GPUOpen

GPUOpen
Original author(s)Advanced Micro Devices
Developer(s)Advanced Micro Devices
Initial releaseJanuary 26, 2016 (2016-01-26)[1]
Repository
Written inC, C++, GLSL
Operating systemLinux, Microsoft Windows
TypeGame effects libraries, GPU debugging, CPU & GPU profiling
LicenseMIT License
Websitegpuopen.com

GPUOpen is a middleware software suite originally developed by AMD's Radeon Technologies Group that offers advanced visual effects for computer games. It was released in 2016. GPUOpen serves as an alternative to, and a direct competitor of Nvidia GameWorks. GPUOpen is similar to GameWorks in that it encompasses several different graphics technologies as its main components that were previously independent and separate from one another.[2] However, GPUOpen is partially open source software, unlike GameWorks which is proprietary and closed.

History

[edit]

GPUOpen was announced on December 15, 2015,[3][4][2][5][6] and released on January 26, 2016.

Rationale

[edit]

Nicolas Thibieroz, AMD's Senior Manager of Worldwide Gaming Engineering, argues that "it can be difficult for developers to leverage their R&D investment on both consoles and PC because of the disparity between the two platforms" and that "proprietary libraries or tools chains with "black box" APIs prevent developers from accessing the code for maintenance, porting or optimizations purposes".[7] He says that upcoming architectures, such as AMD's RX 400 series "include many features not exposed today in PC graphics APIs".

AMD designed GPUOpen to be a competing open-source middleware stack released under the MIT License. The libraries are intended to increase software portability between video game consoles, PCs and also high-performance computing.[8]

Components

[edit]

GPUOpen unifies many of AMD's previously separate tools and solutions into one package, also fully open-sourcing them under the MIT License.[4] GPUOpen also makes it easy for developers to get low-level GPU access.[9]

Additionally AMD wants to grant interested developers the kind of low-level "direct access" to their GCN-based GPUs, that surpasses the possibilities of Direct3D 12 or Vulkan. AMD mentioned e.g. a low-level access to the Asynchronous Compute Engines (ACEs). The ACE implement "Asynchronous Compute", but they cannot be freely configured under either Vulkan or Direct3D 12.

GPUOpen is made up of several main components, tools, and SDKs.[2]

Games and CGI

[edit]

Software for computer-generated imagery (CGI) used in development of computer games and movies alike.

Visual effects libraries

[edit]
GPUOpen Visual Effects Libraries[10]
Name API Source Description
TressFX DirectX 12, Vulkan GitHub This visual effects library allows the creation of realistic hair, fur, and grass.
GeometryFX DirectX 11 GitHub This library allows easy access to compute-based triangle filtering.
DepthOfFieldFX DirectX 11 GitHub This library grants access to a depth of field implementation optimized for the GCN GPU architecture via a compute shader.
ShadowFX DirectX 11, DirectX 12 GitHub This library grants access to an implementation for deferred shadow filtering that is optimized for the GCN GPU architecture.
FidelityFX DirectX 11, DirectX 12, Vulkan GitHub FidelityFX is a suite of visual effects and effects-helper libraries.

FidelityFX

[edit]
FidelityFX Components[11]
Name Algorithm Source Description
FidelityFX CAS Contrast Adaptive Sharpening GitHub This algorithm adaptively sharpens an image or scene while minimizing artifacts.
FidelityFX CACAO Combined Adaptive Compute Ambient Occlusion GitHub This algorithm is an optimized implementation of adaptive sampling ambient occlusion.
FidelityFX LPM Luminance Preserving Mapper GitHub This algorithm is used to tone map the luma of an RGB pixel rather than tone mapping the color of the pixel.
FidelityFX SPD Single Pass Downsampler GitHub This algorithm, optimized for the RDNA GPU architecture, is used to generate 12 MIP levels for a given texture.
FidelityFX SSSR Stochastic Screen Space Reflections GitHub This algorithm is used to add screen space reflections to a frame or scene.
FidelityFX VS Variable Shading GitHub This algorithm is used to generate image-based variable rate shading using the luminance of samples in the prior frame.
FidelityFX Parallel Sort Radix Sort GitHub This algorithm provides a compute-based radix sort.
FidelityFX Denoiser Shadow & Reflection Denoiser GitHub This algorithm provides denoising functionality for ray-traced shadows and ray-traced or screen-space reflections.
FidelityFX Super Resolution 1 Spatial Upsampler GitHub This algorithm is used to upsample an image or frame into a higher resolution using only the spatial information provided in the input frame.
FidelityFX Super Resolution 2 Temporal Upscaler GitHub This algorithm is used to upscale frame(s) into a higher resolution using the temporal information provided by input frames.

