GraphicsResearch » History » Version 51

dl lud, 07/21/2020 04:10 PM
kmsro: link to Mesa's source code in order to exemplify scanout buffer create and PRIME sharing.

1 1 Wolfgang Wiedmeyer
h1. Research on free graphics-related software
3 12 Wolfgang Wiedmeyer
5 5 Wolfgang Wiedmeyer
On this page, information is collected that could help solving graphics issues in Replicant (see #1539). Besides evaluating free implementations that are relevant for currently supported devices, other implementations should also be listed if they are useful for potential future target devices.
6 1 Wolfgang Wiedmeyer
7 46 dl lud
External resources:
"Free and open-source graphics device drivers - Wikipedia":
"Mobile drivers - Debian Wiki":
10 8 Wolfgang Wiedmeyer
11 21 Denis 'GNUtoo' Carikli
h2. Multiple backends
12 35 Jeremy Rand
13 46 dl lud
"Replicant 6.0 0004 RC1": includes a [[Graphics#Choosing-the-software-renderer-for-a-specific-app|mechanism]] that allows choosing between different OpenGL ES implementations for each app or process. This came into place as means to achieve a balance between fully-compliant but slower implementations (e.g. llvmpipe) and non-compliant but fast implementations (e.g. libagl).
14 1 Wolfgang Wiedmeyer
15 46 dl lud
A similar mechanism may be used in future Replicant versions, to take advantage of GPU backed implementations, even if they do not achieve full OpenGL ES compliance.
16 1 Wolfgang Wiedmeyer
h2. Software rendering
18 21 Denis 'GNUtoo' Carikli
19 46 dl lud
Software rendering uses the CPU and not a dedicated graphics processor for graphics rendering. It is slower than per-GPU implementations and is mostly used as a fallback, when GPU acceleration is not available. An advantage is that the same software renderer can work across many different types of hardware. Therefore, improving a software renderer benefits many different devices, regardless of the SoC and graphics unit that they have. Furthermore, a software renderer doesn't require a kernel driver. This makes it easier to work on mainline Linux kernel support for a device, until the graphics driver is in mainline.
20 1 Wolfgang Wiedmeyer
21 46 dl lud
h3. OpenGL ES software renderers
22 2 Wolfgang Wiedmeyer
23 46 dl lud
h4. libagl
24 1 Wolfgang Wiedmeyer
25 46 dl lud
* Source code:
27 45 dl lud
libagl is the fastest software renderer available for Replicant devices. It is used by default on all Replicant-supported devices up until Replicant 6.0 0003 (0004 switched to llvmpipe).
28 1 Wolfgang Wiedmeyer
29 45 dl lud
libagl was developed specifically for Android and it is part of the AOSP source code. The renderer includes optimizations for ARM via "libpixelflinger": and "codeflinger": that do JIT compilation into platform optimized code. Development ceased in 2013 and no work was done to support newer OpenGL ES (GLES) versions (which causes #705).
31 26 dl lud
Until ca. 2011 (Android 4.0), another library with the name @libagl2@ existed and @surfaceflinger2@ was developed based on Mesa. The source code was later removed and no further development is known. At the time, the work was done to support a newer GLES version for the software renderer. It was abandoned later, probably when Google made it mandatory for Android 4.0 and later devices to pack their own hardware GPU with OpenGL ES 2.0 support. It is questionable if it is worth it to port the old @libagl2@ library to a recent Replicant version, also given that we would be the only ones using and maintaining the code.
32 1 Wolfgang Wiedmeyer
33 46 dl lud
h4. Mesa's llvmpipe
34 1 Wolfgang Wiedmeyer
35 46 dl lud
* Replicant llvmpipe support source code:
* Replicant documentation: [[Graphics]]
* Upstream documentation:
* Upstream performance improvement documentation:
39 20 Denis 'GNUtoo' Carikli
40 1 Wolfgang Wiedmeyer
Mesa's llvmpipe is the default software renderer for Replicant since version 6.0 0004. It has a GLES implementation that is more complete than libagl, although slower when used by some system components (e.g. SurfaceFlinger).
