summaryrefslogtreecommitdiffstats
path: root/llvm
Commit message (Collapse)AuthorAgeFilesLines
* Add functions to enable adding a single attribute to a function andEric Christopher2008-05-164-0/+27
| | | | | | its associated call site. llvm-svn: 51204
* Record weak external linkage in a case where we wereDale Johannesen2008-05-161-0/+12
| | | | | | | missing it. gcc.dg/darwin-weakimport-2.c. Handle common and weak differently for darwin ppc32. llvm-svn: 51201
* API change for {BinaryOperator|CmpInst|CastInst}::create*() --> Create. ↵Gabor Greif2008-05-1631-524/+617
| | | | | | Legacy interfaces will be in place for some time. (Merge from use-diet branch.) llvm-svn: 51200
* If the result of a BIT_CONVERT is a v1* vector, it doesn't mean its source ↵Evan Cheng2008-05-161-3/+6
| | | | | | is a v1* vector. llvm-svn: 51192
* Fix test.Evan Cheng2008-05-161-1/+1
| | | | llvm-svn: 51191
* Bill pointed out that system headers should beDuncan Sands2008-05-161-1/+1
| | | | | | included after local headers. llvm-svn: 51187
* Silence the compiler warning differently. TheDuncan Sands2008-05-161-1/+1
| | | | | | original method caused gcc-4.2 to complain. llvm-svn: 51186
* Re-enable tail duplication pass (now with default threshold down to 1 ↵Evan Cheng2008-05-161-0/+1
| | | | | | instruction). llvm-svn: 51184
* Do not dup malloc, vector instructions, etc. Throttle the default theshold ↵Evan Cheng2008-05-161-5/+15
| | | | | | way down. llvm-svn: 51183
* Remove ADCE's ability to delete loops. This ability is now implemented in aOwen Anderson2008-05-161-4/+21
| | | | | | safer manner by loop deletion. llvm-svn: 51182
* Move this test from ADCE to loop deletion, where it is more appropriate.Owen Anderson2008-05-161-1/+1
| | | | llvm-svn: 51181
* Use loop deletion instead of ADCE in these tests.Owen Anderson2008-05-162-2/+2
| | | | llvm-svn: 51180
* Clean ups for loop deletion based on Chris' feedback.Owen Anderson2008-05-161-28/+32
| | | | | | | Also, use SCEV to determine the trip count of the loop, which is more powerful and accurate that Loop::getTripCount. llvm-svn: 51179
* Use loop deletion instead of ADCE for removing loops.Owen Anderson2008-05-164-4/+4
| | | | llvm-svn: 51178
* Fix this test. It was testing broken behavior in that it required ADCE to ↵Owen Anderson2008-05-161-3/+2
| | | | | | | | | eliminate a potentially infinite loop, which is undesirable. Instead, test the LICM behavior that we're really interested in. llvm-svn: 51177
* implement PR2328.Chris Lattner2008-05-162-2/+16
| | | | llvm-svn: 51176
* Use common where we mean common, not weak.Dale Johannesen2008-05-161-1/+1
| | | | llvm-svn: 51173
* Treat common as distinct from weak global on Darwin x86.Dale Johannesen2008-05-161-0/+10
| | | | llvm-svn: 51172
* Revert the change from r51157 inDan Gohman2008-05-162-4/+4
| | | | | | | | test/Verifier/2002-11-05-GetelementptrPointers.ll, which was incorrect. Instead, fix getIndexedType to not follow pointer types, as PointerType is a subclass of CompositeType. llvm-svn: 51171
* Update comments based on Duncan's feedback.Dan Gohman2008-05-151-23/+23
| | | | llvm-svn: 51170
* Fix a bug in LoopStrengthReduce that caused it to emit IR withDan Gohman2008-05-152-3/+42
| | | | | | | | | | | | | | | | | use-before-def. The problem comes up in code with multiple PHIs where one PHI is being rewritten in terms of the other, but the other needs to be casted first. LLVM rules requre the cast instruction to be inserted after any PHI instructions, but when instructions were inserted to replace the second PHI value with a function of the first, they were ended up going before the cast instruction. Avoid this problem by remembering the location of the cast instruction, when one is needed, and inserting the expansion of the new value after it. This fixes a bug that surfaced in 255.vortex on x86-64 when instcombine was removed from the middle of the loop optimization passes. llvm-svn: 51169
* Remove the S92 code, which really has nothing to doDale Johannesen2008-05-151-7/+0
| | | | | | with what the test is testing; makes it pass again on ppc32. llvm-svn: 51167
* Evan has implemented this on ppc, so run the test there.Dale Johannesen2008-05-151-2/+2
