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

src/libANGLE/renderer/vulkan/ContextVk.cpp

Log

Author Commit Date CI Message
Shahbaz Youssefi 2d964a47 2020-10-30T16:46:30 Vulkan: Defer clears even if following command is scissored Take the following scenario: 1. glClear 2. glScissor(half of framebuffer) 3. glDrawArrays The clear in step 1 is deferred. When FramebufferVk::syncState is called in step 3, the deferred clear was applied using vkCmdClearColorImage because the draw call is scissored. This causes loadOp=LOAD to be used after the clear because the render pass is started too small (the same size as the scissor). This change makes scissored operations also take advantage of loadOp=LOAD with deferred clears. A number of changes are made to this effect: - FramebufferVk::syncState no longer limits collecting deferred clears to no-scissor. - FramebufferVk::startNewRenderPass automatically expands the render area to full size if it's clearing any attachment. - A number of bugs are fixed where FramebufferVk::flushDeferredClears is called with the scissor area. Instead, flushDeferredClears now unconditionally uses the complete render area. Note that these bugs didn't have symptoms as "scissor" and "deferred clears" were mutually exclusive. Bug: angleproject:4988 Change-Id: I24fc3d88bf9c8998869b36c863692d0f0acce994 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2511371 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi a0e91016 2020-10-30T10:01:36 Vulkan: Don't break the render pass on scissor change Prior to this change, the render area was decided when the render pass was started, and remained fixed. If a small scissor was initially used, this created a render pass with a small area. If then the scissor region was expanded, the render pass was broken. This change instead expands the render area on scissor change to avoid breaking the render pass. If glInvalidateSubFramebuffer previously successfully resulted in storeOp=DONT_CARE, this optimization may need to undo that. As a result, the invalidate area is stored in the render pass and if the render area grows beyond that, invalidate is undone. Bug: angleproject:4988 Change-Id: I4e8039dec53a95a193a97cb40db3f71e397568d6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2508983 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 265c5fa9 2020-11-02T21:50:25 Vulkan: Fix scissor update in FramebufferVk::syncState A previous change [1] made FramebufferVk::syncState update scissor and rasterization samples only when the DRAW framebuffer is synced. This is incorrect as the READ framebuffer is synced before the DRAW framebuffer, and if the two are the same, the latter is discarded. Very few functions sync both READ and DRAW framebuffers when they are identical. A test is tailored to expose this bug. [1]: https://chromium-review.googlesource.com/c/angle/angle/+/2510013 Bug: angleproject:4988 Change-Id: I6123ac18dded938171bc90a04d4d81f1b42a1694 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2515742 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Courtney Goeltzenleuchter 6bc362c4 2020-10-18T19:09:47 Allow single threaded CommandProcessor In replacing the legacy CommandQueue code with the threading capable CommandProcessor it would be good to be able to run the CommandProcessor in a single-thread environment. This CL changes the meaning of the feature flags for the commandProcessor and asynchronousCommandProcessing so that enabling commandProcess only changes the code paths to use the command processor but work it still done as part of the submitting thread (e.g. ContextVk). Enabling asynchronousCommandProcessing will cause a separate worker thread to be spawned which will asynchronously process the commands. This allows us to switch to the CommandProcessor without threading and then enable threading once performance issues are resolved. Bug: b/161912801 Bug: b/170329600 Change-Id: I534862b109a7e7708108190b7c3e894071d4c2ed Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2483580 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Mohan Maiya 3fcf17e6 2020-11-02T12:51:35 Vulkan: Bug fix in sRGBDecode logic When a sampler's GL_TEXTURE_SRGB_DECODE_EXT state was toggled between GL_DECODE_EXT and GL_SKIP_DECODE_EXT VkImageViews of the TextureVk object were not being updated. Add sRGB_decode state as part of ImageViewSubresourceSerial so we retrieve the correct VkImageView from the texture cache. Bug: angleproject:3609 Tests: angle_end2end_tests --gtest_filter=SRGBTextureTestES3.SRGBDecodeSamplerParameterToggle*Vulkan Change-Id: I897e461957d408b5a5b4f03fefc05f2e9684c7b7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2514900 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Tim Van Patten fcd35965 2020-09-29T14:15:51 Vulkan: Make DescriptorPoolHelper a Resource Descriptor pools need to live as long as the descriptor sets that are allocated from them. Using Serials while building a command to judge a pool's lifetime is prone to errors, since a command's Serial value isn't known until the command is submitted, leading to deleting pools too early relative to when the descriptor set is actually used. This CL updates DescriptorPoolHelper to inherit from Resource, so the descriptor pools can be retain()'ed. This allows the Resource's counter to indicate that a pool is in use until the command's Serial is known and can be recorded to indicate when the command completes. This prevents descriptor pools from being destroyed before the command completes (while the descriptor sets are still in use), or even before the command has been submitted. Destroying a descriptor pool resets all of the descriptors that were allocated from it, which can trigger a variety of VVL errors depending on when it's erroneously performed. This CL also adds the necessary retain() calls for the descriptor pools. In particular, the pools need to be retained each time a cached descriptor set that was allocated from it is re-used. This is relatively simple with the current design, since we always clear the descriptor set caches whenever a new pool is allocated, so the descriptor pool binding is always accurate. Bug: angleproject:5030 Test: VulkanMultithreadingTest::MultiContextDrawSmallDescriptorPools() Change-Id: Iac9e7efef338f169a6bf8ac3b2140e03dd326641 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2504457 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Tim Van Patten ee4e0866 2020-09-24T11:39:49 Vulkan: Add descriptor set allocation counters Add descriptor set allocation counters for the following: - ContextVk - Driver uniform allocations for graphics and compute pipelines. - ProgramExecutableVk - ANGLE driver uniforms - Uniforms - Textures - Other shader resources - UtilsVk - All of the UtilsVk::Function types increment the same counter Each object's counters live within the object itself and the cumulative total is output as part of that object's destruction. On Present, all of the descriptor set counts are collected into a single total which is used to update the overlay each frame. In order to see the cumulative total output for each object, the following GN args must be enabled: is_debug = true angle_enable_perf_counter_output = true To see the descriptor set allocation overlay: ANGLE_OVERLAY=VulkanDescriptorSetAllocations Bug: angleproject:5067 Test: Manual verification with angle_perftests Change-Id: Ie45fda56ade3e68bfba7bf6da9554eb05a02c6b6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2429487 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill 9e3eec54 2020-10-25T15:44:09 Revert "Vulkan: Make DescriptorPoolHelper a Resource" This reverts commit 5dcd29a6e532e4bd617af8767d488120b57f3b2c. Reason for revert: Breaking the ANGLE -> Chromium roller: https://chromium-review.googlesource.com/c/chromium/src/+/2496281 Original change's description: > Vulkan: Make DescriptorPoolHelper a Resource > > Descriptor pools need to live as long as the descriptor sets that are > allocated from them. Using Serials while building a command to judge a > pool's lifetime is prone to errors, since a command's Serial value > isn't known until the command is submitted, leading to deleting pools > too early relative to when the descriptor set is actually used. > > This CL updates DescriptorPoolHelper to inherit from Resource, so the > descriptor pools can be retain()'ed. This allows the Resource's counter > to indicate that a pool is in use until the command's Serial is known > and can be recorded to indicate when the command completes. This > prevents descriptor pools from being destroyed before the command > completes (while the descriptor sets are still in use), or even before > the command has been submitted. Destroying a descriptor pool resets all > of the descriptors that were allocated from it, which can trigger a > variety of VVL errors depending on when it's erroneously performed. > > This CL also adds the necessary retain() calls for the descriptor pools. > In particular, the pools need to be retained each time a cached > descriptor set that was allocated from it is re-used. This is relatively > simple with the current design, since we always clear the descriptor set > caches whenever a new pool is allocated, so the descriptor pool binding > is always accurate. > > Bug: angleproject:5030 > Test: MultithreadingTest::MultiContextDrawSmallDescriptorPools() > Change-Id: I5fdeeb46159448dfd679d7169e423048348be5ab > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2437609 > Commit-Queue: Tim Van Patten <timvp@google.com> > Reviewed-by: Jamie Madill <jmadill@chromium.org> > Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> TBR=courtneygo@google.com,timvp@google.com,jmadill@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: angleproject:5030 Change-Id: I0fd6d9a0e1b0989b22368ef98652281288699deb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2497222 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 96714af8 2020-10-22T17:28:20 Vulkan: Dirty pipeline and desc set after mid-RP clear with draw