Skip to content
Games & CGI
  • GPUOpen
  • Games & CGI
  • Professional Compute
  • Archive
  • Like
    Porting your engine to Vulkan or DX12

    Adam Sawicki, a member of AMD RTG’s Game Engineering team, has spent the best part of a year assisting one of the world’s biggest game …

    9 1 07/06/2018
    Understanding GPU context rolls

    If you’ve ever heard the term “context roll” in the context of AMD GPUs — I’ll do that a lot in this post, sorry in …

    12 1 06/29/2018
    Microsoft PIX Introduces AMD-Integrated Plug-In with Occupancy Data Graph

    Microsoft PIX is the premiere integrated performance tuning and debugging tool for Windows game developers using DirectX 12.  PIX enables developers to debug and analyze …

    2 0 06/26/2018
    GDC 2018 Presentation Links

    With GDC 2018 done and dusted, we thought it’d be valuable to reemphasise that all of the presented content from the Advanced Graphics Techniques Tutorial …

    4 0 06/07/2018
    AMD GPU Services 5.2.0

    The AMD GPU Services (AGS) library provides game and application developers with the ability to query information about installed AMD GPUs and their driver, in …

    4 0 05/31/2018
    Radeon GPU Profiler 1.2

    Radeon GPU Profiler 1.2 At GDC 2018 we talked about a new version of RGP that would interoperate with RenderDoc, allowing the two tools to …

    7 0 05/25/2018
    Compressonator V3.0 Release Brings Powerful New 3D Model Features

    Compressonator is a set of tools that allows artists and developers to work easily with compressed assets and easily visualize the quality impact of various …

    5 0 05/14/2018
    TrueAudio Next Version 1.2 Now Posted to Github

    We have posted the version 1.2 update to the TrueAudio Next open-source library to Github. It is available here. This update has a number of …

    4 2 05/10/2018
    Reducing Vulkan API call overhead

    Vulkan™ is designed to have significantly smaller CPU overhead compared to other APIs like OpenGL®. This is achieved by various means – the API is …

    23 0 04/26/2018
    First steps when implementing FP16

    Introduction Half-precision (FP16) computation is a performance-enhancing GPU technology long exploited in console and mobile devices not previously used or widely available in mainstream PC …

    11 0 04/20/2018
    GDC 2018 Presentation: Real-Time Ray-Tracing Techniques for Integration into Existing Renderers

    Real Time Ray Tracing was one of the hottest topics last week at GDC 2018. In this presentation, AMD Software Development Engineer and architect of Radeon …

    12 0 03/29/2018
    Real-Time Ray Tracing with Radeon ProRender

    The level of visual detail required of CAD models for the automotive industry or the most advanced film VFX requires a level of visual accuracy …

    26 4 03/20/2018
    GDC 2018 Presentations

    If you’re into the state of the art in games, especially real-time gaming graphics, your eyes will undoubtedly be on Moscone Center in San Francisco, …

    13 5 02/19/2018
    TrueAudio Next is Now Integrated into Steam Audio

    The long wait is over. The GPU processing power of TrueAudio Next (TAN) has now been integrated into Steam Audio from Valve (Beta 13 release). …

    8 3 02/09/2018
    Radeon GPU Profiler 1.1.1

    Radeon GPU Profiler 1.1.1 With GDC 2018 getting ever closer, we wanted to get one last minor release of RGP out before things get hectic …

    5 0 02/07/2018
    Radeon GPU Profiler 1.1.0

    Radeon GPU Profiler 1.1.0 It feels like just last week that we released Radeon GPU Profiler (RGP) 1.0.3 but my calendar says almost 2 months …

    19 0 12/12/2017
    Deferred Path Tracing By Enscape

    Insights from Enscape as to how they designed a renderer that produces path traced real time global illumination and can also converge to offline rendered image quality

    15 0 12/06/2017
    Compressonator V2.7 Release adds cross platform support and 3D Model compression with glTF v2.0

