RegisterPressure: Fix default lanemask for missing regunit intervals
In case of missing live intervals for a physical registers
getLanesWithProperty() would report 0 which was not a safe default in
all situations. Add a parameter to pass in a safe default.
No testcase because in-tree targets do not skip computing register unit
live intervals.
Also cleanup the getXXX() functions to not perform the
RequireLiveIntervals checks anymore so we do not even need to return
safe defaults.
llvm-svn: 267977
diff --git a/llvm/lib/CodeGen/RegisterPressure.cpp b/llvm/lib/CodeGen/RegisterPressure.cpp
index e07efd3..a21d6c1 100644
--- a/llvm/lib/CodeGen/RegisterPressure.cpp
+++ b/llvm/lib/CodeGen/RegisterPressure.cpp
@@ -382,7 +382,7 @@
static LaneBitmask getLanesWithProperty(const LiveIntervals &LIS,
const MachineRegisterInfo &MRI, bool TrackLaneMasks, unsigned RegUnit,
- SlotIndex Pos,
+ SlotIndex Pos, LaneBitmask SafeDefault,
bool(*Property)(const LiveRange &LR, SlotIndex Pos)) {
if (TargetRegisterInfo::isVirtualRegister(RegUnit)) {
const LiveInterval &LI = LIS.getInterval(RegUnit);
@@ -402,7 +402,7 @@
// Be prepared for missing liveranges: We usually do not compute liveranges
// for physical registers on targets with many registers (GPUs).
if (LR == nullptr)
- return 0;
+ return SafeDefault;
return Property(*LR, Pos) ? ~0u : 0;
}
}
@@ -411,7 +411,7 @@
const MachineRegisterInfo &MRI,
bool TrackLaneMasks, unsigned RegUnit,
SlotIndex Pos) {
- return getLanesWithProperty(LIS, MRI, TrackLaneMasks, RegUnit, Pos,
+ return getLanesWithProperty(LIS, MRI, TrackLaneMasks, RegUnit, Pos, ~0u,
[](const LiveRange &LR, SlotIndex Pos) {
return LR.liveAt(Pos);
});
@@ -570,11 +570,11 @@
AddFlagsMI != nullptr && (LiveAfter & ~I->LaneMask) == 0)
AddFlagsMI->setRegisterDefReadUndef(RegUnit);
- LaneBitmask LaneMask = I->LaneMask & LiveAfter;
- if (LaneMask == 0) {
+ LaneBitmask ActualDef = I->LaneMask & LiveAfter;
+ if (ActualDef == 0) {
I = Defs.erase(I);
} else {
- I->LaneMask = LaneMask;
+ I->LaneMask = ActualDef;
++I;
}
}
@@ -790,9 +790,11 @@
}
// Discover live outs if this may be the first occurance of this register.
- LaneBitmask LiveOut = getLiveThroughAt(Reg, SlotIdx);
- if (LiveOut != 0)
- discoverLiveOut(RegisterMaskPair(Reg, LiveOut));
+ if (RequireIntervals) {
+ LaneBitmask LiveOut = getLiveThroughAt(Reg, SlotIdx);
+ if (LiveOut != 0)
+ discoverLiveOut(RegisterMaskPair(Reg, LiveOut));
+ }
}
increaseRegPressure(Reg, PreviousMask, NewMask);
@@ -875,10 +877,12 @@
LiveRegs.insert(RegisterMaskPair(Reg, LiveIn));
}
// Kill liveness at last uses.