If UtilsVk::clearFramebuffer issues a draw call on the currently open render pass (as opposed to starting one itself), ContextVk would be unaware of the fact that the graphics pipeline and descriptor sets have been changed. This change sets the necessary dirty bits for ContextVk to recover from a UtilsVk::clearFramebuffer call. Bug: chromium:1141040 Change-Id: I865220fb5b3b78bf4c6b6b2896e57d8a7490c270 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2493184 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill e86b4257 2020-10-21T13:28:10 Vulkan: Refactor how finishToSerial() retires commands. This cleans up some of the logic around command completion. We no longer query each Fence in order after a finishToSerial call. Instead we clear out all the commands that have been completed up to the latest Fence. Also cleans up ordering code. Because we now have a linear ordering of Serials we can validate the ordering is "good". Attempt at solving a suspicious ASSERT firing on AMD Win7. Bug: angleproject:5198 Change-Id: I7750eff7f8d3aff397e7880f997ad86c9bf7d7a8 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2490342 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill 33e30205 2020-10-17T19:03:45 Vulkan: sRGB cleanups. A few fixes to how we check for the sRGB override in the TextureVk class. In at least one instance there was a potential edge case where in syncState we might not create the Texture with the mutable bit the second time through the function. Bug: angleproject:5176 Change-Id: I4f1ca6e469b10514c3a0de3120be9ade62568084 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2482292 Reviewed-by: Mohan Maiya <m.maiya@samsung.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 5dcd29a6 2020-09-29T14:15:51 Vulkan: Make DescriptorPoolHelper a Resource Descriptor pools need to live as long as the descriptor sets that are allocated from them. Using Serials while building a command to judge a pool's lifetime is prone to errors, since a command's Serial value isn't known until the command is submitted, leading to deleting pools too early relative to when the descriptor set is actually used. This CL updates DescriptorPoolHelper to inherit from Resource, so the descriptor pools can be retain()'ed. This allows the Resource's counter to indicate that a pool is in use until the command's Serial is known and can be recorded to indicate when the command completes. This prevents descriptor pools from being destroyed before the command completes (while the descriptor sets are still in use), or even before the command has been submitted. Destroying a descriptor pool resets all of the descriptors that were allocated from it, which can trigger a variety of VVL errors depending on when it's erroneously performed. This CL also adds the necessary retain() calls for the descriptor pools. In particular, the pools need to be retained each time a cached descriptor set that was allocated from it is re-used. This is relatively simple with the current design, since we always clear the descriptor set caches whenever a new pool is allocated, so the descriptor pool binding is always accurate. Bug: angleproject:5030 Test: MultithreadingTest::MultiContextDrawSmallDescriptorPools() Change-Id: I5fdeeb46159448dfd679d7169e423048348be5ab Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2437609 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Courtney Goeltzenleuchter ed876984 2020-10-03T11:00:36 Vulkan: functionally complete worker thread Working on enhancing worker thread to completely own primary command buffers. This will include not only processing SCBs from main thread into a primary, but also submitting those command buffers to the queue. The CommandProcessor is a vk::Context so it can handle errors in the worker thread. When the main thread submits tasks to the worker thread it also syncs any outstanding errors from the worker. Include asynchronousCommandProcessing feature that will control whether the worker thread task does it's work in parallel or not. If false, we wait for the thread to complete it's work before letting the main thread continue. If true, the thread can execute in parallel with the main thread. Bug: b/154030730 Bug: b/161912801 Change-Id: I00f8f013d6cbb2af12a172c4f7927855db2f0ebf Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2328992 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Mohan Maiya 7bbe497a 2020-10-16T14:46:45 Vulkan: Implement EXT_texture_sRGB_decode Implement EXT_texture_sRGB_decode. This builds on the existing functionality from EXT_texture_sRGB_override, with 2 major edge cases: 1. sRGB_decode allows the texture state to be overridden by sampler state, which is implemented by forcing a a texture state sync during updateActiveTextures if a texture is bound to the same unit as a sampler with that state 2. texelFetch calls require us to reenable decoding, regardless of decode state. We add a new compiler pass (FlagSamplersWithTexelFetch) to mark samplers that are used with texelFetch in order to support this. This change also re-enables EXT_texture_sRGB_R8, which was disabled due to a dEQP bug that this change will bypass. Bug: angleproject:3609 Bug: angleproject:4503 Test: dEQP.GLES31/functional_srgb_texture_decode_skip_decode_* Test: GLES31/functional_state_query_texture_*_srgb_decode_* Test: GLES31/functional_state_query_sampler_*_srgb_decode_* Test: GLES31/functional_debug_negative_coverage_*_srgb_decode_* Test: GLES31/functional_android_extension_pack_extensions_ext_texture_srgb_decode Test: angle_end2end_tests --gtest_filter=SRGBTextureTest.*Vulkan* Change-Id: I4a67e487dc82e2f57c8c87d4bcd8ef442b6fe220 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2359481 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Tim Van Patten 16102e8b 2020-10-17T01:15:05 Reland "Vulkan: Fold deferred clears into current clears" This reverts commit 37c400146e59f718b516eb58e16fa53c8a88bf21. Reason for revert: Need to clear the package cache when bisecting. Original change's description: > Revert "Vulkan: Fold deferred clears into current clears" > > This reverts commit e416c92a81c9ef01d633ec5c05e81c2551b6c5d6. > > Reason for revert: Reverted parent: > https://chromium-review.googlesource.com/c/angle/angle/+/2481612 > > Original change's description: > > Vulkan: Fold deferred clears into current clears > > > > If there are clears prior to a glClear() call, those clears were > > flushed (starting a new render pass) and then the clear call's clears > > would be applied (essentially modifying the loadOps of said render > > pass). > > > > The main downside of the above is that the current glClear() clears > > don't get a chance to be deferred. This was observed in Chrome which > > clears an attachment with an emulated format, then switches > > framebuffers. > > > > Additionally, if the render pass had already been started, the deferred > > clears could have become inlined instead of breaking the render pass. > > Although, it's unlikely for there to be deferred clears when the render > > pass is already open. > > > > This change first identifies which clears need to go through the draw > > path (scissored, masked or as workaround for driver bug). It merges the > > rest of the clears (that don't need the draw path) with the deferred > > clears. It then checks deferred clears and applies them by either: > > > > - vkCmdClearAttachments if mid RP > > - Start a new render pass and use loadOps, if any draw-based clear needs > > to follow. > > - Modify current RP loadOps / defer the clear > > > > Afterwards, the draw-based clears are applied. > > > > Bug: angleproject:4836 > > Change-Id: Id4992c78983b199734508c9d4bb18ed3195c91ec > > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2455167 > > Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> > > Reviewed-by: Jamie Madill <jmadill@chromium.org> > > Reviewed-by: Charlie Lao <cclao@google.com> > > TBR=syoussefi@chromium.org,jmadill@chromium.org,cclao@google.com > > Change-Id: I85733b3594409df9b96e3d5b34933522c97c42cf > No-Presubmit: true > No-Tree-Checks: true > No-Try: true > Bug: angleproject:4836 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2481613 > Reviewed-by: Tim Van Patten <timvp@google.com> > Commit-Queue: Tim Van Patten <timvp@google.com> TBR=timvp@google.com,syoussefi@chromium.org,jmadill@chromium.org,cclao@google.com # Not skipping CQ checks because this is a reland. Bug: angleproject:4836 Change-Id: I702cd510f39ee46feab27d4efbf61ae5da10d4e2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2481856 Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Tim Van Patten 37c40014 2020-10-16T22:19:20 Revert "Vulkan: Fold deferred clears into current clears" This reverts commit e416c92a81c9ef01d633ec5c05e81c2551b6c5d6. Reason for revert: Reverted parent: https://chromium-review.googlesource.com/c/angle/angle/+/2481612 Original change's description: > Vulkan: Fold deferred clears into current clears > > If there are clears prior to a glClear() call, those clears were > flushed (starting a new render pass) and then the clear call's clears > would be applied (essentially modifying the loadOps of said render > pass). > > The main downside of the above is that the current glClear() clears > don't get a chance to be deferred. This was observed in Chrome which > clears an attachment with an emulated format, then switches > framebuffers. > > Additionally, if the render pass had already been started, the deferred > clears could have become inlined instead of breaking the render pass. > Although, it's unlikely for there to be deferred clears when the render > pass is already open. > > This change first identifies which clears need to go through the draw > path (scissored, masked or as workaround for driver bug). It merges the > rest of the clears (that don't need the draw path) with the deferred > clears. It then checks deferred clears and applies them by either: > > - vkCmdClearAttachments if mid RP > - Start a new render pass and use loadOps, if any draw-based clear needs > to