kc3-lang/angle/src/libANGLE/renderer/vulkan/vk_cache_utils.h

src/libANGLE/renderer/vulkan/vk_cache_utils.h

Log

Author Commit Date CI Message
Antonio Maiorano 4f343f3e 2020-05-13T14:11:51 Reland "Add support for GL_CHROMIUM_texture_filtering" This is a reland of 38780ae3921d2570316119a881adfb9520e7e296 modulo the changes to disable VK_VALIDATION_FEATURE_DISABLE_UNIQUE_HANDLES_EXT, as this was causing problems. With this landed, the extension will not work on SwiftShader until we find a way to allow this extension through the validation layers. Bug: b/146423360 Bug: b/154620295 Change-Id: Ie09fc507c01a47be3bb227bc78771660170ba5d3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2199639 Commit-Queue: Antonio Maiorano <amaiorano@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tim Van Patten 3b82fdcf 2020-05-12T23:44:07 Revert "Add support for GL_CHROMIUM_texture_filtering" This reverts commit 38780ae3921d2570316119a881adfb9520e7e296. Reason for revert: Breaks SWANGLE-VK: 05-12 23:42:28.612 22836 22861 D libEGL : dlopen_ext from APK (libEGL_angle.so) success at 0x79f4485a10 05-12 23:42:28.617 22836 22861 D libEGL : dlopen_ext from APK (libGLESv1_CM_angle.so) success at 0x79f4485eb0 05-12 23:42:28.619 22836 22861 D libEGL : dlopen_ext from APK (libGLESv2_angle.so) success at 0x79f4485c60 05-12 23:42:28.624 14223 14223 D StatusBar: disable<e i a s b h r c s > disable2<q i n > 05-12 23:42:28.633 14014 14057 V DisplayPowerController: Brightness [0.19986142] reason changing to: 'automatic', previous reason: 'automatic [ dim ]'. 05-12 23:42:28.635 22836 22861 D vulkan : searching for layers in '/data/app/~~0AfWfBsFEO78tqKlnanevg==/org.khronos.gl_cts-NOs3SGclHqlbcqe-08gHrw==/lib/arm64' 05-12 23:42:28.635 22836 22861 D vulkan : searching for layers in '/data/app/~~0AfWfBsFEO78tqKlnanevg==/org.khronos.gl_cts-NOs3SGclHqlbcqe-08gHrw==/base.apk!/lib/arm64-v8a' 05-12 23:42:28.632 22836 22836 W .khronos.gl_cts: type=1400 audit(0.0:218): avc: denied { ptrace } for scontext=u:r:zygote:s0 tcontext=u:r:untrusted_app_25:s0:c512,c768 tclass=process permissive=0 b/77925912 app=org.khronos.gl_cts 05-12 23:42:28.636 14014 14034 I EventSequenceValidator: Transition from ACTIVITY_LAUNCHED to ACTIVITY_FINISHED 05-12 23:42:28.637 22836 22861 F SwiftShader: external/swiftshader/src/Vulkan/libVulkan.cpp:425 vkCreateInstance TRACE_ASSERT: pCreateInfo->pNext sType = 1000247000 --------- beginning of crash 05-12 23:42:28.637 22836 22861 F libc : Fatal signal 6 (SIGABRT), code -1 (SI_QUEUE) in tid 22861 (.khronos.gl_cts), pid 22836 (.khronos.gl_cts) Original change's description: > Add support for GL_CHROMIUM_texture_filtering > > Chromium enables a custom extension, GL_CHROMIUM_texture_filtering, when > using SwiftShaderGL, to enable high precision filtering. This change > makes it so ANGLE also handles this same extension when using the > SwiftShaderVK backend, by enabling the new > VK_GOOGLE_sampler_filtering_precision custom extension. > > Bug: b/146423360 > Bug: b/154620295 > Change-Id: I69cafc1ccf5970a3d220ac7e13ec3c8fdd4a9643 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2185822 > Commit-Queue: Antonio Maiorano <amaiorano@google.com> > Commit-Queue: Jamie Madill <jmadill@chromium.org> > Reviewed-by: Jamie Madill <jmadill@chromium.org> TBR=geofflang@chromium.org,jmadill@chromium.org,amaiorano@google.com # Not skipping CQ checks because original CL landed > 1 day ago. Bug: b/146423360, b/154620295 Change-Id: I803d7a7baac81cf178b59c4bf2789346ec1d3f87 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2197168 Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Antonio Maiorano 38780ae3 2020-05-06T16:05:02 Add support for GL_CHROMIUM_texture_filtering Chromium enables a custom extension, GL_CHROMIUM_texture_filtering, when using SwiftShaderGL, to enable high precision filtering. This change makes it so ANGLE also handles this same extension when using the SwiftShaderVK backend, by enabling the new VK_GOOGLE_sampler_filtering_precision custom extension. Bug: b/146423360 Bug: b/154620295 Change-Id: I69cafc1ccf5970a3d220ac7e13ec3c8fdd4a9643 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2185822 Commit-Queue: Antonio Maiorano <amaiorano@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill fbc2f063 2020-04-24T13:00:11 Vulkan: Refactor AttachmentOpsArray. Will allow for more flexibly setting ops when we defer clears. Bug: angleproject:4517 Change-Id: I7d9116bc92e90eb41a1030fea242eadf1cc74562 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2165629 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 492ec932 2020-04-21T13:38:41 Vulkan: Track allocated sampler counts. Helps to diagnose perf bugs where resources are over-allocated. Also can be useful to evaluate caching strategies. Bug: angleproject:4517 Change-Id: I48df5a09fbc394fa0b1712fa8cf28a179665e6e7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2159293 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com>
Jamie Madill 6fc0066a 2020-04-21T17:40:54 Vulkan: Add a sampler cache. This uses a packed sampler description to re-use samplers for multliple VkImages. The samplers will persist for the lifetime of RendererVk. In the future we could look at doing cache eviction for large object counts. Reduces the active VkSampler cache in Manhattan from over 1200 to 9. Also should reduce the number of VkSamplers used with Chrome. Bug: angleproject:4491 Change-Id: Idca00e4ed8cb660a0865281544aaa57cf884bbdb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2160771 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Tobin Ehlis <tobine@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Charlie Lao 03c18fbd 2020-04-13T16:01:01 Vulkan: Skip load if depth/stencil value are undefined EGL1.5 spec says depth/stencil data are undefined after swap. This CL will track the depth/stencil data and mark it as undefined and skip the load if it is undefined. Bug: b/153885625 Change-Id: Ifb3d88d442da547fa78c1eae091cbae08a20d0ab Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2148179 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Charlie Lao <cclao@google.com>
Charlie Lao 8c67c079 2020-04-16T14:42:15 Vulkan: store ImageLayout in PackedAttachmentOpsDesc The layout in the PackedAttachmentOpsDesc only have 5 bits, not enough for raw VKImageLayout enums. This change will store vk::ImageLayout which are ANGLE defined and packed for 5 bits and able to handle the extension bits of VKImageLayout. Bug: b/153885625 Change-Id: Ida268a974094188d8aeee39c61019eef30a60507 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2153845 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill dd00f16b 2020-04-02T14:04:44 Vulkan: Fix FBO cache when updating disabled attachments. Fix this by consistently checking if the attachment is enabled when we update the serials. Also includes a regression test and more ASSERTs. Bug: angleproject:4540 Change-Id: I154d23cad71f1674d893390f923f45c643a58925 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2134409 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tobin Ehlis <tobine@google.com>
Tim Van Patten abaa3573 2020-02-20T10:23:02 Vulkan: Only calculate variable locations once Variable locations and descriptor set/binding values are calculated multiple times: - Compiling GLSL->SPIR-V - Creating the Vulkan pipeline layout - Updating descriptor sets These values should instead be calculated once and reused throughout since they won't change without recompiling the shader program. Bug: angleproject:3570 Change-Id: I5d8767b3b2e2f741aade7fec9991eea53ee2eb98 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2067101 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 4fb3c28f 2020-03-19T19:10:45 Vulkan: Fix Framebuffer cache depth/stencil issue. The DrawBuffers resetting logic didn't update the depth/stencil serial. Fix this and add a regression test that would crash in this situation. Note that the Manhattan benchmark could trigger this under some cases. Bug: angleproject:4322 Bug: angleproject:4490 Change-Id: Ie4b24efeb5991ecaa85bf8f964f77476bd6f167d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2111333 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Cody Northrop <cnorthrop@google.com>
