similar to: [LLVMdev] MBB Critical edges

> If you don't want critical edges in the machine code CFG, you're going to > have to write a machine code CFG critical edge splitting pass: LLVM > doesn't currently have one. > > -Chris Hey guys, I've coded a pass to break the critical edges of the machine control flow graph. The program works fine, but I am sure it is not the right way of implementing it.

Thanks, Evan. Actually I've solved my problem with some hints from Dale and Anton. At least I think I've solved it. I had to add one method to TargetInstrInfo to tell me when an instruction is an indirect jump - TargetInstrInfo::tii.isIndirectJump(opcode). When that is the case, I update the jump table using: // Change jumps to go to the new basic block:

Sorry about the tardiness of my reply. My mail client has playing tricks with me. :-) I am assuming the issue has nothing to do the branch to jumptable instructions but rather the MachineJumpTableInfo associated with every MachineFunction? If so, please take a look at BranchFoldiing.cpp for an example. Evan On Aug 10, 2007, at 12:30 PM, Fernando Magno Quintao Pereira wrote: > >

Hi all, I have a pass to break critical edges of Machine Basic Blocks, but I just discovered a bug (when compiling code for x86). The problem is 'jumpl *%reg'. I don't know how to update the jump table for this type of instruction. The code that I had (see below) does not update the jump table, and the actual branch keeps jumping to the old basic block, instead of the new.

On Aug 17, 2007, at 3:08 PM, Fernando Magno Quintao Pereira wrote: > > Thanks, Evan. > > Actually I've solved my problem with some hints from Dale and > Anton. Ok, cool. I didn't see their replies. Like I said, my email client is acting funny. > At least I think I've solved it. I had to add one method to > TargetInstrInfo to tell me when an instruction

On Tue, 4 Jul 2006, Fernando Magno Quintao Pereira wrote: > However, it does not remove all the critical edges. I am getting a very > weird dataflow graph (even without the Break Critical edges pass). The > dataflow generated by MachineFunction::dump() for the program below is > given here: > http://compilers.cs.ucla.edu/fernando/projects/soc/images/loop_no_crit2.pdf ... > The

Dear guys, I am having problem to split edges correctly. Mostly because the new basic blocks are creating infinite loops. Could someone help me fixing the code below? It is creating assembly like this one below. Block LBB1_9 was inserted to break the critical edge between blocks LBB1_3 and LBB1_8. But it changes the semantics of the original program, because, before, LBB1_8 was falling

The problem is that you are inserting block 9 in the wrong spot. mf.getLastBlock() returns the block with the greatest number which may have nothing to do with the ordering. Why not use the end iterator (mf.end) to insert? -Tanya On Sat, 8 Jul 2006, Fernando Magno Quintao Pereira wrote: > > Dear guys, > > I am having problem to split edges correctly. Mostly because the new

I want to be able to do two things with LLVM (both just before code emission): 1. Insert a MachineBasicBlock just before a MachineBasicBlock. There is a function called AddPredecessor(). However, the comment says that it does not update the actual CFG. I want to redirect all CFG edges that are incoming to this MachineBasicBlock to the new one I create, and add just one outgoing edge (no branch)

Fernando, The code there should be more or less functional, though it's not currently used by anything. Eventually it should probably be moved to a method on MachineBasicBlock. The API breakage you're seeing is because some methods moved around. Feel free to fix it. :-) --Owen On Jan 26, 2008, at 6:31 PM, Fernando Magno Quintao Pereira wrote: > > Hi LLVMers, > >

Hi LLVMers, what is the status of breaking critical edges in machine functions? I just compiled the top of the LLVM tree, and I found llvm/CodeGen/BreakCriticalMachineEdge.h. But this file seems not to be up-to-date with the other classes in the top of the tree. For instance, it calls isTerminatorInstr on llvm::TargetInstrInfo, but this method is no longer there. If I want to break

On Tue, 3 Oct 2006, Fernando Magno Quintao Pereira wrote: > I am having some problems when compiling huge programs (like gcc). > I have a machine pass to destroy critical edges, and I have to add > new basic blocks, but sometimes the program is too big, and the new > basic block gets out of addressing range. What is the proper way to > address this problem? I am producing code

Dear guys, I am having some problems when compiling huge programs (like gcc). I have a machine pass to destroy critical edges, and I have to add new basic blocks, but sometimes the program is too big, and the new basic block gets out of addressing range. What is the proper way to address this problem? I am producing code for the PowerPC, and I had to implement the insert go-to method. I am

I've just subscribed, but a search of the archive doesn't indicate this has been handled before... Is there a way to get rsync to not create a new file while transferring and then rename it, but to instead update the existing file in place, i.e. simply write those blocks that have been updated and leave the rest alone? That would be ideal for what I wanted rsync for, namely updating

Thanks. Can you file bugzilla reports? I'll look at the first one soon. Evan On Feb 5, 2009, at 8:08 AM, Roman Levenstein wrote: > Hi, > > While testing my new register allocators on some test-cases, I've > noticed that LLVM misses sometimes some optimization opportunities: > > 1) LocalSpiller::RewriteMBB seems not to propagate the information > about e.g. Spills

Hi all, I've been investigating a machine verifier failure on the attached testcase, and I'm tempted to say the verifier is wrong and should accept it. Skip the description for the proposed change. On AArch64, the verifier complains with: *** Bad machine code: MBB exits via unconditional branch but doesn't have exactly one CFG successor! *** - function: t4 - basic

Hi, While testing my new register allocators on some test-cases, I've noticed that LLVM misses sometimes some optimization opportunities: 1) LocalSpiller::RewriteMBB seems not to propagate the information about e.g. Spills between MBBs.In many cases, where MBB B1 has only one predecessor MBB B2, B1 could reuse the information about the physical registers that are in the live-out set of B2.

Dear llvm-dev-list, I have created my own custom section to be added at the end into a binary upon compilation which contains address of all basic blocks. As the final address of the basic block is not known until link time, I collect the MCSymbol* Symbol Values per BB in a temp array and at the in the custom section and emit it (emitSymbolValue) into my section within EmitEndOfAsmFile() I have

Guys, two questions: 1) where, in the LLVM source tree, is the code that creates jump tables? 2) is it possible to change the destiny of a jump table with a MachineFunction pass? best, Fernando

I've ran across an issue in BranchFolding.cpp where it is incorrectly folding a branch to the wrong fallthrough location. This is in LLVM 2.4 and seems to be in 2.5 also. The code in question is: void BranchFolder::OptimizeBlock(MachineBasicBlock *MBB) { MachineFunction::iterator FallThrough = MBB; ++FallThrough; // If this block is empty, make everyone use its fall-through, not