FidelityFX Super Resolution

[edit]

FidelityFX Super Resolution (FSR) is used to upsample an input image into a higher resolution. There are multiple versions of FSR with distinctive upscaling technique and image quality:

  • FSR 1 is a spatial upscaler based on or similar to the Lanczos algorithm, requiring an anti-aliased lower resolution image. It also performs edge reconstruction and gradient reversal. This is then followed by a contrast adaptive sharpening pass (RCAS) to reintroduce detail into the final image. (see the cited quotation below for more information)
    • "FSR is composed of two main passes:
      • An upscaling pass called EASU (Edge-Adaptive Spatial Upsampling) that also performs edge reconstruction. In this pass the input frame is analyzed and the main part of the algorithm detects gradient reversals – essentially looking at how neighboring gradients differ – from a set of input pixels. The intensity of the gradient reversals defines the weights to apply to the reconstructed pixels at display resolution.
      • A sharpening pass called RCAS (Robust Contrast-Adaptive Sharpening) that extracts pixel detail in the upscaled image."[12]
  • FSR 2 is a temporal upscaler based on a modified Lanczos requiring an aliased lower resolution image and utilising the temporal data (such as motion vectors and frame history) and then applies its own antialiasing pass which replaces the game's temporal antialiasing solution.
  • FSR 3 adds frame generation and "native antialiasing". Frame generation increases the perceived frame rate of a game. "Native antialiasing", similar to Nvidia's DLAA, can be used without upscaling for improved antialiasing; it can also be combined with frame generation and Anti-Lag+.[13][14]

The standard presets for FSR by AMD can be found in the table below. Note that these presets are not the only way in which the algorithm can be used, they are simply presets for input/output resolutions. Certain titles, such as Dota 2 have offered resolution sliders to fine tune the scaling percentage or dynamically scaling the internal render resolution depending on the FPS cap. AMD has also created a command line interface tool which allows the user to upscale any image using FSR1/EASU as in addition to other upsampling methods such as Bilinear Interpolation. It also allows the user to run various stages of the FSR pipeline, such as RCAS independently.[15]

Release history
Release[a] Release date Highlights
1.0 / 1.0.1 Jun 2021 FidelityFX Super Resolution (FSR) launch, source code available July 2021.[16][17]
1.0.2 Nov 2021 Robust Contrast-Adaptive Sharpening (RCAS) oversharpening hotfix.[18]
1.1 Jul 2023 Available as part of FidelityFX SDK.[19]
2.0.1 / 2.0.1a Mar 2022 FidelityFX Super Resolution 2.0 (FSR 2) launch, source code available June 2022.[20][21]
2.1.0 Sep 2022 Reduced ghosting and improved upscaling quality. Farming Simulator 2022 was one of early adopters with patch 1.7.1.[22]
2.1.1 Sep 2022 [23]
2.1.2 Oct 2022 [24]
2.2.0 / 2.2.0a Nov 2022 HDR range improvements, ghosting and flickering artefacts reduction. Source code available February 2023.[25]
2.2.1 Jun 2023 [26]
2.2.2 Jul 2023 Available as part of FidelityFX SDK.[19][27]
3.0 / 3.0.3 Sep 2023 FSR 3 adds frame generation combined with FSR 2 and Anti-Lag+ and supports GPUs from AMD, Nvidia, and Intel. FSR 3 is also compatible with the ninth generation of video game consoles.[13]
Source code available December 2023 as part of FidelityFX SDK.[28]
3.0.4 Mar 2024 [29]
Standard FSR presets[30][31][32]
Quality Preset[b] Scale Factor[c] Render Scale[d]
Native AA(since v3.0) 1.00x 100%
Ultra Quality(v1.0 only) 1.30x 77.0%
Quality 1.50x 66.6%
Balanced 1.70x 58.8%
Performance 2.00x 50.0%
Ultra Performance(since v2.0) 3.00x 33.3%

FSR 2 can also be modded into nearly any game supporting DLSS by swapping the DLSS DLL with a translation layer DLL that maps the DLSS API calls to FSR 2 API calls.[33]


  1. ^ FSR versions stated in italic present hotfixes or minor updates.
  2. ^ The algorithm does not necessarily need to be implemented using these presets; it is possible for the implementer to define custom input and output resolutions.
  3. ^ The linear scale factor used for upsampling the input resolution to the output resolution. For example, a scene rendered at 540p with a 2.00x scale factor would have an output resolution of 1080p.
  4. ^ The linear render scale, compared to the output resolution, that the technology uses to render scenes internally before upsampling. For example, a 1080p scene with a 50% render scale would have an internal resolution of 540p.