Tobin Ehlis 175227ba 2020-03-06T14:30:06 Vulkan:FramebufferVk cache draw buffer handling Dirty draw buffers are causing cache clears on Manhattan. Instead of clearing out the entire FramebufferVk cache when draw buffers are dirty, reset the cache signature and regenerate it. This change also contains a fix to make sure that we're not generating serials for a draw buffer that's disabled. Thanks Jamie Madill! Bug: angleproject:4442 Change-Id: I0d48a2d6d95e74898a11bdde0fedbce77c82a3a3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2091862 Reviewed-by: Tobin Ehlis <tobine@google.com> Commit-Queue: Tobin Ehlis <tobine@google.com>
Tobin Ehlis cf2ec3b1 2020-01-31T16:55:50 Vulkan:Add FramebufferVk cache This adds a FramebufferVk cache. Cache signature is based on unique Serial values that are assigned to ImageHelper objects backing all color and DS rendertargets as well as level/layer values unique to the imageView. Update the Serials and cache signature at FramebufferVk::syncState() time. L0 cache is a currently active framebuffer. L1 cache retrieves previously created framebuffer from new cache. If neither of those hit, create new FramebufferVk and add to L1. Bug: angleproject:4322 Change-Id: I3f585271798ddfb9e5f194020adca8cf8a6b19dd Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2033869 Commit-Queue: Tobin Ehlis <tobine@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Shahbaz Youssefi b36e46ab 2020-01-08T15:49:18 Vulkan: Line raster emulation through specialization constant In preparation for compiling shaders early at link time, this change reworks line raster emulation such that it uses specialization constants instead of a preprocessor condition. This means drawing both triangles and lines with this program will still result in a one-time shader compilation. The compilation is still done at draw time in this change. Bug: angleproject:3394 Change-Id: I0bf91398868d7f7147456533b728906b505192b2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1992365 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Ian Elliott c8d7f047 2020-01-07T10:06:07 Vulkan: Update number of samples when changing draw framebuffer This fixes a crash (preceeded by Vulkan validation errors) when running the following test: --gtest_filter=dEQP.KHR_GLES31/core_texture_storage_multisample_FunctionalTests_blitting_multisampled_depth_attachment* --use-angle=vulkan The problem is that GraphicsPipelineDesc::updateRasterizationSamples() is rarely called (e.g. glBindFramebuffer). The above dEQP test changes between draw attachments that do and don't have multisampling enabled. The number of samples is set (the first time) to have multisampling enabled, and never changed, even though other attachments don't have multisampling. FramebufferVk::syncState() now calls updateRasterizationSamples(). That fix uncovered a latent problem: when an application destroys all framebuffers (including the draw framebuffer), the ContextVk still points at the old draw framebuffer's FramebufferVk (now freed). This is fixed with new plubming to clean-up ContextVk::mDrawFramebuffer. Bug: angleproject:4240 Change-Id: I151ec40ab821efc07c26323c34275c9165ad2ed5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1983393 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Ian Elliott <ianelliott@google.com>
Shahbaz Youssefi 806d812e 2020-01-03T13:48:30 Vulkan: Reorder descriptor sets This change moves driver uniforms to a set with a lower id than the one that includes (emulation) transform feedback buffers. Imagine the following: - Program 1 with xfb: UniformsSet: 1 VS uniforms buffer, 1 FS uniforms buffer, 1 xfb buffer - Program 2 without xfb: UniformsSet: 1 VS uniforms bufer, 1 FS uniforms buffer. Previously, UniformsSet was index 0, and DriverUniformsSet was index 3. When switching from Program 1 to Program 2, the layout of UniformsSet changes, which means every subsequent set needs to be rebound. This is a Vulkan pipeline layout compatibility rule. This is done with invalidateCurrentTextures() and invalidateCurrentShaderResources() already when handling gl::State::DIRTY_BIT_PROGRAM_EXECUTABLE. The bug is that the driver uniforms are not invalidated. This is normally not an issue, because usually when switching from Program 1 to Program 2, transform feedback is paused, and this state change does invalidate driver uniforms. However, the following scenario doesn't do this: - Begin Xfb - Pause Xfb - Use Program 1 - Draw - Use Program 2 - Draw - End Xfb There is no driver state change between the two draw calls, which means the second draw will attempt to draw using the driver uniforms bound for the first draw call. There is a Vulkan validation error here due to the above pipeline layout validation rule. The issue manifests itself only when the second draw call actually uses driver uniforms, as otherwise that set is inactive and not validated; i.e. when line raster emulation is used. In summary, the validation error manifests itself when: - Transform feedback and line raster both use emulation - Transform feedback is paused - A draw with an xfb program is followed by a non-xfb program - The second draw is a line draw A test is added for this. The solution is to reorder the sets so that DriverUniformsSet is placed before UniformsSet. This way, changes to the layout of UniformsSet don't invalidate DriverUniformsSet. In fact, based on the above, any change in the layout of the program should have required an invalidation of the driver uniforms. This bug is probably masked by the fact that ContextVk::handleDirtyDescriptorSetsImpl() always rebinds the graphics driver every time any descriptor set needs rebinding. That should be removed in a follow up change. Bug: angleproject:4261 Change-Id: I21ad4152b454a1fe70554be02e18a9c94fb3e7a8 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1986927 Reviewed-by: Tobin Ehlis <tobine@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Tobin Ehlis 71c1138d 2019-08-16T12:23:04 Vulkan: Emulate instanced attrib divisor This sets instancedArrays[ANGLE|EXT] extenstions as always supported regardless of underlying Vulkan HW's max vertex attrib divisor. Then detect instances where app sets a divisor that isn't supported by hardware and emulate those cases. Emulations is accomplished by copying the instanced attribs to a new buffer where each attrib is present once per instance, using the attrib divisor value as a factor to replicate the attribs, and then setting the actual divisor value for the draw to "1". Also, we only store 8 bits for the divisor used in the PSO, so this code also handles emulation of the case where divisor is > 255. This is passing all of the drawInstanced/Elements dEQP tests where divisor has to be emulated. Also enabled end2end InstancingTestES3 for Vulkan backend. Bug: angleproject:2672 Change-Id: I9932f9eab49b16a19e8bbd35dacaf3b5a27a213f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1758689 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Tobin Ehlis <tobine@google.com>
Jaedon Lee a0159c03 2019-09-02T14:49:07 Vulkan: Implement basic geometry shader feature Enable the default behavior of the geometry shader Bug: angleproject:3571 Test: dEQP-GLES31.functional.geometry_shading.input.basic_primitive.points dEQP-GLES31.functional.geometry_shading.input.basic_primitive.lines dEQP-GLES31.functional.geometry_shading.input.basic_primitive.line_loop dEQP-GLES31.functional.geometry_shading.input.basic_primitive.line_strip dEQP-GLES31.functional.geometry_shading.input.basic_primitive.triangles dEQP-GLES31.functional.geometry_shading.input.basic_primitive.triangle_strip dEQP-GLES31.functional.geometry_shading.input.basic_primitive.triangle_fan Change-Id: I65708d19bbfe6a0ad8ca392a1d6b3609b1410ef4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1793753 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tobin Ehlis 197005d8 2019-09-18T12:58:31 Vulkan: Pass ContextVk to various helpers This is groundwork for some upcoming changes where data from ContextVk is needed in GraphicsPipelineDesc::initializePipeline(). Passing ContextVk ptr all the way down instead of the reduced vk::Context ptr. Bug: angleproject:2672 Change-Id: I29f580c3503777085355f1b79f4ae4552a394557 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1811433 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Tobin Ehlis <tobine@google.com>
