summaryrefslogtreecommitdiffstats
path: root/src/gui/rhi/qrhivulkan.cpp
Commit message (Collapse)AuthorAgeFilesLines
* Merge remote-tracking branch 'origin/5.14' into 5.15Qt Forward Merge Bot2019-10-101-151/+246
|\ | | | | | | Change-Id: Iadeca81f499d6b19e86ceae1edd7960db2575e90
| * rhi: Fix non-base level copy and readback wrt source sizeLaszlo Agocs2019-10-091-5/+5
| | | | | | | | | | | | | | | | | | | | | | | | When the source size is not explicitly specified, we take the entire subresource. However, just using the texture's size is wrong: when the source level in a copy or readback is not 0, the size for the corresponding mip level has to be used instead. This fixes occasional crashes with Metal in the autotest. Change-Id: I99f689feef93ec86dffdc9e82d6bfdaf5c1eb041 Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
| * rhi: Autotest rendering a triangleLaszlo Agocs2019-10-091-0/+3
| | | | | | | | | | | | | | | | | | Also improve (docs and runtime checks) and test the minimum set of required data to create a graphics pipeline. Task-number: QTBUG-78971 Change-Id: If5c14f1ab1ff3cf70f168fde585f05fc9d28ec91 Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
| * rhi: gl: Add a feature flag for reading back non-zero mip levelsLaszlo Agocs2019-10-091-0/+2
| | | | | | | | | | | | | | | | | | The joys of "level - Specifies the mipmap level of the texture image to be attached, which must be 0." for glFramebufferTexture2D in OpenGL ES 2.0. Change-Id: Iaf19502f48d7ba73b26abb72535bfa6696a1e182 Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
| * rhi: Add support for buffer readbacksLaszlo Agocs2019-10-091-149/+239
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This also marks the beginnings of significantly extending autotesting of the resource and rendering functionality in QRhi. Also involves fixing up the buffer operation lists like we did for textures before. This is to ensure updates and reads on the same batch execute in the correct order. So just have two lists: one with buffer, one with texture operations. Also simplify the struct layouts. No need for those inner structs with many duplicate members. This reduces the size even, since using a union was never an option here. Also switch to a VLA, the size is around 253 KB per batch. The Null backend now keeps track of the QRhiBuffer data so it can return valid results in readbacks. Task-number: QTBUG-78984 Task-number: QTBUG-78986 Task-number: QTBUG-78971 Task-number: QTBUG-78883 Change-Id: I9694bd7fec523a4e71cf8a5c77c828123ebbb3bd Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
* | Vulkan: Add platform hooks before presentingJohan Klokkhammer Helsing2019-10-081-0/+4
|/ | | | | | | | | | This allows the Wayland plugin to circumvent the frame callback handling of the driver (which blocks until the frame is presented, potentially forever, if the window is minimized). Task-number: QTBUG-78000 Change-Id: Ia7d347019dfeae3bfcfad3d0cca3f4fffdc8c7a9 Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
* Convert a few sizeof(array)/sizeof(element0) fors to range forsAlbert Astals Cid2019-10-041-8/+8
| | | | | | | | Increases readability Change-Id: I81ea915517fd2cd6bc2780f37ba8d8097c63f44b Reviewed-by: Thiago Macieira <thiago.macieira@intel.com> Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Add TriangleFan pipeline topologyPaul Olav Tvete2019-10-031-0/+4
| | | | | | | | | | | | Triangle fan drawing is only supported on some platforms. The feature flag QRhi::TriangleFanTopology is set accordingly. In general, TriangleStrip topology will be more efficient on most GPUs. However, if polygon data is already stored as triangle fans, the CPU savings may be more significant. Change-Id: I9704fad751d2119eac8791074f58942dd9315601 Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
* rhi: vulkan: Disable vkmemalloc's own synchronizationLaszlo Agocs2019-10-031-0/+3
| | | | | | | | No need for those mutexes. Task-number: QTBUG-78908 Change-Id: I60c32df1f8729098f1894eff7c71e7fbcbd23ecf Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: Remove QVectors from the data description structs as wellLaszlo Agocs2019-10-031-50/+53
| | | | | | | | | | | | | | | As usual, keep some QVector overloads around to allow Qt Quick to compile. Color attachments and vertex input bindings get an at(index) type of accessor, unlike any other of similar lists. This is because there the index is significant, and sequential iteration is not the only type of operation that is performed. Sometimes a lookup based on an index will be needed as well. Task-number: QTBUG-78883 Change-Id: I3882941f09e94ee2f179e0e9b8161551f0d5dae7 Reviewed-by: Christian Strømme <christian.stromme@qt.io> Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
