search for: pred_siz

...) { if( seen[i] ) continue; seen[i] = true; MachineBasicBlock * start, *block; start = block = mf.getBlockNumbered(i); std::vector< MachineBasicBlock* > blocks; while( block->succ_size() == 1 && (*block->succ_begin())->pred_size() == 1 ) { block = *block->succ_begin(); seen[block->getNumber()] = true; blocks.push_back( block ); } // TODO: // For each basic block bb in blocks in order of insersion: // 1. Remove basic blocks in the block vector from the mac...

Has anyone written a pass at the MachineFunction level which combines machine basic blocks which is guaranteed to be the single predecessor to another block? Or is there a reason not to combine them? - Thanks Jeff Kunkel

On Oct 6, 2010, at 4:31 PM, Jeff Kunkel wrote: > Has anyone written a pass at the MachineFunction level which combines > machine basic blocks which is guaranteed to be the single predecessor > to another block? Or is there a reason not to combine them? I'm not sure exactly what transformation you're referring to, but BranchFolder::OptimizeBranches does a lot of things like that.

...;>        MachineBasicBlock * start, *block; >>>>        start = block = mf.getBlockNumbered(i); >>>>        std::vector< MachineBasicBlock* > blocks; >>>>        while( block->succ_size() == 1 && >>>> (*block->succ_begin())->pred_size() == 1 ) { >>>>          block = *block->succ_begin(); >>>>          seen[block->getNumber()] = true; >>>>          blocks.push_back( block ); >>>>        } >>>>        // TODO: >>>>        // For each basic block bb in...

...in the entry block. The shrink wrap version probably should go to its own function. Otherwise, it should exit early when the non-shrink wrapping version is done. That reduces nesting and please those of us who are a bit picky. 6. + // Save entry block, return blocks. + if (MBB->pred_size() == 0) + entryBlock = MBB; Entry block is the Fn.front(). 7. + if (!MBB->empty() && MBB->back().getDesc().isReturn()) + returnBlocks.push_back(MBB); PEI::insertPrologEpilogCode also traverse MBBs and get at return blocks. So these probably ought to be sha...

Here is an updated patch for shrink wrapping with: - spills/restores done with stack slot stores/loads - stack adjustment removed - refactoring (but still in need of more) - spill/restore insertion code unified with spill/restore placement code Documentation available here<http://wiki.github.com/jdmdj/llvm-work/shrink-wrapping-work> illustrates shrink wrapping with loops and discusses a

On Mon, Mar 2, 2009 at 10:35 AM, Evan Cheng <echeng at apple.com> wrote: > > On Mar 1, 2009, at 2:57 PM, John Mosby wrote: > > Obviously, all of this applies only when spills are done with push/pop, > which is the case on x86. I used this issue to start looking at generalizing > how spills and restores are handled, before looking too closely at other > targets, and

...the two behaviors around the idea of placing spills/restores. I think the versions need to be separate for the reasons you state but the placement idea should be retained. I will reintroduce the separation. > 6. > + // Save entry block, return blocks. > > > + if (MBB->pred_size() == 0) > > > + entryBlock = MBB; > > Entry block is the Fn.front(). > > 7. > + if (!MBB->empty() && MBB->back().getDesc().isReturn()) > > > + returnBlocks.push_back(MBB); > > PEI::insertPrologEpilogCode also traverse MBBs...

On Wed, Oct 31, 2012 at 09:13:43PM +0100, Bjorn De Sutter wrote: > Hi all, > > I am working on a CGRA backend (something like a 2D VLIW), and we also absolutely need predication. I extended the IfConversion pass to allow it to be executed multiple times and to predicate already predicated code. This is necessary to predicate code with nested conditional statements. At this point, we

...ing spills/restores. > I think the versions need to be separate for the reasons you state but the > placement idea should be retained. I will reintroduce the separation. > >> >> 6. >> +      // Save entry block, return blocks. >> >> >> +      if (MBB->pred_size() == 0) >> >> >> +        entryBlock = MBB; >> Entry block is the Fn.front(). >> 7. >> +        if (!MBB->empty() && MBB->back().getDesc().isReturn()) >> >> >> +          returnBlocks.push_back(MBB); >> PEI::insertPrologEpil...

...MachineBasicBlock * mbb_succ = *succ; // necessary and sufficient condition that characterizes a critical // edge: predecessor with many successors, successor with many // predecessors. if( (mbb->succ_size() > 1) && (mbb_succ->pred_size() > 1) ) { src_blocks.push_back(mbb); dst_blocks.push_back(mbb_succ); } } } for(unsigned u = 0; u < src_blocks.size() > 0; u++) { MachineBasicBlock * src = src_blocks[u]; MachineBasicBlock * dst = dst_blocks[u];...

Hi all, I am working on a CGRA backend (something like a 2D VLIW), and we also absolutely need predication. I extended the IfConversion pass to allow it to be executed multiple times and to predicate already predicated code. This is necessary to predicate code with nested conditional statements. At this point, we support or, and, and conditional predicates (see Scott Mahlke's papers on this