Promotion from AMD internal branch for 2023.Q4.3 #2891
Conversation
* Update the extension VK_KHR_shader_quad_scope to VK_KHR_shader_quad_control * Add a cross-module inlining helper that does not require the caller to create a call instruction first * Add an llpc_version interface "library" * Add an end-to-end continufy mode raytracing lit test * Add OpenGL deprecated built-in variables * Add LoweringPointerTupleMap and a step towards multi-llvm::Value translations of SPIRVValues * [LLPC/builder] Add support for packed accumulator matrices * [Continuations][NFC] Add space before filename * [Continuations] Add argument to continuation.continue * [Continuations] Add builder to legacy cleanup pass impl class * [Continuations] Mark traversal with shaderstage * [Continuations] Generalize pointer address space rewrite helper * [Continuations] Use llvm::forEachCall in DXILContPostProcess. * [Continuations] Use default alignment for CpsStackLowering. * [Continuations] Merge PreCoroutineLowering into LowerRaytracingPipeline * [Continuations] Always inline GetContinuationStackGlobalMemBase * [Continuations] Remove now useless assert * [Continuations] Rename MDLgcCpsModule variable. * [Continuations] Remove unused constants * [Continuations] Remove unused parts of SaveContinuationState pass * [Continuations] Remove CONTINUATION_STATE global * [Continuations] Refactor LegacyCleanupContinuations * [Continuations] Ensure stable address of continuation state * [Continuations] Use impl class for LegacyCleanupContinuations * [Continuations] Improve GpuRt handling, and entry point detection * [Continuations] RegBuffer: Improve handling of globals without registers * [Continuations] Refactor pass. * [Continuations] Ignore non-rt shader stages * [Continuations] Fix detecting new _cont_Traversal name * [Continuations] Fix register count determination * [Continuations] Fix use-after-free on function * llpc/continuations: don't run SpirvLowerRaytracing in continuations mode * llpc/continuations: adjust the linkage of _cont_* functions * llpc/continuations: Link GPURT into all shader modules * Use cross-module inlining in continuations * Change bitwidth of bool type to 32 * Extract vertex id from vertex info bitmask * VK.pipeline.*.extended_dynamic_state.*.color_blend_dual_index_* * Preserve functions in GpuRt module * Set individual resource binding for separate shader object * Set gl_BaseVertex to zero if the draw command has no baseVertex parameter * compute_shader_derivatives support * Move SpirvLowerRaytracing towards the end of the pass pipeline * llpc: move AmdExt* functions from LowerGraphLibrary to the general lowering * llpcfe dxil: support processWaveMatch * llpcfe dxil: exclude helperlane for fragment shader * llpcfe: Depend on iscpc target and fix standalone build * lgc: support SubgroupClusteredMultiExclusive * lgc: fix build with old llvm * Fix a threading issue in createDirectory * Fix and use lit flags for lgc * Fix the amdllpc check error on PipelineVsFs_EnableColorExport.pipe * Fix the regression issue caused by the fix on dynamicDualSourceBlend * Fix the spell check via tool 'typos-1.16.23' * Fix the register 'psLoadProvokingVtx'
|
15df8aa Jenkins build error. |
Test summary for commit 6e5cb84CTS tests (Failed: 0/138381)
Ubuntu navi3x, SrdcvkUbuntu navi2x, Srdcvk |
No description provided.