follow. > - Modify current RP loadOps / defer the clear > > Afterwards, the draw-based clears are applied. > > Bug: angleproject:4836 > Change-Id: Id4992c78983b199734508c9d4bb18ed3195c91ec > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2455167 > Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> > Reviewed-by: Jamie Madill <jmadill@chromium.org> > Reviewed-by: Charlie Lao <cclao@google.com> TBR=syoussefi@chromium.org,jmadill@chromium.org,cclao@google.com Change-Id: I85733b3594409df9b96e3d5b34933522c97c42cf No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: angleproject:4836 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2481613 Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi e416c92a 2020-10-06T23:29:02 Vulkan: Fold deferred clears into current clears If there are clears prior to a glClear() call, those clears were flushed (starting a new render pass) and then the clear call's clears would be applied (essentially modifying the loadOps of said render pass). The main downside of the above is that the current glClear() clears don't get a chance to be deferred. This was observed in Chrome which clears an attachment with an emulated format, then switches framebuffers. Additionally, if the render pass had already been started, the deferred clears could have become inlined instead of breaking the render pass. Although, it's unlikely for there to be deferred clears when the render pass is already open. This change first identifies which clears need to go through the draw path (scissored, masked or as workaround for driver bug). It merges the rest of the clears (that don't need the draw path) with the deferred clears. It then checks deferred clears and applies them by either: - vkCmdClearAttachments if mid RP - Start a new render pass and use loadOps, if any draw-based clear needs to follow. - Modify current RP loadOps / defer the clear Afterwards, the draw-based clears are applied. Bug: angleproject:4836 Change-Id: Id4992c78983b199734508c9d4bb18ed3195c91ec Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2455167 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Courtney Goeltzenleuchter f51e99fb 2020-10-13T11:38:28 Vulkan: Remove serial from RenderPassHelper We don't delete RenderPass's so no need to keep a serial. This simplifies things for coming threading code. Bug: b/169788986 Change-Id: I2577b17bc1f6af163725389589d7cd62d09eea13 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2468538 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Alexey Knyazev a2ef8144 2020-10-14T02:25:37 Vulkan: fix undefined right shift in GetMask Bug: chromium:1137170 Change-Id: I95efdc7888e44f2b4227f9c0d892b9102cead0bb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2469616 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 426fa735 2020-10-13T11:51:50 Vulkan: Fix command reordering on release to external When releasing buffer/image to external, the necessary barrier was recorded on the "outside render pass command buffer". However, if the resource was used in the current render pass, that render pass should have closed before this, otherwise the barrier was reordered before it which is incorrect. Bug: chromium:1136367 Bug: chromium:1135792 Bug: angleproject:5002 Change-Id: I237d5e6bb46189a4ce61c2f4602e356955cc87a9 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2468456 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Shahbaz Youssefi 68bd685a 2020-10-10T22:58:41 Reland: "4 Vulkan content defined CLs." Reland "Vulkan: Avoid content restore by detecting no-op stencil" This relands commit 243d0f899e443cd931c78aba7489382dff79edbb. Reland "Vulkan: Restore at the end of RP if write-after-invalidate" This relands commit e5d52ac3b9a00656acdd912ee8cd62dd14784075. Reland "Vulkan: Invalidate/restore depth/stencil separately." This relands commit 61fa0878964a796f6d3b3c13bc3a3849403ecdbd. Reland "Vulkan: Move content-defined tracking to ImageHelper" This relands commit 2392e6b34c0ddfbfd7b4c3cb67323ba463e11a57. Reason for revert: Caused crashes in Fuchsia x64 and on ARM. Reland fixes content defined for external images. Original CL message: Content-defined tracking was done in render targets prior to this change. This had multiple drawbacks: - When a framebuffer attachment is changed (including the first time it's set), it's unknown whether the contents of the attachment is defined. - Invalidate takes effect at the end of render pass, at which point the render target objects may be gone. Attachment ImageHelpers are however correctly tracked. This change moves content-defined tracking to the ImageHelper itself, and tracks it per subresource. ImageHelper::onWrite() now receives the subresource that is being written, and marks it as having defined content. A future optimization can make use of this change to ImageHelper::onWrite to track "dirty" subresources. This can lead to the removal of unnecessary barriers when same-kind writes are done on different subresources of the image. See http://anglebug.com/3347#c15 Bug: b/167275320 Bug: angleproject:4836 Bug: angleproject:5159 Change-Id: If5c1ae7152657fd7c94db7d55bea4fb9ddf835ba Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2464825 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill d6b1c17b 2020-10-10T14:29:15 Revert 4 Vulkan content defined CLs. Revert "Vulkan: Avoid content restore by detecting no-op stencil" This reverts commit 243d0f899e443cd931c78aba7489382dff79edbb. Revert "Vulkan: Restore at the end of RP if write-after-invalidate" This reverts commit e5d52ac3b9a00656acdd912ee8cd62dd14784075. Revert "Vulkan: Invalidate/restore depth/stencil separately." This reverts commit 61fa0878964a796f6d3b3c13bc3a3849403ecdbd. Revert "Vulkan: Move content-defined tracking to ImageHelper" This reverts commit 2392e6b34c0ddfbfd7b4c3cb67323ba463e11a57. Causes crashes in Fuchsia x64 and on ARM. Original CL message: Content-defined tracking was done in render targets prior to this change. This had multiple drawbacks: - When a framebuffer attachment is changed (including the first time it's set), it's unknown whether the contents of the attachment is defined. - Invalidate takes effect at the end of render pass, at which point the render target objects may be gone. Attachment ImageHelpers are however correctly tracked. This change moves content-defined tracking to the ImageHelper itself, and tracks it per subresource. ImageHelper::onWrite() now receives the subresource that is being written, and marks it as having defined content. A future optimization can make use of this change to ImageHelper::onWrite to track "dirty" subresources. This can lead to the removal of unnecessary barriers when same-kind writes are done on different subresources of the image. See http://anglebug.com/3347#c15 Bug: b/167275320 Bug: angleproject:4836 Bug: angleproject:5159 Change-Id: I93d9dfe973caa7ce70aefa46b5b7d04a8637efb3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2464822 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 243d0f89 2020-10-08T21:54:45 Vulkan: Avoid content restore by detecting no-op stencil Previously, as long as stencil was enabled, it was considered that it is also being modified. This caused stencil invalidate to be undone in a number of situations, such as: - glEnable(GL_STENCIL_TEST); // with func/ops default - glDrawArrays(); - glInvalidateFramebuffer([GL_STENCIL_ATTACHMENT]); - glClear(GL_DEPTH_BUFFER_BIT); - Close render pass In the above scenario, invalidation of stencil was undone at the end of render pass. In this change, the following cases are considered read-only stencil: - Func = GL_NEVER, stencilFail = GL_KEEP - Func = GL_ALWAYS, stencilPassDepth* = GL_KEEP - stencilFail = GL_KEEP, stencilPassDepth* = GL_KEEP Note that while the above scenario is fixed for no-op stencil, a similar issue persists if stencil was not no-op. The reason stencil invalidate is undone in that case is due to the fact that it's assumed any command after the invalidate call will be a draw call that outputs to stencil, but that is not the case with the glClear call in this example. Bug: angleproject:4836 Change-Id: Ie2ea2d52b7c8ee2394f5456773a7ef434e2b2b16 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2461465 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Ian Elliott <ianelliott@google.com>
Shahbaz Youssefi e5d52ac3 2020-10-08T14:26:22 Vulkan: Restore at the end of RP if write-after-invalidate If a depth/stencil attachment is invalidated, but subsequently drawn to in the same render pass, undo the invalidate when the render pass is closed. Adapted from https://chromium-review.googlesource.com/c/angle/angle/+/2386478. Bug: b/167275320 Bug: angleproject:4836 Change-Id: I17a35bfd692ddc403ceaa6ec44b5c4f16ff9eed6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2461464 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Ian Elliott <ianelliott@google.com>
Shahbaz Youssefi 61fa0878 2020-10-08T11:35:48 Vulkan: Invalidate/restore depth/stencil separately. Depth/stencil content defined is already tracked separately in the ImageHelper. This change exposes this tracking from RenderTargetVk. Bug: b/167275320 Bug: angleproject:4836 Change-Id: Ie6520e7a4ab557eb233c60c6ab0d4a8f8f098bf6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2462039 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Ian Elliott <ianelliott@google.com>
Shahbaz Youssefi 2392e6b3 2020-10-07T23:59:43 Vulkan: Move content-defined tracking to ImageHelper Content-defined tracking was done in render targets prior to this change. This had multiple drawbacks: - When a framebuffer attachment is changed (including the first time it's set), it's unknown whether the contents of the attachment is defined. - Invalidate takes effect at the end of render pass, at which point the render target objects may be gone. Attachment ImageHelpers are however correctly tracked. This change moves content-defined tracking to the ImageHelper itself, and tracks it per subresource. ImageHelper::onWrite() now receives the subresource that is being written, and marks it as having defined content. A future optimization can make use of this change to ImageHelper::onWrite to track "dirty" subresources. This can lead to the removal of unnecessary barriers when same-kind writes are done on different subresources of the image. See http://anglebug.com/3347#c15 Bug: b/167275320 Bug: angleproject:4836 Change-Id: Iabd1dace4eae9eb379453a9eb7ec6eafc9db1aef Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2462036 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Ian Elliott <ianelliott@google.com>