Tobin Ehlis 4c833efb 2019-07-11T11:29:35 Vulkan: Enable sampler objects This is the initial implementation to get GLES3.0 Sampler Objects working on the Vulkan backend. When samplers are dirty, a corresponding Vulkan sampler object will be created with associated state and textures are flagged as dirty anytime sampler bindings change. Then when handling textures dirty, any bound sampler objects are pulled from active texture units. When sampler objects are bound, their state is used instead of the built-in texture's sampler state. This change also adds a test that uses the same texture object with different sampler objects to test a corner case that dEQP misses. Bug: angleproject:3208 Change-Id: I643d9c9d5cb92fecc02dad815a07bcf349534c70 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1706897 Commit-Queue: Tobin Ehlis <tobine@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi c13ca2af 2019-07-17T15:46:29 Vulkan: Allow more than one atomic counter buffer binding dEQP assumes there are more than one atomic counter buffers available. This is technically not a requirement by the standard, but nevertheless could be what applications expect as well. This change adds support for multiple atomic counter buffer bindings. This is done by declaring an array of storage buffers for the atomic counter buffers (instead of declaring only one) and passing the (binding, offset) pair around to functions instead of just the offset. The atomic counter is found by indexing `binding` into the storage buffer array first before indexing `offset` into its `uint[]`. ProgramVk's default uniform collection is also fixed not to include atomic counter uniforms. A remaining issue is that atomic counter buffer offsets don't have alignment requirements in GLES, but Vulkan does for storage buffers. Similar to emulated transform feedback buffer offsets, these should be sent to the shader through uniform values. This will be done in a follow up change. Bug: angleproject:3566 Change-Id: I5600225c24c38f1a8ecf5c64388073055733197d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1707931 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jiacheng Lu 758b12fa 2019-07-17T14:55:38 Filter out redundant depth&stencil state updates 1. Compare updating values with active ones at frontend when depth&stencil related gl call happens and do not set any unnecessary dirty bits. 2. Remove d3d and gl backends' checking of depth&stencil states when detect any dirty bits, as it is now being done at frontend. 3. Modification to graphics pipeline description update, make sure it syncs with gl::State initial value. 4. Change gl_vk::CullMode return type to VkCullModeFlagBits, as it only represents single value. Bug: angleproject:3700 Change-Id: Id3aa5186455ee3a10a9c147edad13944e3e41098 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1706903 Commit-Queue: Jiacheng Lu <lujc@google.com> Reviewed-by: Tobin Ehlis <tobine@google.com>
Shahbaz Youssefi 87dc17b5 2019-07-22T11:41:27 Vulkan: Generalize buffers desc set name to include images And atomic counter buffers which are emulated with storage buffers. Bug: angleproject:3566 Change-Id: I102763900c779c3a84661491bad53aec384461f7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1713086 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
James Dong a9ec8749 2019-07-11T11:31:47 Vulkan: override format for mismatched attribs Prevents Vulkan validation error by replacing the input format for any mismatched vertex attributes with a format compatible to what the shader expects. Bug: angleproject:3436 Change-Id: Ia52f29c084d82bbc4e9149102cd4b5fc25ccb9b3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1698567 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
James Dong 9917988a 2019-07-10T13:47:32 Vulkan: change packed attrib to ANGLE format ID This prevents later issues with some VkFormat values being over 256, as well as providing more information to pipeline creation. A preliminary step towards handling mismatched vertex attributes. Bug: angleproject:3634 Change-Id: Idb15a14088a2d73b43b4b92d3cfdb12587c5f711 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1696212 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 29fba5e0 2019-07-04T17:08:41 Vulkan: Prepare for variable-stage pipelines Compute (single-stage pipeline) is upcoming, but this change prepares GlslangWrapper to handle any number of stages (mostly). Additionally, this change binds each resource to each stage based on whether it's active, so that we don't hit the per-stage limit of resources by binding every resource to every stage. Bug: angleproject:3633 Bug: angleproject:3562 Change-Id: Ifebf691482846e0371c6e314f514226a4cfee258 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1689330 Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 1cde0eab 2019-07-03T10:58:32 Vulkan: Add storage buffer support The storage buffers are placed in the same descriptor set as uniform buffers. Some refactoring is done to reuse code that handles UBOs to handle SSBOs as well. A good number of tests still fail as they test SSBOs in conjunction with compute shaders. Bug: angleproject:3561 Change-Id: Ia33c1f68e6f6402c746f5919ede87b2c308cf81c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1687126 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi 0bfa5504 2019-06-03T10:40:10 Vulkan: Emulate Transform Feedback with vertex shader output In ES 3.0 and 3.1, only non-indexed GL_POINTS, GL_LINES and GL_TRIANGLES is supported for transform feedback. Without tessellation and geometry shaders, we can calculate the exact location where each vertex transform output should be written on the CPU, and have each vertex shader invocation write its data separately to the appropriate location in the buffer. This depends on the vertexPipelineStoresAndAtomics Vulkan feature. Bug: angleproject:3205 Change-Id: I68ccbb80aece597cf20c557a0aee842360fea593 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1645678 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
James Dong 2806a898 2019-06-07T12:10:46 Vulkan: implement primitive restart Implements ES 3.0 feature GL_PRIMITIVE_RESTART_FIXED_INDEX, which allows the application to use a fixed "restart" index to restart the primitive during a single draw call. This change does't handle GL_LINE_LOOP primitives, which requires a bit of special handling. Bug: angleproject:3215 Change-Id: I2388852683fd17328a6a76c48d70a24d67ce8b67 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1650301 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill 88596bea 2019-06-13T14:17:48 Vulkan: Implement a texture descriptor cache. We noticed a significant hotspot in vkAllocateDesctiptorSets. The app was repeatedly cycling through a few combinations of active textures. For each state change in ANGLE we were allocating a new desctiptor set. This in turn would trigger internal driver memory allocation and cause jank. Using a cache avoids allocations entirely since the application is rotating through a stable set of textures. The descriptor cache is stored in each program. It is indexed by a set of 32-bit serials. Each texture generates a unique serial for every combination of VkImage and VkSampler that the texture owns. The texture descriptor is refreshed every time a texture changes or is rebound. The descriptor cache is accessed via an unoredered map with the texture serial sets as the hash key. We also store the maximum active texture index in the cache key so we don't need to hash and memcmp on all 64 active textures. This will currently fail if more than MAX_UINT serials are generated. But that number is high enough that it shouldn't be possible to hit in practice in a practical amount of time. Requires shifting the texture sync to ContextVk so we can get the new serial after the textures are updated. And to make sure to update the image layouts even if the descriptors are not dirty. Improves performance of the T-Rex demo. Also improves the score of the texture state change microbenchmark by about 40%. Bug: angleproject:3117 Change-Id: Ieb9bec1e8c1a7619814afab767a1980b959a8241 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1642226 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 70642e42 2019-06-05T14:55:49 Vulkan: Implement ES 3.0 rasterizer discard Rasterizer discard is a feature that allows the graphics pipeline to skip the fragment shader stage. Implementing rasterizer discard in Vulkan is as easy as not binding a fragment shader binary. Tests for rasterizer discard live in dEQP-GLES3.functional.rasterizer_discard.