Commits

Show all
Author Commit Message Labels Comments Date
Bob Wilson
Reapply SSA updater fixes to Pertwee -- the performance regression that was previously exposed by the SSA updater changes is no longer a problem. --- Reverse-merging r109874 into '.': A test/Transforms/GVN/2010-03-31-RedundantPHIs.ll U include/llvm/Transforms/Utils/SSAUpdater.h A include/llvm/Transforms/Utils/SSAUpdaterImpl.h U include/llvm/CodeGen/MachineSSAUpdater.h U lib/CodeGen/MachineSSAUpdater.cpp U lib/Transforms/Utils/SSAUpdater.cpp
Tags
llvmCore-2366.1
Branches
Apple/Pertwee
Bob Wilson
--- Merging r109879 into '.': U lib/Target/ARM/ARMAsmPrinter.cpp
Tags
llvmCore-2365
Branches
Apple/Pertwee
Bob Wilson
Revert the new SSA updater from Pertwee: we still don't have the live range splitting that ought to let us get good performance and good SSA updating at the same time. --- Reverse-merging r103407 into '.': U lib/CodeGen/MachineSSAUpdater.cpp --- Reverse-merging r103184 into '.': U include/llvm/Transforms/Utils/SSAUpdaterImpl.h --- Reverse-merging r103060 into '.': U include/llvm/Transforms/Utils/SSAUpdater.h D include/llvm/Transforms/Utils/SSAUpdaterImpl.h U include/llvm/CodeGen/MachineSSAUpdater.h G lib/CodeGen/MachineSSAUpdater.cpp U lib/T…
Branches
Apple/Pertwee
Jim Grosbach
$ svn merge -c 109842 https://grosbach@llvm.org/svn/llvm-project/llvm/trunk --- Merging r109842 into '.': U test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll A test/CodeGen/Thumb2/thumb2-badreg-operands.ll U utils/TableGen/EDEmitter.cpp U lib/Target/ARM/ARMInstrThumb2.td U lib/Target/ARM/ARMBaseInstrInfo.cpp U lib/Target/ARM/ARMRegisterInfo.td U lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp U lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h U lib/Target/ARM/Thumb2InstrInfo.cpp
Branches
Apple/Pertwee
Bob Wilson
Creating Pertwee from ToT.
Branches
Apple/Pertwee
Bob Wilson
Fix the temporary hack.
Daniel Dunbar
Temporary hack to ensure stale Analysis dir is cleaned up on llvm-i686-linux.
Benjamin Kramer
Remove XFAIL, test doesn't leak anymore.
Bob Wilson
Refactor ARM-specific DAG combining in preparation for adding some more transformations.
Dale Johannesen
Implement vector constants which are splat of integers with mov + vdup. 8003375. This is currently disabled by default because LICM will not hoist a VDUP, so it pessimizes the code if the construct occurs inside a loop (8248029).
Bob Wilson
Don't assert on an unrecognized BrMiscFrm instruction. PR7745.
John McCall
Transcribe IRC to svn. Also don't print basic block names twice if they match.
Eric Christopher
Grammar.
John McCall
Make the header self-contained and follow #include guidelines.
Eli Friedman
PR7750: !CExpr->isNullValue() only properly computes whether CExpr is nonnull if CExpr is a ConstantInt.
Owen Anderson
Make un-named values legible in certain vim configurations.
Nate Begeman
Add intrinsics __builtin_arm_qadd & __builtin_arm_qsub to allow access to the QADD & QSUB instructions. Behave identically to __qadd & __qsub RealView instruction intrinsics.
John McCall
Switch to using the LLVM CommandLine library so that our help message is properly contaminated with nonsense about timing passes that doesn't apply at all to this utility. :)
Benjamin Kramer
Plug the remaining MC leaks by giving MCObjectStreamer/MCAsmStreamer ownership of the TargetAsmBackend and the MCCodeEmitter.
Dale Johannesen
Comment typo.
Jakob Stoklund Olesen
Revert r109652, and remove the offending assert in loadRegFromStackSlot instead. We do sometimes load from a too small stack slot when dealing with x86 arguments (varargs and smaller-than-32-bit args). It looks like we know what we are doing in those cases, so I am going to remove the assert instead of artifically enlarging stack slot sizes. The assert in storeRegToStackSlot stays in. We don't want to write beyond the bounds of a stack slot.
Daniel Dunbar
CrashRecovery: Use ThreadLocal::erase() instead of set(0).
Rafael Avila de Espindola
Make the test while reducing blocks functional. This avoids accessing freed memory when one of the original BB is destroyed.
Benjamin Kramer
Stop leaking std::strings in GetDwarfFile.
Benjamin Kramer
COFFObjectWriter: Don't leak COFFSymbols and COFFSections.
John McCall
Centralize the logic to permanently unify two instructions and make sure it establishes a context and does a complaining diff. Also make sure we unify the prelude and postlude of a diff after a block-diff call.
John McCall
Diagnose non-structural differences in the case where blocks were structurally identical.
John McCall
When unifying instructions during a block diff, actually complain about any differences we see. This should only happen if there are "non-structural" differences between the instructions, i.e. differences which wouldn't cause diff to return true.
John McCall
Somehow I was getting reasonable results for the test cases I was interested in despite not ever incrementing any path costs, so that the only nonzero costs arose from the all-left path in the first column. Anyway. Perform the diff starting from the beginning of the block to avoid capturing (say) loads of allocas. Vastly improves diff results on code that hasn't been mem2reg'ed.
John McCall
Cache the result of errs() and implement formatted logging.
  1. Prev
  2. Next