* rhi: Speed up buffer and texture trackingLaszlo Agocs2019-09-301-14/+12
| | | | | | Task-number: QTBUG-78862 Change-Id: If278bd55530081cbbdbab8dd6e14d86e28da558e Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* Remove QVector in the API of QRhiResource subclassesLaszlo Agocs2019-09-301-1/+2
| | | | | | | | | | | | | | | | | | | | | | | | | Forcing users to go through a QVector, when in practice they almost always want to source the data from an initializer list, a QVarLengthArray, or a plain C array, is not ideal. Especially since we can reason about the maximum number of elements in the vast majority of use cases for all the affected lists. QRhiResource is also not copyable so we do not need the usual machinery offered by containers. So switch to a QVarLengthArray. Note that a resource is not a container. The only operations we are interested in is to be able to source data either via an initializer list or by iterating on something, and to be able to extract the data, in case a user wishes to set up another resource based on the existing one. In some cases a QVector overload is kept for source compatibility with other modules (Qt Quick). These may be removed in the future. Also do a similar QVector->QVarLengthArray change in the srb-related data in the backends. Change-Id: I6f5b2ebd8e75416ce0cca0817bb529446a4cb664 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* De-d-pointer QRhiShaderResourceBindingLaszlo Agocs2019-09-301-7/+7
| | | | | | | | | | | | | | | | | Sad to see this go since the d pointer pattern with implicit sharing would have been perfect for this class, had this been a public API. However, as binary compatibility will not be a concern for QRhi classes, it is wasteful to allocate memory on every QRhiShaderResourceBinding. This allows users, such as Qt Quick, to use QRhiShaderResourceBinding as a cheap, simple, value class, without having to invent their own alternatives in performance critical places. The change brings a not insignficant improvement in certain qmlbench scenes (the ones with thousands of unbatched geometry nodes). Change-Id: I6d1dced6498d9ad625f90ead78bc0a417ea99ed8 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* Ensure drawable size atomicity within a frameLaszlo Agocs2019-09-291-3/+10
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Revert surfacePixelSize() to be a getter only. With Metal this will mean returning the "live" layer size (and so not the layer.drawableSize), which is in line with what we expect with other backends. Instead, we leave it to the swapchain's buildOrResize() to "commit" the size by setting drawableSize on the layer. With typical application or Qt Quick logic this ensures that layer.drawableSize is set once and stays static until we get to process the next resize - on the rendering thread. This of course would still mean that there was a race when a client queries surfacePixelSize() to set the depth-stencil buffer size that is associated with a swapchain. (because that must happen before calling buildOrResize() according to the current semantics) That can however be solved in a quite elegant way, it turns out, because we already have a flag that indicates if a QRhiRenderBuffer is used in combination with (and only in combination with) a swapchain. If we simply say that setting the UsedWithSwapChainOnly flag provides automatic sizing as well (so no setPixelSize() call is needed), clients can simply get rid of the problematic surfacePixelSize() query and everything works. Task-number: QTBUG-78641 Change-Id: Ib1bfc9ef8531bcce033d1f1e5d4d5b4984d6d69f Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
* rhi: gl: Pick up context lossLaszlo Agocs2019-09-131-2/+3
| | | | | | | | ...and change the return value of makeThreadLocalNativeContextCurrent() to a bool since we expect this to mirror QOpenGLContext::makeCurrent(). Change-Id: I339507152e461fe28fcf7fe777165e6d0072f055 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: Add a flag to indicate preferring a software adapterLaszlo Agocs2019-09-121-2/+22