*. Bug: angleproject:3214 Test: dEQP-GLES3.functional.rasterizer_discard.* Change-Id: I0815df3b70f6f9cdc6c8c87ec4813bb629d8bd5f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1646692 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 266a9e8d 2019-06-05T15:15:10 Vulkan: Move descriptor pools into ProgramVk. Previously ContextVk owned the descriptor pools. We were trying to maximize descriptor reuse to conserve memory. However the default uniforms would have no possible sharing. And because uniform buffers are usually unique to a program it's likely there would be less reuse. Image descriptors could be shared. But with the advent of a descriptor cache in the Program it becomes difficult to manage the cache through descriptor pool recycling. Moving the pools into the Program simplifies the cache management. We could look at adding back more reuse in the future. Also shifts driver uniforms back into the end of the descriptor sets to make indexing into the Program's descriptor pools simpler. Bug: angleproject:3117 Change-Id: I52bb49cf322d944ad7cf08791efdf24b7fe573ce Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1644775 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Geoff Lang ee244c77 2019-05-06T10:30:18 Vulkan: Move command graph and garbage to ContextVk. To support multithreading, contexts should manage their own command graphs and garbage. This allows safe access to vulkan resources such as command pools without thread synchronization. BUG=angleproject:2464 Change-Id: I930149bc9f0793028761ee05ab50b8c0a4dec98a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1516515 Commit-Queue: Geoff Lang <geofflang@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi fb8e1b25 2019-05-17T13:18:40 Vulkan: Fix multisampled clear UtilsVk::clearFramebuffer was not setting the sample count. Bug: angleproject:3204 Change-Id: Iab7d385ec58b12f21d888c5dfc9e94b7846ac40c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1616858 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi b16d69c3 2019-05-13T16:28:27 Vulkan: Add support for surface multisampling A multisample image is created for the surface if multisampling is enabled. Prior to present, this multisample image is resolved into the swapchain image. FramebufferVk::readPixelsImpl similarly has got the ability to resolve the region of interest into a temporary image prior to readback. Tests are added to render a point, line and a triangle on a 4x multisampled surface. Bug: angleproject:3204 Change-Id: I34aca502fa1918b5cbf000ff11521c350372e051 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1610188 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 6a4cad87 2019-05-01T13:04:00 Vulkan: Tighten descriptor stage usage hints Per-stage uniform buffers are marked with the specific stage. Program uniform buffers and images are specified to be used in all graphics stages. Descriptors used in internal shaders are marked for use in either compute or fragment stages. Bug: angleproject:3220 Change-Id: Ifcac36a1224f0392ba5fba50660514e498256401 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1595439 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com>
Shahbaz Youssefi 9fa248e1 2019-05-06T14:55:18 Vulkan: Implement EXT_draw_buffers In GLES, color attachments are referenced by their indices. These indices match between the API and GLSL. For example, if a shader has: layout(location=0) out color; layout(location=3) out roughness; Then GLES would bind and enable GL_COLOR_ATTACHMENT0 and GL_COLOR_ATTACHMENT3. In Vulkan, the framebuffer object and the corresponding renderpass define the color attachments, and they don't allow gaps in color attachments as GLES does. A render subpass creates the mapping between the color attachments as defined in the framebuffer and the attachments used by the shader (with possible gaps). This change packs the enabled GL color attachments for the sake of the framebuffer, and sets the subpass up in such a way that the shaders continue to use the same color output indices as GLES. In the example above, we have the attachment indices as follows: Status | GLES | GLSL | RenderPass | Subpass enabled 0 0 0 0 disabled 1 - VK_ATTACHMENT_UNUSED disabled 2 - VK_ATTACHMENT_UNUSED enabled 3 3 1 1 That is, the array of color attachments in the Vulkan framebuffer/renderpass is: [0] = GL color attachment 0 [1] = GL color attachment 3 And the array of color attachment references in the Vulkan render subpass is: [0] = 0 (index 0 of the renderpass attachment array) [1] = VK_ATTACHMENT_UNUSED [2] = VK_ATTACHMENT_UNUSED [3] = 1 (index 1 of the renderpass attachment array) Bug: angleproject:2394 Change-Id: Ib6cd2b60882643ea152986eee453270d09cd4aed Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1595442 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 216f73d0 2019-04-12T13:32:30 Vulkan: add uniform buffer object support Support for layout qualifiers in interface blocks are added. All interface blocks are adjusted to either be in std140 or std430. In the Vulkan backend, a new descriptor set is added for UBOs. A dirty bit is added for UBO updating and pipeline layouts and descriptor bindings are updated. Bug: angleproject:3199, angleproject:3220 Change-Id: I271fc34ac2e1e8b76dee75e54a7cff0fe15fe4ee Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1565061 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi c7565353 2019-04-03T14:03:56 Vulkan: break dependency to the depthClamp feature Image clear's masked clear path using a draw call was using this feature to clear the depth buffer, but this feature is not available on ARM and some Qualcomm devices. This change adds a push constant to the vertex shader used in this call to export the depth clear value, removing the need to rely on depth clamping. Bug: angleproject:3241 Change-Id: I565cd5f731c441820e0702e51dfdf02d0bc7de06 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1551522 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com>
Shahbaz Youssefi f6c937f8 2019-04-02T17:04:08 Vulkan: fix masked stencil clear Previously, masked stencil clear was done by clearing every stencil bit to the ClearValue & Mask. The correct behavior as implemented in this change is to clear only the bits that are set in Mask. This can only be done through a draw call, with ClearValue as the stencil reference, and Mask as the stencil write mask. Note: this change relies on the depthClamp Vulkan feature which is not available on ARM. Bug: angleproject:3241 Change-Id: I0a181c32f881ee813f144e7bdd6f42c8ea6f1966 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1548442 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com>
