kc3-lang/angle/src/libANGLE/renderer/vulkan/ProgramVk.cpp

src/libANGLE/renderer/vulkan/ProgramVk.cpp

Log

Author Commit Date CI Message
Jamie Madill 0c77f3ad 2021-03-10T15:58:00 Vulkan: Implement shader buffers descriptor cache. Implements a descriptor set cache for UBOs, SSBOs, and atomic counter buffers. Storage Images and framebuffer fetch input attachments are not yet included. Requires moving the buffer barrier handling into ContextVk, similarly to how we handle the barriers for Textures. The packed description key for the descriptors uses a "fast" vector with a basic minimum size. For most cases of a few buffers this will fit easily in stack memory, but for larger programs with many buffers we fit this into heap memory. The key has a large upper bound due to the high ES 3.2 requirements and the need to index several values such as the offset and binding size. We use dynamic offsets for uniform buffers when possible. This ensures applications like Manhattan 3.1 that use sets of common buffers with changing offsets hit the cache most of the time. Because of resource limits we pick at compilation time whether to use dynamic or static descriptor sets. Mostly this applies to tests that use a large number of uniform buffers. A future implementation could be smart and would recompile the program with heuristics to use a minimal number of dynamic indices. Reduces the number of descriptor set updates from ~300 -> ~30 per frame in Manhattan 3.1 and in Asphalt 9 from 900+ to as low as 0 per frame. Bug: angleproject:5736 Change-Id: I5c2a3881bec90d301dab15cc86c8a70e60674ad7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2757515 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill efc53d0f 2021-04-03T21:30:15 Vulkan: Refactor to descriptor offset array. This is in preparation for using dynamic offsets for other sets than the default descriptors. Bug: angleproject:5736 Change-Id: I546e688ad57a35b380f8d7f303ffc0c62773ac18 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2804813 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi 776c6015 2021-02-26T00:00:57 Vulkan: Call glslang at compile time With this change, the ANGLE translator immediately compiles the generated GLSL into SPIR-V with glslang and discards the source. This is in preparation for generating SPIR-V directly, by making the frontend and backend already able to digest it. This change also allows the expensive glslang calls to be parallelized, improving the following perf test by about 20%: LinkProgramBenchmark.Run/vulkan_compile_and_link_multi_thread Previously, the test was run as such in the Vulkan backend: Main Thread 1 Thread 2 Compile1 ---> Compile2 ---------------------> Translator Translator <--- <--------------------- Link glslang for shader1 glslang for shader2 Done With this change, it is run as such: Main Thread 1 Thread 2 Compile1 ---> Compile2 ---------------------> Translator Translator glslang glslang <--- <--------------------- Link Done glslang_wrapper_utils no longer interacts with glslang! A rename will follow. Bug: angleproject:4889 Change-Id: If4303e8ba0ba43b1a2f47f8c0a9133d0bee1a19a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2721195 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 572fd801 2021-03-22T16:14:34 Vulkan: Rename Uniforms/XFB descriptor desc. The new name is more consistent with the other names. Refactoring change only. Bug: angleproject:5736 Change-Id: Idc47fef29040e3a422267795c4536163a6f8eb4e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2779953 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 77637f2d 2021-02-19T15:18:52 Vulkan: Generate xfb support code in SPIR-V for emulation path This change moves the code generation at link time from source code to SPIR-V. As a result, transform feedback extension and emulation paths are more similarly handled before SPIR-V transformation (they both store information identically in the ShaderInterfaceVariableInfoMap). This change gets rid of the @@ XFB-OUT @@ marker. With no source code generation at link time, shader compilation can be moved to glCompileShader time. Bug: angleproject:4888 Change-Id: I8cdb89c22b57ce48cf5d226b8e41622d9d550d46 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2713269 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Mohan Maiya 550f2a3e 2021-02-24T09:49:42 Vulkan: Shader support for EXT_shader_framebuffer_fetch_non_coherent Translator can accept gl_LastFragData and 'inout' variable to gain access to framebuffer attachment data. The Vulkan translator replaces it with the SubpassInput type variable. Note that this works only for the noncoherent version of the extension. Bug: angleproject:5454 Test: *EXTShaderFramebufferFetchNoncoherent*.* Change-Id: I392f84ee3ad3eb9fbd09d0b7ff83731a9a3f33f6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2598060 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Mohan Maiya <m.maiya@samsung.com>
Jamie Madill 093250e0 2020-12-23T12:07:46 Vulkan: Make shader variable info maps a class. Instead of using a map type directly we abstract the info maps into an encapsulated class. We can enforce a specific API set instead of using the same API as the map class. This also cleans up a few of the APIs related to these maps. This change will allow future changes to the way the variables are stored in the class without drastically changing the interface. Bug: angleproject:3572 Bug: angleproject:4524 Change-Id: Ic1a63e1776c39f49b895a1274bae8282d7a6b9b5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2600080 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 98c5ff6b 2020-10-25T16:26:55 BinaryStream: Preserve 64-bit integer data. Previously the code would truncate 64-bit data to fit in 32-bits. This ran into a serialization bug when expanding a 64-bit mask. The new blend state masks for extended range were out of range for the 32-bit promotion that was happening before. Also refactors how we capture bools and enums to be more consistent. size_t is now correctly saved and loaded as 64-bits. Bug: angleproject:5247 Change-Id: I452a98c1b0add4c0cf45493032e9310e7d8321b2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2497561 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 0f2542e1 2020-10-21T00:38:40 Vulkan: Validate SPIR-V transformer at link time Using an ASSERT, the SPIR-V transformer is run at link time to catch bugs in tests that link but don't draw. Bug: angleproject:4249 Change-Id: Icbeeb6e968f9901ba003918b1dd1d2eef0792469 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2488227 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Tim Van Patten 233cc31b 2020-09-15T11:05:17 Vulkan: Use linkedShaderStages in GlslangGetShaderSpirvCode() The parameter linkedShaderStages is passed to GlslangGetShaderSpirvCode(), but is unused. This CL uses the parameter, rather than gl::AllShaderTypes(). This CL also removes the unused parameter variableInfoMap from the call chain. Bug: angleproject:3570 Change-Id: Ic60084e87cb2aa2a245e3e963598851331379e89 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2412364 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Tim