| | | | | | | ...if there is one and the concept is applicable in the first place. Change-Id: Iab202c1c1cdd229f4910159de4cae7ce30805ea9 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: vulkan: Report device lostLaszlo Agocs2019-09-121-30/+37
| | | | | | | | | | | | Typically caught in vkQueueSubmit(). The WaitIdles that can be hit upon cleanup must be guarded by !deviceLost because they inexplicably cause an infinite blocking wait when the device was already reported as lost. (with NVIDIA at least) Change-Id: I7142e2461e1aed9ee3068b2b963cdf2c678ca4e0 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: Better handling of device lossLaszlo Agocs2019-09-121-0/+5
| | | | | | | Starting with D3D11. The other backends will follow later. Change-Id: I4f165c9f1743df0fb00bdce1e898917575bf5f6e Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: Make the clang code model happy as much as we canLaszlo Agocs2019-09-121-142/+143
| | | | | | | | | | | | | | | | When interfacing with reality (i.e. native platform APIs provided in C, ObjC, or COM), each of the APIs has its own idea of what types it likes to use for sizes, points, rects, etc. Out of the hundreds of warnings Qt Creator throws at us with the default clang check level when opening one of the rhi backends not a single one is useful. Regardless, let's try getting rid of what we can. This mostly involves throwing in int/uint conversions in order to get the signedness change warnings to shut up. The things that are either unacceptable to change or are beyond our control are left untouched. Change-Id: I6e4cb7cd373bf48dc990eaf83344242bbf30bd66 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: gl, metal, d3d: Reuse shader objects when source is the sameLaszlo Agocs2019-09-091-0/+5
| | | | | | | | | | | | | | | | | | Now that Qt Quick's batch renderer misses one level of shader source caching due to the nature of pipeline state objects, it can be useful to keep and reuse shader objects when the hash of the source code matches. The goal here is to allow Qt Quick to be on par with what the direct OpenGL path has when it comes to caching shader sources and compilation results. The program binary disk cache is not in scope in this patch. Also adds QRhi::releaseCachedResources(), similarly to what the scenegraph has. This can be called to clear caches such as the shader object cache we keep here. Change-Id: Ie3d81d823f61fa65ec814439e882c498f7774d43 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: vulkan: Introduce secondary command buffer usageLaszlo Agocs2019-08-291-126/+384
| | | | | | | | | | | As an option. Must opt in via setting ExternalContentsInPass in the flags for beginFrame(). It is somewhat unfortunate to require declaring this up front, but forcing using secondary command buffers always, even though beginExternal() may not be used in many applications, would be an overkill. Change-Id: I8d52bcab40c96f89f140c4c7877b6c459925e3c7 Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* rhi: vulkan: Add missing VK_QUERY_RESULT_WAIT_BITLaszlo Agocs2019-08-291-1/+2
| | | | | | | Mainly to get the validation layer from newer Vulkan SDKs to shut up. Change-Id: I3a00d2e7b5617eb1656625b1b2a919bb3c07feb9 Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* rhi: Attempt to fix up the logic around beginExternal()Laszlo Agocs2019-08-291-6/+24
| | | | | | | | | | | | | | | | | | | | | | | | | Mainly for Vulkan where it lacked the recording of the still queued commands. Uncovered by Qt Quick examples that integrate custom Vulkan rendering. This still has an issue that needs to be tackled separately. (we probably will switch to using a dedicated secondary command buffer with RENDER_PASS_CONTINUE_BIT for the external commands, and then just have a vkCmdExecuteCommands in our own queue instead of recording everything in beginExternal). The possibility of losing glMemoryBarrier() calls due to begin/endExternal() with the OpenGL backend is fixed too. The logic here mirrors Vulkan to some extent except that we do not have a concept of (and so the trouble with) renderpass instances. Clean up around the implementations of finish() as well. Attempting to share code via a "flushCommandBuffer" function is admirable but is not worth it since some semantics are different. (finish() cannot be called within a begin/endPass, unlike begin/endExternal). Change-Id: I5137db598d6a40d484e53678f5c919abf750d9ed Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* rhi: vulkan: Expose the VkRenderPass via the usual mechanismsLaszlo Agocs2019-08-291-0/+13
