1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 20:23:11 +01:00
llvm-mirror/lib/CodeGen/LiveInterval.cpp

430 lines
14 KiB
C++
Raw Normal View History

//===-- LiveInterval.cpp - Live Interval Representation -------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements the LiveRange and LiveInterval classes. Given some
// numbering of each the machine instructions an interval [i, j) is said to be a
// live interval for register v if there is no instruction with number j' > j
// such that v is live at j' abd there is no instruction with number i' < i such
// that v is live at i'. In this implementation intervals can have holes,
// i.e. an interval might look like [1,20), [50,65), [1000,1001). Each
// individual range is represented as an instance of LiveRange, and the whole
// interval is represented as an instance of LiveInterval.
//
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/LiveInterval.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/Target/MRegisterInfo.h"
#include <algorithm>
#include <iostream>
#include <map>
using namespace llvm;
// An example for liveAt():
//
// this = [1,4), liveAt(0) will return false. The instruction defining this
// spans slots [0,3]. The interval belongs to an spilled definition of the
// variable it represents. This is because slot 1 is used (def slot) and spans
// up to slot 3 (store slot).
//
bool LiveInterval::liveAt(unsigned I) const {
Ranges::const_iterator r = std::upper_bound(ranges.begin(), ranges.end(), I);
if (r == ranges.begin())
return false;
--r;
return r->contains(I);
}
// overlaps - Return true if the intersection of the two live intervals is
// not empty.
//
// An example for overlaps():
//
// 0: A = ...
// 4: B = ...
// 8: C = A + B ;; last use of A
//
// The live intervals should look like:
//
// A = [3, 11)
// B = [7, x)
// C = [11, y)
//
// A->overlaps(C) should return false since we want to be able to join
// A and C.
//
bool LiveInterval::overlapsFrom(const LiveInterval& other,
const_iterator StartPos) const {
const_iterator i = begin();
const_iterator ie = end();
const_iterator j = StartPos;
const_iterator je = other.end();
assert((StartPos->start <= i->start || StartPos == other.begin()) &&
StartPos != other.end() && "Bogus start position hint!");
This patch makes use of the infrastructure implemented before to safely and aggressively coallesce live ranges even if they overlap. Consider this LLVM code for example: int %test(int %X) { %Y = mul int %X, 1 ;; Codegens to Y = X %Z = add int %X, %Y ret int %Z } The mul is just there to get a copy into the code stream. This produces this machine code: (0x869e5a8, LLVM BB @0x869b9a0): %reg1024 = mov <fi#-2>, 1, %NOREG, 0 ;; "X" %reg1025 = mov %reg1024 ;; "Y" (subsumed by X) %reg1026 = add %reg1024, %reg1025 %EAX = mov %reg1026 ret Note that the life times of reg1024 and reg1025 overlap, even though they contain the same value. This results in this machine code: test: mov %EAX, DWORD PTR [%ESP + 4] mov %ECX, %EAX add %EAX, %ECX ret Another, worse case involves loops and PHI nodes. Consider this trivial loop: testcase: int %test2(int %X) { entry: br label %Loop Loop: %Y = phi int [%X, %entry], [%Z, %Loop] %Z = add int %Y, 1 %cond = seteq int %Z, 100 br bool %cond, label %Out, label %Loop Out: ret int %Z } Because of interactions between the PHI elimination pass and the register allocator, this got compiled to this code: test2: mov %ECX, DWORD PTR [%ESP + 4] .LBBtest2_1: *** mov %EAX, %ECX inc %EAX cmp %EAX, 100 *** mov %ECX, %EAX jne .LBBtest2_1 ret Or on powerpc, this code: _test2: mflr r0 stw r0, 8(r1) stwu r1, -60(r1) .LBB_test2_1: addi r2, r3, 1 cmpwi cr0, r2, 100 *** or r3, r2, r2 bne cr0, .LBB_test2_1 *** or r3, r2, r2 lwz r0, 68(r1) mtlr r0 addi r1, r1, 60 blr 0 With this improvement in place, we now generate this code for these two testcases, which is what we want: test: mov %EAX, DWORD PTR [%ESP + 4] add %EAX, %EAX ret test2: mov %EAX, DWORD PTR [%ESP + 4] .LBBtest2_1: inc %EAX cmp %EAX, 100 jne .LBBtest2_1 # Loop ret Or on PPC: _test2: mflr r0 stw r0, 8(r1) stwu r1, -60(r1) .LBB_test2_1: addi r3, r3, 1 cmpwi cr0, r3, 100 bne cr0, .LBB_test2_1 lwz r0, 68(r1) mtlr r0 addi r1, r1, 60 blr 0 Static numbers for spill code loads/stores/reg-reg copies (smaller is better): em3d: before: 47/25/26 after: 44/22/24 164.gzip: before: 433/245/310 after: 403/231/278 175.vpr: before: 3721/2189/1581 after: 4144/2081/1423 176.gcc: before: 26195/8866/9235 after: 25942/8082/8275 186.crafty: before: 4295/2587/3079 after: 4119/2519/2916 252.eon: before: 12754/7585/5803 after: 12508/7425/5643 256.bzip2: before: 463/226/315 after: 482:241/309 Runtime perf number samples on X86: gzip: before: 41.09 after: 39.86 bzip2: runtime: before: 56.71s after: 57.07s gcc: before: 6.16 after: 6.12 eon: before: 2.03s after: 2.00s llvm-svn: 15194
2004-07-25 09:11:19 +02:00
if (i->start < j->start) {
i = std::upper_bound(i, ie, j->start);
if (i != ranges.begin()) --i;
} else if (j->start < i->start) {
++StartPos;
if (StartPos != other.end() && StartPos->start <= i->start) {
assert(StartPos < other.end() && i < end());
j = std::upper_bound(j, je, i->start);
if (j != other.ranges.begin()) --j;
}
} else {
return true;
}
if (j == je) return false;
while (i != ie) {
if (i->start > j->start) {
std::swap(i, j);
std::swap(ie, je);
}
if (i->end > j->start)
return true;
++i;
}
return false;
}
/// NontrivialOverlap - Check to see if the two live ranges specified by i and j
/// overlap. If so, check to see if they have value numbers that are not
/// iIdx/jIdx respectively. If both conditions are true, return true.
static inline bool NontrivialOverlap(const LiveRange &I, const LiveRange &J,
unsigned iIdx, unsigned jIdx) {
if (I.start == J.start) {
// If this is not the allowed value merge, we cannot join.
if (I.ValId != iIdx || J.ValId != jIdx)
return true;
} else if (I.start < J.start) {
if (I.end > J.start && (I.ValId != iIdx || J.ValId != jIdx)) {
return true;
}
} else {
if (J.end > I.start && (I.ValId != iIdx || J.ValId != jIdx))
return true;
}
return false;
}
/// joinable - Two intervals are joinable if the either don't overlap at all
/// or if the destination of the copy is a single assignment value, and it
/// only overlaps with one value in the source interval.