Van Patten f12e4123 2020-08-25T12:01:12 Vulkan: Match descriptor pool sizes to descriptor set layouts When a descriptor pool is created, a list of descriptor types and counts are given to vkCreateDescriptorPool(). Later, when allocating a descriptor set from the pool, we pass along a descriptor set layout to vkAllocateDescriptorSets() which is used to determine how many of each type of descriptor (i.e. binding) to allocate from the pool. In order for our "free descriptor set" counts to be accurate for each pool, the descriptor pools need to be created with descriptor counts that match the descriptor set layout binding counts. This change fixes a bug where the descriptor set layouts were created with more bindings than the descriptor pool sizes, causing the "free descriptor set" count to be inaccurate, leading to allocating too many descriptor sets from a pool. Bug: angleproject:3570 Test: VulkanDescriptorSetTest Change-Id: I660bf02d29a1291391fb15f39e6479bf348d0f83 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2391114 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Courtney Goeltzenleuchter ae09e889 2020-09-10T17:00:13 Fix mismatch issue with precision qualifiers. GLSL allows varyings passed from one stage to another to not match in precision (e.g. float & half-float). Vulkan doesn't allow that so adjust those mismatches to use the higher precision. To fix we keep track of the precision of varyings and in the Vulkan backend if we see they are different patch up the SPIR-V to make them match. Bug: angleproject:3078 Change-Id: I385d31e082da46ccdd4817b6612f5f9d9cbce17c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2337755 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Courtney Goeltzenleuchter c5b5cf6c 2020-09-10T16:58:18 Refactor to pass ProgramMergedVaryings to link impl Follow-on CL needs the ProgramMergedVaryings in the Vulkan backend to generate valid SPIRV. Bug: angleproject:3078 Change-Id: Ic442a3e0bd713fec36bd6b9420f67f3b1118e5ad Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2404336 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Charlie Lao 8765b46a 2020-07-15T17:10:35 Vulkan: Make mDefaultUniformStorage per ContextVk Right now the dynamic buffer for default uniform is per program. Most of time the buffer is unused. That is a huge waste of memory (and these memory are wired memory). This CL moves the mDefaultUniformStorage from per ProgramVk to ContextVk so that we all share with each other. Bug: b/161391337 Change-Id: I1fe8523b2b2dbc39bec3509a3432e38e34bd5713 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2274870 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Charlie Lao <cclao@google.com>
Charlie Lao cc551385 2020-06-29T10:25:45 Vulkan: Add cache for default uniform descriptor set Instead of free and allocate a descriptor set whenever buffer changes, add it into the descriptorset cache. Every time you get the buffer, we also look in the cache first to see if there is a descriptorset for that buffer or not. If yes, we use the cached descriptor set instead of allocate one. Bug: b/159457348 Change-Id: I64c252dfe22db66ea8188415e1b384af371b82ab Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2273698 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Charlie Lao 22e1f3e2 2020-06-26T14:26:50 Vulkan: Let all shader stages share one buffer for default uniform Right now each shader stage has its own vk::DynamicBuffer for default uniform storage. This is less efficient than just share one buffer. This CL moves the storage from per shader stage into its ProgramVk object. Bug: b/159457348 Change-Id: If47248ea23c4e48407d3b211583ae2b048d4d10f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2265281 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi b549be96 2020-07-02T16:17:10 Vulkan: Warm up glslang on startup As soon as glslang is initialized (which is done when EGL is initialized), try to compile a simple shader. This will warm up glslang's internals to avoid hitches on first shader compilation when a WebGL application starts. Potential future optimization is to use a worker thread to do this. Bug: chromium:1091440 Change-Id: I6ac22e982d748af52a6e8b61239889425edc2edb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2280401 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Tim Van Patten 68e60a92 2020-06-26T11:10:07 Vulkan: Don't store transformed SPIR-V Since the original SPIR-V is never store after being modified, we don't need to keep the transformed SPIR-V either. Bug: angleproject:4513 Test: CQ Change-Id: I4b70cfa355a08107e6a27377f6258cbf9e4072be Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2270718 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 7993f34d 2020-06-24T18:39:08 Vulkan: Store transformed SPIR-V in ProgramExecutableVk The original SPIR-V is specific to a ProgramVk while the transformed SPIR-V is specific to a ProgramExecutableVk. This CL keeps the original SPIR-V in the ProgramVk::mOriginalShaderInfo and the transformed SPIR-V in ProgramExecutableVk::mTransformedShaderInfo. This allows us to skip an extra copy that was occurring during PPO link(). Bug: angleproject:4513 Test: CQ Change-Id: I796dcd6bcfb2df08d9b8f1a5295b2821a4096f16 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2265484 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten d10ba166 2020-05-19T18:00:38 Rename ProgramState::getProgramExecutable() to getExecutable() The "Program" part of ProgramState::getProgramExecutable() is already implied by being part of a ProgramState, so it can be removed. Bug: angleproject:3570 Change-Id: I35ffb2af81196fa7f189f7d3a37158f5f1951141 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2209317 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Tim Van Patten <timvp@google.com>
Courtney Goeltzenleuchter a48f3c38 2020-04-29T10:03:14 Implement setBinaryRetrievableHint Remove the UNIMPLEMENTED() macro. Doing nothing is an acceptable implementation at this time. Test: Angry Birds 2 Bug: b/154930384 Change-Id: Ie46c084feb2015aa5fecdbbf2295d46e5087e3dc Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2172711 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Cody Northrop <cnorthrop@google.com>