    We are excited to announce the release of Compressonator V2.7! This version contains several new features and optimizations, including: Cross Platform Support Due to popular demand, …

    5 1 11/27/2017
    Radeon GPU Profiler 1.0.3

    Radeon GPU Profiler 1.0.3 A couple of months on from the release of 1.0.2, we’ve fully baked and sliced 1.0.3 for your low-level DX12- and …

    11 0 10/18/2017
    AMD GPU Services 5.1.1

    The AMD GPU Services (AGS) library provides game and application developers with the ability to query information about installed AMD GPUs and their driver, in …

    7 1 09/27/2017
    CPU core count detection on Windows

    Due to architectural differences between Zen and our previous processor architecture, Bulldozer, developers need to take care when using the Windows® APIs for processor and core enumeration. …

    18 2 09/14/2017
    Stable barycentric coordinates

    The AMD GCN Vulkan extensions allow developers to get access to some additional functionalities offered by the GCN architecture which are not currently exposed in the Vulkan API. One of these is the ability to access the barycentric coordinates at the fragment-shader level.

    8 0 08/30/2017
    Radeon GPU Profiler 1.0.2

    Thanks (again!) Before we dive into a run over the release notes for the 1.0.2 release of Radeon GPU Profiler, we’d like to thank everyone …

    14 0 08/22/2017
    AMD Vega Instruction Set Architecture documentation

    Understanding the instruction-level capabilities of any processor is a worthwhile endeavour for any developer writing code for it, even if the instructions that get executed …

    21 6 08/10/2017
    Understanding Vulkan objects

    An important part of learning the Vulkan API – just like any other API – is to understand what types of objects are defined in it, what they represent and how they relate to each other. To help with this, we’ve created a diagram that shows all of the Vulkan objects and some of their relationships, especially the order in which you create one from another.

    32 1 08/07/2017
    Open-source Radeon ProRender

    Summary In this blog post we are announcing the open-source availability of the Radeon™ ProRender renderer, an implementation of the Radeon ProRender API. We will give …

    22 1 07/30/2017
    Radeon GPU Profiler 1.0

    Introduction and thanks Effective GPU performance analysis is a more complex proposition for developers today than it ever has been, especially given developments in how …

    26 8 07/26/2017
    TressFX 4 Simulation Changes

    TressFX 4 introduces a number of improvements. This blog post focuses on three of these, all of which are tied to simulation: Bone-based skinning Signed distance …

    15 1 07/20/2017
    Vulkan Memory Allocator 1.0

    Full application control over GPU memory is one of the major differentiating features of the newer explicit graphics APIs such as Vulkan® and Direct3D® 12. …

    12 0 07/10/2017
    Compressonator V2.6 Release Adds HDR Tonemapping Compression, New Image Analysis Features

    We are excited to announce the release of Compressonator V2.6. This version contains several new features and optimizations, including: Adaptive Format Conversion for general transcoding operations …

    5 0 07/03/2017
    Vega Frontier : How to for developers

    When getting a new piece of hardware, the first step is to install the driver. You can see how to install them for the Radeon …

    4 0 06/29/2017
    Vega Frontier : How to install the driver

    In this blog we will go through the installation process of the driver for your new Radeon Vega Frontier card. We will go through the …

    4 5 06/29/2017
    Optimizing GPU occupancy and resource usage with large thread groups

    When using a compute shader, it is important to consider the impact of thread group size on performance. Limited register space, memory latency and SIMD occupancy each affect shader performance in different ways. This article discusses potential performance issues, and techniques and optimizations that can dramatically increase performance if correctly applied.

    50 6 05/24/2017
    DirectX12 Hardware Counter Profiling with Microsoft PIX and the AMD Plugin

    The AMD Developer Tools team is thrilled to announce the availability of the AMD plugin for Microsoft’s PIX for Windows tool. PIX is a performance …

    9 0 05/17/2017
    CodeXL 2.3 is released!