bool LiveInterval::joinable(const LiveInterval &other, unsigned CopyIdx) const {
This patch makes use of the infrastructure implemented before to safely and aggressively coallesce live ranges even if they overlap. Consider this LLVM code for example: int %test(int %X) { %Y = mul int %X, 1 ;; Codegens to Y = X %Z = add int %X, %Y ret int %Z } The mul is just there to get a copy into the code stream. This produces this machine code: (0x869e5a8, LLVM BB @0x869b9a0): %reg1024 = mov <fi#-2>, 1, %NOREG, 0 ;; "X" %reg1025 = mov %reg1024 ;; "Y" (subsumed by X) %reg1026 = add %reg1024, %reg1025 %EAX = mov %reg1026 ret Note that the life times of reg1024 and reg1025 overlap, even though they contain the same value. This results in this machine code: test: mov %EAX, DWORD PTR [%ESP + 4] mov %ECX, %EAX add %EAX, %ECX ret Another, worse case involves loops and PHI nodes. Consider this trivial loop: testcase: int %test2(int %X) { entry: br label %Loop Loop: %Y = phi int [%X, %entry], [%Z, %Loop] %Z = add int %Y, 1 %cond = seteq int %Z, 100 br bool %cond, label %Out, label %Loop Out: ret int %Z } Because of interactions between the PHI elimination pass and the register allocator, this got compiled to this code: test2: mov %ECX, DWORD PTR [%ESP + 4] .LBBtest2_1: *** mov %EAX, %ECX inc %EAX cmp %EAX, 100 *** mov %ECX, %EAX jne .LBBtest2_1 ret Or on powerpc, this code: _test2: mflr r0 stw r0, 8(r1) stwu r1, -60(r1) .LBB_test2_1: addi r2, r3, 1 cmpwi cr0, r2, 100 *** or r3, r2, r2 bne cr0, .LBB_test2_1 *** or r3, r2, r2 lwz r0, 68(r1) mtlr r0 addi r1, r1, 60 blr 0 With this improvement in place, we now generate this code for these two testcases, which is what we want: test: mov %EAX, DWORD PTR [%ESP + 4] add %EAX, %EAX ret test2: mov %EAX, DWORD PTR [%ESP + 4] .LBBtest2_1: inc %EAX cmp %EAX, 100 jne .LBBtest2_1 # Loop ret Or on PPC: _test2: mflr r0 stw r0, 8(r1) stwu r1, -60(r1) .LBB_test2_1: addi r3, r3, 1 cmpwi cr0, r3, 100 bne cr0, .LBB_test2_1 lwz r0, 68(r1) mtlr r0 addi r1, r1, 60 blr 0 Static numbers for spill code loads/stores/reg-reg copies (smaller is better): em3d: before: 47/25/26 after: 44/22/24 164.gzip: before: 433/245/310 after: 403/231/278 175.vpr: before: 3721/2189/1581 after: 4144/2081/1423 176.gcc: before: 26195/8866/9235 after: 25942/8082/8275 186.crafty: before: 4295/2587/3079 after: 4119/2519/2916 252.eon: before: 12754/7585/5803 after: 12508/7425/5643 256.bzip2: before: 463/226/315 after: 482:241/309 Runtime perf number samples on X86: gzip: before: 41.09 after: 39.86 bzip2: runtime: before: 56.71s after: 57.07s gcc: before: 6.16 after: 6.12 eon: before: 2.03s after: 2.00s llvm-svn: 15194
2004-07-25 09:11:19 +02:00
const LiveRange *SourceLR = other.getLiveRangeContaining(CopyIdx-1);
const LiveRange *DestLR = getLiveRangeContaining(CopyIdx);
assert(SourceLR && DestLR && "Not joining due to a copy?");
unsigned OtherValIdx = SourceLR->ValId;
unsigned ThisValIdx = DestLR->ValId;
Ranges::const_iterator i = ranges.begin();
Ranges::const_iterator ie = ranges.end();
Ranges::const_iterator j = other.ranges.begin();
Ranges::const_iterator je = other.ranges.end();
if (i->start < j->start) {
i = std::upper_bound(i, ie, j->start);
if (i != ranges.begin()) --i;
} else if (j->start < i->start) {
j = std::upper_bound(j, je, i->start);
if (j != other.ranges.begin()) --j;
}
while (i != ie && j != je) {
if (NontrivialOverlap(*i, *j, ThisValIdx, OtherValIdx))
return false;
