summaryrefslogtreecommitdiffstats
path: root/llvm/lib
Commit message (Collapse)AuthorAgeFilesLines
...
* Fix thinko.Richard Osborne2010-03-102-5/+7
| | | | llvm-svn: 98158
* Fold add(add(mul(x,y),a),b) -> lmul(x,y,a,b) if the intermediateRichard Osborne2010-03-101-0/+56
| | | | | | results are unused elsewhere. llvm-svn: 98157
* Remove duplicated code. No functionality change.Benjamin Kramer2010-03-101-9/+0
| | | | llvm-svn: 98156
* Fix a bug in DEBUG_VALUE handling Devang ran into.Dale Johannesen2010-03-101-6/+3
| | | | | | I'll get this loop right yet. llvm-svn: 98155
* Prefer LMUL to MACCU as LMUL has no tied operands.Richard Osborne2010-03-103-3/+13
| | | | llvm-svn: 98153
* Custom lower (S|U)MUL_LOHI -> MACC(S|U)Richard Osborne2010-03-102-0/+40
| | | | llvm-svn: 98152
* Fix indentationRichard Osborne2010-03-101-1/+1
| | | | llvm-svn: 98151
* Lower add (mul a, b), c into MACCU / MACCS nodes which translateRichard Osborne2010-03-103-0/+86
| | | | | | | | directly to the maccu / maccs instructions. We handle this in ExpandADDSUB since after type legalisation it is messy to recognise these operations. llvm-svn: 98150
* move three lowering hooks from MAI to TLOF and make one of themChris Lattner2010-03-105-20/+18
| | | | | | | semantic instead of syntactic. This completes MCization of darwin/x86[-64]! llvm-svn: 98145
* This survived a bootstrap, so let's try 98104 again.Dale Johannesen2010-03-101-4/+69
| | | | llvm-svn: 98137
* Enable machine cse pass.Evan Cheng2010-03-101-5/+1
| | | | llvm-svn: 98132
* mcize the rest of EH emission, only one more directive missingChris Lattner2010-03-101-12/+17
| | | | | | for darwin/x86 to be completely mcized. llvm-svn: 98130
* add missing filename!Chris Lattner2010-03-101-0/+2
| | | | llvm-svn: 98125
* set the temporary bit on MCSymbols correctly.Chris Lattner2010-03-1016-53/+83
| | | | llvm-svn: 98124
* Fix a comment.Dan Gohman2010-03-101-1/+1
| | | | llvm-svn: 98122
* Add a couple more heuristics to neuter machine cse some more.Evan Cheng2010-03-101-13/+57
| | | | | | | | | | | 1. Be careful with cse "cheap" expressions. e.g. constant materialization. Only cse them when the common expression is local or in a direct predecessor. We don't want cse of cheap instruction causing other expressions to be spilled. 2. Watch out for the case where the expression doesn't itself uses a virtual register. e.g. lea of frame object. If the common expression itself is used by copies (common for passing addresses to function calls), don't perform the cse. Since these expressions do not use a register, it creates a live range but doesn't close any, we want to be very careful with increasing register pressure. Note these are heuristics so machine cse doesn't make register allocator unhappy. Once we have proper live range splitting and re-materialization support in place, these should be evaluated again. Now machine cse is almost always a win on llvm nightly tests on x86 and x86_64. llvm-svn: 98121
* MC/Mach-O: Resolve a FIXME; these relocation types are no longer semanticaly ↵Daniel Dunbar2010-03-101-3/+5
| | | | | | different. llvm-svn: 98120
* eliminate MCContext::CreateSymbol and CreateTemporarySymbol.Chris Lattner2010-03-102-17/+12
| | | | | | | Add a new GetOrCreateTemporarySymbol method and a version that takes a twine. llvm-svn: 98118
* inline away a form of IsPCRelative, eliminating the Chris Lattner2010-03-104-29/+14
| | | | | | dead IsPCRel argument. llvm-svn: 98117
* add some fixme's for MCizing. EH still has a few things thatChris Lattner2010-03-103-13/+9
| | | | | | | | need to be MCized, but the last debug info thing are LEB and cygwin specific (which the MC api doesn't support yet) and one specific form of EmitReference which I'll tackle next. llvm-svn: 98116
* MC/Mach-O: Use the SECTDIFF relocation type for (A - B + constant) where A ↵Daniel Dunbar2010-03-101-8/+11
| | | | | | | | is external. - I'm not sure why, but this is what 'as' does. llvm-svn: 98115
* Clear up the last (famous last words) frame index value reuse issues for Thumb1.Jim Grosbach2010-03-103-8/+12
| | | | llvm-svn: 98109
* Speculatively revert 98104; could be what's causing crashesDale Johannesen2010-03-101-56/+3
| | | | llvm-svn: 98108
* mcize uses of PrintRelDirective and eliminate it.Chris Lattner2010-03-103-21/+9
| | | | llvm-svn: 98107
* inline the bool form of PrintRelDirective away, leaving just the unsigned form.Chris Lattner2010-03-092-9/+4
| | | | llvm-svn: 98106
* eliminate EOL, adding all comments with the OutStreamer.AddComment Chris Lattner2010-03-094-131/+143
| | | | | | | | | | | method. With this, comments should end up on the same lines as the .byte directives (for example) and we now get no output with: $ llc CodeGen/X86/2009-02-12-DebugInfoVLA.ll -o - -filetype=null -asm-verbose woot. llvm-svn: 98105
* Ever more complicated DEBUG_VALUE fixes for branch folding.Dale Johannesen2010-03-091-3/+56
| | | | llvm-svn: 98104
* Avoid analyzing instructions in blocks not reachable from the entry block.Dan Gohman2010-03-091-2/+9
| | | | | | They are lots of trouble, and they don't matter. This fixes PR6559. llvm-svn: 98103
