R600: Schedule copy from phys register at beginning of block

It allows regalloc pass to remove them by trivially assigning associated reg

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183336 91177308-0d34-0410-b5e6-96231b3b80d8

12 files changed, 42 insertions, 11 deletions

diff --git a/lib/Target/R600/R600MachineScheduler.cpp b/lib/Target/R600/R600MachineScheduler.cpp
index 9469e0f..8524fe3 100644
--- a/lib/Target/R600/R600MachineScheduler.cpp
+++ b/lib/Target/R600/R600MachineScheduler.cpp
@@ -71,6 +71,10 @@ SUnit* R600SchedStrategy::pickNode(bool &IsTopNode) {
       (!AllowSwitchFromAlu && CurInstKind == IDAlu))) {
     // try to pick ALU
     SU = pickAlu();
+    if (!SU && !PhysicalRegCopy.empty()) {
+      SU = PhysicalRegCopy.front();
+      PhysicalRegCopy.erase(PhysicalRegCopy.begin());
+    }
     if (SU) {
       if (CurEmitted >= InstKindLimit[IDAlu])
         CurEmitted = 0;
@@ -118,7 +122,22 @@ SUnit* R600SchedStrategy::pickNode(bool &IsTopNode) {
   return SU;
 }
 
+bool IsUnScheduled(const SUnit *SU) {
+  return SU->isScheduled;
+}
+
+static
+void Filter(std::vector<SUnit *> &List) {
+  List.erase(std::remove_if(List.begin(), List.end(), IsUnScheduled), List.end());
+}
+
 void R600SchedStrategy::schedNode(SUnit *SU, bool IsTopNode) {
+  if (IsTopNode) {
+    for (unsigned i = 0; i < AluLast; i++) {
+      Filter(Available[i]);
+      Filter(Pending[i]);
+    }
+  }
 
   if (NextInstKind != CurInstKind) {
     DEBUG(dbgs() << "Instruction Type Switch\n");
@@ -157,13 +176,24 @@ void R600SchedStrategy::schedNode(SUnit *SU, bool IsTopNode) {
   }
 }
 
+static bool
+isPhysicalRegCopy(MachineInstr *MI) {
+  if (MI->getOpcode() != AMDGPU::COPY)
+    return false;
+
+  return !TargetRegisterInfo::isVirtualRegister(MI->getOperand(1).getReg());
+}
+
 void R600SchedStrategy::releaseTopNode(SUnit *SU) {
   DEBUG(dbgs() << "Top Releasing ";SU->dump(DAG););
-
 }
 
 void R600SchedStrategy::releaseBottomNode(SUnit *SU) {
   DEBUG(dbgs() << "Bottom Releasing ";SU->dump(DAG););
+  if (isPhysicalRegCopy(SU->getInstr())) {
+    PhysicalRegCopy.push_back(SU);
+    return;
+  }
 
   int IK = getInstKind(SU);
 
diff --git a/lib/Target/R600/R600MachineScheduler.h b/lib/Target/R600/R600MachineScheduler.h
index 4dedf70..c5024d2 100644
--- a/lib/Target/R600/R600MachineScheduler.h
+++ b/lib/Target/R600/R600MachineScheduler.h
@@ -54,6 +54,7 @@ class R600SchedStrategy : public MachineSchedStrategy {
   std::vector<SUnit *> AvailableAlus[AluLast];
   std::vector<SUnit *> UnscheduledARDefs;
   std::vector<SUnit *> UnscheduledARUses;
+  std::vector<SUnit *> PhysicalRegCopy;
 
   InstKind CurInstKind;
   int CurEmitted;
diff --git a/test/CodeGen/R600/fabs.ll b/test/CodeGen/R600/fabs.ll
index b876c0e..85f2882 100644
--- a/test/CodeGen/R600/fabs.ll
+++ b/test/CodeGen/R600/fabs.ll
@@ -1,6 +1,6 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
-;CHECK: MOV * T{{[0-9]+\.[XYZW], \|PV\.[XYZW]\|}}
+;CHECK: MOV * T{{[0-9]+\.[XYZW], \|T[0-9]+\.[XYZW]\|}}
 