This patch makes use of the infrastructure implemented before to safely and aggressively coallesce live ranges even if they overlap. Consider this LLVM code for example: int %test(int %X) { %Y = mul int %X, 1 ;; Codegens to Y = X %Z = add int %X, %Y ret int %Z } The mul is just there to get a copy into the code stream. This produces this machine code: (0x869e5a8, LLVM BB @0x869b9a0): %reg1024 = mov <fi#-2>, 1, %NOREG, 0 ;; "X" %reg1025 = mov %reg1024 ;; "Y" (subsumed by X) %reg1026 = add %reg1024, %reg1025 %EAX = mov %reg1026 ret Note that the life times of reg1024 and reg1025 overlap, even though they contain the same value. This results in this machine code: test: mov %EAX, DWORD PTR [%ESP + 4] mov %ECX, %EAX add %EAX, %ECX ret Another, worse case involves loops and PHI nodes. Consider this trivial loop: testcase: int %test2(int %X) { entry: br label %Loop Loop: %Y = phi int [%X, %entry], [%Z, %Loop] %Z = add int %Y, 1 %cond = seteq int %Z, 100 br bool %cond, label %Out, label %Loop Out: ret int %Z } Because of interactions between the PHI elimination pass and the register allocator, this got compiled to this code: test2: mov %ECX, DWORD PTR [%ESP + 4] .LBBtest2_1: *** mov %EAX, %ECX inc %EAX cmp %EAX, 100 *** mov %ECX, %EAX jne .LBBtest2_1 ret Or on powerpc, this code: _test2: mflr r0 stw r0, 8(r1) stwu r1, -60(r1) .LBB_test2_1: addi r2, r3, 1 cmpwi cr0, r2, 100 *** or r3, r2, r2 bne cr0, .LBB_test2_1 *** or r3, r2, r2 lwz r0, 68(r1) mtlr r0 addi r1, r1, 60 blr 0 With this improvement in place, we now generate this code for these two testcases, which is what we want: test: mov %EAX, DWORD PTR [%ESP + 4] add %EAX, %EAX ret test2: mov %EAX, DWORD PTR [%ESP + 4] .LBBtest2_1: inc %EAX cmp %EAX, 100 jne .LBBtest2_1 # Loop ret Or on PPC: _test2: mflr r0 stw r0, 8(r1) stwu r1, -60(r1) .LBB_test2_1: addi r3, r3, 1 cmpwi cr0, r3, 100 bne cr0, .LBB_test2_1 lwz r0, 68(r1) mtlr r0 addi r1, r1, 60 blr 0 Static numbers for spill code loads/stores/reg-reg copies (smaller is better): em3d: before: 47/25/26 after: 44/22/24 164.gzip: before: 433/245/310 after: 403/231/278 175.vpr: before: 3721/2189/1581 after: 4144/2081/1423 176.gcc: before: 26195/8866/9235 after: 25942/8082/8275 186.crafty: before: 4295/2587/3079 after: 4119/2519/2916 252.eon: before: 12754/7585/5803 after: 12508/7425/5643 256.bzip2: before: 463/226/315 after: 482:241/309 Runtime perf number samples on X86: gzip: before: 41.09 after: 39.86 bzip2: runtime: before: 56.71s after: 57.07s gcc: before: 6.16 after: 6.12 eon: before: 2.03s after: 2.00s llvm-svn: 15194
2004-07-25 09:11:19 +02:00
if (i->end < j->end)
++i;
else
++j;
}
return true;
}
/// getOverlapingRanges - Given another live interval which is defined as a
/// copy from this one, return a list of all of the live ranges where the
/// two overlap and have different value numbers.
void LiveInterval::getOverlapingRanges(const LiveInterval &other,
unsigned CopyIdx,
std::vector<LiveRange*> &Ranges) {
const LiveRange *SourceLR = getLiveRangeContaining(CopyIdx-1);
const LiveRange *DestLR = other.getLiveRangeContaining(CopyIdx);
assert(SourceLR && DestLR && "Not joining due to a copy?");
unsigned OtherValIdx = SourceLR->ValId;
unsigned ThisValIdx = DestLR->ValId;
Ranges::iterator i = ranges.begin();
Ranges::iterator ie = ranges.end();
Ranges::const_iterator j = other.ranges.begin();
Ranges::const_iterator je = other.ranges.end();
if (i->start < j->start) {
i = std::upper_bound(i, ie, j->start);
if (i != ranges.begin()) --i;
} else if (j->start < i->start) {
j = std::upper_bound(j, je, i->start);
if (j != other.ranges.begin()) --j;
}
while (i != ie && j != je) {
if (NontrivialOverlap(*i, *j, ThisValIdx, OtherValIdx))
Ranges.push_back(&*i);
if (i->end < j->end)
++i;
else
++j;
}
}
/// extendIntervalEndTo - This method is used when we want to extend the range
/// specified by I to end at the specified endpoint. To do this, we should
/// merge and eliminate all ranges that this will overlap with. The iterator is
/// not invalidated.
void LiveInterval::extendIntervalEndTo(Ranges::iterator I, unsigned NewEnd) {
assert(I != ranges.end() && "Not a valid interval!");
unsigned ValId = I->ValId;
// Search for the first interval that we can't merge with.