Tim Van Patten 913f4f42 2020-04-15T00:54:37 Vulkan: Support VS, FS, and CS in the same PPO This CL adds support for a Program Pipeline Object to have a VS, FS, and CS attached to the same PPO and then using that PPO for both draw and dispatch calls. Bug: angleproject:3570 Test: KHR-GLES31.core.compute_shader.sso* Change-Id: I262cdbdfd442f6db5ba2b45d1308003102b237cb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2150078 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Tim Van Patten 405f8e7b 2020-02-24T17:38:10 Vulkan: Support Program Pipeline Objects Add support for PPOs to the Vulkan back end. Bug: angleproject:3570 Change-Id: I5403456929847c185467b008d810f31ecfcb60cc Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2072652 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Tim Van Patten b387ce93 2020-02-21T16:05:41 Vulkan: Move ProgramInfo/ShaderInfo to ProgramExecutableVk The classes ProgramInfo and ShaderInfo are being moved into ProgramExecutableVk along with the ProgramVk members mDefaultProgramInfo and mLineRasterProgramInfo. This refactor is necessary since these members are common between ProgramVks and ProgramPipelineVks. Bug: angleproject:3570 Change-Id: I94cdb1096c6a0c007d858135af694da58d7897ff Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2068901 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 85c076ee 2020-02-20T17:54:35 Vulkan: Make mVariableInfoMap a gl::ShaderMap<> The ShaderInterfaceVariableInfoMap information is specific to each shader stage, since the locations are determined for each shader stage. This change makes mVariableInfoMap a gl::ShaderMap<> to handle this, which will make it easier to compile separable programs, determine the variable locations, and transform the SPIR-V. Bug: angleproject:3570 Change-Id: I28b71a37efe54bbcfe1dcd445fa03ee71e74f0a6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2062741 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@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>
Tim Van Patten 10ade024 2020-02-13T18:11:54 Create the ProgramExecutableVk Class The ProgramExecutableVk class is being created to collect data structures that are common to both ProgramVks and ProgramPipelineVks, as well as any shared functions. This allows callers to make Program-/ProgramPipeline- specific queries without needing to know exactly which responded by querying the current ProgramExecutableVk. This will also allow the necessary data structures to only be populated and stored within the ProgramExecutableVk when necessary and reused as often as necessary. A few things are being moved into this class: - mVariableInfoMap This information will be required to defer translating the SPIR-V until when the vulkan pipeline layout is determined and the actual locations are known.   This will also allow removing determining these locations twice (during GLSL->SPIR-V compilation and pipeline layout). - createPipelineLayout() - update*DescriptorSet() - Descriptor Set Layout lists - Pipeline Layout - Various other functions/members related to pipeplines, descriptor sets, etc. Bug: angleproject:3570 Change-Id: I4b5ababeafec865148783c8ffd4c15f659f4856d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2055656 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 68083e89 2020-03-04T15:55:53 Vulkan: Move cached samplers/images and has*() to ProgramExecutable The active samplers and images are being moved from Program into ProgramExecutable to unify interacting with them for Programs and ProgramPipelines Also, create some helper functions for gl::Program that ProgramExecutable can call to make it easier for ProgramPipeline to respond to similar queries for each of the Programs in the ProgramPipeline. Bug: angleproject:3570 Change-Id: I0b37f1a379e56b9659d82d92f6d7a546beee11cd Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2087648 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 7e0699a2 2020-02-05T17:04:06 Create the ProgramExecutable Class The ProgramExecutable class is being created to collect data structures that are common to both Programs and ProgramPipelines, as well as any shared functions. This allows callers to request the current ProgramExecutable from the State and make Program-/ProgramPipeline-specific queries without needing to know exactly which responded. This will also allow the necessary data structures to only be populated and stored within the ProgramExecutable when necessary and reused as often as necessary. Bug: angleproject:3570 Change-Id: I101f08ab03421894667b4a426a04d2147489f0e1 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2040512 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Geoff Lang 07467b4a 2020-03-20T10:40:56 Remove GL_CHROMIUM_path_rendering Bug: chromium:1063193 Bug: angleproject:4270 Change-Id: I35b24b7d8d892181955e49dd2495655bc57cb0df Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2112275 Reviewed-by: Zhenyao Mo <zmo@chromium.org> Reviewed-by: Jonah Ryan-Davis <jonahr@google.com> Commit-Queue: Geoff Lang <geofflang@chromium.org>
Jamie Madill 9d812d35 2020-03-18T15:31:14 Vulkan: Remove public FramebufferVk::getFramebuffer. Goal: enable rendering feedback loop detection for Manhattan. Making getFramebuffer private prevents direct access to the Framebuffer internals. Hiding access lets the FramebufferVk class switch up its implementation to support different VkFramebuffers for feedback loop cases. Also this is a general code cleanup. Bug: angleproject:4490 Change-Id: Ib4f5e34d476ccd8fc43a1adb031933ab347ca7b7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2109333 Reviewed-by: Cody Northrop <cnorthrop@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill a741abb9 2020-02-21T16:37:37 Vulkan: Rename CommandGraphResource to Resource. Also renames the h and cpp files to ResourceVk (to keep distinct from other resource.h/cpp files) and renames 'onResourceAccess' to 'retain'. Cleans up a few remaining mentions of the command graph in comments. Bug: angleproject:4029 Change-Id: Ifc8e880c8cea3fc48a4aec4730191c88aa35a076 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2065920 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill c58458e6 2020-02-19T14:51:41 Vulkan: Remove CommandGraph code. Also updates relevant comments to no longer refer to a graph. Bug: angleproject:4029 Change-Id: Ic29716e9ae4926870f902947d49d8fee7af98662 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2057804 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 3b82506a 2020-02-11T09:35:38 Vulkan: Command graph linearization (Step 5). Implements ES 3.1 support. Bug: angleproject:4029 Change-Id: I3c08602cf4ec64621bfdb40cd40d142b666e6edf Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2046052 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Cody Northrop <cnorthrop@google.com>
Jamie Madill 3c066404 2020-02-05T12:43:12 Vulkan: Command graph linearization (Step 4). Implements GLES 3.0 functionality for the new command recording style. Tested with angle_end2end_tests. dEQP/CTS testing will come later. Bug: angleproject:4029 Change-Id: Ibdc4f1355a7b880946699ec227fd9807f4a27af9 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2040197 Reviewed-by: Cody Northrop <cnorthrop@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi e5385ea9 2020-01-27T13:40:18 Vulkan: Compile shaders at link time Since line raster emulation was changed to use specialization constants, it has been possible to compile shaders at link time. However, program pipeline objects would have required keeping the shader sources around for recompilation. Now that all necessary decorations are modified directly in SPIR-V, it's possible to compile the shaders at link time and forget about their sources. Program pipeline objects then simply "reconfigure" the generated SPIR-V. A next step could be to also create the Vulkan pipeline object at link time. A number of failures due to gaps in CTS testing prevent that work currently. In particular, in some situations the generated SPIR-V is not per spec, for example