    A new version of the CodeXL open-source developer tool is out! Here are the major new features in this release: CPU Profiling Support for AMD …

    6 0 05/10/2017
    Content Creation Tools and Multi-GPU

    When it comes to multi-GPU (mGPU), most developers immediately think of complicated Crossfire setups with two or more GPUs and how to make their game …

    8 0 05/05/2017
    Capsaicin and Cream developer talks at GDC 2017

    Introduction Shortly after our Capsaicin and Cream event at GDC this year where we unveiled Radeon RX Vega, we hosted a developer-focused event designed to …

    6 0 04/05/2017
    Compressonator V2.5 Release Adds Enhanced HDR Support

     BC6 HDR Compression The BC6H codec has been improved and now offers better quality then previous releases, along with support for both 16 bit Half …

    4 0 03/29/2017
    Live VGPR Analysis with Radeon GPU Analyzer

    This article explains how to use Radeon GPU Analyzer (RGA) to produce a live VGPR analysis report for your shaders and kernels. Basic RGA usage …

    10 5 03/21/2017
    The Radeon Loom Stitching Pipeline

    I’m Mike Schmit, Director of Software Engineering with the Radeon Technologies Group at AMD. I’m leading the development of a new open-source 360-degree video-stitching framework …

    9 3 03/15/2017
    AMD LiquidVR MultiView Rendering in Serious Sam VR

    AMD LiquidVR MultiView Rendering in Serious Sam VR with the GPU Services (AGS) Library AMD’s MultiView Rendering feature reduces the number of duplicated object draw …

    9 1 02/27/2017
    TrueAudio Next Demo and Paper at GameSoundCon

    In 2016, AMD brought TrueAudio Next to GameSoundCon. GameSoundCon was held Sept 27-28 at the Millennium Biltmore Hotel in Los Angeles. GameSoundCon caters to game …

    4 2 02/24/2017
    Profiling video memory with Windows Performance Analyzer

    Budgeting, measuring and debugging video memory usage is essential for the successful release of game titles on Windows. As a developer, this can be efficiently achieved with the …

    12 0 02/09/2017
    GDC 2017 Presentations

    Another year, another Game Developer Conference! GDC is held earlier this year (27 February – 3 March 2017) which is leaving even less time for …

    16 5 02/01/2017
    AGS 5.0 – Shader Compiler Controls

    With the launch of AGS 5.0 developers now have access to the shader compiler control API.  Here’s a quick summary of the how and why…. Background …

    18 0 01/12/2017
    Optimizing Terrain Shadows

    There are many games out there taking place in vast environments. The basic building block of every environment is height-field based terrain – there’s no …

    12 2 12/15/2016
    Leveraging asynchronous queues for concurrent execution

    Understanding concurrency (and what breaks it) is extremely important when optimizing for modern GPUs. Modern APIs like DirectX® 12 or Vulkan™ provide the ability to …

    37 12 12/01/2016
    Selecting the Best Graphics Device to Run a 3D Intensive Application

    Summary Many Gaming and workstation laptops are available with both (1) integrated power saving and (2) discrete high performance graphics devices. Unfortunately, 3D intensive application …

    10 2 11/16/2016
    Vulkan and DOOM

    This post is taking a look at some of the interesting bits of helping id Software with their DOOM® Vulkan™ effort, from the perspective of …

    17 1 11/10/2016
    Implementing LiquidVR™ Affinity Multi-GPU support in Serious Sam VR

    This blog is guest authored by Croteam developer Karlo Jez and he will be giving us a detailed look at how Affinity Multi-GPU support was …

    12 0 10/31/2016
    AMD Driver Symbol Server

    When opening a 64-bit crash dump you will find that you will not necessarily get a sensible call stack. This is because 64-bit crash dumps …

    38 7 10/27/2016
    Vulkan barriers explained

    Vulkan™’s barrier system is unique as it not only requires you to provide what resources are transitioning, but also specify a source and destination pipeline …

