diff options
author | Nadav Rotem <nrotem@apple.com> | 2013-04-12 00:48:32 +0000 |
---|---|---|
committer | Nadav Rotem <nrotem@apple.com> | 2013-04-12 00:48:32 +0000 |
commit | 07706e5506fb1bf494667e9fd16795b6c0b4d0fb (patch) | |
tree | 904fea6f8e94e70005c42efc315e29b35f1983ad | |
parent | e292347503cd7598429c08f9984ab3e0a44ab8a3 (diff) | |
download | llvm-07706e5506fb1bf494667e9fd16795b6c0b4d0fb.tar.gz llvm-07706e5506fb1bf494667e9fd16795b6c0b4d0fb.tar.bz2 llvm-07706e5506fb1bf494667e9fd16795b6c0b4d0fb.tar.xz |
Add a flag to align all basic blocks in the function.
When debugging performance regressions we often ask ourselves if the regression
that we see is due to poor isel/sched/ra or due to some micro-architetural
problem. When comparing two code sequences one good way to rule out front-end
bottlenecks (and other the issues) is to force code alignment. This pass adds
a flag that forces the alignment of all of the basic blocks in the program.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179353 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r-- | lib/CodeGen/MachineBlockPlacement.cpp | 14 | ||||
-rw-r--r-- | test/CodeGen/X86/code_placement_align_all.ll | 22 |
2 files changed, 36 insertions, 0 deletions
diff --git a/lib/CodeGen/MachineBlockPlacement.cpp b/lib/CodeGen/MachineBlockPlacement.cpp index cd948e24a6..774df81b4d 100644 --- a/lib/CodeGen/MachineBlockPlacement.cpp +++ b/lib/CodeGen/MachineBlockPlacement.cpp @@ -39,6 +39,7 @@ #include "llvm/CodeGen/MachineLoopInfo.h" #include "llvm/CodeGen/MachineModuleInfo.h" #include "llvm/Support/Allocator.h" +#include "llvm/Support/CommandLine.h" #include "llvm/Support/Debug.h" #include "llvm/Target/TargetInstrInfo.h" #include "llvm/Target/TargetLowering.h" @@ -52,6 +53,11 @@ STATISTIC(CondBranchTakenFreq, STATISTIC(UncondBranchTakenFreq, "Potential frequency of taking unconditional branches"); +static cl::opt<unsigned> AlignAllBlock("align-all-blocks", + cl::desc("Force the alignment of all " + "blocks in the function."), + cl::init(0), cl::Hidden); + namespace { class BlockChain; /// \brief Type for our function-wide basic block -> block chain mapping. @@ -1083,6 +1089,14 @@ bool MachineBlockPlacement::runOnMachineFunction(MachineFunction &F) { TLI = F.getTarget().getTargetLowering(); assert(BlockToChain.empty()); + if (AlignAllBlock) { + // Align all of the blocks in the function to a specific alignment. + for (MachineFunction::iterator FI = F.begin(), FE = F.end(); + FI != FE; ++FI) + FI->setAlignment(AlignAllBlock); + return true; + } + buildCFGChains(F); BlockToChain.clear(); diff --git a/test/CodeGen/X86/code_placement_align_all.ll b/test/CodeGen/X86/code_placement_align_all.ll new file mode 100644 index 0000000000..2a36a3a3af --- /dev/null +++ b/test/CodeGen/X86/code_placement_align_all.ll @@ -0,0 +1,22 @@ +; RUN: llc -march=x86 -align-all-blocks=16 < %s | FileCheck %s + +;CHECK: foo +;CHECK: .align 16, 0x90 +;CHECK: .align 16, 0x90 +;CHECK: .align 16, 0x90 +;CHECK: ret +define i32 @foo(i32 %t, i32 %l) nounwind readnone ssp uwtable { + %1 = icmp eq i32 %t, 0 + br i1 %1, label %4, label %2 + +; <label>:2 ; preds = %0 + %3 = add nsw i32 %t, 2 + ret i32 %3 + +; <label>:4 ; preds = %0 + %5 = icmp eq i32 %l, 0 + %. = select i1 %5, i32 0, i32 5 + ret i32 %. +} + + |