Ranges::iterator MergeTo = next(I);
for (; MergeTo != ranges.end() && NewEnd >= MergeTo->end; ++MergeTo) {
assert(MergeTo->ValId == ValId && "Cannot merge with differing values!");
}
// If NewEnd was in the middle of an interval, make sure to get its endpoint.
I->end = std::max(NewEnd, prior(MergeTo)->end);
// Erase any dead ranges.
ranges.erase(next(I), MergeTo);
// If the newly formed range now touches the range after it and if they have
// the same value number, merge the two ranges into one range.
Ranges::iterator Next = next(I);
if (Next != ranges.end() && Next->start <= I->end && Next->ValId == ValId) {
I->end = Next->end;
ranges.erase(Next);
}
}
/// extendIntervalStartTo - This method is used when we want to extend the range
/// specified by I to start at the specified endpoint. To do this, we should
/// merge and eliminate all ranges that this will overlap with.
LiveInterval::Ranges::iterator
LiveInterval::extendIntervalStartTo(Ranges::iterator I, unsigned NewStart) {
assert(I != ranges.end() && "Not a valid interval!");
unsigned ValId = I->ValId;
// Search for the first interval that we can't merge with.
Ranges::iterator MergeTo = I;
do {
if (MergeTo == ranges.begin()) {
I->start = NewStart;
ranges.erase(MergeTo, I);
return I;
}
assert(MergeTo->ValId == ValId && "Cannot merge with differing values!");
--MergeTo;
} while (NewStart <= MergeTo->start);
// If we start in the middle of another interval, just delete a range and
// extend that interval.
if (MergeTo->end >= NewStart && MergeTo->ValId == ValId) {
MergeTo->end = I->end;
} else {
// Otherwise, extend the interval right after.
++MergeTo;
MergeTo->start = NewStart;
MergeTo->end = I->end;
}
ranges.erase(next(MergeTo), next(I));
return MergeTo;
}
LiveInterval::Ranges::iterator
LiveInterval::addRangeFrom(LiveRange LR, Ranges::iterator From) {
unsigned Start = LR.start, End = LR.end;
Ranges::iterator it = std::upper_bound(From, ranges.end(), Start);
// If the inserted interval starts in the middle or right at the end of
// another interval, just extend that interval to contain the range of LR.
if (it != ranges.begin()) {
Ranges::iterator B = prior(it);
if (LR.ValId == B->ValId) {
if (B->start <= Start && B->end >= Start) {
extendIntervalEndTo(B, End);
return B;
}
} else {
// Check to make sure that we are not overlapping two live ranges with
// different ValId's.
assert(B->end <= Start &&
"Cannot overlap two LiveRanges with differing ValID's"
" (did you def the same reg twice in a MachineInstr?)");
}
}
// Otherwise, if this range ends in the middle of, or right next to, another
// interval, merge it into that interval.
if (it != ranges.end())
if (LR.ValId == it->ValId) {
if (it->start <= End) {
it = extendIntervalStartTo(it, Start);
// If LR is a complete superset of an interval, we may need to grow its
// endpoint as well.
if (End > it->end)
extendIntervalEndTo(it, End);
return it;
}
} else {
// Check to make sure that we are not overlapping two live ranges with
// different ValId's.
assert(it->start >= End &&
"Cannot overlap two LiveRanges with differing ValID's");
}
// Otherwise, this is just a new range that doesn't interact with anything.
