[XRay] Detect loops in functions being lowered

Summary:
This is an implementation of the loop detection logic that XRay needs to
determine whether a function might take time at runtime. Without this
heuristic, XRay will tend to not instrument short functions that have
loops that might have runtime dependent on inputs or external values.

While this implementation doesn't do any further analysis than just
figuring out whether there is a loop in the MachineFunction being
code-gen'ed, we're paving the way for being able to perform more
sophisticated analysis of the function in the future (for example to
determine whether the trip count for the loop might be constant, and
make a decision on that instead). This enables us to cover more
functions with the default heuristics, and potentially identify ones
that have variable runtime latency just by looking for the presence of
loops.

Reviewers: chandlerc, rnk, pelikan

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D32274

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@302103 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/XRayInstrumentation.cpp b/lib/CodeGen/XRayInstrumentation.cpp
index 7d2848bdc13b1c3d6243290ca74dee8bc7ebb73e..2df3602733f3e068f035e2877bdc715df3c6caa9 100644 (file)
--- a/lib/CodeGen/XRayInstrumentation.cpp
+++ b/lib/CodeGen/XRayInstrumentation.cpp
@@ -18,6 +18,8 @@
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/CodeGen/MachineLoopInfo.h"
+#include "llvm/CodeGen/MachineDominators.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/Support/TargetRegistry.h"
 #include "llvm/Target/TargetInstrInfo.h"
@@ -33,6 +35,14 @@ struct XRayInstrumentation : public MachineFunctionPass {
     initializeXRayInstrumentationPass(*PassRegistry::getPassRegistry());
   }
 
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.setPreservesCFG();
+    AU.addRequired<MachineLoopInfo>();
+    AU.addPreserved<MachineLoopInfo>();
+    AU.addPreserved<MachineDominatorTree>();
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+
   bool runOnMachineFunction(MachineFunction &MF) override;
 
 private:
@@ -43,7 +53,7 @@ private:
   // This is the approach to go on CPUs which have a single RET instruction,
   //   like x86/x86_64.
   void replaceRetWithPatchableRet(MachineFunction &MF,
-    const TargetInstrInfo *TII);
+                                  const TargetInstrInfo *TII);
 
   // Prepend the original return instruction with the exit sled code ("patchable
   //   function exit" pseudo-instruction), preserving the original return
@@ -54,13 +64,12 @@ private:
   //   have to call the trampoline and return from it to the original return
   //   instruction of the function being instrumented.
   void prependRetWithPatchableExit(MachineFunction &MF,
-    const TargetInstrInfo *TII);
+                                   const TargetInstrInfo *TII);
 };
 } // anonymous namespace
 
-void XRayInstrumentation::replaceRetWithPatchableRet(MachineFunction &MF,
-  const TargetInstrInfo *TII)
-{
+void XRayInstrumentation::replaceRetWithPatchableRet(
+    MachineFunction &MF, const TargetInstrInfo *TII) {
   // We look for *all* terminators and returns, then replace those with
   // PATCHABLE_RET instructions.
   SmallVector<MachineInstr *, 4> Terminators;
@@ -91,9 +100,8 @@ void XRayInstrumentation::replaceRetWithPatchableRet(MachineFunction &MF,
     I->eraseFromParent();
 }
 
-void XRayInstrumentation::prependRetWithPatchableExit(MachineFunction &MF,
-  const TargetInstrInfo *TII)
-{
+void XRayInstrumentation::prependRetWithPatchableExit(
+    MachineFunction &MF, const TargetInstrInfo *TII) {
   for (auto &MBB : MF) {
     for (auto &T : MBB.terminators()) {
       unsigned Opc = 0;
@@ -106,7 +114,7 @@ void XRayInstrumentation::prependRetWithPatchableExit(MachineFunction &MF,
       if (Opc != 0) {
         // Prepend the return instruction with PATCHABLE_FUNCTION_EXIT or
         //   PATCHABLE_TAIL_CALL .
-        BuildMI(MBB, T, T.getDebugLoc(),TII->get(Opc));
+        BuildMI(MBB, T, T.getDebugLoc(), TII->get(Opc));
       }
     }
   }
@@ -125,8 +133,13 @@ bool XRayInstrumentation::runOnMachineFunction(MachineFunction &MF) {
       return false; // XRay threshold attribute not found.
     if (Attr.getValueAsString().getAsInteger(10, XRayThreshold))
       return false; // Invalid value for threshold.
-    if (F.size() < XRayThreshold)
-      return false; // Function is too small.
+
+    // Check if we have a loop.
+    // FIXME: Maybe make this smarter, and see whether the loops are dependent
+    // on inputs or side-effects?
+    MachineLoopInfo &MLI = getAnalysis<MachineLoopInfo>();
+    if (MLI.empty() && F.size() < XRayThreshold)
+      return false; // Function is too small and has no loops.
   }
 
   // We look for the first non-empty MachineBasicBlock, so that we can insert
@@ -142,12 +155,10 @@ bool XRayInstrumentation::runOnMachineFunction(MachineFunction &MF) {
 
   if (!MF.getSubtarget().isXRaySupported()) {
     FirstMI.emitError("An attempt to perform XRay instrumentation for an"
-      " unsupported target.");
+                      " unsupported target.");
     return false;
   }
 
-  // FIXME: Do the loop triviality analysis here or in an earlier pass.
-
   // First, insert an PATCHABLE_FUNCTION_ENTER as the first instruction of the
   // MachineFunction.
   BuildMI(FirstMBB, FirstMI, FirstMI.getDebugLoc(),
@@ -176,5 +187,8 @@ bool XRayInstrumentation::runOnMachineFunction(MachineFunction &MF) {
 
 char XRayInstrumentation::ID = 0;
 char &llvm::XRayInstrumentationID = XRayInstrumentation::ID;
-INITIALIZE_PASS(XRayInstrumentation, "xray-instrumentation", "Insert XRay ops",
-                false, false)
+INITIALIZE_PASS_BEGIN(XRayInstrumentation, "xray-instrumentation",
+                      "Insert XRay ops", false, false)
+INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo)
+INITIALIZE_PASS_END(XRayInstrumentation, "xray-instrumentation",
+                    "Insert XRay ops", false, false)

diff --git a/test/CodeGen/X86/xray-loop-detection.ll b/test/CodeGen/X86/xray-loop-detection.ll
new file mode 100644 (file)
index 0000000..3cd6b4a
--- /dev/null
+++ b/test/CodeGen/X86/xray-loop-detection.ll
@@ -0,0 +1,23 @@
+; RUN: llc -filetype=asm -o - -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+; RUN: llc -filetype=asm -o - -mtriple=x86_64-darwin-unknown    < %s | FileCheck %s
+
+define i32 @foo(i32 %i) nounwind noinline uwtable "xray-instruction-threshold"="1" {
+entry:
+  br label %Test
+Test:
+  %indvar = phi i32 [0, %entry], [%nextindvar, %Inc]
+  %cond = icmp eq i32 %indvar, %i
+  br i1 %cond, label %Exit, label %Inc
+Inc:
+  %nextindvar = add i32 %indvar, 1
+  br label %Test
+Exit:
+  %retval = phi i32 [%indvar, %Test]
+  ret i32 %retval
+}
+
+; CHECK-LABEL: xray_sled_0:
+; CHECK-NEXT:  .ascii "\353\t"
+; CHECK-NEXT:  nopw 512(%rax,%rax)
+; CHECK-LABEL: Ltmp0:
+