    27 10 10/18/2016
    VDR Follow Up – Tonemapping for HDR Signals

    This is the third post in the follow up series to my prior GDC talk on Variable Dynamic Range. Prior posts covered dithering, today’s topic …

    7 0 10/05/2016
    Using RapidFire for Virtual Desktop and Cloud Gaming

    Virtual desktop infrastructure systems and cloud gaming are increasingly gaining popularity thanks to an ever more improved internet infrastructure. This gives more flexibility to the …

    7 0 09/27/2016
    AMD TrueAudio Next and CU Reservation – What is the Context?

    As noted in my previous blog, new innovations in virtual reality have spearheaded a renewed interest in audio processing, and many new as well as …

    28 0 09/26/2016
    Anatomy Of The Total War Engine: Part V

    This week marks the last in the series of our regular Warhammer Wednesday blog posts. We’d like to extent our thanks to Creative Assembly’s Lead …

    13 0 08/22/2016
    The Importance of Audio in VR

    Audio Must be Consistent With What You See Virtual reality demands a new way of thinking about audio processing. In the many years of history …

    10 8 08/16/2016
    Anatomy Of The Total War Engine: Part IV

    Happy Warhammer Wednesday! This week Creative Assembly’s Lead Graphics Programmer Tamas Rabel talks about how Total War: Warhammer utilized asynchronous compute to extract some extra …

    8 0 08/16/2016
    Anatomy Of The Total War Engine: Part III

    It’s Wednesday, so we’re continuing with our series on Total War: Warhammer. Here’s Tamas Rabel again with some juicy details about how Creative Assembly brought …

    16 2 08/10/2016
    Blazing CodeXL 2.2 is here!

    A new release of the CodeXL open-source developer tool is out! Here’s the hot new stuff in this release: New platforms support Support Linux systems …

    21 5 08/08/2016
    Anatomy Of The Total War Engine: Part II

    We’re back again on this fine Warhammer Wednesday with more from Tamas Rabel, Lead Graphics Programmer on the Total War series. In last week’s post …

    6 1 08/03/2016
    Anatomy Of The Total War Engine: Part I

    For the next few weeks we’ll be having a regular feature on GPUOpen that we’ve affectionately dubbed “Warhammer Wednesdays”. We’re extremely lucky to have Tamas Rabel, …

    12 10 07/27/2016
    Texel Shading

    Game engines do most of their shading work per-pixel or per-fragment. But there is another alternative that has been popular in film for decades: object …

    24 3 07/21/2016
    Vulkan Device Memory

    EDIT: 2016/08/08 – Added section on Targeting Low-Memory GPUs This post serves as a guide on how to best use the various Memory Heaps and …

    34 2 07/21/2016
    Performance Tweets Series: Root signature & descriptor sets

    Before Direct3D® 12 and Vulkan™, resources were bound to shaders through a “slot” system. Some of you might remember when hardware did have only very …

    4 8 07/14/2016
    Performance Tweets Series: Multi-GPU

    Multi-GPU systems are much more common than you might think. Most of the time, when someone mentions mGPU, you think about high-end gaming machines with …

    14 2 07/05/2016
    Compressonator v2.3 Release Delivers ASTC, ETC2 Codec Support and GPU Rendered Image Views

    Compressonator is a set of tools to allow artists and developers to more easily create compressed texture image assets and easily visualize the quality impact …

    10 0 06/27/2016
    Performance Tweets Series: Debugging & Robustness

    Prior to explicit graphics APIs a lot of draw-time validation was performed to ensure that resources were synchronized and everything set up correctly. A side-effect of this robustness …

    6 0 06/22/2016
    Performance Tweets Series: Rendering and Optimizations

    Direct3D® 12 and Vulkan™ significantly reduce CPU overhead and provide new tools to better use the GPU. For instance, one common use case for the …

