Previously, the NLM kernels would be launched once per offset with one thread per pixel.
However, with the smaller tile sizes that are now feasible, there wasn't enough work to fully occupy GPUs which results in a significant slowdown.
Therefore, the kernels are now launched in a single call that handles all offsets at once.
This has two downsides: Memory accesses to accumulating buffers are now atomic, and more importantly, the temporary memory now has to be allocated for every shift at once, increasing the required memory.
On the other hand, of course, the smaller tiles significantly reduce the size of the memory.
The main bottleneck right now is the construction of the transformation - there is nothing to be parallelized there, one thread per pixel is the maximum.
I tried to parallelize the SVD implementation by storing the matrix in shared memory and launching one block per pixel, but that wasn't really going anywhere.
To make the new code somewhat readable, the handling of rectangular regions was cleaned up a bit and commented, it should be easier to understand what's going on now.
Also, some variables have been renamed to make the difference between buffer width and stride more apparent, in addition to some general style cleanup.
Goal is to reduce OpenCL kernel recompilations.
Currently viewport renders are still set to use 64 closures as this seems to
be faster and we don't want to cause a performance regression there. Needs
to be investigated.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D2775
This is a prequisite for getting host memory allocation to work. There appears
to be no support for 3D textures using host memory. The original version of
this code was written by Stefan Werner for D2056.
Some drivers may report very large allocation sizes, which could cause
unnecessary memory usage. This is now limited to 2gb which should
still be enough to get the needed performance benefits without waste.
* Remove tex_* and pixels_* functions, replace by mem_*.
* Add MEM_TEXTURE and MEM_PIXELS as memory types recognized by devices.
* No longer create device_memory and call mem_* directly, always go
through device_only_memory, device_vector and device_pixels.
CPU rendering will be restricted to a BVH2, which is not ideal for raytracing
performance but can be shared with the GPU. Decoupled volume shading will be
disabled to match GPU volume sampling.
The number of CPU rendering threads is reduced to leave one core dedicated to
each GPU. Viewport rendering will also only use GPU rendering still. So along
with the BVH2 usage, perfect scaling should not be expected.
Go to User Preferences > System to enable the CPU to render alongside the GPU.
Differential Revision: https://developer.blender.org/D2873
This change affects CUDA GPUs not connected to a display or connected to a
display but supporting compute preemption so that the display does not
freeze. I couldn't find an official list, but compute preemption seems to be
only supported with GTX 1070+ and Linux (not GTX 1060- or Windows).
This helps improve small tile rendering performance further if there are
sufficient samples x number of pixels in a single tile to keep the GPU busy.
Best guess is that cuInit() somehow interferes with the AMD graphics driver
on Windows, and switching the initialization order to do OpenCL first seems
to solve the issue.
* Use common TextureInfo struct for all devices, except CUDA fermi.
* Move image sampling code to kernels/*/kernel_*_image.h files.
* Use arrays for data textures on Fermi too, so device_vector<Struct> works.
The work size is still very conservative, and this doesn't help for progressive
refine. For that we will need to render multiple tiles at the same time. But this
should already help for denoising renders that require too much memory with big
tiles, and just generally soften the performance dropoff with small tiles.
Differential Revision: https://developer.blender.org/D2856
This is a bit confusing, especially when one mixes OpenCL code where ulong equals
to uint64_t with CPU side code where ulong is expected to be something else from
the naming.
This commit makes it so we use explicit name, common on all platforms.
Problem was that some code checks to see if device_pointer is null or
not and the new allocator wasn't even setting the pointer to anything
as it tracks memory location separately. Setting the pointer to non
null keeps all users of device_pointer happy.
- Apparently MSVC does not support compound literals
in C++ (at least by the looks of it).
- Not sure how opencl_device_assert was managing to
set protected property of the Device class.
Common folks, nobody considered master a C++11 only branch. Such decision is to
be done officially and will involve changes in quite a few infrastructure related
areas.
Image textures were being packed into a single buffer for OpenCL, which
limited the amount of memory available for images to the size of one
buffer (usually 4gb on AMD hardware). By packing textures into multiple
buffers that limit is removed, while simultaneously reducing the number
of buffers that need to be passed to each kernel.
Benchmarks were within 2%.
Fixes T51554.
Differential Revision: https://developer.blender.org/D2745
