Welcome to mirror list, hosted at ThFree Co, Russian Federation.

git.blender.org/blender.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLukas Stockner <lukas.stockner@freenet.de>2020-09-24 01:37:23 +0300
committerLukas Stockner <lukas.stockner@freenet.de>2020-10-31 03:57:39 +0300
commit517ff40b124bc9d1324ccf7561a59ac51bf86602 (patch)
tree96295b1b6a11a597f7927cc61ce9371077bc7e54 /intern/cycles/device
parent523414dda2bf81b69b1c04e1145ac21758fa4268 (diff)
Cycles: Implement tile stealing to improve CPU+GPU rendering performance
While Cycles already supports using both CPU and GPU at the same time, there currently is a large problem with it: Since the CPU grabs one tile per thread, at the end of the render the GPU runs out of new work but the CPU still needs quite some time to finish its current times. Having smaller tiles helps somewhat, but especially OpenCL rendering tends to lose performance with smaller tiles. Therefore, this commit adds support for tile stealing: When a GPU device runs out of new tiles, it can signal the CPU to release one of its tiles. This way, at the end of the render, the GPU quickly finishes the remaining tiles instead of having to wait for the CPU. Thanks to AMD for sponsoring this work! Differential Revision: https://developer.blender.org/D9324
Diffstat (limited to 'intern/cycles/device')
-rw-r--r--intern/cycles/device/device_cpu.cpp5
-rw-r--r--intern/cycles/device/device_memory.h8
-rw-r--r--intern/cycles/device/device_task.h1
3 files changed, 14 insertions, 0 deletions
diff --git a/intern/cycles/device/device_cpu.cpp b/intern/cycles/device/device_cpu.cpp
index 23aedcd0c48..6912ac1e638 100644
--- a/intern/cycles/device/device_cpu.cpp
+++ b/intern/cycles/device/device_cpu.cpp
@@ -932,6 +932,11 @@ class CPUDevice : public Device {
break;
}
+ if (tile.stealing_state == RenderTile::CAN_BE_STOLEN && task.get_tile_stolen()) {
+ tile.stealing_state = RenderTile::WAS_STOLEN;
+ break;
+ }
+
if (tile.task == RenderTile::PATH_TRACE) {
for (int y = tile.y; y < tile.y + tile.h; y++) {
for (int x = tile.x; x < tile.x + tile.w; x++) {
diff --git a/intern/cycles/device/device_memory.h b/intern/cycles/device/device_memory.h
index 32654e62a6f..00b2aa864aa 100644
--- a/intern/cycles/device/device_memory.h
+++ b/intern/cycles/device/device_memory.h
@@ -450,6 +450,14 @@ template<typename T> class device_vector : public device_memory {
device_zero();
}
+ void move_device(Device *new_device)
+ {
+ copy_from_device();
+ device_free();
+ device = new_device;
+ copy_to_device();
+ }
+
protected:
size_t size(size_t width, size_t height, size_t depth)
{
diff --git a/intern/cycles/device/device_task.h b/intern/cycles/device/device_task.h
index fd380788282..f819f84eb43 100644
--- a/intern/cycles/device/device_task.h
+++ b/intern/cycles/device/device_task.h
@@ -159,6 +159,7 @@ class DeviceTask {
function<void(RenderTile &)> update_tile_sample;
function<void(RenderTile &)> release_tile;
function<bool()> get_cancel;
+ function<bool()> get_tile_stolen;
function<void(RenderTileNeighbors &, Device *)> map_neighbor_tiles;
function<void(RenderTileNeighbors &, Device *)> unmap_neighbor_tiles;