Frame Generation

[edit]

FSR 3 adds frame generation, a technique that creates new frames in between existing ones by using motion interpolation. Launching in September 2023, FSR 3 uses a combination of FSR 2 and optical flow analysis, which runs using asynchronous compute (as opposed to Nvidia's DLSS 3 which uses dedicated hardware). Because FSR 3 uses a software-based solution, it is compatible with GPUs from AMD, Nvidia, and Intel as well as the ninth generation of video game consoles. To combat additional latency inherent to the frame generation process, AMD has a driver-level feature called Anti-Lag, which only runs on AMD GPUs.[13]

AMD Fluid Motion Frames (AFMF) is a driver-level frame generation technology launching in Q1 2024 which is compatible with all DirectX 11 and DirectX 12 games, however it runs on RDNA 2 and RDNA 3 GPUs. AFMF uses optical flow analysis but not motion vectors, so it can only interpolate a new frame between two traditionally rendered frames. AFMF currently is not compatible with VSYNC.[13]

Tools

[edit]

The official AMD directory lists:[34]

Name source-code API OS Task
CodeXL CodeXL Direct3D, OpenGL, OpenCL, Vulkan Linux
Windows
software development tool suite that includes a GPU debugger, a GPU profiler, a CPU profiler, a static OpenCL kernel analyzer and various plugins.[35]
static analyzer for AMD CodeXL amd-codexl-analyzer Direct3D, OpenGL, OpenCL Linux
Windows 64bit
Off-line compiler and performance analysis CLI-tool for processing: OpenCL kernels, HLSL shaders and GLSL shaders
part of the AMD CodeXL tools suite
Requires either Radeon Software Crimson Edition or AMD Catalyst to be installed to run this tool.[36]
D3D 12 plug-in for GPU PerfStudio amd-gpuperfstudio-dx12 Direct3D 12 Windows a plug-in to GPU PerfStudio GPU perfstudio[37]
Tootle amd-tootle agnostic Linux
Windows
Triangle Order Optimization Tool; originally developed in 2006; can be easily integrated as part of a rendering or mesh pre-processing tool chain[38] Cf. http://mgarland.org/files/papers/quadrics.pdf

Having been released by ATI Technologies under the BSD license in 2006 HLSL2GLSL is not part of GPUOpen. Whether similar tools for SPIR-V will be available remains to be seen, as is the official release of the Vulkan (API) itself. Source-code that has been defined as being part of GPUOpen is also part of the Linux kernel (e.g. amdgpu and amdkfd[39]), Mesa 3D and LLVM.

Software development kits

[edit]
Name Source API OS Task
Advanced Media Framework (AMF) SDK GitHub DirectX 12 Linux, Windows 64-bit Light-weight, portable multimedia framework that abstracts away most of the platform and API-specific details.
AMD GPU Services (AGS) SDK GitHub DirectX Windows 64-bit
LiquidVR SDK GitHub Direct3D 11 Windows improves the smoothness of virtual reality.[40] The aim is to reduce latency between hardware so that the hardware can keep up with the user's head movement, eliminating the motion sickness. A particular focus is on dual GPU setups where each GPU will now render for one eye individually of the display
Radeon Machine Learning (RML) SDK GitHub DirectX 12, Metal, OpenCL Linux, OS X, Windows
Radeon ProRender SDK (formerly FireRender) GitHub OpenCL Linux, macOS, Windows physically-based rendering engine
RadeonRays SDK (formerly FireRays) GitHub DirectX 12, Vulkan Linux 64-bit, OS X, Windows 64-bit A high efficiency, high performance heterogeneous ray tracing intersection library for GPU and CPU or APU on any platform.
RapidFire SDK GitHub DirectX, OpenGL Windows facilitates the use of AMD's video compression acceleration SIP blocks VCE (H.264 encoder) and UVD (H.264 decoder) for "Cloud gaming"/off-site rendering
True Audio Next (TAN) SDK GitHub OpenCL Windows 64-bit SDK for Radeon GPU accelerated and multi-core high-performance audio signal processing.

Professional Compute

[edit]

As of 2022, AMD compute software ecosystem is regrouped under the ROCm metaproject.