Shahbaz Youssefi 43997017 2019-03-30T23:24:01 Vulkan: fix non-float clear with draw Instead of using one draw call that clears all attachments, multiple draw calls are issued that clear a single attachment each. This allows us to have a manageable number of variations for the ImageClear.frag shader, now that non-float format support is introduced. Bug: angleproject:3187 Change-Id: Ic0c1067a396250bd80f31d00cad5a272acff8be8 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1545523 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi f9f18ef0 2019-04-03T10:07:18 Vulkan: Allow vertex-only pipelines This allows issuing draw calls which only manipulate depth/stencil. Bug: angleproject:3241 Change-Id: I62ab18a185ea5b234d3559f30c5b2b8ecb317bbb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1550900 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi db4ed317 2019-03-29T00:32:45 Vulkan: glClearBuffer* implementation Refactors FramebufferVk::clear such that specific render targets could be cleared, with clear values not necessarily set through glClearColor etc. FramebufferVk::clearWithRenderPassOp is modified so that loadOp and clear values are set after the render pass has been registered in the graph. This allows multiple glClearBuffer calls to coalesce into the same render pass. glClearBuffer calls are then implemented simply as calls to the refactored clear function with the appropriate parameters. Bug: angleproject:3187 Change-Id: I2fdfcbea5bf244f63ec981b91caca47f5ee3cd3a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1545204 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi 0c128e15 2019-03-25T23:50:14 Vulkan: Use render pass ops to clear images when possible On tiling GPUs, render pass loadOp and stencilLoadOp can be used to very cheaply clear an image as it is being render to. This change uses this feature to clear render targets when possible. Bug: angleproject:2361 Change-Id: Ic4bdc908873f4802760d549f4893f84a47beac0f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1500576 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi e397949e 2019-03-01T13:06:50 Vulkan: shave off 60 bytes from AttachmentOpsArray Bug: angleproject:2361 Change-Id: I39eb34b3c415fa165fa7803b2bc09338833f6773 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1496039 Reviewed-by: Jamie Madill <jmadill@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi d838178d 2019-03-04T11:07:47 Vulkan: Rename vk::Shared* to vk::RefCounted* To be specific regarding what being "shared" entails. Also, avoids confusion w.r.t to an upcoming vk::Shared class. Bug: angleproject:2464 Change-Id: Ib9c112bbb822ae30dab39c75a8cde25dd79b2258 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1499693 Reviewed-by: Jamie Madill <jmadill@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Frank Henigman 52047de4 2018-11-13T17:22:36 Vulkan: support instanced draws. (reland) Enable instanced draws with the Vulkan backend. So far it only works when Vulkan has VK_EXT_vertex_attribute_divisor. BUG=angleproject:2672 Change-Id: Ib6655625776344305911a1a742c85f17638cee8f Reviewed-on: https://chromium-review.googlesource.com/c/1469263 Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Frank Henigman <fjhenigman@chromium.org>
Shahbaz Youssefi 341304d8 2019-02-12T20:58:54 Revert "Vulkan: support instanced draws." This reverts commit 199a9f385f5489b957fe1e42bf08f3f40edd38ca. Reason for revert: Causes a validation error when the extension is not present. See failures here: https://ci.chromium.org/p/chromium/builders/luci.chromium.try/win-angle-rel/693 Original change's description: > Vulkan: support instanced draws. > > Enable instanced draws with the Vulkan backend. > So far it only works when Vulkan has VK_EXT_vertex_attribute_divisor. > > BUG=angleproject:2672 > > Change-Id: I9445ba64282fa00a6eaee207b15efa2c7a9abbd3 > Reviewed-on: https://chromium-review.googlesource.com/c/1334973 > Commit-Queue: Frank Henigman <fjhenigman@chromium.org> > Reviewed-by: Jamie Madill <jmadill@chromium.org> TBR=fjhenigman@chromium.org,jmadill@chromium.org,syoussefi@chromium.org Change-Id: Iffccc2cca259bcd19c068a87a415d4e196901f45 No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: angleproject:2672 Reviewed-on: https://chromium-review.googlesource.com/c/1468201 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Frank Henigman 199a9f38 2018-11-13T17:22:36 Vulkan: support instanced draws. Enable instanced draws with the Vulkan backend. So far it only works when Vulkan has VK_EXT_vertex_attribute_divisor. BUG=angleproject:2672 Change-Id: I9445ba64282fa00a6eaee207b15efa2c7a9abbd3 Reviewed-on: https://chromium-review.googlesource.com/c/1334973 Commit-Queue: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 7403dd2c 2019-01-16T13:27:17 Vulkan: Inline transition query. Slight speed improvement in the Vulkan vertex array state change test. Bug: angleproject:3014 Change-Id: I4a5cd26849cd247b7e67cd6bda083aabeb4e34c0 Reviewed-on: https://chromium-review.googlesource.com/c/1406890 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill ab2bfa81 2019-01-15T19:06:47 Enable Chromium clang style plugin for libANGLE. This fixes a few style warnings: * auto should not deduce to raw pointer type * inlined virtual methods are not allowed * non-trivial constructors and destructors should be explicit * inlined non-trivial constructors should not be in-class * missing override keywords Bug: angleproject:3069 Change-Id: I3b3e55683691da3ebf6da06a5d3c729c71b6ee53 Reviewed-on: https://chromium-review.googlesource.com/c/1407640 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Jamie Madill 633d5e69 2018-12-23T19:58:01 Vulkan: Put viewport and scissor back in pipeline desc. Turns out this is much faster than using dynamic state. When we support multiple viewports it might be easier to use dynamic state since we won't need to make an overly large pipeline description. We could support both methods using a flag to indicate the viewport and/or scissor regions are invalid. Until then we can remove the pipeline and scissor dirty bits. Improves perf by about 15% in the Vulkan VBO state change test. Bug: angleproject:3013 Change-Id: I4572250626a9a0f0ca3451b17e8f0de186416cae Reviewed-on: https://chromium-review.googlesource.com/c/1390359 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 2450d8dd 2019-01-11T14:41:04 Vulkan: Squash Vertex Pipeline info. Instead of using 8 bytes per vertex we can reduce the space used for the divisor to 8 bytes. For larger values than 255 we can emulate the divisor by unrolling the draw call. We will likely need to do this in any case for instanced draws when the instancing extension isn't available. The tighter packing will allow for us to move the viewport and scissor back into the pipeline description. It seems this is much faster than using dynamic state. Every state change that would pull in a new Pipeline would need the viewport and scissor re-applied. It seems these driver calls are costly. Does not improve perf significantly but enables future improvements. Bug: angleproject:3013 Change-Id: I1a41c3acadc6fbd47c7a7b961c706e82f78de936 Reviewed-on: https://chromium-review.googlesource.com/c/1390358 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Jamie Madill 3f0c4a56 2019-01-10T10:20:35 Vulkan: Faster state transitions. Implements a transition table from Pipeline Cache entry to state change neighbouring Pipeline Cache entries. We use a 64-bit mask to do a quick scan over the pipeline desc. This ends up being a lot faster than doing a full hash and memcmp over the pipeline description. Note that there could be future optimizations to this design. We might keep a hash map of the pipeline transitions instead of a list. Or use a sorted list. This could speed up the search when there are many transitions for cache entries. Also we could skip the transition table and opt to do a full hash when there are more than a configurable number of dirty states. This might be a bit faster in some cases. Likely this will be something we can add performance tests for in the future. Documentation is also added in a README file for the Vulkan back end. This will be extended over time. Improves performance about 30-35% on the VBO state change test. Bug: angleproject:3013 Change-Id: I793f9e3efd8887acf00ad60e4ac2502a54c95dee Reviewed-on: https://chromium-review.googlesource.com/c/1369287 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Jamie Madill 80766cfa 2019-01-10T10:20:34 Vulkan: Update individual VAO attribs when dirty. This simplifies the dirty state updates in VertexArrayVk. It also lets us use a smaller dirty bit mask when a single attribute is marked dirty in a vertex array. Improves performance by about 1-2% in the VBO state change test. Will allow for better performance using a pipeline transition table. Bug: angleproject:3013 Change-Id: I25c5172b3f41b7abac6b8273c8f9cd42eb46cc9f Reviewed-on: https://chromium-review.googlesource.com/c/1403958 Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill dbc605ce 2019-01-04T16:39:14 Vulkan: Optimize VBO state changes. Also has some minor optimizations for the front-end. 12% improvement on the Vulkan VBO change test. Bug: angleproject:3014 Change-Id: I38e1a8194edfc14bfe57424be348cb9688e928f4 Reviewed-on: https://chromium-review.googlesource.com/c/1369286 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi b51eb09f 2018-11-29T10:48:36 Vulkan: Allow bindings to compute stage Includes both binding of resources and updating push constants on the compute shader stage. Bug: angleproject:2958 Change-Id: I5686cbac81e0dd83d0e938cb40f9f9ac7d2ef48a Reviewed-on: https://chromium-review.googlesource.com/c/1355500 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill b980c563 2018-11-27T11:34:27 Reformat all cpp and h files. This applies git cl format --full to all ANGLE sources. Bug: angleproject:2986 Change-Id: Ib504e618c1589332a37e97696cdc3515d739308f Reviewed-on: https://chromium-review.googlesource.com/c/1351367 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill dc65c5bd 2018-11-21T11:07:26 Vulkan: Cache pipelines with Shader Programs. This allows for a few nice things. First and foremost it reduces the size of the PipelineDesc, which is now 232 bytes. It also allows us to completely forego pipeline caches for compute shaders. We also allow sharing vertex and fragment shaders among multiple programs for internal shaders. This is good for memory savings. To allow this we keep the shaders as ref counted objects. Bug: angleproject:2522 Change-Id: I2322be5061979d9669a0b25c152359561eeb80ee Reviewed-on: https://chromium-review.googlesource.com/c/1344449 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi c30f45d3 2018-11-12T16:37:59 Vulkan: Rename PipelineDesc/Cache to Graphics& PipelineDesc describes a Vertex-Fragment pipeline and PipelineCache (not to be confused with vk::PipelineCache) implements a cache of such pipeline objects. In preparation for Compute support, these data structures are prefixed with Graphics. Bug: angleproject:2959 Change-Id: I9181586fb946b787216ca0b2ad6340f90c3ab55f Reviewed-on: https://chromium-review.googlesource.com/c/1333971 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill bc5834cd 2018-11-06T11:13:50 Vulkan: Use dynamic state for Viewport and Scissor. This reduces the size of the pipeline cache descriptor under 256 bytes. Further improves the speed of cache query. Has the minor cost of needing more state application during a new command buffer or render pass. Bug: angleproject:2522 Change-Id: I3d71e457a36084ac4748d04fe3c9bab4caad503c Reviewed-on: https://chromium-review.googlesource.com/c/1316888 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 77b2436e 2018-11-05T22:39:29 Vulkan: Reduce PipelineDesc size. This packs the desc into 288 bytes. Down from over 400. The hashing and memcmp functions are sped up considerably. Improves performance in the VulkanPipelineCachePerf test by 50% and also improves performance in the state changing draw call test by 20%. Bug: angleproject:2522 Change-Id: I72ed191a93721b875684f647f806c09be4e66821 Reviewed-on: https://chromium-review.googlesource.com/c/1308460 Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 502d2e21 2018-11-01T11:06:23 Vulkan: Crunch RenderPassDesc. This reduces the size of the RenderPass desc from 64 to 12 bytes. Bug: angleproject:2522 Change-Id: Iff2df87ba65be0bd976bba81c76c285cb0fa1ceb Reviewed-on: https://chromium-review.googlesource.com/c/1308459 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 763cec07 2018-10-26T14:19:54 Vulkan: Add warnings for cache struct packing. This adds diagnostic warnings for packed cache structures. It ensures the packed versions of the structures don't have any unexpected misalignments or inserted members. This gives us consistent behaviour and ensures all memory is initialized. Implemented for Clang/GCC/MSVC. Bug: angleproject:2522 Change-Id: I6ec453a40d292e4a498319ffa767988a502d225e Reviewed-on: https://chromium-review.googlesource.com/c/1302533 Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 78bcd2be 2018-10-16T15:05:20 Vulkan: Fix deleting in-use descriptor sets. Sequences of many frames with uniform updates could lead to a bug where we attempt to delete descriptor sets that are still in use. To trigger the bug we need to write enough uniform data to trigger a descriptor set to be freed. This would correctly trigger refresh sets in the Program. But if there was a second program idle in the background that also allocated descriptors from the old pool, the bug would manifest. Fix this by storing a shared handle to the descriptor pool in the Program. The dynamic descriptor pool won't recycle descriptor pools internally unless there are zero outstanding references to the pool. We could also improve this in a resource sharing situation by keeping a single shared dynamic descriptor pool per share group. Includes a contribution from tobine@google.com that adds a test to cover the bug. Bug: angleproject:2863 Change-Id: Id585b85f33f8cfa3772ceff3af512d1e4fb0b75a Reviewed-on: https://chromium-review.googlesource.com/c/1271919 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 996628a4 2018-09-24T16:39:26 Vulkan: Add support for VkPipelineCache The cache is initialized from the application's blob cache and is occasionally written back to it for disk storage. Bug: angleproject:2516 Change-Id: I4cba4b00a7b9641c2983ef07159bc62cd10a5519 Reviewed-on: https://chromium-review.googlesource.com/1241373 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill d511948b 2018-09-18T11:32:42 Use FastVector in angle::Subject. This improves object binding performance. We no longer need to check two containers when manipulating the bound subjects/observers. Bug: angleproject:2763 Change-Id: I0ccb2e76be47bd8a28fba5a0c3eff857aa166bb7 Reviewed-on: https://chromium-review.googlesource.com/1227795 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Frank Henigman d731ff82 2018-08-13T18:18:51 Vulkan: support depth offset. Update pipeline description with depth offset state. Enable test. BUG=angleproject:2353 Change-Id: Ib087bbff8e145b0e6e862eb6616a4943a989a7b2 Reviewed-on: https://chromium-review.googlesource.com/1173447 Commit-Queue: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Geoff Lang c16f518e 2018-07-18T10:40:03 Vulkan: Disable the depth/stencil states when there are no depth/stencil buffers We sometimes emulate depth-only or stencil-only buffers with depth-stencil buffers. Disable depth-stencil states that allow reading or writing to these buffers that should not exist. BUG=angleproject:2739 Change-Id: I4f54800404f340eb53f04176e208f19a83a2899c Reviewed-on: https://chromium-review.googlesource.com/1141932 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Commit-Queue: Geoff Lang <geofflang@chromium.org>