- LaneBitmask LastUseMask = getLastUsedLanes(Reg, SlotIdx);
- if (LastUseMask != 0) {
- LiveRegs.erase(RegisterMaskPair(Reg, LastUseMask));
- decreaseRegPressure(Reg, LiveMask, LiveMask & ~LastUseMask);
+ if (RequireIntervals) {
+ LaneBitmask LastUseMask = getLastUsedLanes(Reg, SlotIdx);
+ if (LastUseMask != 0) {
+ LiveRegs.erase(RegisterMaskPair(Reg, LastUseMask));
+ decreaseRegPressure(Reg, LiveMask, LiveMask & ~LastUseMask);
+ }
}
}
@@ -1197,10 +1201,8 @@
LaneBitmask RegPressureTracker::getLiveLanesAt(unsigned RegUnit,
SlotIndex Pos) const {
- if (!RequireIntervals)
- return 0;
-
- return getLanesWithProperty(*LIS, *MRI, TrackLaneMasks, RegUnit, Pos,
+ assert(RequireIntervals);
+ return getLanesWithProperty(*LIS, *MRI, TrackLaneMasks, RegUnit, Pos, ~0u,
[](const LiveRange &LR, SlotIndex Pos) {
return LR.liveAt(Pos);
});
@@ -1208,11 +1210,9 @@
LaneBitmask RegPressureTracker::getLastUsedLanes(unsigned RegUnit,
SlotIndex Pos) const {
- if (!RequireIntervals)
- return 0;
-
+ assert(RequireIntervals);
return getLanesWithProperty(*LIS, *MRI, TrackLaneMasks, RegUnit,
- Pos.getBaseIndex(),
+ Pos.getBaseIndex(), 0,
[](const LiveRange &LR, SlotIndex Pos) {
const LiveRange::Segment *S = LR.getSegmentContaining(Pos);
return S != nullptr && S->end == Pos.getRegSlot();
@@ -1221,10 +1221,8 @@
LaneBitmask RegPressureTracker::getLiveThroughAt(unsigned RegUnit,
SlotIndex Pos) const {
- if (!RequireIntervals)
- return 0;
-
- return getLanesWithProperty(*LIS, *MRI, TrackLaneMasks, RegUnit, Pos,
+ assert(RequireIntervals);
+ return getLanesWithProperty(*LIS, *MRI, TrackLaneMasks, RegUnit, Pos, 0u,
[](const LiveRange &LR, SlotIndex Pos) {
const LiveRange::Segment *S = LR.getSegmentContaining(Pos);
return S != nullptr && S->start < Pos.getRegSlot(true) &&
@@ -1251,12 +1249,12 @@
if (TrackLaneMasks)
RegOpers.adjustLaneLiveness(*LIS, *MRI, SlotIdx);
- for (const RegisterMaskPair &Use : RegOpers.Uses) {
- unsigned Reg = Use.RegUnit;
- LaneBitmask LastUseMask = getLastUsedLanes(Reg, SlotIdx);
- if (LastUseMask == 0)
- continue;
- if (RequireIntervals) {
+ if (RequireIntervals) {
+ for (const RegisterMaskPair &Use : RegOpers.Uses) {
+ unsigned Reg = Use.RegUnit;
+ LaneBitmask LastUseMask = getLastUsedLanes(Reg, SlotIdx);
+ if (LastUseMask == 0)
+ continue;
// The LastUseMask is queried from the liveness information of instruction
// which may be further down the schedule. Some lanes may actually not be
// last uses for the current position.
@@ -1267,11 +1265,11 @@
= findUseBetween(Reg, LastUseMask, CurrIdx, SlotIdx, *MRI, LIS);
if (LastUseMask == 0)
continue;
- }
- LaneBitmask LiveMask = LiveRegs.contains(Reg);
- LaneBitmask NewMask = LiveMask & ~LastUseMask;
- decreaseRegPressure(Reg, LiveMask, NewMask);
+ LaneBitmask LiveMask = LiveRegs.contains(Reg);
+ LaneBitmask NewMask = LiveMask & ~LastUseMask;
+ decreaseRegPressure(Reg, LiveMask, NewMask);
+ }
}
// Generate liveness for defs.
diff --git a/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll b/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
index f1ab48b..6cf1967 100644
--- a/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
+++ b/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
@@ -1,10 +1,10 @@
; RUN: llc -march=amdgcn -mattr=-promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GCN-LABEL: {{^}}stored_fi_to_lds:
-; GCN-DAG: s_load_dword [[LDSPTR:s[0-9]+]]
-; GCN-DAG: v_mov_b32_e32 [[ZERO1:v[0-9]+]], 0{{$}}
-; GCN-DAG: buffer_store_dword v{{[0-9]+}}, [[ZERO1]]
+; GCN: s_load_dword [[LDSPTR:s[0-9]+]]
; GCN: v_mov_b32_e32 [[ZERO0:v[0-9]+]], 0{{$}}
+; GCN: v_mov_b32_e32 [[ZERO1:v[0-9]+]], 0{{$}}
+; GCN: buffer_store_dword v{{[0-9]+}}, [[ZERO1]]
; GCN: v_mov_b32_e32 [[VLDSPTR:v[0-9]+]], [[LDSPTR]]
; GCN: ds_write_b32 [[VLDSPTR]], [[ZERO0]]
define void @stored_fi_to_lds(float* addrspace(3)* %ptr) #0 {
diff --git a/llvm/test/CodeGen/AMDGPU/valu-i1.ll b/llvm/test/CodeGen/AMDGPU/valu-i1.ll
index ffa0110..c9776dd 100644
--- a/llvm/test/CodeGen/AMDGPU/valu-i1.ll
+++ b/llvm/test/CodeGen/AMDGPU/valu-i1.ll
@@ -137,7 +137,7 @@
; SI: BB#4:
; SI: buffer_store_dword
-; SI: v_cmp_ge_i64_e64 [[CMP:s\[[0-9]+:[0-9]+\]]]
+; SI: v_cmp_ge_i64_e32 [[CMP:s\[[0-9]+:[0-9]+\]|vcc]]
; SI: s_or_b64 [[TMP:s\[[0-9]+:[0-9]+\]]], [[CMP]], [[COND_STATE]]
; SI: BB3_5: