Connor Abbott 8df9403bb9 AMDGPU: Fix handling of infinite loops in fragment shaders ... Summary: Due to the fact that kill is just a normal intrinsic, even though it's supposed to terminate the thread, we can end up with provably infinite loops that are actually supposed to end successfully. The AMDGPUUnifyDivergentExitNodes pass breaks up these loops, but because there's no obvious place to make the loop branch to, it just makes it return immediately, which skips the exports that are supposed to happen at the end and hangs the GPU if all the threads end up being killed. While it would be nice if the fact that kill terminates the thread were modeled in the IR, I think that the structurizer as-is would make a mess if we did that when the kill is inside control flow. For now, we just add a null export at the end to make sure that it always exports something, which fixes the immediate problem without penalizing the more common case. This means that we sometimes do two "done" exports when only some of the threads enter the discard loop, but from tests the hardware seems ok with that. This fixes dEQP-VK.graphicsfuzz.while-inside-switch with radv. Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D70781		2020-01-29 17:13:25 +01:00
..
AsmParser	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
Disassembler	CMake: Make most target symbols hidden by default	2020-01-14 19:46:52 -08:00
MCTargetDesc	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
TargetInfo	CMake: Make most target symbols hidden by default	2020-01-14 19:46:52 -08:00
Utils	AMDGPU/R600: Emit rodata in text segment	2020-01-22 14:31:51 -05:00
AMDGPU.h	[AMDGPU] Bundle loads before post-RA scheduler	2020-01-24 11:33:38 -08:00
AMDGPU.td
AMDGPUAliasAnalysis.cpp
AMDGPUAliasAnalysis.h
AMDGPUAlwaysInlinePass.cpp	AMDGPU: Simplify getAddressSpace calls	2019-10-31 07:51:38 -07:00
AMDGPUAnnotateKernelFeatures.cpp
AMDGPUAnnotateUniformValues.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
AMDGPUArgumentUsageInfo.cpp
AMDGPUArgumentUsageInfo.h
AMDGPUAsmPrinter.cpp	CMake: Make most target symbols hidden by default	2020-01-14 19:46:52 -08:00
AMDGPUAsmPrinter.h	[AMDGPU] separate accounting for agprs	2019-10-02 00:26:58 +00:00
AMDGPUAtomicOptimizer.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
AMDGPUCallingConv.td	AMDGPU: Allow i16 shader arguments	2020-01-27 06:55:32 -08:00
AMDGPUCallLowering.cpp	AMDGPU/GlobalISel: Fix argument lowering for vectors of pointers	2020-01-09 16:29:44 -05:00
AMDGPUCallLowering.h
AMDGPUCodeGenPrepare.cpp	AMDGPU: Implement FDIV optimizations in AMDGPUCodeGenPrepare	2020-01-23 16:57:43 -08:00
AMDGPUCombine.td	AMDGPU/GlobalISel: Add pre-legalize combiner pass	2020-01-22 10:16:39 -05:00
AMDGPUFeatures.td
AMDGPUFixFunctionBitcasts.cpp
AMDGPUFrameLowering.cpp	Use Align for TFL::TransientStackAlignment	2019-10-21 08:31:25 +00:00