    3 0 06/14/2016
    Performance Tweets Series: Streaming & Memory Management

    As promised, we’re back and today I’m going to cover how to get resources to and from the GPU. In the last post, we learned …

    5 0 06/07/2016
    CodeXL 2.1 is out and Searing hot with Vulkan

    A new CodeXL release is out! For the first time the AMD Developer Tools group worked on this release on the CodeXL GitHub public repository, …

    20 0 05/31/2016
    ShadowFX Effect Library for DirectX 12

    Today, we are excited to announce that we are releasing an update for ShadowFX that adds support for DirectX® 12. Features Different shadowing modes Union of …

    12 1 05/26/2016
    Turbocharge your Graphics and GPU Compute Applications with GPUPerfAPI

    Achieving high performance from your Graphics or GPU Compute applications can sometimes be a difficult task. There are many things that a shader or kernel …

    12 3 05/25/2016
    GCN Shader Extensions for Direct3D and Vulkan

    The GCN architecture contains a lot of functionality in the shader cores which is not currently exposed in current APIs like Vulkan™ or Direct3D® 12. One …

    21 11 05/24/2016
    AMD DOPPEngine – Post Processing on Your Desktop in Practice

    A Complete Tool to Transform Your Desktop Appearance After introducing our Display Output Post Processing (DOPP) technology, we are introducing a new tool to change …

    32 0 05/23/2016
    Fast compaction with mbcnt

    Compaction is a basic building block of many algorithms – for instance, filtering out invisible triangles as seen in Optimizing the Graphics Pipeline with Compute. …

    13 8 05/20/2016
    TressFX 3.1

    We are releasing TressFX 3.1. Our biggest update in this release is a new order-independent transparency (OIT) option we call “ShortCut”. We’ve also addressed some of …

    10 4 05/19/2016
    GeometryFX 1.2 – Cluster Culling

    Today’s update for GeometryFX introduces cluster culling. Previously, GeometryFX worked on a per-triangle level only. With cluster culling, GeometryFX is able to reject large chunks …

    7 7 05/18/2016
    Unlock the Rasterizer with Out-of-Order Rasterization

    Full-speed, out-of-order rasterization If you’re familiar with graphics APIs, you’re certainly aware of the API ordering guarantees. At their core, these guarantees mean that if …

    15 6 05/17/2016
    AMD FireRays 2.0 – Open Sourcing and Customizing Ray Tracing for Efficient Hardware Platforms Support

    A New Milestone After the success of the first version, FireRays is moving to another major milestone. We are open sourcing the entire library which …

    29 2 05/16/2016
    Slides from our “The most common Vulkan mistakes” talk

    Last week, we organized a two hours-long talk at University of Lodz in Poland where we discussed the most common mistakes we come across in Vulkan applications. Dominik Witczak, …

    8 5 05/13/2016
    Compressonator (AMD Compress) is Going Open Source

    We are very pleased to be announcing that AMD is open-sourcing one of our most popular tools and SDKs.  Compressonator (previously released as AMD Compress …

    17 0 05/12/2016
    AMD Crossfire API

    Gaming at optimal performance and quality at high screen resolutions can sometimes be a demanding task for a single GPU. 4K monitors are becoming mainstream and gamers …

    18 5 05/05/2016
    AMD GPU Services, an introduction

    If you have supported Crossfire™ or Eyefinity™ in your previous titles, then you have probably already used our AMD GPU Services (AGS) library.  A lot of …

    17 1 04/28/2016
    Performance Tweets Series: Resource Creation

    Resource creation and management has changed dramatically in Direct3D® and Vulkan™ compared to previous APIs. In older APIs, memory is managed transparently by the driver. …

    8 0 04/20/2016
    CodeXL 2.0 is Here and Open Source

    CodeXL major release 2.0 is out! It is chock-full of new features and a drastic change in the CodeXL development model: CodeXL is now open …

