- 54f7c59 Better diagnostics when inline asm fails to allocate. by Jakob Stoklund Olesen · 14 years ago
- 6cc775f - Rename TargetInstrDesc, TargetOperandInfo to MCInstrDesc and MCOperandInfo and by Evan Cheng · 14 years ago
- a1dceb0 Print registers by name instead of by number. by Jakob Stoklund Olesen · 14 years ago
- d88b8ba After register is spilled there should not be any DBG_VALUE referring the same register. by Devang Patel · 14 years ago
- 0ab7767 There could be more than one DBG_VALUE instructions for variables where all of them have offset based on one register. by Devang Patel · 14 years ago
- fb03a92 Be less aggressive about hinting in RAFast. by Jakob Stoklund Olesen · 14 years ago
- 4b0bb83 Avoid calling TRI->getAllocatableSet in RAFast. by Jakob Stoklund Olesen · 14 years ago
- 50663b7 Use RegisterClassInfo::getOrder in RAFast. by Jakob Stoklund Olesen · 14 years ago
- 8ea3af4 Typo by Evan Cheng · 15 years ago
- c37aa0b Fix a bug where we were counting the alias sets as completely used by Eric Christopher · 15 years ago
- 28f4c72 Temporarily revert r129408 to see if it brings the bots back. by Eric Christopher · 15 years ago
- d829f43 Fix a bug where we were counting the alias sets as completely used by Eric Christopher · 15 years ago
- de9d585 Add more comments... err debug statements to the fast allocator. by Eric Christopher · 15 years ago
- c378336 Fix typo. by Eric Christopher · 15 years ago
- d650b30 Mark that the return is using EAX so that we don't use it for some other by Nick Lewycky · 15 years ago
- 2fb5b31 Simplify a bunch of isVirtualRegister() and isPhysicalRegister() logic. by Jakob Stoklund Olesen · 15 years ago
- d82ac37 Remove MachineRegisterInfo::getLastVirtReg(), it was giving wrong results by Jakob Stoklund Olesen · 15 years ago
- 1331a15 Replace TargetRegisterInfo::printReg with a PrintReg class that also works without a TRI instance. by Jakob Stoklund Olesen · 15 years ago
- 66a8bf5 Fix comment. by Eric Christopher · 15 years ago
- 6c18d1a Get rid of static constructors for pass registration. Instead, every pass exposes an initializeMyPassFunction(), which by Owen Anderson · 15 years ago
- 6095d81 Add DEBUG message. by Devang Patel · 15 years ago
- 005155e previous patch was a little too tricky for its own good. Don't try to by Jim Grosbach · 15 years ago
- df6b67b The register allocator shouldn't consider allocating reserved registers. by Jim Grosbach · 15 years ago
- cb2e56f tidy up a few 80-column and trailing whitespace bits. by Jim Grosbach · 15 years ago
- 2c325dc Ignore unallocatable registers in RegAllocFast. by Jakob Stoklund Olesen · 15 years ago
- ac305d2 Delete dead comment. by Eli Friedman · 15 years ago
- a7aed18 Reapply r110396, with fixes to appease the Linux buildbot gods. by Owen Anderson · 15 years ago
- 8a18aee While emitting DBG_VALUE for registers spilled at the end of a block do not use location of MBB->end(). If a block does not have terminator then incoming iterator points to end(). by Devang Patel · 15 years ago
- bda59bd Revert r110396 to fix buildbots. by Owen Anderson · 15 years ago
- 755aceb Don't use PassInfo* as a type identifier for passes. Instead, use the address of the static by Owen Anderson · 15 years ago
- d71bc1a While spilling live registers at the end of block check whether they are used by DBG_VALUE machine instructions or not. If a spilled register is used by DBG_VALUE machine instruction then insert a new DBG_VALUE machine instruction to encode variable's new location on stack. by Devang Patel · 15 years ago
- 36cf119 Fix a bug in the -regalloc=fast handling of exotic two-address instruction with by Jakob Stoklund Olesen · 15 years ago
- d61b735 Fix memory leak reported by valgrind. by Devang Patel · 15 years ago
- 37c42a3 Remove many calls to TII::isMoveInstr. Targets should be producing COPY anyway. by Jakob Stoklund Olesen · 15 years ago
- 57e7237 Update DBG_VALUE to refer appropriate stack slot in case of a spill. by Devang Patel · 15 years ago
- 4c82a9e Detect and handle COPY in many places. by Jakob Stoklund Olesen · 15 years ago
- dadea5b Fix the handling of partial redefines in the fast register allocator. by Jakob Stoklund Olesen · 15 years ago
- 0d94d7a Add more special treatment for inline asm in RegAllocFast. by Jakob Stoklund Olesen · 15 years ago
- 246e9a0 Avoid processing early clobbers twice in RegAllocFast. by Jakob Stoklund Olesen · 15 years ago
- 864827a Keep track of the call instructions whose clobber lists were skipped during fast by Jakob Stoklund Olesen · 15 years ago