Besides llvmpipe, Mesa has two other software rasterizers: swrast/swr and softpipe, but both are of no interest. swrast's GLES implementation is incomplete and this driver is mostly deprecated in favor of those built upon Gallium (softpipe and llvmpipe). softpipe on the other hand, "is as complete as llvmpipe": but is slower than it.
43 45 dl lud
44 1 Wolfgang Wiedmeyer
The "Android-x86 project": is using llvmpipe. A few of their Android-specific framework patches are applied in Replicant 6.0. Their Mesa source code fork is also used in Replicant 6.0. A lot of porting work of llvmpipe to Android was done by Jide while Intel is contributing as well. So there is an interest from different parties to have llvmpipe working on Android. Android patches are upstreamed to mainline Mesa.
llvmpipe is still not ported to ARM which makes it slow. Also for Android, it is mostly used on the x86 platform in other projects. See #705 for more information. Optimizing llvmpipe for ARM seems currently the most promising approach to fix graphics-related issues with Replicant.
48 46 dl lud
h5. Tuning llvmpipe
50 36 Jeremy Rand
The following environment variables might speed up llvmpipe somewhat:
52 46 dl lud
53 1 Wolfgang Wiedmeyer
LP_PERF=no_mipmap,no_linear,no_mip_linear,no_tex,no_blend,no_depth,no_alphatest MESA_NO_DITHER=1
54 46 dl lud
55 1 Wolfgang Wiedmeyer
56 46 dl lud
@LP_PERF@ seems to be undocumented; @MESA_NO_DITHER@ is documented in the Mesa performance tips. According to adjtm, @MESA_NO_DITHER=1@ improved glxgears performance by 3% on a GNU/Linux PC without breaking any apps that they tested. @LP_PERF=no_mipmap,no_linear,no_mip_linear@ also didn't break any apps on GNU/Linux in tests but there was no noticeable increase in performance. @no_tex,no_blend,no_depth,no_alphatest@ broke rendering of all tested apps in GNU/Linux.
57 1 Wolfgang Wiedmeyer
58 46 dl lud
To test with those environment variables on a per-app basis, you can disable SELinux, and then run the following, substituting the name of the app (here @info.guardianproject.orfox@) for the one you wish to test; then run the app from the launcher:
59 1 Wolfgang Wiedmeyer
60 46 dl lud
61 1 Wolfgang Wiedmeyer
setprop "LP_PERF=no_mipmap,no_linear,no_mip_linear,no_tex,no_blend,no_depth,no_alphatest MESA_NO_DITHER=1"
62 46 dl lud
63 1 Wolfgang Wiedmeyer
64 46 dl lud
Unfortunately, wrap seems to crash Replicant sometimes. There was allegedly a fix at .
65 1 Wolfgang Wiedmeyer
Setting those environment variables globally might be possible by editing .
Benchmarking performance in real-world apps might be feasible via this script:
70 46 dl lud
h4. SwiftShader
71 1 Wolfgang Wiedmeyer
72 46 dl lud
* Source code:
73 1 Wolfgang Wiedmeyer
74 46 dl lud
"Google released SwiftShader as free software in mid 2016": It supports x86 and ARM architectures with SDIV/UDIV support. It is used in the Chromium project but also with Android, for example in the Android-x86 project. Swiftshader doesn't seem to depend on any external libraries besides what are provided in its Git repository, therefore it is very easy to compile it and use it as a software renderer for Replicant. All the Android build files are provided so you just need to the add the following packages to PRODUCT_PACKAGES in order to be able to use it:
75 1 Wolfgang Wiedmeyer
76 46 dl lud
 * libGLESv2_swiftshader
 * libEGL_swiftshader
78 1 Wolfgang Wiedmeyer
79 46 dl lud
SwiftShader features a broadly compliant GLES implementation like llvmpipe.