| | | | | | | | | Qt Quick in turn will expose it via QSGRendererInterface. Essential when adding custom Vulkan rendering into a Qt Quick application because the custom pipeline state objects will need to reference a VkRenderPass. Change-Id: Idf4092cfc3937830fb8123164081059b0d8d030e Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* Mark all rhi docs as internalLaszlo Agocs2019-08-271-4/+8
| | | | | | | | This is private API for the time being. Without this the QRhi* classes show up in the class list in the docs. Change-Id: I662abb9cc8eaae13ffe9266bd6313faa8e138353 Reviewed-by: Christian Strømme <christian.stromme@qt.io>
* rhi: Move info prints to categorized loggingLaszlo Agocs2019-08-021-12/+13
| | | | | | | | | | | Following the Quick scenegraph (qt.scenegraph.*), we now have qt.rhi.general. Other categories may get added later. This does not change the printing of real errors, those will continue to use qWarning(). Change-Id: Id95416fc82ba8add9527212e431bcbd47d416f1a Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* rhi: gl: Add support for computeLaszlo Agocs2019-07-311-32/+4
| | | | | | | ...and storage buffers and images. Change-Id: If38a51322e3187088a13cf4e9b88cb40c8af8621 Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* rhi: vulkan: Fix glitches on resize on X11Laszlo Agocs2019-07-311-0/+4
| | | | | | | | ...by notifying QVulkanInstance about the present. With the xcb platform this then gets turned into updating the sync request counter. Change-Id: Iecfb6d10ead3befcb24c19433d4712ed73a84bb4 Reviewed-by: Andy Nichols <andy.nichols@qt.io>
* rhi: Improve base vertex/instance supportLaszlo Agocs2019-07-011-2/+6
| | | | | | | | | Have feature flags as appropriate. OpenGL is causing a mess here but let's support what we can since some of this will become relevant in more sophisticated mesh drawing cases with 3D in particular. Change-Id: Idfa7b4642ec87147978e03d78d6233efbd151491 Reviewed-by: Lars Knoll <lars.knoll@qt.io>
* rhi: Enhance line width and point size supportLaszlo Agocs2019-07-011-7/+11
| | | | | | | | | | | | | | | | | | | | | ...but this will vary between backends, or in some cases even between implementations of the same API. Point size is settable only via the vertex shader (gl_PointSize). It is silently ignored with D3D and HLSL. Line widths other than 1 are supported only on OpenGL and Vulkan. (but this is in fact deprecated with GL and optional with Vulkan) Add QRhi::Feature values for both. The line width is now settable on QRhiGraphicsPipeline. It is not a dynamic state since the static, per-pipeline width is good enough for most cases. (and the feature is not supported on half of the backends anyways so it will get limited use in practice). Change-Id: I6d3a32269527c452b794b2cb8b0f03101eab40b2 Reviewed-by: Lars Knoll <lars.knoll@qt.io>
* rhi: Add compute api and implement for Vulkan and MetalLaszlo Agocs2019-06-171-97/+474
| | | | | | | | | | | | | D3D11 and GL (4.3+, ES 3.1+) will come separately at a later time. Change-Id: If30f2f3d062fa27e57e9912674669225b82a7b93 Reviewed-by: Lars Knoll <lars.knoll@qt.io>
* rhi: Make it possible to request making the context current on GLLaszlo Agocs2019-06-171-0/+5
| | | | | | | | | | | | | | | | | Needed by Qt Quick to handle cases where the application (or other Qt) code contains OpenGL calls, and Qt Quick facilitates this by ensuring the scenegraph's GL context is current. The expectation is that when running with the GL backend of the rhi, all such code remains fully functional. So add a makeCurrent type of call into the QRhi API that is a no-op with anything other than OpenGL. Change-Id: I6f774bf828e31802bdab0c3fef9421cdc0cebe5c Reviewed-by: Lars Knoll <lars.knoll@qt.io>
* Introduce the Qt graphics abstraction as private QtGui helpersLaszlo Agocs2019-06-131-0/+5681
Comes with backends for Vulkan, Metal, Direct3D 11.1, and OpenGL (ES). All APIs are private for now. Shader conditioning (i.e. generating a QRhiShader in memory or on disk from some shader source code) is done via the tools and APIs provided by qt-labs/qtshadertools. The OpenGL support follows the cross-platform tradition of requiring ES 2.0 only, while optionally using some (ES) 3.x features. It can operate in core profile contexts as well. Task-number: QTBUG-70287 Change-Id: I246f2e36d562e404012c05db2aa72487108aa7cc Reviewed-by: Lars Knoll <lars.knoll@qt.io>
generated by cgit v1.2.3 (git 2.25.1) at 2024-06-07 21:52:38 +0000