| | | | llvm-svn: 51166
* Fix typos and comments.Evan Cheng2008-05-154-6/+6
| | | | llvm-svn: 51165
* Allow an extra bit for CommonLinkage.Dale Johannesen2008-05-151-1/+1
| | | | | | | | This changes the .bc file format, but if I understand how it works correctly, old .bc files continue to be readable. llvm-svn: 51161
* Actually scalarize the operand to BIT_CONVERT instead of asking someone to doNate Begeman2008-05-151-1/+2
| | | | | | something with a v1 type. llvm-svn: 51160
* Reverting r51100...I need to use VC++ 2005.Steve Naroff2008-05-151-11/+172
| | | | llvm-svn: 51159
* IR support for extractvalue and insertvalue instructions. Also, beginDan Gohman2008-05-1515-360/+946
| | | | | | moving toward making structs and arrays first-class types. llvm-svn: 51157
* Use static and anonymous namespaces consistently with other passes.Dan Gohman2008-05-151-8/+8
| | | | llvm-svn: 51156
* Revert 51067 check-in for now.Devang Patel2008-05-1531-229/+220
| | | | llvm-svn: 51155
* Remove useless check.Devang Patel2008-05-151-6/+2
| | | | | | Patch by Matthijs Kooijman. llvm-svn: 51154
* Disable JIT symbol table for now.Evan Cheng2008-05-151-1/+1
| | | | llvm-svn: 51152
* Use of UINT_MAX requires climits, at least whenDuncan Sands2008-05-151-0/+1
| | | | | | compiling with gcc 4.3. llvm-svn: 51145
* Fix a bunch of 80col violations that arose from the Create API change. Tweak ↵Gabor Greif2008-05-1523-78/+134
| | | | | | makefile targets to find these better. llvm-svn: 51143
* Make use of vector load and store operations to implement memcpy, memmove, ↵Evan Cheng2008-05-155-106/+204
| | | | | | and memset. Currently only X86 target is taking advantage of these. llvm-svn: 51140
* Remove tail duplication pass.Evan Cheng2008-05-151-1/+0
| | | | llvm-svn: 51139
* Move the operator new and operator delete out of line. This fixes an issue withNate Begeman2008-05-152-17/+21
| | | | | | | operator new() referring to the static initTags function, which has to be in the same linkage unit as any file including User.h. llvm-svn: 51136
* Don't assume underlying APInt type is limitedDale Johannesen2008-05-141-1/+1
| | | | | | to 64 bits. llvm-svn: 51135
* Situations can arise when you have a function called that returns a 'void', butBill Wendling2008-05-143-4/+30
| | | | | | | | | | is bitcast to return a floating point value. The result of the instruction may not be used by the program afterwards, and LLVM will happily remove all instructions except the call. But, on some platforms, if a value is returned as a floating point, it may need to be removed from the stack (like x87). Thus, we can't get rid of the bitcast even if there isn't a use of the value. llvm-svn: 51134
* Remove undefined behavior in hex string->APFloatDale Johannesen2008-05-141-1/+1
| | | | | | conversion. Try 0x1.0000a4p+0f. Neil, please review. llvm-svn: 51132
* Use a better idiom to silence compiler warnings.Evan Cheng2008-05-141-20/+5
| | | | llvm-svn: 51131
* rename SimplifyCFG.cpp -> SimplifyCFGPass.cppChris Lattner2008-05-141-1/+1
| | | | llvm-svn: 51130
* Silence warnings.Evan Cheng2008-05-141-2/+2
| | | | llvm-svn: 51129
* Add support to IR builder for new vicmp, vfcmp routinesNate Begeman2008-05-141-4/+19
| | | | llvm-svn: 51127
* Really silence compiler warnings.Evan Cheng2008-05-141-2/+2
| | | | llvm-svn: 51126
* Don't generate unused variables in a no-assert buildNate Begeman2008-05-141-16/+28
| | | | | | Add some checks to the new vicmp,vfcmp instructions llvm-svn: 51125
* Really silence compiler warnings.Evan Cheng2008-05-141-5/+20
| | | | llvm-svn: 51123
* CommonLinkage (missed a file)Dale Johannesen2008-05-141-1/+2
| | | | llvm-svn: 51120
* Generated files for CommonLinkage.Dale Johannesen2008-05-143-1740/+1752
| | | | llvm-svn: 51119
* Add CommonLinkage; currently tentative definitionsDale Johannesen2008-05-1422-31/+70
| | | | | | | | | | are represented as "weak", but there are subtle differences in some cases on Darwin, so we need both. The intent is that "common" will behave identically to "weak" unless somebody changes their target to do something else. No functional change as yet. llvm-svn: 51118
OpenPOWER on IntegriCloud