Jamie Madill 21061026 2018-07-12T23:56:30 Vulkan: Use angle::Result error handling. Introduces a vk::Context class to contain an error handler and Renderer pointer. This abtracts the common code path for ContextVk + DisplayVk. Removes vk::Error in favor of the POD angle::Result class. There are a few remaining usages of gl::Error that will have to be cleaned up when we can change the front-end APIs. Bug: angleproject:2713 Change-Id: I5e68f223d595c6c561b59d6a85759e5738ed43c6 Reviewed-on: https://chromium-review.googlesource.com/1128924 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Luc Ferron b70ad524 2018-07-09T16:06:26 Vulkan: Fix all depth/stencil related failures due to Y flip - Fixes all deqp functional_fragment_ops_depth_stencil_* with Y flipping enabled. Bug: angleproject:2673 Change-Id: I94a4225dec8adf9113309e8b8b2c8aa61f6a2bb9 Reviewed-on: https://chromium-review.googlesource.com/1129857 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 834a3a10 2018-07-09T13:32:39 Vulkan: Add driver uniforms set. This will be used to specify the applied Vulkan. We will use this to implement OpenGL line rasterization rules. Bug: angleproject:2717 Change-Id: I3395bf620a01c4b84b19a00037d05f148e5523f3 Reviewed-on: https://chromium-review.googlesource.com/1120151 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Luc Ferron <lucferron@chromium.org>
Luc Ferron 1a135adc 2018-07-04T10:35:31 Vulkan: Fix how the viewport is calculated with Y flip - This fixes all ViewportTest.* and MipmapTest.* - Tests left to fix in end2end: - BlitFramebuffer* - MaxTextureSizeTest.* - PointSpritesTest.* Every other test in angle_end2end_tests are working. Bug: angleproject:2673 Change-Id: I162083bc847c15fa5490ab524ad4c22747d232ea Reviewed-on: https://chromium-review.googlesource.com/1126333 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Luc Ferron bf6dc379 2018-06-28T15:24:19 Vulkan: Flip viewport on y for the backbuffer only - Hide the implementation behind a feature flag, currently disabled permanently as I'm working on fixing the different failures. - SimpleOperationTest.* passing Bug: angleproject:2673 Change-Id: Ic86520c3cc478d62bebbaeaf4c6b33c439a67b0f Reviewed-on: https://chromium-review.googlesource.com/1119089 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 8a4c49fb 2018-06-21T15:43:06 Vulkan: Refactor DynamicDescriptorPool. Changes the set counting to use a free count by descriptor type. Also changes the init to take a pool size count that more closely mirrors Vulkan. Will lead to a fix for descriptor set recreation. Bug: angleproject:2678 Change-Id: I26816befec28ebb866fb0f3aaba27d2bc1d3be43 Reviewed-on: https://chromium-review.googlesource.com/1110661 Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Luc Ferron <lucferron@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Luc Ferron 5fd3693e 2018-06-19T14:55:50 Vulkan: Mask the alpha channel for draw when needed When the angleFormat we have has no alpha channel, but the actual texture underneath has one, we shouldn't be drawing over the alpha channel, so we apply a mask on it when we're in this situation. Bug: angleproject:2597 Change-Id: Ia7110709e6ee32bb61988d08f5049e4e80e7e24e Reviewed-on: https://chromium-review.googlesource.com/1106759 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill c7918ceb 2018-06-13T13:25:31 Vulkan: Add PipelineLayout and DescriptorSetLayout caches. This will be necessary to support more than one PipelineLayout per instance of ANGLE. Sampler array handling requires different layouts for different sampler uses. Bug: angleproject:2462 Change-Id: I1d8b4919eed1a589002ad1898b05186f420061c7 Reviewed-on: https://chromium-review.googlesource.com/1089806 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Luc Ferron <lucferron@chromium.org>
Jamie Madill aaca96ee 2018-06-12T10:19:48 Vulkan: Minor style fixes. Makes some methods const or static. Fixes parameter names. explicit single argument constructors. Other small style fixes. Bug: angleproject:2462 Change-Id: Ie2bf96603e9eda81d6f2aa788867978de10e9992 Reviewed-on: https://chromium-review.googlesource.com/1089805 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Luc Ferron <lucferron@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Jamie Madill 493f9571 2018-05-24T19:52:15 Add PrimitiveMode packed GLenum. Bug: angleproject:2574 Change-Id: I3d7bd7ca0d69a364a611dc04799ea34906fc4a6c Reviewed-on: https://chromium-review.googlesource.com/1067114 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Jamie Madill 9aef3670 2018-04-27T11:45:06 Vulkan: Implement masked color clears. This implements masked color clear using clear shaders. The shaders themselves were introduced in a prior patch. In order to get the right setup for the draw call to trigger the shaders, we create an internal pipeline from the pipeline cache. We also use a special pipeline layout with only uniform buffers. The masked out color channels are disabled via settings on the pipeline. This fixes the dEQP masked color clear tests. It doesn't handle masked color clears combined with the depth clear bit. It's likely we don't have test coverage for this case. Bug: angleproject:2455 Change-Id: I513248cc0f7e58f490fc16ac9afb40119d730ccc Reviewed-on: https://chromium-review.googlesource.com/1031373 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 78feddc2 2018-04-27T11:45:05 Vulkan: Take serials in PipelineDesc::updateShaders. This makes the API easy to use with internal shaders and pipelines. This is useful for the implementation of masked color clear. Also renames the serials as shader serials. This is more precise than program serials. Bug: angleproject:2455 Change-Id: Ie6247d1212ed4df856b561a5e9f16c0378202588 Reviewed-on: https://chromium-review.googlesource.com/1032857 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 33318de4 2018-05-01T11:22:54 Vulkan: Use ShaderType enum. This re-uses the same machinery as the GL front-end. It saves a lot of custom casting and array sizing. Currently we only support vertex and fragment shaders in Vulkan. Because of the Pipeline cache sizing, it's easier to stick with just VS/FS and introduce the full set of shaders when we move to packing the cache better. Bug: angleproject:2522 Bug: angleproject:2455 Change-Id: I21432a335c741885af87970d8ee52b4a36338304 Reviewed-on: https://chromium-review.googlesource.com/1036927 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jiawei Shao <jiawei.shao@intel.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Luc Ferron 0986f1cc 2018-04-16T13:47:23 Vulkan: Implement depth_range dirty bit and enable tests - Additional fix in GlslWrapper.cpp to remove the @@ markers for unused attributes. - Enables 28 dEQP gles2 tests in functional.depth_range.* Bug: angleproject:2454 Change-Id: I1a6f72d846b476ba681140d4b242208d24e18b95 Reviewed-on: https://chromium-review.googlesource.com/1014262 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Luc Ferron 14f4817c 2018-04-11T08:43:28 Vulkan: Simplify viewport / scissor updates Stop tying the viewport and the scissor together. Instead, we simply use a very large (0->maxInt) scissor when scissor isn't enabled and we use the clipped scissor to the renderArea size when its enabled. Bug: angleproject:2443 Change-Id: If7454793a050b1833c7d3166ea6b380192085c8f Reviewed-on: https://chromium-review.googlesource.com/1006996 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Luc Ferron d17bdfe5 2018-04-05T13:50:10 Vulkan: 3 Final bugfixes to enable all dEQP depth_stencil tests - The scissor when clearing should not be clipped against the viewport. - The render pass was created with a render area == to the current viewport, but the viewport can be changed by the tests multiple times per render pass, so we should always keep the renderArea to the full framebuffer instead. - Enables an additional 163 dEQP tests. - We should clip the scissor to the framebuffer dimensions instead of the viewport. Its valid to do a cmdClearAttachments outside the viewport, but not outside the full framebuffer's dimension. Bug: angleproject:2443 Change-Id: I79168e9f0c782d6dec77470fef938b85ad7a8794 Reviewed-on: https://chromium-review.googlesource.com/998448 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 6c7ab7fe 2018-03-31T14:19:15 Vulkan: Reorganize helper classes. This renames ResourceVk to vk::CommandGraphResource, which should help clarify its usage. This also moves LineLoopHandler, ImageHelper, and the DynamicBuffer and DynamicCommandPool classes into a new vk_helpers module. This file contains helper classes that manage other resources. Also this makes DynamicBuffer and DynamicDescriptorPool no longer inherit from CommandGraphResource. In the future, only Impl objects will be allowed to be graph resources. Bug: angleproject:2318 Change-Id: I0fa23da2ac853d90f3c822547a4a314f247cc757 Reviewed-on: https://chromium-review.googlesource.com/985200 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Luc Ferron dd196e0b 2018-04-04T11:41:44 Vulkan: Implement color mask and depth mask bits These two features are heavily used in the functional.depth_stencil_clear.