Alexey Knyazev d5fa6ea9 2020-04-29T04:13:54 Vulkan: Implement OES_draw_buffers_indexed Bug: angleproject:4394 Change-Id: I7db9c695c233b2daf740acc654b1b2e546a8b681 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2172739 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Courtney Goeltzenleuchter 9413c402 2020-10-03T10:58:12 Vulkan: Move CommandBatch for threading support Will need access to CommandBatch class in threading worker. Bug: b/154030730 Change-Id: Ia79eab77a81b135c22bdeecbaf65bf3c301dc987 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2447442 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Mohan Maiya 9f5eb0b8 2020-10-05T08:18:05 Vulkan: Add OES_sample_shading extension support Support OES_sample_shading extension if the underlying Vulkan ICD supports sampleRateShading. Bug: angleproject:3587 Tests: dEQP-GLES31.functional.sample_shading.* Change-Id: I1b324c1ad3ea3b2157d3cbe0abcdf7085aa4231b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2444213 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Ian Elliott bc4ec4c1 2020-10-05T17:10:18 Vulkan: Plumb OpenGL ES commands to AGI/debuggers Use vkCmd*DebugUtilsLabelEXT() calls to delimit all GLES commands, and group them under GLES draw/dispatch calls. Plumb calls from the EVENT macro (for every GL entrypoint) to vkCmd*DebugUtilsLabelEXT() via a newly-implemented DebugAnnotatorVk class. Use a new dirty bit so that cached entrypoints are associated with a triggering draw/dispatch command. The DebugAnnotatorVk::beginEvent() method saves a string in a vector of all GL commands in ContextVk. The dirty bit converts the strings into begin-end vkCmd*DebugUtilsLabelEXT() pairs. The DebugAnnotatorVk::endEvent() method makes the final vkCmdEndDebugUtilsLabelEXT() call for a draw/dispatch command. Enable the OGL->VK mapping feature by setting "angle_enable_trace = true" in GN args. Bug: b/162068318 Bug: b/169243237 Change-Id: I61b6a8d113168c0ce578d6efd002d8a393659aba Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2451517 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Ian Elliott <ianelliott@google.com>
Charlie Lao 77e3d0ae 2020-09-25T14:12:04 Vulkan: Defer the depthStencil buffer layout change to endRenderPass Depth stencil layout may change while we build the render pass, depending on the read/write access been made. Right now we are always inserting a layout change barrier at the start of render pass. Later on when the read/write property changes, we insert another layout change barrier. Similarly, we maintain the attachmentOps and RenderPassDesc::mPackedColorAttachmentRangeAndDSAccess as we changes read/write access. This makes code quite commplicated. This CL moves mReadOnlyDepthStencilMode from FramebufferVK to CommandBufferHelper object and we only maintain that boolean while we updating the read/write access. Then at the end of render pass or when depthStencil image is deleted, we update attachmentOps and mRenderPassDesc and layout transition all at once and only done once. This simplifies the read only depth stencil mode implementation a lot. Bug: b/168953278 Change-Id: Ie263b4526c82a9858e5d1f141ea58f499187a3ca Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2432075 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Charlie Lao 86ca5d2b 2020-10-01T11:56:05 Vulkan: Add plumbing to render pass when ImageHelper gets deleted ImageHelper object is not refcounted and garbage collected and endRenderPass call is deferred until next render pass starts. This caused a situation that an ImageHelper object gets deleted while still referenced in the open render pass. This CL make sure that we call into all shared context's open renderpass when an image goes away so that they can take appropriate action for this. Bug: b/169618408 Change-Id: I5075e805980084db82ca3e699462272eee5d2d59 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2443571 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Charlie Lao <cclao@google.com>
Shahbaz Youssefi 01641c7a 2020-09-30T15:25:28 Vulkan: Fix UtilsVk clear in non-zero subpass Mid-render-pass clears (through UtilsVk) run on the current subpass, which in the presence of multisampled-render-to-texture unresolve would be subpass 1. The graphics pipeline for that draw call should set the correct subpass index. Bug: angleproject:4836 Change-Id: Iba4a03ea96a63b0f5d09c27e5283ff8a8b534e05 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2441509 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 6534a6f8 2020-10-02T01:10:54 Revert "Vulkan:Add debug labels for OpenGL calls" This reverts commit 3705fc41315b50a64299ecb1d75015a33fefd6aa. Reason for revert: Causes errors on Debug bots. See bug. Bug: angleproject:5121 Original change's description: > Vulkan:Add debug labels for OpenGL calls > > Implement the DebugAnnotatorVk class, and plumb the EVENT macro in the > GL entrypoints to save a string of call info in the vector of all > GL calls in ContextVk. > > Then add a vkCmdBeginDebugUtilsLabelEXT() call that includes the > OpenGL draw/dispatch call prior to any Vulkan Draw or Dispatch calls. > Also embedded under that label add a second > vkCmdBeginDebugUtilsLabelEXT() call labeled "OpenGL Commands" that > includes all of the OpenGL calls leading up to the draw/dispatch. > Each individual OpenGL call is then given its own > vkCmdBegin/EndDebugUtilsLabelEXT() pair so that the complete sequence > of GL calls leading up to a draw call is visible for each Draw. > > Enable the OGL->VK mapping feature by setting > "angle_enable_trace = true" in GN args. > > Note: This will create an ANGLE APK on Android that generally won't > work with games, unless launched by AGI (which provides the debug > utils extension). A future version will disable these labels unless > the debug utils extension is found. > > Bug: b/162068318 > Bug: b/169243237 > Change-Id: I09886f17fa9287528c12552698738ea1fe2a4b8c > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2427557 > Commit-Queue: Ian Elliott <ianelliott@google.com> > Reviewed-by: Tim Van Patten <timvp@google.com> > Reviewed-by: Jamie Madill <jmadill@chromium.org> > Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> > Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> TBR=courtneygo@google.com,ianelliott@google.com,timvp@google.com,syoussefi@chromium.org,jmadill@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: b/162068318 Bug: b/169243237 Change-Id: I772d549213e1ad64ae58a1937e5de0f7ea740084 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2444094 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Commit-Queue: Yuly Novikov <ynovikov@chromium.org>
Brandon Schade 22d7db6e 2020-09-25T18:09:23 Vulkan: Change sampleCoverage calculation When emulating the sample coverage to a mask, truncate the value. Test: dEQP-GLES3.functional.multisample.fbo*sample_coverage* Bug: angleproject:4568 Change-Id: Ie0f9fee7dc7cf08e2289ff24f0fa69f9929c4ae3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2433069 Commit-Queue: Brandon Schade <b.schade@samsung.com> Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Ian Elliott 3705fc41 2020-09-28T21:40:57 Vulkan:Add debug labels for OpenGL calls Implement the DebugAnnotatorVk class, and plumb the EVENT macro in the GL entrypoints to save a string of call info in the vector of all GL calls in ContextVk. Then add a vkCmdBeginDebugUtilsLabelEXT() call that includes the OpenGL draw/dispatch call prior to any Vulkan Draw or Dispatch calls. Also embedded under that label add a second vkCmdBeginDebugUtilsLabelEXT() call labeled "OpenGL Commands" that includes all of the OpenGL calls leading up to the draw/dispatch. Each individual OpenGL call is then given its own vkCmdBegin/EndDebugUtilsLabelEXT() pair so that the complete sequence of GL calls leading up to a draw call is visible for each Draw. Enable the OGL->VK mapping feature by setting "angle_enable_trace = true" in GN args. Note: This will create an ANGLE APK on Android that generally won't work with games, unless launched by AGI (which provides the debug utils extension). A future version will disable these labels unless the debug utils extension is found. Bug: b/162068318 Bug: b/169243237 Change-Id: I09886f17fa9287528c12552698738ea1fe2a4b8c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2427557 Commit-Queue: Ian Elliott <ianelliott@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Courtney Goeltzenleuchter b156a753 2020-09-28T16:43:50 Move LayoutCaches to ShareGroup Testing with TSN found a race condition with RefCounted objects (DescriptorSetLayout and PipelineLayout). Rather than add more lock calls to protect accesses to mRefCount and mObject recommendation was to put these caches in the ShareGroup (basically part of the context). Locking at the GL level will ensure that two threads that share the same context will not access the ShareGroup at the same time. The ShareGroup also works because these layouts are not destroyed until the context is destroyed so don't have to worry about other threads (e.g. command processor thread) accessing them. Bug: b/168744561 Change-Id: Icc0aa07bf4787a69572d6ec62da2f21d286232c3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2437509 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Courtney Goeltzenleuchter 34f66126 2020-09-10T15:59:22 Vulkan: Check that its okay to add commands It can be hard to tell sometimes when the mRenderPassCommands or mOutSideRenderPassCommands command buffers have changed and there have been some issues with code that