it may contain vertex attributes with aliasing locations, or have transform feedback capture of array elements misconfigured. Bug: angleproject:3394 Bug: angleproject:4253 Change-Id: I54c0884cf056b511a4a306225cc6ed2cef84d257 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2023186 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 20b1259a 2020-02-05T17:08:05 Vulkan: Command graph linearization (Step 2). ES 2.0 is feature complete. Passes all of the angle_end2end_tests with the new linear command recording method. Also runs the T-Rex benchmark without any obvious glitches. Likely has issues with creating too many RenderPasses. ES3 is mostly untouched. Bug: angleproject:4029 Change-Id: Ic5acf3d768495fbffd07b07bf0a6f2b5787c51f8 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2012900 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi f1f082e1 2020-01-17T14:53:38 Vulkan: Set varying location & xfb decorations in SPIR-V The shader translator outputs arbitrary location indices. Once compiled by glslang, the SPIR-V transformer modifies these decorations. If the transform feedback extension is used, it will also add the relevant decorations to the varyings that are captured. Bug: angleproject:3394 Change-Id: I5ecafd0536408612a5d4b920dbabbfabe650657c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2008468 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 5ca050d1 2020-01-20T14:59:41 Vulkan: Set fragment output locations in SPIR-V The shader translator outputs arbitrary location indices for fragment outputs. Once compiled by glslang, the SPIR-V transformer modifies these decorations. Bug: angleproject:3394 Change-Id: Ib9d8336bccc392e789e4d93031fdcce9c466b7a6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2011214 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 71e6afb1 2020-01-14T14:12:31 Vulkan: Set set/binding in SPIR-V This change introduces a SPIR-V transformer that modifies shader interface variable decorations directly in SPIR-V instead of manipulating the input GLSL. Currently, descriptor set and binding indices are set by the transformer. The shader translator outputs arbitrary set and binding indices. Once compiled by glslang, the SPIR-V transformer modifies these decorations. The ultimate goal is to be able to modify the SPIR-V again when program pipeline objects decide a different set/binding is necessary. Bug: angleproject:3394 Change-Id: If358265a72bf1fe9f5676562b39a632cb2e05dc4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2001477 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 3712b2e4 2020-01-24T17:08:30 Vulkan: Command graph linearization (Step 1). This initial prototype introduces a new feature to the Vulkan back-end that disables the deferred command recording. The intent is to have a lower CPU overhead during submission calls which currently walk a DAG. The feature is not complete. Currently it only passes the ANGLE SimpleOperationTests. Moreover it is extremely simple and only allows use of one command buffer at a time. In the future we'll allow open command buffers for recording outside and inside render pass commands at the same time. We'll also support collapsing RenderPasses together for some use cases. Currently the prototype only passes "SimpleOperationTest". There are quite a few unimplemented features like queries, XFB, etc. Bug: angleproject:4029 Change-Id: I82760986683f55e37ac4ea559de6f4cffb6ef84e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1953485 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com>
Shahbaz Youssefi 135f8fcb 2020-01-13T16:18:41 Vulkan: Remove inactive uniforms in the translator By removing inactive uniforms in the translator, glslang wrapper doesn't need to comment them out. Additionally, inactive uniforms don't find their way in the default uniform block, reducing its size if there's a mix of active and inactive uniforms. As collateral, it also fixes a bug where inactive uniforms of struct type were not correctly removed by glslang wrapper. Bug: angleproject:3394 Bug: angleproject:4211 Bug: angleproject:4248 Change-Id: I874747070e875fe24bf59d39d1322e319e280a16 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1999278 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
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>
Jamie Madill 579c5940 2020-01-17T13:20:21 Vulkan: Rename onGraphAccess to onResourceAccess. Clarifies the function for the upcoming command graph linearization. Bug: angleproject:4029 Change-Id: Ib50997c007053757de533f69c155a92d7555e0b2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2003235 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 441b72f0 2020-01-16T14:48:39 Vulkan: Make ContextVk own ResourceUseList. This moves the resource use tracking functionality out of CommandGraph. Making the list a separate class helps the implementation avoid tricky circular include problems. Bug: angleproject:4029 Change-Id: I3288fc685b21e949f12b0796109a2b7bb117c249 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2002931 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 4ff1651e 2020-01-16T12:42:27 Vulkan: Add ResourceUseList helper. This abstracts away the ResourceUse tracking from CommandGraph. Pulling it out will allow us to more easily swap resource use tracking into ContextVk. The command graph is eventually going away so we need to move functionality out of CommandGraph. Bug: angleproject:4029 Change-Id: Icc3bcc824510b03e91f8ff010a82751a81516bfa Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2002930 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
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>
Shahbaz Youssefi 49135f68 2019-12-10T11:27:09 Vulkan: Fix descriptor pool init in presence of atomic counters The descriptor pool was initialized with the number of atomic counter buffers, but we always use 8 storage buffers for atomic counter buffer emulation regardless of how many atomic counter buffers there. This change fixes the descriptor pool initialization by requesting 8 storage buffer descriptors if there are any atomic counter buffers. Bug: angleproject:4199 Change-Id: Ide58c013cba8ac46228dba795a82304b97ee83ac Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1959489 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Hyunchang Kim caa6eccd 2019-10-27T13:46:56 Vulkan: Implement Transform Feedback support via extension Implemented transform feedback extension path. Where VK_EXT_transform_feedback is supported, extension path will be taken over an emulation path. Extension path has advantages in terms of performance. BUG=angleproject:3206 Test: dEQP-GLES3.functional.transform_feedback.* angle_end2end_tests --gtest_filter=TransformFeedbackTest* Change-Id: Ia07c23afb289d9c67073469a97b714ec96f5265a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1882767 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Cody Northrop <cnorthrop@google.com>