80 36 Jeremy Rand
81 46 dl lud
h3. Vulkan software renderers
82 36 Jeremy Rand
83 46 dl lud
h4. SwiftShader
84 36 Jeremy Rand
85 46 dl lud
Apart from GLES, SwiftShader also supports Vulkan. Actually, Vulkan is SwiftShader's main focus now.
h4. Vallium
* Source code:
Vallium is a Vulkan software renderer based on Mesa's llvmpipe and Gallium. It is still a work-in-progress, led single-handedly by David Airlie from Red Hat. Hopefully it will get merged into Mesa once deemed mature enough.
h4. Kazan
* Source code:
Kazan is another work-in-progress Vulkan software renderer. It is an independent stack, not relying on any underlying graphics library. However, much like Vallium and SwiftShader, it uses LLVM for its shader compiler. Kazan is programmed in Rust.
h3. kms_swrast
"kms_swrast": is a Mesa driver, built upon Gallium, that uses DRM nodes for memory allocation and to present images on the display, but does the actual OpenGL rendering through a Mesa software renderer such as llvmpipe or softpipe.
It may allow noticeable performance improvements by avoiding expensive memory copy operations between the software renderer and DRM.
Exynos based devices have a working free-software DRM driver that can be used with kms_swrast. Devices that do not have DRM driver can still benefit from kms_swrast by usage of the "VGEM (Virtual GEM) driver":
106 34 dl lud
h3. Pixman
108 46 dl lud
* Source code:
109 34 dl lud
> "Pixman": is a low-level software library for pixel manipulation, providing features such as image compositing and trapezoid rasterization.
It is highly optimized for ARM processors and has a "fast path for ARM NEON": It may be worthwhile to write an OpenGL ES backend for Replicant that detects 2D operations and translates them to Pixman. This could provide a considerable performance improvement over llvmpipe or SwiftShader and, if complete enough, could even replace them.
114 1 Wolfgang Wiedmeyer
On the other hand, writing such translation layer may prove to be an enormous task, as the OpenGL ES 2.0 API is quite extensive. In this scenario we can still benefit from Pixman by reproducing it's ARM NEON fast paths on llvmpipe.
116 9 Wolfgang Wiedmeyer
h2. Per-GPU implementations
117 1 Wolfgang Wiedmeyer
118 23 dl lud
In the following, free software implementations are listed that should make it possible to use the respective GPU with free software.
119 7 Wolfgang Wiedmeyer
120 42 Kurtis Hanna
h3. ARM Mali-4xx (Utgard) with Lima
121 7 Wolfgang Wiedmeyer
122 22 dl lud
Supported devices that use Mali 400: Galaxy S 2, Galaxy S 3, Galaxy S 3 4G, Galaxy Note, Galaxy Note 2, Galaxy Note 8.0
123 1 Wolfgang Wiedmeyer
124 47 dl lud
* Project page:
* Source code:
126 7 Wolfgang Wiedmeyer
127 47 dl lud
Lima is now merged into Mesa and should be compatible with GLES 2.0. GLES 3 is not planned because Utgard's "programmable pipeline" is not much more flexible than a fixed function pipeline, and cannot cope with the GLES 3 features which usually require a unified shader model.
128 1 Wolfgang Wiedmeyer
Lima saw first light with the reverse-engineering efforts by Luc Verhaegen, that produced an experimental driver with it's original page at  
130 16 Jeremy Rand . Luc's development stopped around 2014, but in 2017 Qiang Yu took on the task and started development on top of Mesa's Gallium3D driver as "reported by Phoronix": 
131 47 dl lud
The code was then hosted at "'s GitLab": and later merged into mainline Mesa. It gets contributions by many developers besides Qiang Yu.
