search for: removedeadnodes

Hi, For my research, I need to reshape the current ARM backend to support armv2a. Zero-extend half word load (ldrh) is not supported by armv2a, so I need to make the code generation to not generate ldrh instructions. I want to replace all those instances with a 32-bit load (ldr) and then and the result with 0xffff to mask out the upper bits. These are the modifications that I have made to

...ME: Handle some special cases: FNEG->FSUB + Result = UnrollVectorOp(SDValue(Result, 0)).getNode(); + break; + } + break; + } + if (&*I != Result) { + Changed = true; + DAG.ReplaceAllUsesWith(I, Result); + } + } + + // Remove dead nodes now. + DAG.RemoveDeadNodes(); + + return Changed; +} + +SDValue VectorLegalizer::PromoteVectorOp(SDValue Op) { + MVT VT = Op.getValueType(); + assert(Op.getNode()->getNumValues() == 1 && + "Can't promote a vector with multiple results!"); + MVT NVT = TLI.getTypeToPromoteTo(Op.getOpcode(),...

Hi, During instruction selection, I have the following code for certain LOAD instructions: const LoadSDNode *LD = cast<LoadSDNode>(N); SDNode* LDW = CurDAG->getMachineNode(AVR::LDWRdPtr, SDLoc(N), VT, PtrVT, MVT::Other, LD->getBasePtr(), LD->getChain()); // Honestly, I have no idea what this does, but other memory // accessing instructions

...ME: Handle some special cases: FNEG->FSUB + Result = UnrollVectorOp(SDValue(Result, 0)).getNode(); + break; + } + break; + } + if (&*I != Result) { + Changed = true; + DAG.ReplaceAllUsesWith(I, Result); + } + } + + // Remove dead nodes now. + DAG.RemoveDeadNodes(); + + return Changed; +} + +SDValue VectorLegalizer::PromoteVectorOp(SDValue Op) { + MVT VT = Op.getValueType(); + assert(Op.getNode()->getNumValues() == 1 && + "Can't promote a vector with multiple results!"); + MVT NVT = TLI.getTypeToPromoteTo(Op.getOpcode(),...

...t; > ... the common node in this case never gets unified. This is one of the > top priorities for me when I get back to Illinois (after thanksgiving > break), but it is unlikely to make much progress in the mean time. If you > would like to disable this behavior, look in the DSGraph::removeDeadNodes > (or similar) in DataStructure.cpp. In it you will see the code that marks > nodes as being "alive" in the current graph. All you have to do is add > code to mark all globals as alive, and they will be propogated up to main. > Alternatively, take a look at the CVS revisions...

On Friday 03 October 2008 12:06, Dan Gohman wrote: > On Fri, October 3, 2008 9:10 am, David Greene wrote: > > On Thursday 02 October 2008 19:32, Dan Gohman wrote: > >> Looking at your dump() output above, it looks like the pre-selection > >> loads have multiple uses, so even though you've managed to match a > >> larger pattern that incorporates them, they

I have some questions regarding how globals are represented in DS graph. Specifically, I wrote the following simple program: List *g; void alloc_func(){ g = ( List* ) malloc( sizeof( List ) ); } void free_func(){ free( g ); } int main(){ alloc_func(); free_func(); } I noticed that the DSnode for g in alloc_func is different from that of free_func and NEITHER of them had GlobalNode

"Martin J. O'Riordan" <martin.oriordan at movidius.com> writes: > Thanks Ahmed and also Alex for your replies. > > This is more or less what I was realising, but it is a great > confidence booster to know that it is the correct way also. I can > replace all of my various 'Select*' specialisations with version that > use

On 7/29/2017 1:28 AM, Dr. ERDI Gergo via llvm-dev wrote: > Hi, > > During instruction selection, I have the following code for certain > LOAD instructions: > > const LoadSDNode *LD = cast<LoadSDNode>(N); > SDNode* LDW = CurDAG->getMachineNode(AVR::LDWRdPtr, SDLoc(N), > VT, PtrVT, MVT::Other, > LD->getBasePtr(), LD->getChain()); >

...uot; global node of g? ... the common node in this case never gets unified. This is one of the top priorities for me when I get back to Illinois (after thanksgiving break), but it is unlikely to make much progress in the mean time. If you would like to disable this behavior, look in the DSGraph::removeDeadNodes (or similar) in DataStructure.cpp. In it you will see the code that marks nodes as being "alive" in the current graph. All you have to do is add code to mark all globals as alive, and they will be propogated up to main. Alternatively, take a look at the CVS revisions a few checkins back...