Jamie Madill 61c6aecc 2019-11-28T14:30:18 Vulkan: Fold read access into write flags on buffer writes This could lead to a subtle bug: Say we use a buffer as a storage buffer and write to it. We then must issue a memory barrier before using it as a uniform buffer. However we would set the SHADER_READ bit as a read mask on a storage buffer access. This seems like it could lead to a missing barrier. Bug: angleproject:4178 Change-Id: I486002739b7fb000ffacc0a1e996784b7875e7ba Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1943034 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Le Quyen a48f95ab 2019-10-14T14:49:49 Move Vulkan GlslangWrapper code to a shared location. Metal backend will reuse Vulkan's GlslangWrapper code. The Metal backend will use this code to translate glsl to spirv then cross compile to Metal Shading Language using spirv-cross. So the source code of GlslangWrapper should be moved outside vulkan folder. Bug: angleproject:2634 Change-Id: I208062daf0d77756c9d32cfdab925b7dfdf83e05 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1858042 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 68591eff 2019-10-13T15:05:23 Vulkan: Store ImageView access in the graph. This will ensure we don't destroy the image views when they are still in use by other Contexts. Bug: angleproject:2464 Change-Id: I1d3ba2ad241250e31ea32873446c4cb23971750d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1843236 Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 0b9ebe58 2019-10-15T11:15:42 Vulkan: Add "ImageViewHelper". This allows views to track a different lifetime than vk::ImageHelper. This in turn will fix the race condition on ContextVk destruction when releasing ImageViews owned by TextureVk and RenderbufferVk. For now this is a refactoring change only. Bug: angleproject:2464 Change-Id: I9581975bd5d4913233bbed8439dd4a632cc78a2a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1843231 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Ian Elliott <ianelliott@google.com>
Tim Van Patten 616a4dc0 2019-09-12T16:10:47 Vulkan: Full support for program interface queries Program interface queries are a generic way to query attributes of the program like uniforms, samplers, attributes, etc. This change supports those queries for program outputs. Bug: angleproject:3596 Test: dEQP-GLES31.functional.program_interface_query.* Test: ProgramInterfaceTest.cpp Change-Id: I0f13692949073b45988b6f930eee9eaa6411bbe2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1801998 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill f10bf6bf 2019-09-26T10:27:18 Vulkan: Implement multi-threaded GL. The main component of this change is to make vk::BufferHelper, vk::ImageHelper and vk::SyncHelper use a common path. We introduce a new "vk::SharedGarbage" helper class that stores small lists of garbage from individual objects like an ImageHelper or BufferHelper. The SharedGarbage is stored in the RendererVk with the ResourceUse of the helper object. The ResourceUse tells RendererVk when it is safe to destroy the GarbageObjects. New "onGraphAccess" commands are added in a few places to enable the common garbage collection path. A couple Context-only resources like default attributes now are referenced where they were not before. Also reorganizes some functions so we can add a few helpful ASSERTs to our graph dependencies. Added "updateCurrentAccessNodes" for this. Also adds a "RendererScoped" helper to replace many uses of "ContextScoped". The multithreading EGL tests mostly pass but have some remaining flakiness so cannot yet be enabled. Bug: angleproject:2464 Change-Id: Ia3e3ae8848d731abf3f21ebe04c33e381e130be0 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1808444 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill b540da89 2019-09-19T14:19:12 Vulkan: Use ResourceUse to track object lifetime. With the new resource tracking scheme the CommandGraph, tracking a "Context serial" aka "current" serial is no longer necessary for CommandGraphResources. Serial tracking has been moved to the shared ResourceUse struct that gets updated on a command submission. Thus we don't need to store the serial as a current separate piece of info in BufferHelper/ImageHelper. Will lead to further redesign for the multi-threading support for Vulkan. Bug: angleproject:2464 Change-Id: I1ae4bcc27fcfb93422b4ab4c9682a458e482f295 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1785990 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 8be7a4c7 2019-09-19T12:48:31 Re-land "Vulkan: SwiftShader integration." Re-land fixes Win7 configs and placement of the SwiftShader module for ASAN/TSAN configs. Adds a new EGL extension for picking SwiftShader when using the Vulkan back-end. Also cleans up ICD enabling code RendererVk. Also includes a change to a buffer size necessary to support SwiftShader's minimum limits. Bug: angleproject:3876 Bug: b/140251624 Change-Id: I5e16057ac4de07bbdbbd248542b1b9103133294f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1810065 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill d7f28aae 2019-09-19T14:19:10 Vulkan: Pass CommandGraph when updating serials. This is in preparation for storing a pointer to a shared resource use structure. Bug: angleproject:2464 Change-Id: I8f4ba1c71de6ad6a27ac06fc8012a0e94267cc16 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1785988 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Austin Kinross bf4268a3 2019-09-17T13:33:56 Fix misc VS2019 bool conversion warnings BUG=angleproject:3921 Change-Id: I06de5131f98b27c2556ed60dd7228c9cfa154802 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1811858 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Zhenyao Mo db7a36f4 2019-09-17T00:22:32 Revert "Vulkan: SwiftShader integration." This reverts commit 1224802c214d8f3d5b1f31a35d4461ada39b94c1. Reason for revert: https://bugs.chromium.org/p/angleproject/issues/detail?id=3912 Original change's description: > Vulkan: SwiftShader integration. > > Adds a new EGL extension for picking SwiftShader when using the Vulkan > back-end. Also cleans up ICD enabling code RendererVk. Also includes a > change to a buffer size necessary to support SwiftShader's minimum > limits. > > 32-bit is currently left disabled to work around an issue on AMD Win7. > > Bug: angleproject:3876 > Bug: b/140251624 > Change-Id: I33c55d994801d6154caca0cad0a608c1e808d517 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1776231 > Commit-Queue: Jamie Madill <jmadill@chromium.org> > Reviewed-by: Geoff Lang <geofflang@chromium.org> TBR=geofflang@chromium.org,jonahr@google.com,jmadill@chromium.org Change-Id: Id7464250f6941b0228a6b2a9bd0349823727275b No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: angleproject:3876, b/140251624 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1808101 Reviewed-by: Zhenyao Mo <zmo@chromium.org> Commit-Queue: Zhenyao Mo <zmo@chromium.org>