132 1 Wolfgang Wiedmeyer
133 47 dl lud
Apart from the driver in Mesa, Lima also has a "corresponding driver in the Linux kernel":
134 1 Wolfgang Wiedmeyer
h3. Imagination PowerVR
136 5 Wolfgang Wiedmeyer
137 47 dl lud
Supported devices that use PowerVR SGX540: Nexus S, Galaxy S, Galaxy Nexus, Galaxy Tab 2 7.0, Galaxy Tab 2 10.1
Supported devices that use PowerVR SGX530: GTA04
139 10 Wolfgang Wiedmeyer
140 8 Wolfgang Wiedmeyer
A reverse-engineering project exists:
141 47 dl lud
* Website:
* Mailing list:
* Libreplanet group:
144 23 dl lud
Despite initial reverse-engineering progress until 2013, no further development updates seem to be available. The project website provides documentation.
h3. ARM Mali-T6xx/T7xx/T8xx (Midgard) and G7x (Bifrost) with Panfrost
149 47 dl lud
Not used by a supported device.
Used on several Samsung Galaxy S series phones that are now supported by LineageOs and may be potential targets for Replicant ports: S6 (Mali T760MP8), S7 (Mali-T880 MP12) and S9 (Mali-G72 MP18).
151 23 dl lud
152 47 dl lud
* Project page:
* Source code:
154 33 dl lud
155 47 dl lud
Panfrost is another reverse-engineered driver for ARM Mali GPUs based on Mesa's Gallium3D, but aimed at the latest architectures (Midgard and Bifrost). It is merged into Mesa and under active development with a strong community (as of July 2020). On Midgard it is now "mostly GLES 2.0 conformant": and has "initial GLES 3 support":, while Bifrost support is still lagging behind. According to the developers, Vulkan support should be achievable with both Midgard and Bifrost architectures.
Panfrost's main focus are the ARM SoCs present on some laptops and single-board computers (SBC), but it is "intended to work on as many SoCs as possible to make everyone's lives easier":
157 1 Wolfgang Wiedmeyer
Panfrost is developed by Alyssa Rosenzweig and Lyude Paul, and recently attracted other contributors such as "Tomeu Vizoso": and "Rob Herring":
158 33 dl lud
Development can be followed on "Alyssas's blog":
159 23 dl lud
160 1 Wolfgang Wiedmeyer
h3. Qualcomm Adreno with freedreno
161 24 dl lud
162 47 dl lud
Not used by any supported device.
163 1 Wolfgang Wiedmeyer
164 47 dl lud
* Wiki:
165 9 Wolfgang Wiedmeyer
freedreno is actively developed. Linux kernel component is available in mainline since version 3.12. It needs to be investigated how well freedreno could work on potential target devices. However, even when using freedreno, non-free firmware is very likely still needed.
Generally speaking, Qualcomm devices have a lot of blobs and no modem isolation which is the reason why no Qualcomm-based device is yet supported by Replicant. See [[TargetsEvaluation]] for some analysis. We have not yet identified a Qualcomm-based device that would be a promising target for Replicant and where freedreno could be used on.
170 15 Wolfgang Wiedmeyer
h3. Vivante GCxxxx with Etnaviv
171 24 dl lud
172 47 dl lud
Not used by any supported device.
173 15 Wolfgang Wiedmeyer
174 47 dl lud
* Project page:
* Source code:
Etnaviv is a driver for some Vivante GCxxxx GPU variants based on Mesa's Gallium3D. It would be interesting to investigate whether devices using such a GPUs are good Replicant targets. We also have check if non-free firmware need to be loaded to the GPU.
178 29 dl lud
179 49 dl lud
h3. kmsro
181 51 dl lud
"kmsro": (kernel mode-setting render-only) is a Mesa's component that glues display-only (kms) with render-only drivers. It allows PRIME buffer sharing between these two counterparts.
It can also "allocate scanout buffers using the display-only driver (kms node)":, catering for cases when only the display driver can allocate the appropriate memory (e.g. contiguous memory). These buffers are then shared through PRIME and imported at the render-only driver (e.g. "lima_bo.c":
183 50 dl lud
h4. Background
185 1 Wolfgang Wiedmeyer
The graphics cards present on most desktop and laptop computers include both a computing unit (GPU) and display controller circuitry, i.e., the display is directly connected to the graphics card. However, on the SoCs embedded on most mobile devices, these two functionalities are split between two different hardware components, there is:
187 51 dl lud
* a render-only GPU, with no connections for displays;
* a display controller, where the displays are attached, that has no computing capabilities except for hardware compositing (blending of images/planes).