    26 6 04/19/2016
    VDR Follow Up – Grain and Fine Details

    The prior post in this series established a base technique for adding grain, and now this post is going to look at very subtle changes to …

    3 0 04/13/2016
    Performance Tweets Series: Shaders, Threading, Compiling

    Welcome back to our performance & optimization series. Today, we’ll be looking more closely at shaders. On the surface, it may look as if they …

    12 1 04/06/2016
    VDR Follow Up – Fine Art of Film Grain

    This is the first of a series of posts expanding on the ideas presented at GDC in the Advanced Techniques and Optimization of VDR Color …

    8 1 04/04/2016
    GDC 2016 Presentations Available

    The Game Developer Conference 2016 was an event of epic proportions. Presentations, tutorials, round-tables, and the show floor are only one part of the story …

    8 2 03/30/2016
    GCN Memory Coalescing

    This post describes how GCN hardware coalesces memory operations to minimize traffic throughout the memory hierarchy. The post uses the term “invocation” to describe one …

    13 1 03/21/2016
    Delta Color Compression Overview

    Bandwidth is always a scarce resource on a GPU. On one hand, hardware has made dramatic improvements with the introduction of ever faster memory standards …

    24 10 03/14/2016
    Using the Vulkan™ Validation Layers

    Vulkan™ provides unprecedented control to developers over generating graphics and compute workloads for a wide range of hardware, from tiny embedded processors to high-end workstation GPUs with wildly different …

    41 6 03/09/2016
    GDC 2016 Presentations

    The Game Developer Conference 2016 (GDC16) is held March 14-18 in the Moscone Center in San Francisco. This is the most important event for game developers, …

    36 0 02/29/2016
    Performance Tweets series: Barriers, fences, synchronization

    Welcome back to our DX12 series! Let’s dive into one of the hottest topics right away: synchronization, that is, barriers and fences! Barriers A barrier is …

    15 1 02/22/2016
    Vulkan Renderpasses

    Vulkan™ is a high performance, low overhead graphics API designed to allow advanced applications to drive modern GPUs to their fullest capacity. Where traditional APIs …

    58 0 02/16/2016
    Say Hello to a New Rendering API in Town!

    Imagine that you were asked one day to design an API with bleeding-edge graphics hardware in mind. It would need to be as efficient as …

    45 3 02/16/2016
    Performance Tweets Series: Command lists

    Hello and welcome to our series of blog posts covering performance advice for Direct3D® 12 & Vulkan™. You may have seen the #DX12PerfTweets on Twitter, and …

    10 1 02/10/2016
    Fetching From Cubes and Octahedrons

    For GPU-side dynamically generated data structures which need 3D spherical mappings, two of the most useful mappings are cubemaps and octahedral maps. This post explores …

    10 0 02/04/2016
    It’s Time to Open up the GPU

    I have met enough game developers in my professional life to know that these guys are among the smartest people on the planet. Those particular individuals will go …

    412 34 01/26/2016
    Up and Running with CodeXL Analyzer CLI

    About CodeXL Analyzer CLI CodeXL Analyzer CLI is an offline compiler and performance analysis tool for OpenCL™ kernels, DirectX® shaders and OpenGL® shaders. Using CodeXL …

    24 0 01/26/2016
    Create Your own GPU PerfStudio DirectX® 12 Plugin

    GPU PerfStudio supports DirectX® 12 on Windows® 10 PCs. The current tool set for DirectX 12 comprises of an API Trace, a new GPU Trace …

    10 0 01/26/2016
    Maxing out GPU usage in nBodyGravity

    Today we’re going to take a look at how asynchronous compute can help you to get the maximum out of a GPU. I’ll be explaining …

    26 6 01/26/2016
    Have You Tootled Your 3D Models?