Jamie Madill 1224802c 2019-09-16T15:17:33 Vulkan: SwiftShader integration. Adds a new EGL extension for picking SwiftShader when using the Vulkan back-end. Also cleans up ICD enabling code RendererVk. Also includes a change to a buffer size necessary to support SwiftShader's minimum limits. 32-bit is currently left disabled to work around an issue on AMD Win7. Bug: angleproject:3876 Bug: b/140251624 Change-Id: I33c55d994801d6154caca0cad0a608c1e808d517 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1776231 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 4fdeb8d2 2019-09-12T10:08:15 Vulkan: Reduce uniform block dynamic buffer size. This is to fit in the minimum required Vulkan limits that SwiftShader exposes. Bug: angleproject:3876 Bug: b/140251624 Change-Id: If1a36fc578675ebb07d3f6546689709882c174f3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1801264 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Alexis Hétu <sugoi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 90a58622 2019-09-04T15:39:58 Refactor ShaderVariable to Remove Specializations The following structs are being refactored and moved into the parent struct ShaderVariable: VariableWithLocation Uniform Attribute OutputVariable InterfaceBlockField Varying Bug: angleproject:3899 Test: CQ Change-Id: I389eb3ab4ed44a360e09fca75ecc78d64a277f83 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1785877 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi 912e52d8 2019-08-23T00:25:09 Vulkan: Storage image support Image bindings are placed after atomic counters in the "resources" descriptor set. There are two issues yet to be addressed: - GL can create a 2D (array) view of a 3D image, but this is not allowed in Vulkan. If this cannot be made possible, emulation needs to be done. https://github.com/KhronosGroup/Vulkan-Docs/issues/1033 - GL can create an image view of a texture with a different format and have the data reinterpreted. This is not currently done. Bug: angleproject:3563 Change-Id: I95c4d92c50bb033212a9a67f3f2d6f97c074c7bf Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1767366 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 5d27a696 2019-08-13T11:29:07 Support separable shader programs The spec states: Shader stages including vertex shaders, fragment shaders, and compute shaders...A single program object can contain all of these shaders, or any subset thereof. This change allows shader programs without a fragment shader. The biggest driver of this change is dEQP since a large number of tests create shader programs without a fragment shader. Bug: angleproject:3803 Test: dEQP-GLES31.functional.program_interface_query Change-Id: Id6cb098c62a1489a14b9ec1b31bd4cd59f655e49 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1752010 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Cody Northrop <cnorthrop@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
James Dong 83a369bb 2019-08-14T10:39:34 Vulkan: Improve cubemap emulation seam handling Changes seamful cubemap emulation to always compute the derivative, emulating the bias parameter by scaling the provided derivatives. This results in more accurate mipmap levels for seams within primitives. There are some artifacts as a result of how derivatives are calculated, but this matches the native driver. Bug: angleproject:3243 Bug: angleproject:3732 Change-Id: Icb976e2a7e14cb4210645571edc037d4e607bd0d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1754383 Commit-Queue: James Dong <dongja@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
James Dong 1d5aaa6c 2019-08-06T11:20:13 Vulkan: support dynamic indices in array of arrays Expands existing struct-sampler rewrite to flatten arrays of arrays. This allows us to support dynamically-uniform array indexing, which is core in ES 3.2. Samplers inside (possibly nested) structs are broken apart as before, and then if the type resulting from merging the array sizes of the field and its containing structs is an array of array, the array is flattened. Also adds an offset parameter to functions taking in arrays to account for this translation. As a result of outer array sizes leaking into function signatures, functions taking arrays of different sizes are duplicated according to how the function is invoked. Bug: angleproject:3604 Change-Id: Ic9373fd12a38f19bd811eac92e281055a63c1901 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1744177 Commit-Queue: James Dong <dongja@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Hyunchang Kim 6d653c02 2019-08-14T14:09:02 Vulkan: Set empty buffer to xfb descriptor set as an initial buffer When we have xfb descriptor set without buffer bindings it causes errors in Vulkan validation layer. So we set empty buffer by default. BUG=angleproject:3681 Tests: dEQP-GLES3.functional.*transform_feedback* dEQP-GLES3.functional.lifetime.attach.deleted_input.buffer_vertex_array Change-Id: I4e7c9c9ee661f1fc92bf01fb78a2212c1a423eaa Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1753387 Commit-Queue: Hyunchang Kim <hckim.kim@samsung.com> Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 5a2553a7 2019-08-07T14:44:12 Vulkan: Emulate subgroup ops in seamful cubemap emulation Where subgroup ops are not available, they are emulated as such: Code with subgroup ops: float lH = subgroupQuadSwapHorizontal(layer); float lV = subgroupQuadSwapVertical(layer); float lD = subgroupQuadSwapDiagonal(layer); bool isHelperH = subgroupQuadSwapHorizontal(gl_HelperInvocation); bool isHelperV = subgroupQuadSwapVertical(gl_HelperInvocation); if (gl_HelperInvocation) { layer = !isHelperH ? lH : !isHelperV ? lV : lD; } Emulated code: float nonHelperLayer = gl_HelperInvocation ? 0.0 : layer; float lH = abs(dFdxFine(nonHelperLayer)); float lV = abs(dFdyFine(nonHelperLayer)); float lD = abs(dFdxFine(lV)); float isHelperDiffH = abs(dFdxFine(float(gl_HelperInvocation))); bool isNonHelperH = isHelperDiffH > 0.5; float isHelperDiffV = abs(dFdyFine(float(gl_HelperInvocation))); bool isNonHelperV = isHelperDiffV > 0.5; if (gl_HelperInvocation) { layer = isNonHelperH ? lH : isNonHelperV ? lV : lD; } Both paths are supported as on nvidia devices the emulated code misbehaves. This change therefore effectively only enables seamful cube map emulation on Android where subgroup operations are not supported. Bug: angleproject:3243 Bug: angleproject:3732 Change-Id: I9664d9760756758748183eb121c626f176789f3a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1742222 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
James Darpinian 7e48c9eb 2019-08-06T17:17:19 Add explicit integer casts WebKit uses the -Wshorten-64-to-32 flag which warns on these cases. Bug: 3439 Change-Id: I8c1de60da0f173ca2036e2120e79b857f5f2775f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1740866 Commit-Queue: James Darpinian <jdarpinian@chromium.org> Reviewed-by: Kenneth Russell <kbr@chromium.org>