// Insert it.
return ranges.insert(it, LR);
}
/// removeRange - Remove the specified range from this interval. Note that
/// the range must already be in this interval in its entirety.
void LiveInterval::removeRange(unsigned Start, unsigned End) {
// Find the LiveRange containing this span.
Ranges::iterator I = std::upper_bound(ranges.begin(), ranges.end(), Start);
assert(I != ranges.begin() && "Range is not in interval!");
--I;
assert(I->contains(Start) && I->contains(End-1) &&
"Range is not entirely in interval!");
// If the span we are removing is at the start of the LiveRange, adjust it.
if (I->start == Start) {
if (I->end == End)
ranges.erase(I); // Removed the whole LiveRange.
else
I->start = End;
return;
}
// Otherwise if the span we are removing is at the end of the LiveRange,
// adjust the other way.
if (I->end == End) {
I->end = Start;
return;
}
// Otherwise, we are splitting the LiveRange into two pieces.
unsigned OldEnd = I->end;
I->end = Start; // Trim the old interval.
// Insert the new one.
ranges.insert(next(I), LiveRange(End, OldEnd, I->ValId));
}
/// getLiveRangeContaining - Return the live range that contains the
/// specified index, or null if there is none.
const LiveRange *LiveInterval::getLiveRangeContaining(unsigned Idx) const {
Ranges::const_iterator It = std::upper_bound(ranges.begin(),ranges.end(),Idx);
if (It != ranges.begin()) {
const LiveRange &LR = *prior(It);
if (LR.contains(Idx))
return &LR;
}
return 0;
}
/// join - Join two live intervals (this, and other) together. This operation
/// is the result of a copy instruction in the source program, that occurs at
/// index 'CopyIdx' that copies from 'Other' to 'this'.
void LiveInterval::join(LiveInterval &Other, unsigned CopyIdx) {
const LiveRange *SourceLR = Other.getLiveRangeContaining(CopyIdx-1);
const LiveRange *DestLR = getLiveRangeContaining(CopyIdx);
assert(SourceLR && DestLR && "Not joining due to a copy?");
unsigned MergedSrcValIdx = SourceLR->ValId;
unsigned MergedDstValIdx = DestLR->ValId;
// Try to do the least amount of work possible. In particular, if there are
// more liverange chunks in the other set than there are in the 'this' set,
// swap sets to merge the fewest chunks in possible.
if (Other.ranges.size() > ranges.size()) {
std::swap(MergedSrcValIdx, MergedDstValIdx);
std::swap(ranges, Other.ranges);
std::swap(NumValues, Other.NumValues);
}
// Join the ranges of other into the ranges of this interval.
Ranges::iterator InsertPos = ranges.begin();
std::map<unsigned, unsigned> Dst2SrcIdxMap;
for (Ranges::iterator I = Other.ranges.begin(),
E = Other.ranges.end(); I != E; ++I) {
// Map the ValId in the other live range to the current live range.
if (I->ValId == MergedSrcValIdx)
I->ValId = MergedDstValIdx;
else {
unsigned &NV = Dst2SrcIdxMap[I->ValId];
if (NV == 0) NV = getNextValue();
I->ValId = NV;
}
InsertPos = addRangeFrom(*I, InsertPos);
}
weight += Other.weight;
}
std::ostream& llvm::operator<<(std::ostream& os, const LiveRange &LR) {
return os << '[' << LR.start << ',' << LR.end << ':' << LR.ValId << ")";
}
void LiveRange::dump() const {
std::cerr << *this << "\n";
}
void LiveInterval::print(std::ostream &OS, const MRegisterInfo *MRI) const {
if (MRI && MRegisterInfo::isPhysicalRegister(reg))
OS << MRI->getName(reg);
else
OS << "%reg" << reg;
OS << ',' << weight;
if (empty())
OS << "EMPTY";
else {
OS << " = ";
for (LiveInterval::Ranges::const_iterator I = ranges.begin(),
E = ranges.end(); I != E; ++I)
OS << *I;
}
}
void LiveInterval::dump() const {
std::cerr << *this << "\n";
}