AMDGPUFrameLowering.h	[Alignment][NFC] Deprecate Align::None()	2020-01-24 12:53:58 +01:00
AMDGPUGenRegisterBankInfo.def	AMDGPU/GlobalISel: Fix RegBankSelect for G_INSERT_VECTOR_ELT	2020-01-22 10:57:50 -05:00
AMDGPUGISel.td	AMDGPU/GlobalISel: Select buffer atomics	2020-01-27 15:16:44 -05:00
AMDGPUGlobalISelUtils.cpp	AMDGPU/GlobalISel: Add new utils file	2020-01-03 15:25:50 -05:00
AMDGPUGlobalISelUtils.h	AMDGPU/GlobalISel: Add new utils file	2020-01-03 15:25:50 -05:00
AMDGPUHSAMetadataStreamer.cpp	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
AMDGPUHSAMetadataStreamer.h
AMDGPUInline.cpp	[NFC] Refactor InlineResult for readability	2020-01-15 13:34:20 -08:00
AMDGPUInstrInfo.cpp
AMDGPUInstrInfo.h
AMDGPUInstrInfo.td	AMDGPU: Remove custom node for exports	2020-01-15 18:33:15 -05:00
AMDGPUInstructions.td	AMDGPU: Don't check constant address space for atomic stores	2020-01-24 12:15:09 -08:00
AMDGPUInstructionSelector.cpp	AMDGPU/GlobalISel: Look through copies for source modifiers	2020-01-29 08:08:13 -08:00
AMDGPUInstructionSelector.h	AMDGPU/GlobalISel: Manually select scalar f64 G_FNEG	2020-01-29 06:49:16 -08:00
AMDGPUISelDAGToDAG.cpp	AMDGPU/GlobalISel: Eliminate SelectVOP3Mods_f32	2020-01-27 17:53:54 -05:00
AMDGPUISelLowering.cpp	[AMDGPU] Allow narrowing muti-dword loads	2020-01-24 11:03:41 -08:00
AMDGPUISelLowering.h	AMDGPU: Remove custom node for exports	2020-01-15 18:33:15 -05:00
AMDGPULegalizerInfo.cpp	AMDGPU/GlobalISel: Select buffer atomics	2020-01-27 15:16:44 -05:00
AMDGPULegalizerInfo.h	AMDGPU/GlobalISel: Select buffer atomics	2020-01-27 15:16:44 -05:00
AMDGPULibCalls.cpp	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
AMDGPULibFunc.cpp	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
AMDGPULibFunc.h	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
AMDGPULowerIntrinsics.cpp
AMDGPULowerKernelArguments.cpp	[Alignement][NFC] Deprecate untyped CreateAlignedLoad	2020-01-23 13:34:32 +01:00
AMDGPULowerKernelAttributes.cpp
AMDGPUMachineCFGStructurizer.cpp	[AMDGPU] Fixes -Wrange-loop-analysis warnings	2019-12-22 19:39:28 +01:00
AMDGPUMachineFunction.cpp	AMDGPU: Refactor treatment of denormal mode	2019-11-19 19:55:43 +05:30
AMDGPUMachineFunction.h	AMDGPU: Refactor treatment of denormal mode	2019-11-19 19:55:43 +05:30
AMDGPUMachineModuleInfo.cpp
AMDGPUMachineModuleInfo.h
AMDGPUMacroFusion.cpp
AMDGPUMacroFusion.h
AMDGPUMCInstLower.cpp	[MC] Add parameter Address to MCInstPrinter::printInst	2020-01-06 20:42:22 -08:00
AMDGPUOpenCLEnqueuedBlockLowering.cpp
AMDGPUPerfHintAnalysis.cpp
AMDGPUPerfHintAnalysis.h
AMDGPUPreLegalizerCombiner.cpp	AMDGPU/GlobalISel: Add pre-legalize combiner pass	2020-01-22 10:16:39 -05:00
AMDGPUPrintfRuntimeBinding.cpp	[AMDGPU] add support for hostcall buffer pointer as hidden kernel argument	2019-11-20 15:53:55 +05:30
AMDGPUPromoteAlloca.cpp	[Alignement][NFC] Deprecate untyped CreateAlignedLoad	2020-01-23 13:34:32 +01:00
