Fix PR3457: Ignore control successors when looking for closest scheduled successor. A control successor doesn't read result(s) produced by the scheduling unit being evaluated.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@64210 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 24 insertions, 3 deletions

diff --git a/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp b/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
index bff85704e0..16102f342f 100644
--- a/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
+++ b/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
@@ -30,7 +30,6 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/STLExtras.h"
 #include <climits>
-#include "llvm/Support/CommandLine.h"
 using namespace llvm;
 
 STATISTIC(NumBacktracks, "Number of times scheduler backtracked");
@@ -1047,6 +1046,7 @@ static unsigned closestSucc(const SUnit *SU) {
   unsigned MaxHeight = 0;
   for (SUnit::const_succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
        I != E; ++I) {
+    if (I->isCtrl()) continue;  // ignore chain succs
     unsigned Height = I->getSUnit()->getHeight();
     // If there are bunch of CopyToRegs stacked up, they should be considered
     // to be at the same position.
diff --git a/test/CodeGen/ARM/memcpy-inline.ll b/test/CodeGen/ARM/memcpy-inline.ll
index 665d3acd89..5d1beea5fc 100644
--- a/test/CodeGen/ARM/memcpy-inline.ll
+++ b/test/CodeGen/ARM/memcpy-inline.ll
@@ -1,6 +1,9 @@
-; RUN: llvm-as < %s | llc -mtriple=arm-apple-darwin | grep ldmia
 ; RUN: llvm-as < %s | llc -mtriple=arm-apple-darwin | grep ldrb
 ; RUN: llvm-as < %s | llc -mtriple=arm-apple-darwin | grep ldrh
+; This used to look for ldmia. But it's no longer lucky enough to
+; have the load / store instructions lined up just right after
+; scheduler change for pr3457. We'll look for a robust solution
+; later.
 
 	%struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
 @src = external global %struct.x
diff --git a/test/CodeGen/X86/fold-pcmpeqd-0.ll b/test/CodeGen/X86/fold-pcmpeqd-0.ll
index 152c12215d..32eddbd329 100644
--- a/test/CodeGen/X86/fold-pcmpeqd-0.ll
+++ b/test/CodeGen/X86/fold-pcmpeqd-0.ll
@@ -1,8 +1,10 @@
-; RUN: llvm-as < %s | llc -mtriple=i386-apple-darwin   | grep pcmpeqd | count 1
+; RUN: llvm-as < %s | llc -mtriple=i386-apple-darwin   | not grep pcmpeqd
+; RUN: llvm-as < %s | llc -mtriple=i386-apple-darwin   | grep orps | grep CPI1_2  | count 2
 ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin | grep pcmpeqd | count 1
 
 ; This testcase shouldn't need to spill the -1 value,
 ; so it should just use pcmpeqd to materialize an all-ones vector.
+; For i386, cp load of -1 are folded.
 
 	%struct.__ImageExecInfo = type <{ <4 x i32>, <4 x float>, <2 x i64>, i8*, i8*, i8*, i32, i32, i32, i32, i32 }>
 	%struct._cl_image_format_t = type <{ i32, i32, i32 }>
diff --git a/test/CodeGen/X86/pr3457.ll b/test/CodeGen/X86/pr3457.ll
new file mode 100644
index 0000000000..36d4a5d554
--- /dev/null
+++ b/test/CodeGen/X86/pr3457.ll
@@ -0,0 +1,16 @@
+; RUN: llvm-as < %s | llc -mtriple=i386-apple-darwin | not grep fstpt
+; PR3457
+; rdar://6548010
+
+define void @foo(double* nocapture %P) nounwind {
+entry:
+	%0 = tail call double (...)* @test() nounwind		; <double> [#uses=2]
+	%1 = tail call double (...)* @test() nounwind		; <double> [#uses=2]
+	%2 = mul double %0, %0		; <double> [#uses=1]
+	%3 = mul double %1, %1		; <double> [#uses=1]
+	%4 = add double %2, %3		; <double> [#uses=1]
+	store double %4, double* %P, align 8
+	ret void
+}
+
+declare double @test(...)