AMD Boltzmann Initiative: amdgpu (Linux kernel 4.2+) and amdkfd (Linux kernel 3.19+)

Software around Heterogeneous System Architecture (HSA), General-Purpose computing on Graphics Processing Units (GPGPU) and High-Performance Computing (HPC)

Radeon Open Compute (ROCm)

[edit]

AMD's "Boltzmann Initiative" (named after Ludwig Boltzmann) was announced in November 2015 at the SuperComputing15[41][42][43][44][45] and productized as the Radeon Open Compute platform (ROCm). It aims to provide an alternative to Nvidia's CUDA which includes a tool to port CUDA source-code to portable (HIP) source-code which can be compiled on both HCC and NVCC.

  • Radeon Open Compute Kernel (ROCK) driver
  • Radeon Open Compute Runtime (ROCR) runtime
  • HCC: Heterogeneous Compute Compiler
  • HIP: C++ Heterogeneous-Compute Interface for Portability

Heterogeneous System Architecture

[edit]

Various deprecated

[edit]
  • clFFT library for Fast Fourier transform written in OpenCL
  • hcFFT library for Fast Fourier transform written in HCC-optimized C++

Availability

[edit]

GPUOpen are available under the MIT license to the general public through GitHub starting on January 26, 2016.[4]

There is interlocking between GPUOpen and well established and widespread free software projects, e.g. Linux kernel, Mesa 3D and LLVM.

See also

[edit]

