.ptr_map[tile.buffer];
if(tile.rng_state) tile.rng_state = sub.ptr_map[tile.rng_state];
}
}
}
int device_number(Device *sub_device)
{
int i = 0;
foreach(SubDevice& sub, devices) {
if(sub.device == sub_device)
return i;
i++;
}
return -1;
}
int get_split_task_count(DeviceTask& task)
{
int total_tasks = 0;
list<DeviceTask> tasks;
task.split(tasks, devices.size());
foreach(SubDevice& sub, devices) {
if(!tasks.empty()) {
DeviceTask subtask = tasks.front();
tasks.pop_front();
total_tasks += sub.device->get_split_task_count(subtask);
}
}
return total_tasks;
}
void task_add(DeviceTask& task)
{
list<DeviceTask> tasks;
task.split(tasks, devices.size());
foreach(SubDevice& sub, devices) {
if(!tasks.empty()) {
DeviceTask subtask = tasks.front();
tasks.pop_front();
if(task.buffer) subtask.buffer = sub.ptr_map[task.buffer];
if(task.rgba_byte) subtask.rgba_byte = sub.ptr_map[task.rgba_byte];
if(task.rgba_half) subtask.rgba_half = sub.ptr_map[task.rgba_half];
if(task.shader_input) subtask.shader_input = sub.ptr_map[task.shader_input];
if(task.shader_output) subtask.shader_output = sub.ptr_map[task.shader_output];
if(task.shader_output_luma) subtask.shader_output_luma = sub.ptr_map[task.shader_output_luma];
sub.device->task_add(subtask);
}
}
}
void task_wait()
{
foreach(SubDevice& sub, devices)
sub.device->task_wait();
}
void task_cancel()
{
foreach(SubDevice& sub, devices)
sub.device->task_cancel();
}
protected:
Stats sub_stats_;
};
Device *device_multi_create(DeviceInfo& info, Stats &stats, bool background)
{
return new MultiDevice(info, stats, background);
}
CCL_NAMESPACE_END