locally caches a pointer to a commandBuffer that then becomes invalid. This change adds checking so that if a command is being added to a commandBuffer that's been closed (e.g. submitted for processing) then we hit an assert. Bug: b/168144059 Change-Id: If5d37c462e3bcb51f6ec2ca44c27a2fad4e57c19 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2405812 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Charlie Lao d2d9e682 2020-09-15T16:06:42 Vulkan: Use DepthStencilReadOnly when it is read only. We are tracking depth and stencil read or write during the renderpass. We can use that to switch to DepthStencilReadOnly layout if both depth and stencil are not writing. This allows drivers to optimize out the storeOp for the renderpass. Bug: b/168953278 Change-Id: Id82e06b4bae1ae8c83d880bb5e58accfa61f8191 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2411336 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 32ed3e71 2020-09-21T23:42:34 Vulkan: Make UtilsVk not contribute to occlusion queries UtilsVk implements some functionality with draw calls. Of these draw calls, clear was accidentally contributing to occlusion query results. Additionally, the copyImage utility creates its own framebuffer and thus directly creates a render pass, bypassing ContextVk::startRenderPass. This change also fixes bugs where occlusion query handling assumed mRenderPassCommandBuffer to be valid and correspond to the started render pass command buffer, which is not true for the copyImage render pass. Bug: angleproject:5042 Change-Id: I19cdbb8fbeabf139791b314a6da265ac5cfd68e5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2423209 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Mohan Maiya a2d8bbb5 2020-09-18T18:18:34 Vulkan: Add GL_EXT_buffer_storage extension support Addition of support for immutable storage to buffer objects. Also adds new end2end tests for these usecases * Basic BufferStorage * SubData update * map/unmap buffer Bug: angleproject:5056 Tests: angle_end2end_tests --gtest_filter=BufferStorageTestES3*Vulkan Change-Id: Iba74b372ad033711927b63c6a04cec0eeb4db699 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2419952 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Mohan Maiya <m.maiya@samsung.com>
Courtney Goeltzenleuchter 54dbd5e7 2020-09-23T16:10:12 Vulkan: Add mutex around queueSubmit related data There are several queueSubmit related members of RendererVk that can be accessed from multiple threads. Adding mutex around thoses accesses resolves race condition flagged by TSAN. Add stress test for QueueSerial handling Add test to catch race issue in garbage collection found by TSAN. Test: angle_end2end_tests MultithreadingTest.MultiContextDrawWithSwapBuffers angle_end2end_tests MultithreadingTest.MultiContextCreateAndDeleteResources Bug: b/168744561 Change-Id: I238cce9052476710778a3b35f8531891d90ddd6e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2415402 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 68a5baeb 2020-09-23T22:13:03 Revert "Vulkan: Implement a SharedResourceUse pool" This reverts commit de335c16855f11d1f0a6f0b37bee30c8a09a6c1d. Reason for revert: Might actually regress CPU overhead perf. Unsure but it's possible the reported perf improvement was due to variance. Original change's description: > Vulkan: Implement a SharedResourceUse pool > > When adding a Resource to the ResourceUseList of ContextVk > we constructed a new SharedResourceUse object for tracking > and update of the Resource's Serial. We would then delete > it after releasing the resource. This incurs repeated > memory operation costs. > > Instead we now allocate a pool of SharedResourceUse objects > and acquire and release from this pool as needed. > > VTune profile of the Manhattan 30 offscreen benchmark > shows the CPU occupancy of bufferRead decrease from an > average of 0.9% -> 0.6% and imageRead decreases from > an average of 0.4% -> 0.3%. The bottleneck for both > these methods is the retain() method that leverages > the new SharedResourceUse pool. > > Bug: angleproject:4950 > Change-Id: Ib4f67c6f101d4b2de118014546e6cc14ad108703 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2396597 > Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> > Reviewed-by: Jamie Madill <jmadill@chromium.org> > Commit-Queue: Mohan Maiya <m.maiya@samsung.com> TBR=syoussefi@chromium.org,jmadill@chromium.org,m.maiya@samsung.com # Not skipping CQ checks because original CL landed > 1 day ago. Bug: angleproject:4950 Change-Id: I40081551c3db67d6e55182fea40119946ed16ac3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2426479 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Tim Van Patten ed899835 2020-09-11T21:21:55 Vulkan: Allocate descriptor pools with layouts Descriptor set layouts and pools are very tightly coupled, since their binding types and counts must match to ensure the number of available descriptor sets within a pool remains accurate. To enforce this, the descriptor pools will now keep a copy of the VkDescriptorSetLayout that the pool was created for, which is verified when a descriptor set is allocated from the pool. If the handles don't match, an ASSERT() will fire. Bug: angleproject:5033 Test: CQ Change-Id: I4faf82c24f31052e57b656c968788bb0c131b619 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2407282 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi f39e0f01 2020-09-07T23:07:37 Vulkan: Use subpass to unresolve render-to-texture attachments GL_EXT_multisampled_render_to_texture allows singlesampled textures to be used with multisampled framebuffers in such a way that the final resolve operation is automatically done. In Vulkan terminology, the render-to-texture GL attachment is used as a Vulkan subpass resolve attachment with an implicit (ideally-)lazy-memory multisampled image as the color attachment. This extension expects that if the texture is drawn to after the automatic resolve, the implicit multisampled image would take its fragment colors from the singlesampled image. In other words, the opposite of a resolve operation should be automatically performed at the start of the render pass. This change refers to this operation as "unresolve". The goal of this extension is to allow tiling GPUs to always keep multisampled data on tile memory and only ever load/store singlesampled data. The latter is achieved by using a subpass resolve attachment and setting storeOp of the multisampled color attachment to DONT_CARE. This change achieves the former by using an initial subpass that uses the resolve attachment as input attachment, draws into the multisampled color attachment and sets loadOp of said attachment to DONT_CARE. Bug: angleproject:4881 Change-Id: I99f410530365963567c77a7d62fc9db1500e5e3e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2397206 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 57193e8b 2020-09-13T23:58:42 Vulkan: Optimize Attribute Change Perf (5/5) This patch series optimizes programs that use the pattern: for (;;) { glVertexAttribPointer(...) glDraw(...) } Change 5: Re-check fragment opts on necessary changes. Check on a program binding change and on a sample coverage enabled chagne. This likely could be optimized further for program binding changes. In total the patch series reduces test iteration time by 25%. Test: DrawCallPerfBenchmark.Run/vulkan_attrib_change Bug: angleproject:5045 Bug: b/168493024 Change-Id: I5116f228d5eeef6606136a9a0cc81a4545df9d33 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2409177 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Mohan Maiya de335c16 2020-09-14T12:04:20 Vulkan: Implement a SharedResourceUse pool When adding a Resource to the ResourceUseList of ContextVk we constructed a new SharedResourceUse object for tracking and update of the Resource's Serial. We would then delete it after releasing the resource. This incurs repeated memory operation costs. Instead we now allocate a pool of SharedResourceUse objects and acquire and release from this pool as needed. VTune profile of the Manhattan 30 offscreen benchmark shows the CPU occupancy of bufferRead decrease from an average of 0.9% -> 0.6% and imageRead decreases from an average of 0.4% -> 0.3%. The bottleneck for both these methods is the retain() method that leverages the new SharedResourceUse pool. Bug: angleproject:4950 Change-Id: Ib4f67c6f101d4b2de118014546e6cc14ad108703 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2396597 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Mohan Maiya <m.maiya@samsung.com>
Jamie Madill d79db259 2020-09-13T23:15:40 Vulkan: Optimize Attribute Change Perf (3/5) This patch series optimizes programs that use the pattern: for (;;) { glVertexAttribPointer(...) glDraw(...) } Change 3: Use FastUnorderedSet for RP XFB buffers. Most of the time this list is empty. This speeds up the count check for empty and non-empty sets considerably. In total the patch series reduces test iteration time by 25%. Test: DrawCallPerfBenchmark.Run/vulkan_attrib_change Bug: angleproject:5045 Bug: b/168493024 Change-Id: Ia43eb11bdd62aa3bf425069b591f9318f519fd4c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2409175 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 357caadb 2020-09-16T21:44:56 Vulkan: Optimize Attribute Change Perf (2/5) This patch series optimizes programs that use the pattern: for (;;) { glVertexAttribPointer(...) glDraw(...) } Change 2: Micro-optimize XFB resume CPU overhead. We don't need to set resume on every new command buffer. We only need to set the dirty bit when we have an unexpected pause. In total the patch series reduces test iteration time by 25%. Test: DrawCallPerfBenchmark.Run/vulkan_attrib_change Bug: angleproject:5045 Bug: b/168493024 Change-Id: I8f6c68ff0513be4f405276e395d80bc1a185a061 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2409174 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: 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>