* eliminate a bunch of \n's that are being printed to O. Next up is to killChris Lattner2010-03-092-35/+19
| | | | | | off "EOL". llvm-svn: 98102
* convert the non-"ispcrel" case of EmitReference to MC,Chris Lattner2010-03-091-0/+8
| | | | | | | significant debug info testcases are now all going through MCStreamer, though they print a lot of extraneous newlines to "O". llvm-svn: 98101
* make the NullStreamer set the section on a label when emitted so that ↵Chris Lattner2010-03-091-1/+5
| | | | | | isDefined() works. llvm-svn: 98100
* Try to keep the cached inliner costs around for a bit longer for big functions.Jakob Stoklund Olesen2010-03-094-6/+54
| | | | | | | | | | | | | | | The Caller cost info would be reset everytime a callee was inlined. If the caller has lots of calls and there is some mutual recursion going on, the caller cost info could be calculated many times. This patch reduces inliner runtime from 240s to 0.5s for a function with 20000 small function calls. This is a more conservative version of r98089 that doesn't break the clang test CodeGenCXX/temp-order.cpp. That test relies on rather extreme inlining for constant folding. llvm-svn: 98099
* MC/X86: Rename alternate spellings of ADD{8,16,32} and mark as "code gen ↵Daniel Dunbar2010-03-091-9/+11
| | | | | | only" so they don't get selected by the asm matcher. llvm-svn: 98098
* MC/X86: Rename alternate spellings of CMP{8,16,32} and mark as "code gen ↵Daniel Dunbar2010-03-091-6/+12
| | | | | | only" so they don't get selected by the asm matcher. llvm-svn: 98097
* Revert r98089, it was breaking a clang test.Jakob Stoklund Olesen2010-03-094-48/+6
| | | | llvm-svn: 98094
* Try to keep the cached inliner costs around for a bit longer for big functions.Jakob Stoklund Olesen2010-03-094-6/+48
| | | | | | | | | | | The Caller cost info would be reset everytime a callee was inlined. If the caller has lots of calls and there is some mutual recursion going on, the caller cost info could be calculated many times. This patch reduces inliner runtime from 240s to 0.5s for a function with 20000 small function calls. llvm-svn: 98089
* Permit inlining into huge functions. This heuristic is ancient, and inliningJakob Stoklund Olesen2010-03-091-5/+0
| | | | | | can sometimes help reduce function size. llvm-svn: 98088
* Change the Value argument to eliminateFrameIndex to a type-tagged value. ThisJim Grosbach2010-03-0929-32/+38
| | | | | | | | | | is preparatory to having PEI's scavenged frame index value reuse logic properly distinguish types of frame values (e.g., whether the value is stack-pointer relative or frame-pointer relative). No functionality change. llvm-svn: 98086
* MSR (Move to Special Register from ARM core register) requires a mask to specifyJohnny Chen2010-03-092-12/+12
| | | | | | what fields of the CPSR or SPSR are affected. llvm-svn: 98085
* MC/Mach-O: For PCrel relocations, we need to compensate for the PCrel ↵Daniel Dunbar2010-03-091-5/+9
| | | | | | adjustment when determining if we need a scattered relocation. llvm-svn: 98082
* MC/Mach-O: Also set the PCrel bit in the second half of paired relocation ↵Daniel Dunbar2010-03-091-4/+2
| | | | | | entries. llvm-svn: 98081
* MC/Mach-O: Don't generate relocations for PCrel fixups to local labels.Daniel Dunbar2010-03-091-9/+19
| | | | llvm-svn: 98080
* The address of an indirect call must be in R12 on Darwin.Dale Johannesen2010-03-091-0/+10
| | | | | | | | Make it so. (This patch is in LowerCall_Darwin, which seems to be used by SVR4 code as well; since that doesn't belong here, I haven't worried about this case.) llvm-svn: 98077
* scavenged frame index value re-use gets confused when more than one baseJim Grosbach2010-03-092-1/+8
| | | | | | | | register is involved for thumb1. Work around this for the moment by only re-using SP-relative offsets. This is temporary 'til the code can distinguish multiple base registers. llvm-svn: 98071
* The ARM EH experiment worked!Bill Wendling2010-03-094-50/+78
| | | | | | | | | | | | Place the LSDA into the TEXT section for ARM platforms. This involves making the encoding indirect, pcrel, and sdata4 instead of an absolute pointer. The references to the type infos are then non-lazy pointers. Revision 98019 changed the encoding of non-lazy pointers to add the symbol to the non-lazy pointer definition if it's a local symbol (otherwise, it's external and set to '0' so that the loader can adjust it to the real value). This paved the way for this change to work on ARM. llvm-svn: 98068
* In cases where the carry / borrow unused converted ladd / lsubRichard Osborne2010-03-091-0/+30
| | | | | | to an add or a sub. llvm-svn: 98059
* Canonicalize ladd constant to RHS.Richard Osborne2010-03-091-0/+4
| | | | llvm-svn: 98058
* Add DAG combine for ladd / lsub.Richard Osborne2010-03-092-0/+70
| | | | llvm-svn: 98057
* Allow more cross-rc coalescing.Evan Cheng2010-03-091-10/+13
| | | | llvm-svn: 98048
* reapply r98035:Chris Lattner2010-03-094-24/+24
| | | | | | | Now that setStartLabel takes an MCSymbol, we can de-ID'ize beginScope and RecordSourceLine. llvm-svn: 98047
OpenPOWER on IntegriCloud