* dEQP tests. Enable a bunch of color/depth/stencil clear tests, however there is still 2 tests in particular that are giving me trouble. I will work on them separately in a subsequent change. Bug: angleproject:2443 Bug: angleproject:2455 Change-Id: Ic93420c7b525b424e9641f78265e264ddb163ab1 Reviewed-on: https://chromium-review.googlesource.com/996035 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Luc Ferron 364a9557 2018-03-29T09:44:51 Vulkan: Implement stencil test support All the fragment_ops.depth and the fragment_ops.depth tests in dEQP are now working, but not the fragment_ops.depth_stencil. Still debugging these separately and will come up with a fix for them in another commit. Bug: angleproject:2443 Change-Id: I84c3a22f612fb6dcf30598434f96c2100fd29f9c Reviewed-on: https://chromium-review.googlesource.com/993654 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Corentin Wallez <cwallez@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill bc54342b 2018-03-30T10:43:19 Vulkan: Make RenderTargetVk use ImageHelper. Bug: angleproject:2318 Change-Id: I9bc4bb7f5bcd3029a31c8570809f253cf5e4b12a Reviewed-on: https://chromium-review.googlesource.com/980773 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Luc Ferron <lucferron@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Jamie Madill 0cec82a5 2018-03-14T09:21:07 Vulkan: Implement basic depth/stencil buffers. This implements basic depth/stencil states and clearing. This also implements "fallback" texture formats in the texture generation. Fallback formats are those that are chosen at runtime for replacements for main formats which lack driver support. They are different from override formats, which are always used because we assume there is no driver support. The Vulkan spec only asserts that one of the two of D32 or D24 has mandatory support. In the case of AMD, D24 is not supported fully, and we need the fallback format support emulation. Bug: angleproject:2357 Change-Id: Ic86cede3c69ff9893a06b3a55c3b5d2d897fa55f Reviewed-on: https://chromium-review.googlesource.com/916701 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Luc Ferron 5242d5bf 2018-02-15T07:14:35 Vulkan: Implement scissored clears. Bug: angleproject:2356 Change-Id: I33888f9b4ebaf4b0b5af4ad59b12ad963325a790 Reviewed-on: https://chromium-review.googlesource.com/921882 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill e4c5a23b 2018-03-02T21:00:31 Vulkan: Always use LOAD for RenderPass attachments. The RenderPass load/store ops allow us to specify how we want to use the data from the attachments. Previously we had the load op set to CLEAR always, which would prevent us from doing multiple kinds of operation. Using LOAD should conversatively work in any situation as long as we can ensure each Image is cleared before we use it. To this effect this patch also inserts a preliminary clear into each Texture or Renderbuffer Image's initialization. We already had this for Surfaces. In the future we'll improve this by inserting proper load/store ops, but this unblocks a lot more functionality in the interim. Bug: angleproject:2361 Change-Id: I7610eaa39d81b23dd74b4a24b7f28a66a6dfffc6 Reviewed-on: https://chromium-review.googlesource.com/948782 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Luc Ferron <lucferron@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Luc Ferron f8be756e 2018-02-06T15:59:11 Vulkan: Implement blend states and add a simple test for it Bug: angleproject:2346 Change-Id: I462a2cb29ceda5563f48b4a3cc1d0aa20f4a49fc Reviewed-on: https://chromium-review.googlesource.com/907169 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Luc Ferron 00155d58 2018-02-06T10:48:47 Vulkan: Implement scissor test and add a simple test for it Bug: angleproject:2338 Change-Id: I699189fcd41feca1656c8553fdf4c1078421524d Reviewed-on: https://chromium-review.googlesource.com/904861 Commit-Queue: Luc Ferron <lucferron@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Luc Ferron ceb7190b 2018-02-05T15:18:47 Vulkan: Fix bug in reuse of vertex input bindings Fix bug that used previously used vertex bindings. Now we use the current program mask to only initialize the vulkan pipeline with the active vertex input bindings. + adjust dEQP expectations Bug: angleproject:2334 Change-Id: Ie6176eee99f87dc7a95f664d28e8312b9cb274bc Reviewed-on: https://chromium-review.googlesource.com/902434 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Luc Ferron <lucferron@chromium.org>
Jamie Madill 5dd4ad89 2018-01-29T13:53:43 Vulkan: Add a perf test for the Pipeline cache. This micro-benchmark can be used to measure the performance impact of changing the Pipeline cache. For instance, we can check if changing the size of the hash key affects performance significantly. Also updates the build files so angle_perftests can see vulkan.h, and makes the Vulkan headers an explicit source set. This test currently shows that a lot of time is spent in PMurmurHash, with some time also spent in memcmp. Bug: angleproject:2163 Change-Id: Ie8bb3e31d58590f373d28cbbb59f7e372b80cc29 Reviewed-on: https://chromium-review.googlesource.com/884882 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Jamie Madill ffa4cbb6 2018-01-23T13:04:07 Vulkan: Implement the Pipeline cache. This currently keeps a cache of every PSO compiled and does not trim the cache or evict old members on memory pressure. This will be done as a follow-up. Improves the speed of the Draw Call microbenchmark 50x when using a single state change. Bug: angleproject:2163 Change-Id: I2cceb38ca57ae639f36a944f4571b627481b92da Reviewed-on: https://chromium-review.googlesource.com/876954 Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Corentin Wallez <cwallez@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 112a3a8e 2018-01-23T13:04:06 Vulkan: De-couple Program from VertexArrayVk dirtyiness. The VertexArrayVk is responsible for filling out the packed shader input info in ANGLE's packed PipelineDesc info structure. This packed info structure is used for Pipeline init and caching lookup. The prior design had this info depend on the active inputs in the current Program. This was undesirable because then, on a Program change, the ContextVk would have to call into the VertexArrayVk to invalidate this info. Instead, keep a working copy of the VertexArrayVk bits and only update the bits corresponding to dirty vertex attributes. This simplifies the cached state management a little bit for ContextVk. This also means we don't have to update the cached copy in the VertexArray on a change in VertexArray binding. Bug: angleproject:2163 Change-Id: I5ba74535367aed74957d17bdc61f882508562d0e Reviewed-on: https://chromium-review.googlesource.com/881703 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Jamie Madill 3c424b48 2018-01-19T12:35:09 Vulkan: Add vk_cache_utils.h. This file contains the Pipeline and RenderPass cache utils. Also renames renderervk_utils.h to vk_utils.h and the format utils file. Refactoring change only. Bug: angleproject:2163 Change-Id: I5113a9a2c6f0b0960d38e6c2d8e391fa2d9f5f6a Reviewed-on: https://chromium-review.googlesource.com/876505 Reviewed-by: Frank Henigman <fjhenigman@chromium.org> Reviewed-by: Corentin Wallez <cwallez@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
kmx.io     mail     discord kmxgit v0.4.0