Reduces memory allocation for split kernel.
This allows for faster rendering due to bigger global size,
specially when GPU memory is limited.
Perfromance results:
R9 290 total render time
Before After Change
BMW 4:37 4:34 -1.1 %
Classroom 14:43 14:30 -1.5 %
Fishy Cat 11:20 11:04 -2.4 %
Koro 12:11 12:04 -1.0 %
Pabellon Barcelona 22:01 20:44 -5.8 %
Pabellon Barcelona(*) 15:32 15:09 -2.5 %
(*) without glossy connected to volume
Decoupled ray marching is not supported yet.
Transparent shadows are always enabled for volume rendering.
Changes in kernel/bvh and kernel/geom are from Sergey.
This simiplifies code significantly, and prepares it for
record-all transparent shadow function in split kernel.
By calculating the size of the state buffer in the kernel rather than the host
less code is needed and the size actually reflects the requested features.
Will also be a little faster in some cases because of larger global work size.
This was only needed for the previous implementation of parallel samples. As
we don't have that any more it can be removed.
Real reason for removal tho is this: `per_sample_output_buffers` was being
calculated too small and artifacts resulted. The tile buffer is already
the correct size and calculating the size for `per_sample_output_buffers`
is a bit difficult with the current layout of the code. As
`per_sample_output_buffers` was only needed for `sum_all_radiance`,
removing that kernel and writing output to the tile buffer directly
fixes the artifacts.
This is to help debug and track memory usage for generic buffers. We
have similar for textures already since those require a name, but for
buffers the name is only for debugging proposes.
Simple workaround for some issues we've been having with AMD drivers hanging
and rendering systems unresponsive. Unfortunately this makes things a bit
slower, but its better than having to do hard reboots. Will be removed when
drivers have been fixed.
Define CYCLES_DISABLE_DRIVER_WORKAROUNDS to disable for testing purposes.
This does a few things at once:
- Refactors host side split kernel logic into a new device
agnostic class `DeviceSplitKernel`.
- Removes tile splitting, a new work pool implementation takes its place and
allows as many threads as will fit in memory regardless of tile size, which
can give performance gains.
- Refactors split state buffers into one buffer, as well as reduces the
number of arguments passed to kernels. Means there's less code to deal
with overall.
- Moves kernel logic out of OpenCL kernel files so they can later be used by
other device types.
- Replaced OpenCL specific APIs with new generic versions
- Tiles can now be seen updating during rendering
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
There is some define conflict between system headers and clew,
so delay include of clew.h as much as possible.]
This is something which needed to be done in the code before
the refactor, hopefully such change will still work.
Previously, it was only possible to choose a single GPU or all of that type (CUDA or OpenCL).
Now, a toggle button is displayed for every device.
These settings are tied to the PCI Bus ID of the devices, so they're consistent across hardware addition and removal (but not when swapping/moving cards).
From the code perspective, the more important change is that now, the compute device properties are stored in the Addon preferences of the Cycles addon, instead of directly in the User Preferences.
This allows for a cleaner implementation, removing the Cycles C API functions that were called by the RNA code to specify the enum items.
Note that this change is neither backwards- nor forwards-compatible, but since it's only a User Preference no existing files are broken.
Reviewers: #cycles, brecht
Reviewed By: #cycles, brecht
Subscribers: brecht, juicyfruit, mib2berlin, Blendify
Differential Revision: https://developer.blender.org/D2338
Previously an error message would be printed whenever the OpenCL build produced output.
However, some frameworks seem to print extra information even if the build succeeded, so now the actual returned error is checked as well.
When --debug-cycles is activated, the build output will always be printed, otherwise it only gets printed if there was an error.
The problem here was, as the title says, that the two kernels were swapped.
Since shader evaluation is only used for building the samling map when World MIS is enabled, rendering without it would still work fine, although baking also was broken.
The previous refactor changed the code to use a separate logging mechanism to support multithreaded compilation.
However, since that's not supported by any frameworks yes, it just resulted in bad logging behaviour.
So, this commit changes the logging to go diectly to stdout/stderr once again by default.
It will discard the whole tile, but it's still kind of more friendly than
fully locked interface (sort of) for until tile is fully sampled.
Sorry if it causes PITA to merge for the opencl split work, but this issue
bothering a lot when collecting benchmarks.
Basically just moves cached kernels from ~/.config/blender/BLENDER_VERSION to
~/.cache/cycles/kernels. This has following benefits:
- Follows XDG specification more closely,
not as if it's totally crucial or measurable by users, but still nice.
- Prevents unexpected sizes of config folder, makes disk space used in more
predictable for users way.
- Allows to share kernels across multiple Blender versions,
which makes it easier debugging at the times close to release.
- "Copy Previous Settings" operator will no longer be copying possibly
gigabytes of cached kernels, which used to lead to really nast disk usage
and annoying delays of copying settings.
- In the future we can have some smart logic to clear old unused cached
kernels.
Currently only done for Linux and OSX. Windows still follows old "cache"
folder logic, but it's not really important for now because we don't
support kernel compilation on this platform yet.
Reviewers: dingto, juicyfruit, brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D2197
Kernels can now be built without patch evaluation when not needed by the
scene (Catmull-Clark subdivision not in use), giving a performance boost
for some devices.
It is possible that compilation will fail without giving anything in the
log buffer. For this cases giving a tip about error code will be really
handy.
Patch by @Ilia, thanks!
This way we can easily switch between toolkits without worrying
whether some kernel was compiled with old or new CUDA toolkit.
It's also now possible to switch machine architecture and have
proper cached kernel detected. Not as if it happens every day,
but i did such a bitness switch back in the days :)
All the changes are mainly giving explicit tips on inlining functions,
so they match how inlining worked with previous toolkit.
This make kernel compiled by CUDA 8 render in average with same speed
as previous kernels. Some scenes are somewhat faster, some of them are
somewhat slower. But slowdown is within 1% so far.
On a positive side it allows us to enable newer generation cards on
buildbots (so GTX 10x0 will be officially supported soon).