References

[edit]
  1. ^ AMD: GPUOpen (2016-01-26). "Welcome to GPUOpen". Archived from the original on 2016-03-05.
  2. ^ a b c Tom's Hardware (2015-12-15). "AMD GPUOpen: Doubling Down On Open-Source Development". Retrieved 2016-01-24.
  3. ^ Maximum PC (2015-12-15). "AMD Radeon Technologies Group Summit: GPUOpen and Software". Retrieved 2016-01-24.
  4. ^ a b c AnandTech (2015-12-15). "AMD's GPUOpen bundle of developer tools in 2016". Retrieved 2016-01-24.
  5. ^ Heinz Heise (2015-12-16). "AMDs Open-Source-Initiative GPUOpen: Direkte GPU-Kontrolle und bessere Treiber" (in German).
  6. ^ PC Games Hardware [in German] (2015-12-16). "AMD GPU Open: Radeon-Software wird bald zu 100 % Open-Source". PC Games Hardware (in German).
  7. ^ "It's Time to Open up the GPU". Archived from the original on 2016-03-05.
  8. ^ wccftech.com (2015-12-15). "AMD's Answer To Nvidia's GameWorks, GPUOpen Announced". Retrieved 2016-01-24.
  9. ^ HotHardware (2015-12-15). "AMD Goes Open Source, Announces GPUOpen Initiative, New Compiler And Drivers For Linux And HPC". Retrieved 2016-01-24.
  10. ^ "GPUOpen Effects". GitHub.
  11. ^ "FidelityFX". GitHub. 20 October 2021.
  12. ^ "AMD FidelityFX™ Super Resolution 1 (FSR 1)". AMD GPUOpen. Retrieved 2024-05-25.
  13. ^ a b c d "AMD reveals long-awaited FSR 3 tech and frame gen for every DX11/DX12 game". EuroGamer. 25 August 2023. Retrieved 6 September 2023. FSR 3 is a frame generation solution that operates along similar lines to Nvidia's DLSS 3 - combining frame generation (Fluid Motion Frames) with super resolution upscaling (FSR 2) and latency reduction (Anti-Lag+) in a small number of supported games, with Forspoken and Immortals of Aveum set to first debut the technology. FSR 3 will work on Radeon graphics cards, as well as Nvidia and Intel GPUs.
  14. ^ " "AMD claims there's nothing stopping Starfield from adding Nvidia DLSS". The Verge. 24 August 2023. Retrieved 6 September 2023.
  15. ^ GPUOpen-Effects/FidelityFX-CLI, GPUOpen Effects, 2024-05-21, retrieved 2024-05-25
  16. ^ "AMD FidelityFX Super Resolution is coming soon to GPUOpen - AMD GPUOpen". AMD GPUOpen. June 1, 2021. Retrieved March 21, 2023.
  17. ^ "FidelityFX Super Resolution (FSR) source code is here, along with Unity and UE4 support too! - AMD GPUOpen". AMD GPUOpen. July 15, 2021. Retrieved March 21, 2023.
  18. ^ Sommefeldt, Rys (November 16, 2021). "Release FidelityFX FSR v1.0.2 · GPUOpen-Effects/FidelityFX-FSR · GitHub". GitHub. Retrieved March 23, 2023.
  19. ^ a b "The AMD FidelityFX SDK 1.0 is now available on GPUOpen - AMD GPUOpen". AMD GPUOpen. July 11, 2023. Retrieved September 11, 2023.
  20. ^ "It's time for AMD FidelityFX Super Resolution 2.0 - AMD GPUOpen". AMD GPUOpen. March 17, 2022. Retrieved March 23, 2023.
  21. ^ "It's time to see the FSR 2 source code! - AMD GPUOpen". AMD GPUOpen. June 22, 2022. Retrieved March 23, 2023.
  22. ^ "It's time to upscale FSR 2 even further: Meet FSR 2.1! - AMD GPUOpen". AMD GPUOpen. September 8, 2022. Retrieved March 23, 2023.
  23. ^ Sommefeldt, Rys (September 15, 2022). "Release FidelityFX FSR2 v2.1.1 · GPUOpen-Effects/FidelityFX-FSR2 · GitHub". GitHub. Retrieved March 23, 2023.
  24. ^ Sommefeldt, Rys (October 19, 2022). "Release FidelityFX FSR2 v2.1.2 · GPUOpen-Effects/FidelityFX-FSR2 · GitHub". GitHub. Retrieved March 23, 2023.
  25. ^ "Don't cross the streams! Bust more ghosts with the source code to FidelityFX Super Resolution 2.2 - AMD GPUOpen". AMD GPUOpen. February 16, 2023. Retrieved March 23, 2023.
  26. ^ "AMD FidelityFX Super Resolution 2.2.1 hotfix! - AMD GPUOpen". AMD GPUOpen. June 9, 2023. Retrieved June 14, 2023.
  27. ^ "FidelityFX Super Resolution 2.2.2 (FSR2) - FidelityFX SDK - AMD GPUOpen". AMD GPUOpen. Retrieved September 11, 2023.
  28. ^ "With a seasonal frame of mind, in one fluid motion we've generated the AMD FSR 3 GitHub source code repo for gamedevs everywhere! - AMD GPUOpen". AMD GPUOpen. December 14, 2023. Retrieved March 22, 2024.
  29. ^ Sommefeldt, Rys (December 14, 2023). "Release FidelityFX SDK for FSR3 v3.0.4 · GPUOpen-LibrariesAndSDKs/FidelityFX-SDK · GitHub". GitHub. Retrieved March 22, 2024.
  30. ^ Walker, Alex (June 23, 2021). "AMD's FSR Only Supports 7 Games, But It's Already Super Promising". Kotaku Australia. Retrieved July 11, 2021.
  31. ^ Mujtaba, Hassan (March 23, 2022). "AMD Details FSR 2.0: NVIDIA GeForce 10 & Up Support, High-Quality Upscaling Without Machine Learning, More Quality Modes". Wccftech. Retrieved March 24, 2022.
  32. ^ Schiesser, Tim (October 11, 2023). "AMD FSR 3 Frame Generation Analyzed". TechSpot. Retrieved October 5, 2024.
  33. ^ Sims, Daniel (July 5, 2022). "Unofficial FSR 2.0 mod arrives to several more games including Dying Light 2, RDR 2 and Death Stranding". TechSpot.
  34. ^ "GPUOpen Libraries & SDKS". GitHub.
  35. ^ AMD GPUOpen (2016-04-19). "CodeXL 2.0 made open-source". Archived from the original on 2018-06-27. Retrieved 2016-04-21.
  36. ^ AMD GPUOpen (2016-01-26). "CodeXL Static Analyzer CLI".
  37. ^ AMD GPUOpen (2016-01-26). "Create Your own GPU PerfStudio Direct3D 12 Plugin". Archived from the original on 2019-01-09. Retrieved 2016-01-27.
  38. ^ AMD GPUOpen (2016-01-26). "Have You Tootled Your 3D Models?". Archived from the original on 2019-01-09. Retrieved 2016-01-27.
  39. ^ "Linux kernel 4.2 /drivers/gpu/drm/amd".
  40. ^ Heinz Heise (2015-03-04). "LiquidVR: Neues Virtual-Reality-SDK von AMD" (in German).
  41. ^ AnandTech (2015-11-16). "AMD@SC15: Boltzmann Initiative Announced - C++ and CUDA Compilers for AMD GPUs".
  42. ^ Heinz Heise (2015-11-17). "Supercomputer: AMD startet Software-Offensive "Boltzmann"" (in German).
  43. ^ 3dcenter.org (2015-11-16). "AMDs Boltzmann-Initiative geht direkt gegen nVidias CUDA" (in German).{{cite web}}: CS1 maint: numeric names: authors list (link)
  44. ^ AMD (2015-11-16). "AMD Launches 'Boltzmann Initiative'".
  45. ^ AMD (2015-11-16). "A Defining Moment for Heterogeneous Computing".
[edit]