    What’s New With the recent adoption of new APIs such as DirectX® 12 and Vulkan™, we are seeing renewed interest in an older tool.  AMD …

    35 0 01/26/2016
    Optimized Reversible Tonemapper for Resolve

    A typical problem with MSAA Resolve mixed with HDR is that a single sample with a large HDR value can over-power all other samples, resulting …

    18 1 01/26/2016

    Understanding GPU context rolls

    Posted on June 29, 2018 by Rys Sommefeldt
    context rolls, optimization, performance, profiling, RGP

    If you’ve ever heard the term “context roll” in the context of AMD GPUs — I’ll do that a lot in this post, sorry in advance — chances are you maybe have an intuitive instinct for what the term means. But I’d bet you’re probably missing some of the detail about the actual GPU- and driver-side mechanics, maybe aren’t sure what “roll” means, or maybe you’re even thinking about a context at the wrong conceptual level.

    Regardless, let me clear it up so that you know definitively what a context roll on our GPUs is, how they apply to the pipeline and how they’re managed, and what you can do to analyse them and find out if they’re a limiting factor in the performance of your game or application. Note that everything I write here applies to modern AMD Radeon™ GPUs based on GCN.

    What’s a context in the context of a context roll?

    First, let’s clear up what we mean by context. If you’ve done any GPU programming, it’s most common that context refers to a single instance of the API mechanism that lets you issue commands to a GPU device. In that context (I’m not sorry), it’s a top-level software construct and you generally only have one held in your game or app while using the GPU.

    That’s not what a context is when we talk about AMD GPU context rolls, though. Context there means what the GPU hardware understands about the current state it needs to be in to correctly draw. You can adjust various bits of per-draw state on a GPU; things like which order the front-end should process vertices in a triangle, or the mode the blender should be in to correctly blend values. Those pieces of mutable GPU state apply to things across the whole GPU pipeline.

    So, when we talk about context in terms of context rolls, we mean the required GPU pipeline state needed to correctly draw something.

    The AMD GPU roll mechanism

    Now that we know that context refers to the operating state of the GPU during a draw, let’s talk about some GPU hardware specifics. The hardware maintains 8 banks of context registers to store operating state, but the count of 8 is just an implementation detail. We reserve bank 0 for clear state, so there are 7 for normal use. Those 7 usable banks are just a logical construct though; the real implementation in hardware is a bit more complex.

    The copies of the context registers are stored in banked RAMs, but they’re distributed across the GPU design and stored in the blocks that need them, rather than being stored all together in a single physical RAM. There’s a hardware block that manages setting new values that talks to the individual blocks that have context registers, broadcasting any updates chip-wide to the blocks that need to know.

    Each individual block is then responsible for managing its copies. In terms of how that management block is driven, it’s all down to the command processor, or CP, which is driven by the driver. I’ll talk about that shortly.

    Now that you know the GPU maintains multiple copies of context registers, a “roll” is when the CP asks the management block to start working on a new set of the registers because some pipeline state needs to be changed to draw correctly. It does that by asking each block to take a copy of the current registers into a free bank, so that everything is ready to apply any new state updates by programming fresh register values.

    The driver roll mechanism

    The CP doesn’t operate autonomously to manage that state. Instead, it just does what the driver tells it to do. One of the advantages about writing this post now, instead of 8 months ago when I first added it to my list, is that we’ve since open sourced our Vulkan driver, including the lower-level platform abstraction layer (PAL) that the Vulkan user-mode driver is built on!

    The driver, via PAL, is responsible for instructing the CP to set the GPU up correctly to work on a draw, and it does so using a packetised command format we call PM4. And now, instead of PM4 being an opaque thing I can’t really talk about, the driver source that builds and issues PM4 packets to the CP is right there for you to read if you want to.

    You don’t need to go read it in order to understand context rolls, but if you’re curious, head over to the PAL repository on GitHub. Check out the GFX9 PM4 opcodes for example, or search for EOP (end-of-pipe, which I’ll come to later) or CmdUtil::BuildLoadContextRegs() in the repository. That’s the function which builds the PM4 packet that uses the LOAD_CONTEXT_REG PM4 opcode to program the context registers.

