index
:
~vlj/llvm
cfg
clause
codesize
codesize2
codesize3
codesize4
codesize5
codesize6
constbuf
constbuf2
indirect-wip
master
native
radeonsi
radeonsi-backup
radeonsi-backup2
radeonsi-scheduling
scheduling
scheduling-backup
scheduling2
textures
vliw5
AMDGPU backend development for LLVM
UNKNOWN
summary
refs
log
tree
commit
diff
log msg
author
committer
range
path:
root
/
lib
/
CodeGen
/
MachineLICM.cpp
Age
Commit message (
Expand
)
Author
Files
Lines
2009-07-25
More migration to raw_ostream, the water has dried up around the iostream hole.
Daniel Dunbar
1
-6
/
+7
2009-02-27
MachineLICM CSE should match destination register classes; avoid hoisting imp...
Evan Cheng
1
-3
/
+13
2009-02-05
Machine LICM increases register pressure and it almost always increase code s...
Evan Cheng
1
-0
/
+4
2009-02-05
Teach machine licm to CSE hoisted instructions.
Evan Cheng
1
-23
/
+88
2009-02-04
Don't call isInvariantLoad twice.
Evan Cheng
1
-9
/
+2
2009-02-04
Don't bother hoisting out a "cheap" instruction if all of its uses are PHIs. ...
Evan Cheng
1
-10
/
+55
2009-02-04
For now, only hoist re-materilizable instructions. LICM will increase registe...
Evan Cheng
1
-2
/
+11
2009-01-15
Simplify the MachineLICM pass by having it only traverse outer
Dan Gohman
1
-122
/
+60
2009-01-05
Tidy up #includes, deleting a bunch of unnecessary #includes.
Dan Gohman
1
-1
/
+0
2008-12-23
Use isTerminator() instead of isBranch()||isReturn() in
Dan Gohman
1
-1
/
+1
2008-12-18
Give MachineLICM a name, for -time-passes etc.
Dan Gohman
1
-0
/
+2
2008-10-03
Switch the MachineOperand accessors back to the short names like
Dan Gohman
1
-1
/
+1
2008-09-04
Tidy up several unbeseeming casts from pointer to intptr_t.
Dan Gohman
1
-1
/
+1
2008-08-31
Cosmetic changes to Machine LICM. No functionality change.
Bill Wendling
1
-31
/
+36
2008-08-20
Don't hoist instructions that define a physical register.
Bill Wendling
1
-1
/
+8
2008-08-18
Don't hoist instructions that have implicit uses or defines. Before, it was
Bill Wendling
1
-0
/
+5
2008-07-07
Prevent option name conflict.
Bill Wendling
1
-1
/
+1
2008-05-13
Clean up the use of static and anonymous namespaces. This turned up
Dan Gohman
1
-4
/
+4
2008-05-12
Revert the previous commit. Go ahead and hoist rematerializable instructions.
Bill Wendling
1
-4
/
+0
2008-05-12
One real change - don't hoist something that's trivially rematerializable. It's
Bill Wendling
1
-18
/
+19
2008-03-10
Minor cleanup. No functionality change.
Bill Wendling
1
-6
/
+8
2008-02-26
Rename PrintableName to Name.
Bill Wendling
1
-2
/
+2
2008-02-26
Change "Name" to "AsmName" in the target register info. Gee, a refactoring tool
Bill Wendling
1
-2
/
+2
2008-02-10
Rename MRegisterInfo to TargetRegisterInfo.
Dan Gohman
1
-6
/
+6
2008-01-10
Simplify the side effect stuff a bit more and make licm/sinking
Chris Lattner
1
-5
/
+22
2008-01-07
rename TargetInstrDescriptor -> TargetInstrDesc.
Chris Lattner
1
-4
/
+4
2008-01-07
Rename MachineInstr::getInstrDescriptor -> getDesc(), which reflects
Chris Lattner
1
-8
/
+6
2008-01-04
Add that this preserves some analyses.
Bill Wendling
1
-1
/
+3
2008-01-04
Move option to enable machine LICM into LLVMTargetMachine.cpp.
Bill Wendling
1
-10
/
+0
2008-01-04
Call the parent's getAnalysisUsage.
Bill Wendling
1
-0
/
+1
2008-01-04
remove dead #includes and reorder the rest.
Chris Lattner
1
-10
/
+6
2008-01-02
Use the correct MachineRegisterInfo object.
Bill Wendling
1
-2
/
+1
2008-01-02
Remove dead code.
Bill Wendling
1
-35
/
+0
2008-01-02
Use the new architecture to get the containing machine basic block for a machine
Bill Wendling
1
-28
/
+28
2007-12-29
Remove attribution from file headers, per discussion on llvmdev.
Chris Lattner
1
-2
/
+2
2007-12-20
Updated comments to reflect what "side effects" means in this situation.
Bill Wendling
1
-10
/
+3
2007-12-18
Add debugging info. Use the newly created "hasUnmodelledSideEffects" method.
Bill Wendling
1
-21
/
+38
2007-12-11
Need to grow the indexed map. Added debug statements.
Bill Wendling
1
-21
/
+27
2007-12-11
Simplify slightly.
Bill Wendling
1
-3
/
+4
2007-12-11
Blark! How in the world did this work without this?!
Bill Wendling
1
-4
/
+9
2007-12-11
- Update the virtual reg to machine instruction map when hoisting.
Bill Wendling
1
-2
/
+12
2007-12-11
Checking for "zero operands" during the "CanHoistInst()" method isn't necessary
Bill Wendling
1
-11
/
+4
2007-12-08
Reverting 44702. It wasn't correct to rename them.
Bill Wendling
1
-13
/
+13
2007-12-08
Renaming:
Bill Wendling
1
-1
/
+1
2007-12-08
Incorporated comments from Evan and Chris:
Bill Wendling
1
-58
/
+54
2007-12-07
Initial commit of the machine code LICM pass. It successfully hoists this:
Bill Wendling
1
-0
/
+336