 define void @test() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/fadd.ll b/test/CodeGen/R600/fadd.ll
index bec5043..9a67232 100644
--- a/test/CodeGen/R600/fadd.ll
+++ b/test/CodeGen/R600/fadd.ll
@@ -1,7 +1,7 @@
 ; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
 ; CHECK: @fadd_f32
-; CHECK: ADD * T{{[0-9]+\.[XYZW], PV\.[XYZW], PV\.[XYZW]}}
+; CHECK: ADD * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @fadd_f32() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/floor.ll b/test/CodeGen/R600/floor.ll
index 7b8aa0a..877d69a 100644
--- a/test/CodeGen/R600/floor.ll
+++ b/test/CodeGen/R600/floor.ll
@@ -1,6 +1,6 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
-;CHECK: FLOOR * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
+;CHECK: FLOOR * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @test() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/fmad.ll b/test/CodeGen/R600/fmad.ll
index eb82b44..75e65d8 100644
--- a/test/CodeGen/R600/fmad.ll
+++ b/test/CodeGen/R600/fmad.ll
@@ -1,6 +1,6 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
-;CHECK: MULADD_IEEE * {{T[0-9]+\.[XYZW], PV\.[XYZW], PV.[XYZW], PV\.[XYZW]}}
+;CHECK: MULADD_IEEE * {{T[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @test() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/fmax.ll b/test/CodeGen/R600/fmax.ll
index 9357287..8b704e5 100644
--- a/test/CodeGen/R600/fmax.ll
+++ b/test/CodeGen/R600/fmax.ll
@@ -1,6 +1,6 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
-;CHECK: MAX * T{{[0-9]+\.[XYZW], PV\.[XYZW], PV\.[XYZW]}}
+;CHECK: MAX * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @test() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/fmin.ll b/test/CodeGen/R600/fmin.ll
index e38269c..5e34b7c 100644
--- a/test/CodeGen/R600/fmin.ll
+++ b/test/CodeGen/R600/fmin.ll
@@ -1,6 +1,6 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
-;CHECK: MIN * T{{[0-9]+\.[XYZW], PV\.[XYZW], PV\.[XYZW]}}
+;CHECK: MIN * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @test() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/fmul.ll b/test/CodeGen/R600/fmul.ll
index fee5ead..a40e818 100644
--- a/test/CodeGen/R600/fmul.ll
+++ b/test/CodeGen/R600/fmul.ll
@@ -1,7 +1,7 @@
 ; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
 ; CHECK: @fmul_f32
-; CHECK: MUL_IEEE * {{T[0-9]+\.[XYZW], PV\.[XYZW], PV\.[XYZW]}}
+; CHECK: MUL_IEEE * {{T[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @fmul_f32() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/fsub.ll b/test/CodeGen/R600/fsub.ll
index 2c88cbc..f784cde 100644
--- a/test/CodeGen/R600/fsub.ll
+++ b/test/CodeGen/R600/fsub.ll
@@ -1,7 +1,7 @@
 ; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
 ; CHECK: @fsub_f32
-; CHECK: ADD * T{{[0-9]+\.[XYZW], PV\.[XYZW], -PV\.[XYZW]}}
+; CHECK: ADD * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}}
 
 define void @fsub_f32() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/llvm.AMDGPU.mul.ll b/test/CodeGen/R600/llvm.AMDGPU.mul.ll
index be14e6f..cc0732b 100644
--- a/test/CodeGen/R600/llvm.AMDGPU.mul.ll
+++ b/test/CodeGen/R600/llvm.AMDGPU.mul.ll
@@ -1,6 +1,6 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
-;CHECK: MUL NON-IEEE * T{{[0-9]+\.[XYZW], PV\.[XYZW], PV\.[XYZW]}}
+;CHECK: MUL NON-IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @test() {
    %r0 = call float @llvm.R600.load.input(i32 0)
diff --git a/test/CodeGen/R600/llvm.pow.ll b/test/CodeGen/R600/llvm.pow.ll
index 532983f..1422083 100644
--- a/test/CodeGen/R600/llvm.pow.ll
+++ b/test/CodeGen/R600/llvm.pow.ll
@@ -1,7 +1,7 @@
 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s
 
 ;CHECK: LOG_IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-;CHECK: MUL NON-IEEE * T{{[0-9]+\.[XYZW], PV\.[XYZW], T[0-9]+\.[XYZW]}}
+;CHECK: MUL NON-IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ;CHECK-NEXT: EXP_IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 
 define void @test() {