AMDGPUPropagateAttributes.cpp	Make llvm::StringRef to std::string conversions explicit.	2020-01-28 23:25:25 +01:00
AMDGPUPTNote.h
AMDGPURegisterBankInfo.cpp	AMDGPU/GlobalISel: Select buffer atomics	2020-01-27 15:16:44 -05:00
AMDGPURegisterBankInfo.h	AMDPGPU/GlobalISel: Select more MUBUF global addressing modes	2020-01-27 07:28:36 -08:00
AMDGPURegisterBanks.td	AMDGPU/GlobalISel: Replace handling of boolean values	2020-01-06 18:26:42 -05:00
AMDGPURegisterInfo.cpp	AMDGPU: Remove outdated comment	2020-01-16 14:54:27 -05:00
AMDGPURegisterInfo.h	AMDGPU/GlobalISel: Handle more G_INSERT cases	2019-10-07 19:16:26 +00:00
AMDGPURegisterInfo.td
AMDGPURewriteOutArguments.cpp	[Alignment][NFC] Use Align with CreateAlignedStore	2020-01-23 17:34:32 +01:00
AMDGPUSearchableTables.td	[AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores	2019-11-20 22:59:30 +01:00
AMDGPUSubtarget.cpp	[AMDGPU] Bundle loads before post-RA scheduler	2020-01-24 11:33:38 -08:00
AMDGPUSubtarget.h	[AMDGPU] Fix vccz after v_readlane/v_readfirstlane to vcc_lo/hi	2020-01-28 10:52:17 +00:00
AMDGPUTargetMachine.cpp	[AMDGPU] Bundle loads before post-RA scheduler	2020-01-24 11:33:38 -08:00
AMDGPUTargetMachine.h
AMDGPUTargetObjectFile.cpp
AMDGPUTargetObjectFile.h
AMDGPUTargetTransformInfo.cpp	Resubmit: [DA][TTI][AMDGPU] Add option to select GPUDA with TTI	2020-01-24 10:39:40 -08:00
AMDGPUTargetTransformInfo.h	Resubmit: [DA][TTI][AMDGPU] Add option to select GPUDA with TTI	2020-01-24 10:39:40 -08:00
AMDGPUUnifyDivergentExitNodes.cpp	AMDGPU: Fix handling of infinite loops in fragment shaders	2020-01-29 17:13:25 +01:00
AMDGPUUnifyMetadata.cpp	[AMDGPU] Fixes -Wrange-loop-analysis warnings	2019-12-22 19:39:28 +01:00
AMDILCFGStructurizer.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
AMDKernelCodeT.h
BUFInstructions.td	AMDGPU/GlobalISel: Select buffer atomics	2020-01-27 15:16:44 -05:00
CaymanInstructions.td
CMakeLists.txt	[AMDGPU] Bundle loads before post-RA scheduler	2020-01-24 11:33:38 -08:00
DSInstructions.td	TableGen/GlobalISel: Add way for SDNodeXForm to work on timm	2020-01-09 17:37:52 -05:00
EvergreenInstructions.td
FLATInstructions.td	AMDGPU/GlobalISel: Fix not using global atomics on gfx9+	2020-01-27 07:42:42 -08:00
GCNDPPCombine.cpp	[AMDGPU][DPP] Corrected DPP combiner	2019-11-20 15:56:45 +03:00
GCNHazardRecognizer.cpp	Make more use of MachineInstr::mayLoadOrStore.	2019-12-19 11:51:52 +00:00
GCNHazardRecognizer.h
GCNILPSched.cpp	Prune a LegacyDivergenceAnalysis and MachineLoopInfo include each	2019-10-19 01:31:09 +00:00
GCNIterativeScheduler.cpp
GCNIterativeScheduler.h
GCNMinRegStrategy.cpp
GCNNSAReassign.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
GCNProcessors.td
GCNRegBankReassign.cpp	AMDGPU: Fix ubsan error	2020-01-23 15:05:47 -05:00
GCNRegPressure.cpp	[AMDGPU] Fix GCN regpressure trackers for INLINEASM instructions.	2020-01-27 17:25:25 +03:00