...to obtain a DSGraph for a complete module, but haven't succeeded. I tried - getting the DSGraph for each function, and copying the nodes to a Module graph. But since there are no nodehandles, apparently there is no merging of nodes. - modifying BUDataStructure Pass so that it will not execute removeDeadNodes() or removeTriviallyDeadNodes(). - couldn't get mergeInGraph() to work, but tried. My main interest is in situations like void *M(int size) { return malloc(size); } void F() { void *p = M(....); } where I can identify that %p and the object returned by M() are the same. nicolas

...::Value* const, llvm::DSNodeHandle>, llv] opt(__gnu_cxx::hashtable<std::pair<llvm::Value* const, llvm::DSNodeHandle>, llv] opt(__gnu_cxx::hash_map<llvm::Value*, llvm::DSNodeHandle, __gnu_cxx::hash<llvm:] opt(llvm::DSScalarMap::find(llvm::Value*)+0x1a)[0x84a998e] opt(llvm::DSGraph::removeDeadNodes(unsigned)+0x3e1)[0x847def5] opt(llvm::IPModRef::getFuncInfo(llvm::Function const&, bool)+0xf3)[0x85db297] opt(llvm::IPModRef::run(llvm::Module&)+0xaf)[0x85db189] opt(llvm::PassManagerTraits<llvm::Module>::runPass(llvm::Pass*, llvm::Module*)+] opt(llvm::PassManagerT<llvm::Module>...

...e common node in this case never gets unified. This is one of the > > top priorities for me when I get back to Illinois (after thanksgiving > > break), but it is unlikely to make much progress in the mean time. If you > > would like to disable this behavior, look in the DSGraph::removeDeadNodes > > (or similar) in DataStructure.cpp. In it you will see the code that marks > > nodes as being "alive" in the current graph. All you have to do is add > > code to mark all globals as alive, and they will be propogated up to main. > > Alternatively, take a look...

Thanks Ahmed and also Alex for your replies. This is more or less what I was realising, but it is a great confidence booster to know that it is the correct way also. I can replace all of my various 'Select*' specialisations with version that use 'ReplaceNode/SelectCode' and return 'void', but what about the places where I currently call 'Select(N)' directly?

On Oct 7, 2008, at 12:04 PM, David Greene wrote: > On Friday 03 October 2008 12:06, Dan Gohman wrote: >> On Fri, October 3, 2008 9:10 am, David Greene wrote: >>> On Thursday 02 October 2008 19:32, Dan Gohman wrote: >>>> Looking at your dump() output above, it looks like the pre- >>>> selection >>>> loads have multiple uses, so even though

On 3/4/13 8:05 AM, Kevin Streit wrote: > Hi, > > during the hunt for a bug causing strange behavior of our automatic > parallelization framework, > I found some, at least for me, unexpected behavior of the > DataStructureAnalysis in Poolalloc. > > Consider the following simplified program: > > ==================== > int ARR[4] = {1, 2, 3, 4}; > > int a(int

On Fri, October 3, 2008 9:10 am, David Greene wrote: > On Thursday 02 October 2008 19:32, Dan Gohman wrote: > >> Looking at your dump() output above, it looks like the pre-selection >> loads have multiple uses, so even though you've managed to match a >> larger pattern that incorporates them, they still need to exist to >> satisfy some other users. > > Yes,

On Thursday 02 October 2008 19:32, Dan Gohman wrote: > Looking at your dump() output above, it looks like the pre-selection > loads have multiple uses, so even though you've managed to match a > larger pattern that incorporates them, they still need to exist to > satisfy some other users. Yes, I looked at that too. It looks like these other uses end up being chains to

Hi, during the hunt for a bug causing strange behavior of our automatic parallelization framework, I found some, at least for me, unexpected behavior of the DataStructureAnalysis in Poolalloc. Consider the following simplified program: ==================== int ARR[4] = {1, 2, 3, 4}; int a(int pos) { return ARR[pos]; } int sum(int op_a, int op_b) { return a(op_a) + a(op_b); } int

...) << "===== Instruction selection begins:\n"); Indent = 0; #endif SelectRoot(*CurDAG); #ifndef NDEBUG - DEBUG(errs() << "===== Instruction selection ends:\n"); + DEBUG(dbgs() << "===== Instruction selection ends:\n"); #endif CurDAG->RemoveDeadNodes(); @@ -853,7 +853,7 @@ bool is64Bit = Subtarget->is64Bit(); DebugLoc dl = N.getDebugLoc(); DEBUG({ - errs() << "MatchAddress: "; + dbgs() << "MatchAddress: "; AM.dump(); }); // Limit recursion. @@ -1711,9 +1711,9 @@ #ifndef N...