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 93560ef5 2019-07-25T16:13:02 Vulkan: Seamful cube map emulation In GLSL, a cube texture is sampled with one of textureCube* functions. This function takes a 3D coordinate which is a vector from the center of the cube and identifies a direction to sample from. GLES2.0 has the following table that translates this 3D coordinate (Rx, Ry, Rz) to a face and ST coordinates within that face. This table can be found in Section 3.7.5 (Cube Map Texture Selection). A compiler pass is implemented in ANGLE that replaces samplerCube declarations with a sampler2DArray. The textureCube* functions are replaced with the corresponding texture* functions with the translated coordinates according to that table. Gradients provided to textureCubeGrad are translated using the same formulae, which is not precise but the spec specifies this projection to be implementation dependent. Helper invocations enabled through WQM (whole quad mode) cause a nuisance in that the extrapolated varyings used as coordinates in a textureCube call could have a different major axis (and therefore face) from the non-helper invocations that lie within the geometry. subgroupQuadSwap* operations are used in conjunction with gl_HelperInvocation to make sure the helper threads calculate texture UVs in the same face as the non-helper invocations. Bug: angleproject:3300 Bug: angleproject:3240 Bug: angleproject:3243 Bug: angleproject:3732 Change-Id: I0cb6a9b1f2e1e6a392b5baca1c7118ed1c502ccf Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1715977 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 44e690ca 2019-07-22T16:42:55 Vulkan: Support unaligned atomic counter buffer binding GLES doesn't require any implementation-specified alignment requirement for atomic counter buffers. They are emulated with Vulkan storage buffers, which do have restrictions. The storage buffers are bound at aligned offsets, and the remaining offsets are passed to the shader as uniform values. This means that the driver uniforms are now also bound to the compute pipeline. Bug: angleproject:3566 Change-Id: I1a3429438f76d95e33cb5c6ef2c9370a10d900d6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1713095 Commit-Queue: 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>
James Dong 593b1a6c 2019-07-30T10:02:22 Vulkan: Fix array handling in default uniforms Changes default uniform block layout generation to only remove the first array subscript, which causes arrays of arrays to be recognized correctly. Bug: angleproject:3604 Test: ./angle_deqp_gles31_no_gtest --deqp-egl-display-type=angle-vulkan -n 'dEQP-GLES31.functional.shaders.arrays*' Test: ./angle_deqp_gles31_no_gtest --deqp-egl-display-type=angle-vulkan -n 'dEQP-GLES31.functional.program_interface_query.*array*array*' Change-Id: Ifacc665cae87c651253e5cb8d929178d7fc92817 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1724906 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: James Dong <dongja@google.com>
Jiacheng Lu 1ebe6014 2019-07-29T16:02:15 Vulkan: Only check sampler when setUniform1i(v) 1. only do isSampler checking when update int uniforms 2. inline some simple methods of LinkUniform This fix reduce CPU time spent on `isSampler` call from 70ms to 30ms in T-Rex bench on specific platform. Bug: angleproject:3743 Change-Id: I98ef3c892df27e08d54ed40946d924b5a50c796c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1725114 Commit-Queue: Tobin Ehlis <tobine@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jiacheng Lu 3a36f306 2019-07-19T12:43:22 Always set matrix uniform to dirty when it updates Remove the dirty checking of matrix uniform update at frontend and always let backend update whenever glUniformMatrix* is called. Profiled with UniformsBenchmark/* with vulkan backend, performance increases around 6% when data being updated is changing and performance increases around 2% when data being updated stays same. Bug: angleproject:3705 Change-Id: I8eaf6a1231e634b69c6dc540db1b9d3a312bf45d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1709725 Commit-Queue: Jiacheng Lu <lujc@google.com> Reviewed-by: Tobin Ehlis <tobine@google.com>
Cody Northrop a2129356 2019-07-23T12:54:13 Vulkan: Add support for 2D array textures Includes changes from jmadill to align with Vulkan backend design. Correctly setting layer count and depth when the texture type is 2Darray. Vulkan requires depth of 1 for 2Darray textures. Bug: angleproject:3189 Change-Id: I0d58c33fcd75b1d768ea0308ac6e54230d8cfcc5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1721169 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jonah Ryan-Davis <jonahr@google.com> Commit-Queue: Cody Northrop <cnorthrop@google.com>
Shahbaz Youssefi b82d8633 2019-07-15T11:23:08 Vulkan: Atomic counter buffer support Vulkan doesn't treat atomic counters especially, and they are emulated with atomic access to storage buffers. A single atomic counter buffer binding per pipeline is supported. All the atomic counters identify an offset within this buffer. The shader is modified to include a storage buffer definition with `uint counters[];` as the only field. A compiler pass replaces atomic counter definitions with variables that hold the corresponding offset parameter, as well as changing atomic_uint types to just uint (as the offset). Where an atomic counter variable is used, it is replaced with the offset variable (plus the array index, if array). At the same time, built-in `atomicCounter*` functions are replaced with a corresponding `atomic*` function and `memoryBarrierAtomicCounter` is replaced with `memoryBarrierBuffer`. Bug: angleproject:3566 Change-Id: Iefb3d47de6a5cb3072bfa0cb94a46ac6a886d369 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1704635 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@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>
Mohan Maiya 9ec3f51d 2019-07-19T13:06:16 Reland "Vulkan: Implement OES_get_program_binary extension" Reason for revert: default uniform initialization was incomplete This change has the following fixes: 1. Add missing default uniform initialization when loading program binaries. 2. Re-enable OES Program Binary capabilities for Vulkan. 3. Added two angle end2end test, - ProgramBinaryES3Test.BinaryWithLargeUniformCount uses several uniforms across the vertex and fragment shaders. - ProgramBinaryES3Test.ActiveUniformShader tests the difference between uniform static and active use Bug: angleproject:3216 Bug: angleproject:3217 Bug: angleproject:3665 Tests: dEQP-GLES3.functional.shader_api.program_binary* angle_end2end_tests --gtest_filter=ProgramBinary* Change-Id: If6886f01241d65bb1e17a21cc3406533021072ee Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1699069 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jiacheng Lu f35f1110 2019-07-19T09:18:55 Add script to apply clang-format on all sources 1. python script wrapper to call clang-format over the whole code base 2. Add clang-format rule `IncludeBlocks: Preserve` to tell clang-format do not merge include blocks 3. Fix existed clang-format issue in code base Bug: angleproject:3532 Change-Id: I289292dc62c2784ff21688065c87c3f3f5538f17 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1709720 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jiacheng Lu ec30d78d 2019-07-03T12:56:58 Optimize uniform matrix update 1. Add a fast matrix update function to do a single memcpy for uniform matrix assignment with same layout. It benefits row-4 no-transpose GLSL matrix and col-4 transpose HLSL matrix. 2. Make boolean IsColumnMajor to be a template parameter in generate uniform matrix updating, which gets rid of the conditional branch in loop and has better performance. 