- e0eddb2 Add support for partial redefs to the fast register allocator. by Jakob Stoklund Olesen · 15 years ago
- 663543b4 Properly handle multiple definitions of a virtual register in the same by Jakob Stoklund Olesen · 15 years ago
- 5857927 Pull the UsedInInstr.test() calls into calcSpillCost() and remember aliases. by Jakob Stoklund Olesen · 15 years ago
- 70563bb Remove debug option. Add comment on spill order determinism. by Jakob Stoklund Olesen · 15 years ago
- 176a9c4 Avoid allocating the same physreg to multiple virtregs in one instruction. by Jakob Stoklund Olesen · 15 years ago
- f5e8c86 Minor optimizations. DenseMap::begin() is surprisingly slow on an empty map. by Jakob Stoklund Olesen · 15 years ago
- 6649cda Extract spill cost calculation to a new method, and use definePhysReg() to clear by Jakob Stoklund Olesen · 15 years ago
- 7d22a81b Only use clairvoyance when defining a register, and then only if it has one use. by Jakob Stoklund Olesen · 15 years ago
- f915d14 Eliminate a hash table probe when killing virtual registers. by Jakob Stoklund Olesen · 15 years ago
- edd3d9d Execute virtreg kills immediately instead of after processing all uses. by Jakob Stoklund Olesen · 15 years ago
- e07a408 Sprinkle superregister <imp-def> and <imp-kill> operands when dealing with subregister indices. by Jakob Stoklund Olesen · 15 years ago
- 1069a09 Now that we don't keep live registers across calls, there is not reason to go by Jakob Stoklund Olesen · 15 years ago
- 397068d Boldly attempt consistent capitalization. Functional changes unintended. by Jakob Stoklund Olesen · 15 years ago
- 8044c98 Spill and kill all virtual registers across a call. by Jakob Stoklund Olesen · 15 years ago
- d2ef1fb Reduce hashtable probes by using DenseMap::insert() for lookup. by Jakob Stoklund Olesen · 15 years ago
- fb43e06 Make MBB a class member instead of passing it around everywhere. by Jakob Stoklund Olesen · 15 years ago
- 75142e6 Fix an GCC warning that seems to have actually caught a bug (!!!) in by Chandler Carruth · 15 years ago
- 84ce290 Calculate liveness on the fly for local registers. by Jakob Stoklund Olesen · 15 years ago
- 089e942 Don't bother spilling before a return by Jakob Stoklund Olesen · 15 years ago
- b160139 Track allocatable instead of reserved regs, and never take an unallocatable hint. by Jakob Stoklund Olesen · 15 years ago
- e68b814 Avoid scanning the long tail of physreg operands on calls by Jakob Stoklund Olesen · 15 years ago
- 6c038e3 Count coalesced copies by Jakob Stoklund Olesen · 15 years ago
- 33af4fc Allow virtreg redefines when verifying for RegAllocFast by Jakob Stoklund Olesen · 15 years ago
- 670492c When verifying two-address instructions, check the following: by Jakob Stoklund Olesen · 15 years ago
- 4d5c106 Simplify the handling of physreg defs and uses in RegAllocFast. by Jakob Stoklund Olesen · 15 years ago
- ceb5a7a Enable opportunistic coalescing by Jakob Stoklund Olesen · 15 years ago
- 68c235b Trust kill flags from isel and later passes. by Jakob Stoklund Olesen · 15 years ago
- 41f8dc8 Fix an embarrassing runtime regression for RegAllocFast. by Jakob Stoklund Olesen · 15 years ago
- d74a564 Clean up RegAllocFast debug output by Jakob Stoklund Olesen · 15 years ago
- 0ba2e2a Take allocation hints from copy instructions to/from physregs. by Jakob Stoklund Olesen · 15 years ago
- 680b749 More asserts around physreg uses by Jakob Stoklund Olesen · 15 years ago
- 955a0e7 Make sure to add kill flags to the last use of a virtreg when it is redefined. by Jakob Stoklund Olesen · 15 years ago
- f98a355 Avoid scoping issues, fix buildbots by Jakob Stoklund Olesen · 15 years ago
- 69b8f42 Make Clang happy. by Daniel Dunbar · 15 years ago
- 11f1ba1 Store the Dirty bit in the LiveReg structure instead of a bit vector. by Jakob Stoklund Olesen · 15 years ago
- 1326681 Keep track of the last place a live virtreg was used. by Jakob Stoklund Olesen · 15 years ago
- 3f0241e Simplify the tracking of used physregs to a bulk bitor followed by a transitive by Jakob Stoklund Olesen · 15 years ago
- f1b3029 Mostly rewrite RegAllocFast. by Jakob Stoklund Olesen · 15 years ago
- 1ef0660 It's not safe eliminate copies where src and dst have different sub-register indices. by Evan Cheng · 15 years ago
- efb126a Add argument TargetRegisterInfo to loadRegFromStackSlot and storeRegToStackSlot. by Evan Cheng · 16 years ago
- 14b1d75 Run LiveVariables instead of computing liveness locally in -regalloc=fast. by Jakob Stoklund Olesen · 16 years ago
- 8a070a5 Add fast register allocator, enabled with -regalloc=fast. by Jakob Stoklund Olesen · 16 years ago