    So when you change any GPU pipeline states, the usermode part of the driver that accepts those calls issues its own calls into PAL. PAL then builds the right PM4 packets with the right register ranges, all in order to make the context state update happen at the right time. Say you issue 5 draws to the API from your game, 2 with one set of GPU pipeline state and 3 with another. You’ll ideally get a PM4 command stream with this kind of (pseudo) annotated packet order:


    IT_SET_CONTEXT_REG; # program some state
    IT_DRAW_INDEX_2; # draw 0
    IT_DRAW_INDEX_2; # draw 1
    IT_SET_CONTEXT_REG; # program some more state
    IT_DRAW_INDEX_2; # draw 2
    IT_DRAW_INDEX_2; # draw 3
    IT_DRAW_INDEX_2; # draw 4

    Performance impact

    Now we know how the GPU manages state per-block, and how the CP is instructed to set it and kick new draws afterwards, what about the performance impact?

    There’s no performance impact as long as there’s a free logical context register bank to roll any new context into. But what if all 7 banks are full with needed states for in-flight draws? If your draws contain enough work in them then there still shouldn’t be any impact, since the GPU will likely still be busy as a draw drains out of the GPU pipeline, and the associated end-of-pipe work happens, freeing the logical context and therefore letting the CP roll a new one for a fresh draw.

    However if your draws don’t contain much work, you could end up incurring overhead from stalls relating to the GPU not being terribly busy, and there being no free logical contexts to start a new draw and fill the GPU up with work again. In a well-written game or app that won’t happen, but you might find yourself in that position during development. How do you spot that and mitigate it by making changes to your renderer?

    The best way to avoid being context starved is to be sensible about how you submit work to the GPU, usually via batching. Batch and submit draws with the same materials and/or rendering states as much as possible, so that back-to-back draws ideally don’t need to roll to a new context at all. Otherwise the multiple banks of context registers are there for a reason!

    Analysing context rolls

    The CP knows when there are no freely available logical context register banks and stalls until one is free if needed. But how do you get notified as the developer? We have context roll analysis in Radeon GPU Profiler!

    RGP will let you interrogate your frame and it’ll show you in a number of ways what’s happening with contexts. There’s explicit notification in the overview panel, plus you can also colour your draws so that there’s a different colour per device context being used. So, if the timeline looks like a unicorn came to visit, check the overview panel to find out if the CP couldn’t start new draws because you rolled too much 🌈.

    Summary

    Hopefully that’s a decent overview of how the GPU views context state in hardware, how the driver works to program it, and how you can analyse what’s happening in your app with RGP so you can avoid unnecessary rolls.

    The driver will try its best to help, but you have the best view of what’s happening during rendering, plus we try very hard to keep the driver light in terms of CPU usage, so spending cycles on context management is best left to the place it can happen most efficiently: your game or app.

    And if you have any further questions about context rolls, just ask!

    Rys Sommefeldt looks after the Game Engineering group in Europe, which is part of the Radeon Technologies Group at AMD. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.

    1 Comment

    Comment
    steve m says:
    July 10, 2018 at 1:18 am

    Thanks for the article! A couple of questions still remain:

    1) If I recall, there are some API calls that don’t roll context, and some that do. Is there a way to form some intuition about which calls will roll?

    2) Just to make it clear, the GPU can be working on several draw calls at the same time with different contexts, correct? Do all units of the same type use the same context at any one time, or can different units of the same type use different contexts as they’re working on different draw calls (let’s say shader units)?

    Reply

    Leave a Reply Cancel reply

    Your email address will not be published. Required fields are marked *

    • Terms and Conditions
    • /
    • Privacy
    • /
    • Cookie Policy
    • /
    • Trademarks
    ©2018 Advanced Micro Devices, Inc. OpenCL™ and the OpenCL™ logo are trademarks of Apple, Inc., used with permission by Khronos.