GCNRegPressure.h
GCNSchedStrategy.cpp	[AMDGPU] Attempt to reschedule withou clustering	2020-01-27 10:27:16 -08:00
GCNSchedStrategy.h	[AMDGPU] Attempt to reschedule withou clustering	2020-01-27 10:27:16 -08:00
LLVMBuild.txt
MIMGInstructions.td	[AMDGPU] deduplicate tablegen predicates	2019-11-04 12:19:17 -08:00
R600.td
R600AsmPrinter.cpp	[NFC] Fix trivial typos in comments	2020-01-06 10:50:26 +00:00
R600AsmPrinter.h
R600ClauseMergePass.cpp
R600ControlFlowFinalizer.cpp
R600Defines.h
R600EmitClauseMarkers.cpp
R600ExpandSpecialInstrs.cpp
R600FrameLowering.cpp
R600FrameLowering.h	[Alignment][NFC] Deprecate Align::None()	2020-01-24 12:53:58 +01:00
R600InstrFormats.td
R600InstrInfo.cpp	Update spelling of {analyze,insert,remove}Branch in strings and comments	2020-01-21 10:15:38 -06:00
R600InstrInfo.h	Use MCRegister in copyPhysReg	2019-11-11 14:42:33 +05:30
R600Instructions.td	AMDGPU: Eliminate more legacy codepred address space PatFrags	2020-01-09 10:29:32 -05:00
R600ISelLowering.cpp	[DAG] Add helper for creating constant vector index with correct type. NFC.	2020-01-18 01:23:36 -05:00
R600ISelLowering.h
R600MachineFunctionInfo.cpp
R600MachineFunctionInfo.h
R600MachineScheduler.cpp
R600MachineScheduler.h
R600OpenCLImageTypeLoweringPass.cpp
R600OptimizeVectorRegisters.cpp
R600Packetizer.cpp
R600Processors.td
R600RegisterInfo.cpp
R600RegisterInfo.h	[TargetRegisterInfo] Default trackLivenessAfterRegAlloc() to true	2020-01-19 14:20:37 -08:00
R600RegisterInfo.td
R600Schedule.td
R700Instructions.td
SIAddIMGInit.cpp
SIAnnotateControlFlow.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SIDefines.h
SIFixSGPRCopies.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SIFixupVectorISel.cpp
SIFixVGPRCopies.cpp
SIFoldOperands.cpp	[amdgpu] Remove unused header. NFC.	2020-01-08 11:32:09 -05:00
SIFormMemoryClauses.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SIFrameLowering.cpp	[AMDGPU] Don't create MachinePointerInfos with an UndefValue pointer	2019-12-23 15:58:19 +00:00
SIFrameLowering.h	[Alignment][NFC] Deprecate Align::None()	2020-01-24 12:53:58 +01:00
SIInsertSkips.cpp	Resubmit: [AMDGPU] Invert the handling of skip insertion.	2020-01-22 13:18:32 +09:00
SIInsertWaitcnts.cpp	[AMDGPU] Fix vccz after v_readlane/v_readfirstlane to vcc_lo/hi	2020-01-28 10:52:17 +00:00
SIInstrFormats.td
SIInstrInfo.cpp	[AMDGPU] override isHighLatencyDef	2020-01-29 08:01:29 -08:00
SIInstrInfo.h	[AMDGPU] override isHighLatencyDef	2020-01-29 08:01:29 -08:00
SIInstrInfo.td	AMDGPU/GlobalISel: Eliminate SelectVOP3Mods_f32	2020-01-27 17:53:54 -05:00
SIInstructions.td	AMDGPU/GlobalISel: Eliminate SelectVOP3Mods_f32	2020-01-27 17:53:54 -05:00
SIISelLowering.cpp	AMDGPU: Fix not using f16 fsin/fcos	2020-01-27 08:59:59 -08:00
SIISelLowering.h	CodeGen: Use LLT instead of EVT in getRegisterByName	2020-01-09 17:37:52 -05:00