3. Add e2e test of uploading multiple 3x4 GLSL matrices at the same time, which adds coverage to this CL. Bug: angleproject:3632 Change-Id: Id1701ef6fbf63ea4b9884254d93ea8eacfe4e16a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1688274 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi f92fc916 2019-07-15T22:39:23 Vulkan: Compute shader support A DispatchHelper class is created as the equivalent of FramebufferHelper as a command graph resource. There's currently a single dispatcher and all dispatch calls are recorded on that. Context dirty bits are set up in such a way that graphics and compute workloads are independently handled, so that issuing a dispatch call wouldn't cause a framebuffer's render pass to rebind resources. Bug: angleproject:3562 Change-Id: Ib96db48297074d99b04324e44b067cfbfd43e333 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1688504 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 998a37c9 2019-07-15T22:13:14 Vulkan: merge bindGraphics/ComputeDescriptorSets The two called bindDescriptorSets with a fixed GRAPHICS or COMPUTE bind point enum value. The differentation however would result in unnecessary code duplication with upcoming compute support. Bug: angleproject:3562 Change-Id: Ica4400c573a90fda168c64ad777d87aa83256e48 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1703524 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@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 55efe37c 2019-07-03T11:52:21 Vulkan: Cleanup buffer dependencies onWrite now sets the dependency to make sure the order of setting dependency and adding memory barrier is correct. onReadByBuffer is added to handle buffer-to-buffer dependency and barrier setting correctly without causing a graph loop. onExternalWrite is added so that BufferVk doesn't have to track write access flags. Additionally, setting write dependencies now include both read and write flags. This is in preparation for SSBO support where the buffer can be used to read data in addition to write. Bug: angleproject:3561 Change-Id: I2028186ea14459cd159cf79f6d640df54538fc62 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1687119 Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Mohan Maiya a8da8668 2019-07-02T12:21:26 Vulkan: Implement OES_get_program_binary extension - Serialize and deserialize completed shader source of program for saving out for glGetProgramBinary(). - Cleaned up some unnecessary includes in cpp files. - Some refactoring within ProgramVk::ShaderInfo to minimize code duplication. - Added ProgramVk::ShaderInfo::saveShaderSource and ProgramVk::ShaderInfo::loadShaderSource. - Updated vk_caps_utils.cpp to enable getProgramBinary and add the GL_PROGRAM_BINARY_ANGLE program binary format. This follows the pattern for other backends. Bug: angleproject:3216 Tests: dEQP-GLES3.functional.shader_api.program_binary* angle_end2end_tests --gtest_filter=ProgramBinaryTest* Change-Id: I927a27aaf9aa3d7fac550819ee80d2676ec1d1be Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1683099 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
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>
Jamie Madill c773ab98 2019-06-25T17:11:58 Vulkan: Recycle dynamic buffer storage. This adds a free list to the dynamic buffer storage. Buffers are added to the free list when the retained buffers are released. They are taken from the free list when we allocate a new buffer. We only allocate a new buffer in the ring when we run out of free buffers. This reduces the amount of time we spend in allocation for frequent updates. Now that we're recycling buffers inside of DynamicBuffer we also need to be a bit more careful about when we allow ourselves to reuse them. If they're still in use by the GPU we should not try to modify them. Bug: angleproject:3082 Change-Id: Ibee5a7e2fe4a17f4a2f7af6bc6bcce54bdc413c2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1646548 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com>
Jamie Madill 7f2520f1 2019-06-26T11:18:33 Vulkan: Refactor DynamicBuffer::init. This will allow us to more easily create a white box test that sets a very small initial size for a dynamic buffer. Bug: angleproject:3082 Change-Id: Ic02bbee83ee8e0f4bfe182e9448c2ce60dea66d5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1667645 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tobin Ehlis <tobine@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>
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 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>
Courtney Goeltzenleuchter eaf2d928 2019-04-18T16:31:25 Add support for OES_depth_texture Note: Includes workaround for http://anglebug.com/3452 - some Android devices do not indicate filtering support on VK_FORMAT_D16_UNORM. Bug: angleproject:3103 Test: angle_end2end_tests --gtest_filter=DepthStencilFormatsTest.DepthTexture/* angle_end2end_tests --gtest_filter=DepthStencilFormatsTest.PackedDepthStencil/* angle_end2end_tests --gtest_filter=DepthStencilFormatsTest.DepthTextureRender/ES2_VULKAN Change-Id: Ic325fb94ab0e619a17c2e149e0e0865fa4142f3a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1575426 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi b783fbc7 2019-05-10T23:16:17 Vulkan: Fix matrix uniforms A piece of code shared with d3d was used to implement this feature. However, we store the uniform data transposed and row-major in d3d to accomodate the fact that matrix indexing in HLSL is the opposite of GLSL (mat[row][col] in HLSL as opposed to mat[col][row] in GLSL). There were two functions that copied the source matrix fields either as column-major or source-major (corresponding to `transpose` false or true respectively in glUniformMatrix*) into a row-major destination. These functions are refactored into one, which copies from any-major source to any-major destination. The HLSL backend uses the "to row-major" variant and the Vulkan backend uses the "to column-major" variant. Bug: angleproject:3198 Change-Id: I7254da4fbe6916af78c5906abcb82ca01674ab9f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1607541 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 9d519ab1 2019-05-09T23:09:46 Vulkan: Rework layout block storage conversion Previously, a pass over the shader was made, converting shared and packed interface blocks with block storage to std140. This resulted in link success between interface blocks with different storage as they were all translated to std140. With this change, this pass is removed. The link step proceeds with the block storage specifiers as seen by GLES, and only upon Vulkan GLSL shader generation "shared" and "packed" are converted to std140. Bug: angleproject:3199 Change-Id: I069415ab9c9b4e1034bc00f64cd2d9e2d73f5956 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1605262 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi fdbece20 2019-05-09T18:06:34 Vulkan: Support unsigned int uniform types Bug: angleproject:3198 Change-Id: Iac5d1389f4f466b7886246d8ca529e8f26a6a5bd Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1603629 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
kmx.io     mail     discord kmxgit v0.4.0