190 50 dl lud
These two components then share image buffers through the main memory (RAM).
On the Linux kernel DRM subsystem, these appear as independent devices, handled by different drivers.
Along the years, Mesa got support for several render-only GPUs, through Gallium drivers like Lima, Panfrost, Freedreno and Etnaviv. The GPUs handled by these drivers are available on many different SoCs, with varying display controllers: exynos, imx, armada, etc.
The number of possible combinations, and thus code duplication, grew in Mesa. kmsro was then introduced as an abstraction layer to get rid of this code duplication.
195 49 dl lud
196 29 dl lud
h2. Gralloc
198 48 dl lud
In order to have a working software rendering on Replicant 10 we will need a gralloc (graphics memory allocator) library that:
199 29 dl lud
* Implements the "Android Gralloc HAL": API version 1.
* Is compatible with "drm-hwcomposer": (a libre implementation, based on Linux's DRM, of Android's Hardware Composer HAL).
* Is compatible with Mesa and particularly its llvmpipe software renderer.
202 30 dl lud
* (optional) Is compatible with SwiftShader.
203 29 dl lud
There are currently 3 free-software grallocs under evaluation:
205 1 Wolfgang Wiedmeyer
* "drm_gralloc": - Directly uses Linux DRM for buffer allocation. Built by the Android-x86 team and now getting phased out in favor of gbm_gralloc.
206 48 dl lud
* "gbm_gralloc": - Uses Mesa's GBM (Generic Buffer Management) for buffer allocation through libgbm. GBM itself will then call DRM.
207 30 dl lud
* "minigbm": - Uses its own embedded GBM implementation. Does not depend on Mesa. Built by Intel for their Android-IA project now dubbed "Project Celadon": Also used by Google "in ChromiumOS":
The table bellow summarizes the capabilities of these 3 gralloc implementations (sourced from the "slides": of "Mauro's Rossi talk at XDC 2018":
|_. project |_. API version |_. GEM/flink names |_. PRIME fd |_. binderization |
|=. drm_gralloc |=. 0 |=. Yes |=. incomplete |=. No |
|=. gbm_gralloc |=. 0 |=. N/A |=. Yes |=. Yes |
|=. minigbm |=. 0, 1 |=. N/A |=. Yes |=. Yes |
* *project*: the implementation name.
* *API version*: lists the supported versions of Android's gralloc HAL.
* *GEM/flink names*: whether there is support for GEM (Graphics Execution Manager) object sharing through GEM names. GEM names are unique 32 bit integers, created via the flink operation, that point to a unique GEM object inside a DRM device. GEM names are mostly deprecated due to security reasons.
220 38 Denis 'GNUtoo' Carikli
* *PRIME fd*: whether there is support for buffer sharing between different DRM drivers through PRIME, the successor of GEM names. PRIME allows the conversion of local GEM handles to DMA-BUF file descriptors and vice-versa, via the DMA Buffer Sharing API.
221 44 dl lud
* *binderization*: whether there is support for using a binder, a structure that allows memory sharing between different processes on the same machine (for instance a Queue).
h2. Composer
h3. Non Android Hardware Composer HAL compatible
226 38 Denis 'GNUtoo' Carikli
* "libliftoff": - Eases the use of KMS planes.
227 39 Denis 'GNUtoo' Carikli
h2. Wayland
229 40 Denis 'GNUtoo' Carikli
| Implementation | Architecture | Advantages | Disadvantages | Sustainability |
231 1 Wolfgang Wiedmeyer
| "SPURV": | Android<->Hwcomposer<->Wayland | | Probably too low in the stack, maybe should replace surfaceflinger ideally?
=> Do some benchmarks |