Charlie Lao 06b4fb92 2020-09-11T21:24:46 Vulkan: Fool proof packed attachment index to vulkan and from OpenGL ANGLE packs FBO attachments from OpenGL and uses packed attachments to create VkFramebuffer and renderpass. When we use attachment index into the attachment array, we must be very careful to use packed index for vulkan objects. It is easy to make mistakes here and introduce hard to debug bugs. This CL defines a PackedAttachmentIndex class that uses that to index into vulkan attachments and pass around APIs so that compiler would catch the error when wrong index is used. This also introduces PackedClearValuesArray that stores clear value in packed attachment index so that it is impossible to mix it with ClearValuesArray that stores clear value in GL attachment index. Bug: b/167301719 Change-Id: I68680522c60beeb5096e5211eaef89da28c7097e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2410366 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Charlie Lao c03c4490 2020-08-28T11:03:20 Vulkan: Defer glFlush issued in middle of renderpass to endRenderpass Manhattan is calling glFlush in middle of a renderpass. This CL defers the flush that issued in the middle of renderpass to the end of renderpass. Bug: b/166475273 Change-Id: I6baa3898d5efc456e2205c44e13c64f3d79d1464 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2381942 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>
Charlie Lao e4e2a016 2020-09-01T11:23:09 Vulkan: Use color mask to handle draw buffer disabled case When draw buffers set to GL_NONE, instead of remove the attachment from renderpass which breaks renderpass, we force vulkan's per buffer color mask to false while keep the disabled draw buffer attached. This CL also always create FrameBuffer with all color attachments regardless it is enabled or not. Bug: b/167301719 Change-Id: Ice9fca9aacf774a47d13b749f822b222cc050174 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2389007 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Charlie Lao <cclao@google.com>
Ian Elliott 4a41204d 2020-09-02T05:04:33 Vulkan: Improve invalidate for depth/stencil Improve state tracking when the depth and/or stencil attachments are invalidated. Since no draw-time tracking is done, we use the number of command-buffer commands to determine when an attachment is drawn to. That allows all cases to be handled for store ops. Still need to handle mContentDefined at endRP time (we have the data, just not the plumbing). Test: angle_white_box_tests --gtest_filter=VulkanPerformanceCounterTest.*Invalidate*/* Test: angle_deqp_gles3_tests --gtest_filter=dEQP.GLES3/functional_fbo_invalidate_* --use-angle=vulkan Bug: b/167276207 Change-Id: Iae10857dbb4d43b934c51ad7e400b71ae0db4f55 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2378670 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Ian Elliott <ianelliott@google.com>
Mohan Maiya f7b607c6 2020-08-29T16:34:30 Vulkan: Optimize atomic counter buffer offset calculations Call writeAtomicCounterBufferDriverUniformOffsets method only if there are valid atomic counter buffers Bug: angleproject:3566 Change-Id: I03c3fd2095e30381597bfc66acae0ccbf828d401 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2383787 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: back sept 10 - Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 167a0f22 2020-08-25T17:11:54 Vulkan: Fixes to transform feedback emulation path Two bugs were affecting the following scenario: - Create program with xfb and no default uniforms - Draw without xfb - Begin xfb - Draw with xfb - End xfb - map and validate The first bug was that the descriptor set that contains the xfb buffers and default uniforms was never allocated for the first draw call resulting in VVL errors and crashes. The second bug was that "Begin xfb" was not setting the xfb buffers dirty bit, which meant that after the first draw call bound dummy buffers to the descriptor set, they would not get updated. This was simultaneously fixed in 5d3a4ca4e, but the added test nevertheless tests this. Bug: angleproject:4660 Change-Id: I16eee315e39ec5a657481e54ecd0491581e886aa Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2376172 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Courtney Goeltzenleuchter a08495d9 2020-08-27T13:47:19 Revert "Vulkan: restore mContentDefined at endRP()" This reverts commit 53ee431e3d36692cbe50a96ebbf0ff898ec4d523. Bug: chromium:1122621 Change-Id: Ifd63aa0694e00ed6ef74b385466b874604355e79 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2380610 Reviewed-by: Ian Elliott <ianelliott@google.com> Commit-Queue: Ian Elliott <ianelliott@google.com>
Jamie Madill c5b9c49c 2020-08-25T18:01:29 Vulkan: Fix optimizeRenderPassForPresent regression. gfxbench clears the depth buffer right before the swap. Even though the last draw call that presents the frame didn't ever read or write to depth, the tracking we added thought this depth write meant we had to keep the LOAD_OP as CLEAR. Instead we can refine our check to treat clears specially when enabling the depth-stencil read-only mode instead of changing how the tracking works for clears. This way the tracking can not affect other apps that don't use depth-stencil read-only loops. Also adds a regression test that counts the clears after a swap. Bug: angleproject:4959 Bug: angleproject:4979 Change-Id: I12ece6474019f7519a467f827110ad817f7d4df7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2370364 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 5d3a4ca4 2020-08-21T13:09:56 Vulkan: Fix dirty state in XFB emulation on EndXFB. This regressed in "Vulkan: Preserve RPs on XFB changes when possible." The bug manifested as incorrect fire rendering in Manhattan. The fix is to ensure we dirty the correct buffer state when we call endXFB. Also adds a regression test. Bug: b/161744596 Bug: angleproject:4622 Change-Id: If16cc22b149526950f300e74c0cc82c0fefae5bc Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2368016 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Tobin Ehlis <tobine@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Tim Van Patten 85ddcc93 2020-08-24T14:22:22 Vulkan: Convert ProgramExecutableVk::mDescriptorSets to std::array Currently ProgramExecutableVk::mDescriptorSets is a vector, but it's size is bound so it can be a std::array (DescriptorSetLayoutArray). To ensure the size grows correctly in the future, the various descriptor set indexes are also being converted from independent constexpr uint32_ts into the enum DescriptorSetIndex. Bug: angleproject:4898 Test: CQ Change-Id: I7ae8ff3455bcfb61e24b73bd16cc3f8cf9873087 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2372664 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Tim Van Patten <timvp@google.com>
Ian Elliott 53ee431e 2020-08-21T12:40:03 Vulkan: restore mContentDefined at endRP() CommandBufferHelper will keep a pointer to the depth-stencil RenderTargetVk, and use this to set RenderTargetVk::mContentDefined to true at the end of a render pass. Test: angle_white_box_tests --gtest_filter=VulkanPerformanceCounterTest.InvalidatingAndUsingDepthDoesNotBreakRenderPass/* Test: angle_deqp_gles3_tests --gtest_filter=dEQP.GLES3/functional_fbo_invalidate_* --use-angle=vulkan Bug: b/163854287 Change-Id: I891381825ee01e141dfa4f9099d07d9ffc943f77 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2368194 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Ian Elliott <ianelliott@google.com>
Shahbaz Youssefi 295d2ccd 2020-08-24T14:46:31 Vulkan: Generate perf warnings on suboptimal paths Using KHR_debug features, this change creates a performance-warning-generation macro and employs it in a handful of locations to provide useful feedback to application developers. The warnings added in this change are not exhaustive. Bug: angleproject:3461 Bug: angleproject:4900 Change-Id: Id62435d170d90c5be9c1c5cab2d6779ccb58345e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2372628 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi c7645886 2020-08-24T16:47:21 Vulkan: Remove the flipViewportY feature This was featurized as a workaround for Intel/Windows drivers that did not render flipped in comparison to GL per Vulkan spec. This issue has long since been fixed, and due to missing parentheses the feature was accidentally unconditionally true for a long time already. Bug: angleproject:4896 Change-Id: I1db55d298bd47df649ff1f03d287d0cfa7453ea7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2372632 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 7bce5194 2020-08-22T23:52:51 Vulkan: Simplify image read barrier necessity check RAW hazards were being tested for if the layouts were identical, but that's impossible. Bug: angleproject:4911 Change-Id: I73f568b1df2cbffe943217e19b115561e48a56c9 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2370862 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Mohan Maiya f9a062c9 2020-08-16T14:09:41 Vulkan: Add FastIntegerSet and FastIntegerMap class Add FastIntegerSet container to enable fast contains operation for a set of integer keys. The class uses a BitSet vector to achieve performance. Add FastIntegerMap container to improve buffer serial tracking performance. FastIntegerMap uses FastIntegerSet container to track buffer serial keys. It also provides an ensureCapacity method to reserve space, for the expected buffer count, upfront. CommandBufferHelper::mUsedBuffers and ContextVk::descriptorSetCache are now FastIntegerMap CommandBufferHelper::mRenderPassUsedImages is now a FastIntegerSet Based on a CL by Jamie Bug: angleproject:4950 Test: angle_unittests.exe --gtest_filter=FastInteger* Change-Id: Ib58be20143f588baab99acadac796f2435f72d54 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2369466 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Mohan Maiya <m.maiya@samsung.com>
Charlie Lao 1d331c91 2020-08-19T10:30:14 Vulkan: Use context staging buffer for copyImageDataToBuffer ImageHelper::copyImageDataToBuffer() is called from glCopyTexture*. It allocate staging buffer and write the copy command into command buffer right away. This uses context staging buffer instead of ImageHelper's staging buffer. This has the benefit of able to share staging buffer with other objects (including buffers etc). Bug: b/164511310 Change-Id: I3f680b1cd95df172a442aac573a8cc8d48972b1b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2364717 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill ea3f501e 2020-08-18T11:13:33 Feedback Loop Redesign 3/3: Remove feedback loop tracking. We now we detect feedback loops by tracking the Framebuffers that the Texture is bound to. We still have the old tracking method that counts sampler and image bindings in the code as well. This CL removes the old front-end tracking for feedback loops. It's no longer used by any back-ends. This removal should reduce CPU overhead around Texture and Program binding changes. Reverts the image binding tracking to the simpler scheme that tracks if a Texture has ever been bound as an Image. This should practically have little or no perf effect and we can reinstate some simpler tracking in the future if required. Bug: angleproject:4500 Bug: angleproject:4959 Change-Id: Idc625d6e4c519919f97a4dc72dd9c35d262706fb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2363210 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tobin Ehlis <tobine@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Ian Elliott 296d3bfd 2020-08-21T10:38:32 Vulkan: do not end render pass when invalidating Initially, FramebufferVk::invalidateImpl() was very conservative and always ended a render pass (if the framebuffer is part of the current render pass). This adversely affects PUBG Mobile, which invalidates the depth buffer every frame, causing the render pass to be split. Test: PUBG MOBILE on Android Test: angle_white_box_tests --gtest_filter=VulkanPerformanceCounterTest.InvalidatingAndUsingDepthDoesNotBreakRenderPass/* Test: angle_deqp_gles3_tests --gtest_filter=dEQP.GLES3/functional_fbo_invalidate_* --use-angle=vulkan Bug: b/163854287 Change-Id: I343dee1db3ebaf039ff92557f9ef25b24bcdcc93 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2352627 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Ian Elliott <ianelliott@google.com>
Charlie Lao 4909d08c 2020-08-19T12:21:44 Vulkan: Rename mStagingBufferStorage to mStagingBuffer Rename only to keep it consistency. Bug: b/164511310 Change-Id: I7b00c48010b76bff0b292e6e75b9aff154cdecc0 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2364727 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill d3e800e9 2020-08-15T17:26:04 Vulkan: Restart RenderPasses in DS read-only mode. We can combine an initial RenderPass with a read-only RP if the first RP never writes to depth. We can check the depth-write tracking in CommandBufferHelper and substitute in a new Framebuffer/RP Desc in this case as well as issue new layout barriers. We need to disable barrier merging in this special case. This reduces the RenderPass count in the Manhattan trace from 15->13. The performance on the Pixel 4 benchmark goes to ~82% of native for the on-screen version and ~88% for off-screen. There's also a ~5% bump in speed for the desktop trace. Bug: angleproject:4959 Change-Id: I70241824f75eaa1e11b50370f826abc36e91686e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2358772 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill e689d316 2020-08-14T22:51:03 Vulkan: Don't end RenderPass on DS feedback loops. Instead of always switching the Framebuffer to mask out depth/stencil loops we now switch the RenderPass to a "read-only" depth/stencil mode. Reduces the RenderPass count in Manhattan from 18->15. There are still a couple extra RenderPasses inserted that we can get rid. We can merge a few RenderPasses by retroactively changing a started RenderPass to "read-only" when there are no prior recorded depth writes or clears. Also adds a test to count the number of RenderPasses ANGLE generates in DS feedback loop situaions. Bug: angleproject:4959 Change-Id: I1855a45959655fc27ccd47a3469c1c672fc8fd9e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2357973 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Tim Van Patten dff47d5f 2020-07-14T19:10:12 Vulkan: Optimize MSAA using subpass resolve attachments If a user is performing a blit to resolve a multisample color buffer into a single attachment, ANGLE can use subpass resolve attachments to resolve directly into the destination buffer as part of the render pass. This allows the data to remain in tiler memory and reduce the extra bandwidth required to write the multisampled data back to perform the copy. This work also requires restoring/reopening a render pass if it has been finished already, assuming the finished render pass was started and for the framebuffer that is the source for the blit command. Other objects that were created when the render pass was started need to be updated as well, such as the source FramebufferVk's resolve attachment, the CommandBufferHelper's vk::Framebuffer and vk::RenderPassDesc, etc. While this is better than performing vkCmdResolveImage(), there is still another major part of optimizing MSAA using resolve attachments not implemented here: discarding the multisampled image rather than writing it to GMEM, which requires the user to invalidate the read FBO after the blit. This CL was verified with AGI to make sure there are no explicit blits to resolve the multiple sampled image. Bug: b/159903491 Test: FramebufferTest_ES31.*Blit* Test: VulkanPerformanceCounterTest_ES31.MultisampleResolveWithBlit Change-Id: I320a26088d8f614a295e7feec275d71310391806 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2298663 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill 27a24f6d 2020-08-17T19:02:09 Vulkan: Check aggregate barrier feature instead of caching. Bug: b/157508684 Change-Id: I6c98cb017d44c1279439d815e77f6ad1a32133e2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2360903 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill d286daf1 2020-08-16T23:53:31 Vulkan: Track RP's read/write access for depth/stencil. This generalizes the read tracking into read/write. Knowing the write access can let us determine if we can switch a RenderPass to a read- only mode. And switching to read-only will let us combine some RenderPasses in Manhattan. Bug: angleproject:4959 Change-Id: Ic97547e84fef4a2670437677000d4525006ef69f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2358771 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Ian Elliott <ianelliott@google.com>
Jamie Madill 5b4f6e31 2020-08-15T23:03:34 Vulkan: Add overlay widget for RP buffer count. Can help evaluate when scenes stress out the resource tracking in the RenderPass command buffer. Bug: angleproject:4950 Bug: angleproject:4965 Change-Id: I7da2ad0101a840c5441f2112db4bb61f564afcef Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2358521 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill ee0a9a34 2020-08-13T15:39:49 Vulkan: Clean up ImageLayout enum. Remove the redundant manual numbering. Also rename AllGraphicsShadersReadWrite to AllGraphicsShadersWrite for consistency. Refactoring change only. Bug: angleproject:4959 Change-Id: I2e7ca00993f192897bbf88f4bdc3f1610bcb345f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2354279 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill 26ae42cf 2020-08-09T00:11:18 Vulkan: Clean up handleDirtyTexturesImpl. Adds a new bitset helper and changes a check to an ASSERT. Refactoring change only. Bug: angleproject:4959 Change-Id: I0de9f1b707c87cfb6fed8a110654783059e55c99 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2345025 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill dceaabb1 2020-08-08T17:03:47 Vulkan: Clean up ImageHelper barrier functions. We don't need to explicitly check if a barrier is required for write barriers. Write barriers always require a barrier and read barriers need the layout change check. We introduce a new enum encoding ReadOnly vs Write layout types and call specialized write/read functions instead. Also renames the helper APIs to be more consistent. Refactoring change only. Bug: angleproject:4959 Change-Id: I0ce39ceaca6be588327c381194a580dc6b11f036 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2344744 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Charlie Lao d63bca58 2020-08-06T17:37:18 Vulkan: Add test for RGB texture not break renderpass This adds a test that will use an empty RGB texture. It should not break renderpass. Bug: b/162603208 Change-Id: I90f5aa2fafdbd70c654c9ad0675d6def32b58313 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2338828 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 2e5647c8 2020-08-14T11:21:14 Vulkan: Switch buffer barrier check. ContextVk::onBufferRead can check for "usesBufferForWrite". Bug: angleproject:4911 Change-Id: Id295c068fd9a4027c3127830c682d62d614718dc Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2355102 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Jamie Madill eb85c021 2020-08-05T15:07:15 Vulkan: Refactor Context CommandBuffer query. getOutsideRenderPassCommandBuffer returns the command buffer directly since it's now stateless and cannot throw an error. All the RenderPass begin/end and flushing are done by the dependency functions (buffer/ image read/write). Bug: angleproject:4911 Change-Id: I5e7806be9d0e1b5e358524bd485298d660fac942 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2339544 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill d201ed8b 2020-08-02T16:29:35 Vulkan: Track used Images in RenderPass. Adding a simple ImageSerial tracking map in our RenderPass allows us to know when we do or do not need to close the RenderPass on a new Image access. This simple tracking scheme improves Manhattan performance by up to 25% on Android. The improved perf comes from reducing our RenderPass count (23->18 RenderPasses in our capture scene). Adds a FastUnorderedSet class to manage the used RP Image serials. Updates the Query helpers to explicitly flush the RP before inserting queries. Bug: angleproject:4911 Change-Id: I0c34fc8e307514ebdf3e81e08d8e5aedb70ebe8f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2334346 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 5921a040 2020-08-06T17:39:56 Vulkan: Refactor image dependency commands. Uses commands similar to the Buffer APIs. Also updates docs. Bug: angleproject:4911 Change-Id: I1f2ec9bdd1d725d4ec3d6601e63bcb0c045e2121 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2342287 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill 3d17084d 2020-08-06T17:53:36 Vulkan: Simpify check for started RenderPass + FB. We can check the RenderPass CB helper directly instead of caching a separate variable. Bug: angleproject:4911 Change-Id: Ic953c101c64fd5142e2dedfd06e3fea287331fd4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2341761 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 23daea2f 2020-08-06T10:40:00 Vulkan: Rename CommandBuffer get method. This method will no longer end the RenderPass in a following CL. Renaming it in a split CL makes the review diff simple. Bug: angleproject:4911 Change-Id: Id48257884dccb7c86f7de2cc9ca95e651fb68df7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2340788 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill e26c5e07 2020-08-02T16:04:17 Vulkan: Add Image serial. This is owned by the ImageHelper. We'll use it in tracking used images in RenderPasses. Bug: angleproject:4911 Change-Id: Ic71cef2b16eefb86c73c9c5e1e38ee7cd6a60adb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2332895 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill f9dd2c15 2020-08-02T18:03:16 Vulkan: Accumulate Buffer barriers. Uses an unordered_map in the CommandBufferHelper to track buffer reads and writes. Buffer barriers are tracked specially in the CommandBufferHelper class as a barrier we execute immediately when we execute the commands into the primary. So when we run into an incompatible buffer access we must start a new command buffer. The rules for an incompatible access are: - when we are reading a buffer, any prior write in the same command buffer is incompatible. - when we are writing a buffer, any prior read or write in the same command buffer is incopatible. Also adds a regression test using a new performance counter. Bug: angleproject:4429 Change-Id: I393a4ed87314f955eb998940b877ba76ea15a7b8 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2334091 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 18dd0c28 2020-08-02T17:50:48 Vulkan: Add command buffer performance counter. Adds a counter for the secondary command buffers (non-RenderPass). We'll use this in an upcoming test that validates that ANGLE only issues a single barrier (CB) in some buffer read/write scenarios. Also adds a PerfCounters struct. Bug: angleproject:4429 Change-Id: Iaf75ca89da3d02753897cb4066e2c56db497417e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2334090 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Tim Van Patten e8789a53 2020-07-29T12:01:49 Vulkan: Don't always end the render pass when updating the scissor We don't always need to end the render pass when updating the scissor, since it will be ended later when necessary. This change is in preparation for optimizing resolving multisample images with glBlit, since the render pass needs to be updated before it's ended. Bug: angleproject:4753 Test: CQ Change-Id: Ie657587ca9f4461dcc03f0f9c251ac2c17398f5b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2327334 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Jamie Madill 25b0de6b 2020-08-01T13:45:11 Vulkan: Squash Texture+ImageView Serial and improve caching. Previously we regenerated TextureSerial on Texture state changes such as base/max level changes. This caused ANGLE to update descriptor sets even though it was using the same image view handles. This change instead uses an ImageViewSubresourceSerial which includes both a serial for the ImageView and a 32-bit packed subresource range. The CL speeds up NBA2k because ANGLE no longer writes new descriptors for Texture max level changes. Local testing showed up to a 40% speedup. Also adds a regression test with a counter for the number of descriptor set writes in a frame. This change will also be useful in upcoming changes that track Image serials in the RenderPass. Bug: angleproject:4911 Change-Id: I66249634aa56288079acf2c0eb8aa3391103533c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2333396 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill b9b5fa55 2020-08-02T00:17:20 Vulkan: Redo Sampler Serials. Instead of refreshing sampler serials every time the SamplerVk or TextureVk has a state change we can give a VkSampler a unique serial. The serial is unique to this VkSampler and repeated state changes will fetch the same Serial from the SamplerCache. This allows for more cache hits. We store the the new Serial together with the VkSampler in a SamplerHelper class and store references to a SamplerHelper in SamplerVk and TextureVk instead of the VkSampler directly. In a follow-up change we will improve image view caching by also improving how we store serials for ImageViews. Bug: angleproject:4911 Change-Id: I9168c2700e383bca796cca925b38cfd30132d982 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2333988 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 11207393 2020-08-01T19:47:21 Vulkan: Add an overlay counter for descriptor writes. Note: currently the trace test shows about 1000 descriptor set writes per frame in NBA2k20. Bug: angleproject:4911 Change-Id: Id50d05fe405249c80a38dbbe3c96e7bd8c66cbc4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2333398 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 22e6fc03 2020-07-31T15:58:28 Vulkan: Move Resource Serial gen into Renderer. Putting Serial allocation in the Renderer allows the Helper classes to manage allocating its own Serial. The init functions for ImageViewHelper only have access to a vk::Context/RendererVk, not a ContextVk. This will be updated in a future CL. Re-uses the Serial Type X-Macro to do more code generation. Serial allocation now uses an atomic because of its now Renderer shared location. Bug: angleproject:4911 Change-Id: I2d5d3d0bbf613d5468de795a700f66164291bc79 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2332884 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Charlie Lao 4e1abc75 2020-07-24T10:38:40 Vulkan: Set depthStencil loadOp to DontCare when not used and stored If depth stencil buffer is disabled during entire renderpass, and at the end of renderpass we are not storing the data back to memory, then force the loadOP to DontCare to avoid unnecessary depth stencil load or clear. Bug: b/162080462 Change-Id: I30905a6d45bec038de68e7f363ec699eb2be09ab Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2317726 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 89ade79a 2020-07-31T16:04:09 Vulkan: Clean up ScopedDescriptorSetUpdates. Matches style guide requirement for types before members. Also moves the implementation of the class entirely into the cpp file. Moves the method implementation in ContextVk so we can more easily alter member variables. Unrelated cleanup done while working on consolidating RenderPasses. Bug: angleproject:4911 Change-Id: Ibe4273fc609b494840f1e86584bcee5bc31397d5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2331950 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Jamie Madill d04a5593 2020-07-31T10:55:38 Vulkan: Add RenderPass count test. Bug: angleproject:4911 Change-Id: I4c4e187b6b048118ce1231ae8924340ad382c43f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2331741 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Charlie Lao 38e7f7a6 2020-07-30T14:35:50 Vulkan: Make CommandBufferHelper::empty() logic less obtuse CommandBufferHelper::empty() logic is simplified and commented to make it easier to read. Basically empty() means there is no work in the CommandBufferHelper object. For RenderpassCommands, that is equivalent to !started(). Bug: b/162521656 Change-Id: Ic9683b392835a677501cb81e5e8a025e031ebf93 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2330379 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Commit-Queue: Charlie Lao <cclao@google.com>
Tobin Ehlis a9504808 2020-07-29T16:33:45 Vulkan: ATrace marker fix-ups Fix a couple of mislabled atrace markers and add explicit markers for inside/outside renderpass flushes. Bug: b/156403378 Change-Id: I4045846e54ff54bc8fc3dd6ef47339f6f5eb8e87 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2327828 Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tobin Ehlis <tobine@google.com> Commit-Queue: Tobin Ehlis <tobine@google.com>
Shahbaz Youssefi dcc56215 2020-07-19T01:12:09 Vulkan: Implement GL_EXT_multisampled_render_to_texture This change allows the use of resolve attachments in the Vulkan backend. GL_EXT_multisampled_render_to_texture is implemented using this feature. The infrastructure for specifying resolve attachments is designed with eventual support for GL_EXT_multisampled_render_to_texture2 in mind as well as optimizations to glBlitFramebuffer() and multisampled backbuffers. Proper support for glRenderbufferStorageMultisampledEXT is still missing from this change. All tests use this for the depth/stencil attachment and don't read back the data. Currently, the depth/stencil attachment is created as a normal multisampled image. Bug: angleproject:4836 Change-Id: I110a7f63312ae61a657b6094adf7d97c92bd5843 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2304170 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi 19aac801 2020-07-28T11:18:53 Vulkan: Remove unnecessary endRenderPass calls. In both onImageRead and onImageWrite, there's a call to endRenderPassAndGetCommandBuffer when necessary to record a command. There was an extraneous call to endRenderPass in both functions for no good reason. Bug: angleproject:4882 Change-Id: I8f3b18dbb3cb9a8b92dedcc603eaf44b1e485986 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2323642 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
kmx.io     mail     discord kmxgit v0.4.0