SILoadStoreOptimizer.cpp	AMDGPU/SILoadStoreOptimizer: Fix uninitialized variable error	2020-01-24 21:53:05 -08:00
SILowerControlFlow.cpp	Resubmit: [AMDGPU] Invert the handling of skip insertion.	2020-01-22 13:18:32 +09:00
SILowerI1Copies.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SILowerSGPRSpills.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SIMachineFunctionInfo.cpp	AMDGPU: Refactor treatment of denormal mode	2019-11-19 19:55:43 +05:30
SIMachineFunctionInfo.h	AMDGPU: Refactor treatment of denormal mode	2019-11-19 19:55:43 +05:30
SIMachineScheduler.cpp	[AMDGPU] override isHighLatencyDef	2020-01-29 08:01:29 -08:00
SIMachineScheduler.h	AMDGPU/SI: make ~SIScheduleBlockCreator trivial	2019-11-11 21:51:59 -08:00
SIMemoryLegalizer.cpp	[AMDGPU] Bundle loads before post-RA scheduler	2020-01-24 11:33:38 -08:00
SIModeRegister.cpp
SIOptimizeExecMasking.cpp	AMDGPU: Use Register	2019-12-27 16:53:21 -05:00
SIOptimizeExecMaskingPreRA.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SIPeepholeSDWA.cpp	AMDGPU: Fixed indeterminate map iteration in SIPeepholeSDWA	2019-12-02 12:08:49 +00:00
SIPostRABundler.cpp	[AMDGPU] Bundle loads before post-RA scheduler	2020-01-24 11:33:38 -08:00
SIPreAllocateWWMRegs.cpp	Sink all InitializePasses.h includes	2019-11-13 16:34:37 -08:00
SIProgramInfo.h	[AMDGPU] separate accounting for agprs	2019-10-02 00:26:58 +00:00
SIRegisterInfo.cpp	[TargetRegisterInfo] Default trackLivenessAfterRegAlloc() to true	2020-01-19 14:20:37 -08:00
SIRegisterInfo.h	[TargetRegisterInfo] Default trackLivenessAfterRegAlloc() to true	2020-01-19 14:20:37 -08:00
SIRegisterInfo.td	AMDGPU: Make VReg_1 only include 1 artificial register	2019-10-28 20:51:51 -07:00
SIRemoveShortExecBranches.cpp	[AMDGPU] SIRemoveShortExecBranches should not remove branches exiting loops	2020-01-22 13:18:40 +09:00
SISchedule.td
SIShrinkInstructions.cpp	AMDGPU: Don't fold S_NOPs with implicit operands	2019-10-30 14:40:56 -07:00
SIWholeQuadMode.cpp	[AMDGPU] Remove unnecessary v_mov from a register to itself in WQM lowering.	2020-01-10 23:01:19 -05:00
SMInstructions.td	[AMDGPU] deduplicate tablegen predicates	2019-11-04 12:19:17 -08:00
SOPInstructions.td	Revert "AMDGPU: Temporary drop s_mul_hi_i/u32 patterns"	2020-01-27 08:07:21 -08:00
VIInstrFormats.td
VIInstructions.td
VOP1Instructions.td	AMDGPU/GlobalISel: Select llvm.amdgcn.mov.dpp	2020-01-22 11:43:53 -05:00
VOP2Instructions.td	AMDGPU/GlobalISel: Fix import of zext of s16 op patterns	2020-01-09 10:29:32 -05:00
VOP3Instructions.td	AMDGPU/GlobalISel: Select V_ADD3_U32/V_XOR3_B32	2020-01-23 12:04:20 -05:00
VOP3PInstructions.td	[AMDGPU] deduplicate tablegen predicates	2019-11-04 12:19:17 -08:00
VOPCInstructions.td	AMDGPU: Remove VOP3Mods0Clamp0OMod	2020-01-07 15:10:08 -05:00
VOPInstructions.td	[AMDGPU] copy OtherPredicates from pseudo to VOP3_Real	2019-09-26 21:06:17 +00:00