summaryrefslogtreecommitdiff
path: root/compiler/optimizing
diff options
context:
space:
mode:
Diffstat (limited to 'compiler/optimizing')
-rw-r--r--compiler/optimizing/bounds_check_elimination.cc58
-rw-r--r--compiler/optimizing/builder.cc293
-rw-r--r--compiler/optimizing/builder.h23
-rw-r--r--compiler/optimizing/code_generator_arm.cc27
-rw-r--r--compiler/optimizing/code_generator_arm64.cc32
-rw-r--r--compiler/optimizing/code_generator_mips64.cc33
-rw-r--r--compiler/optimizing/code_generator_x86.cc32
-rw-r--r--compiler/optimizing/code_generator_x86_64.cc32
-rw-r--r--compiler/optimizing/dead_code_elimination.cc75
-rw-r--r--compiler/optimizing/graph_checker.cc16
-rw-r--r--compiler/optimizing/graph_checker.h1
-rw-r--r--compiler/optimizing/induction_var_analysis.cc201
-rw-r--r--compiler/optimizing/induction_var_analysis.h15
-rw-r--r--compiler/optimizing/induction_var_analysis_test.cc51
-rw-r--r--compiler/optimizing/induction_var_range.cc316
-rw-r--r--compiler/optimizing/induction_var_range.h57
-rw-r--r--compiler/optimizing/induction_var_range_test.cc109
-rw-r--r--compiler/optimizing/nodes.cc160
-rw-r--r--compiler/optimizing/nodes.h50
-rw-r--r--compiler/optimizing/register_allocator.cc6
-rw-r--r--compiler/optimizing/ssa_builder.cc30
21 files changed, 976 insertions, 641 deletions
diff --git a/compiler/optimizing/bounds_check_elimination.cc b/compiler/optimizing/bounds_check_elimination.cc
index 62f5b9aa52..42b3541912 100644
--- a/compiler/optimizing/bounds_check_elimination.cc
+++ b/compiler/optimizing/bounds_check_elimination.cc
@@ -14,8 +14,11 @@
* limitations under the License.
*/
-#include "base/arena_containers.h"
#include "bounds_check_elimination.h"
+
+#include <limits>
+
+#include "base/arena_containers.h"
#include "induction_var_range.h"
#include "nodes.h"
@@ -48,11 +51,11 @@ class ValueBound : public ValueObject {
if (right == 0) {
return false;
}
- if ((right > 0) && (left <= INT_MAX - right)) {
+ if ((right > 0) && (left <= (std::numeric_limits<int32_t>::max() - right))) {
// No overflow.
return false;
}
- if ((right < 0) && (left >= INT_MIN - right)) {
+ if ((right < 0) && (left >= (std::numeric_limits<int32_t>::min() - right))) {
// No underflow.
return false;
}
@@ -120,8 +123,8 @@ class ValueBound : public ValueObject {
return instruction_ == nullptr;
}
- static ValueBound Min() { return ValueBound(nullptr, INT_MIN); }
- static ValueBound Max() { return ValueBound(nullptr, INT_MAX); }
+ static ValueBound Min() { return ValueBound(nullptr, std::numeric_limits<int32_t>::min()); }
+ static ValueBound Max() { return ValueBound(nullptr, std::numeric_limits<int32_t>::max()); }
bool Equals(ValueBound bound) const {
return instruction_ == bound.instruction_ && constant_ == bound.constant_;
@@ -213,7 +216,7 @@ class ValueBound : public ValueObject {
int32_t new_constant;
if (c > 0) {
- if (constant_ > INT_MAX - c) {
+ if (constant_ > (std::numeric_limits<int32_t>::max() - c)) {
*overflow = true;
return Max();
}
@@ -227,7 +230,7 @@ class ValueBound : public ValueObject {
*overflow = true;
return Max();
} else {
- if (constant_ < INT_MIN - c) {
+ if (constant_ < (std::numeric_limits<int32_t>::min() - c)) {
*underflow = true;
return Min();
}
@@ -256,8 +259,8 @@ class ArrayAccessInsideLoopFinder : public ValueObject {
explicit ArrayAccessInsideLoopFinder(HInstruction* induction_variable)
: induction_variable_(induction_variable),
found_array_length_(nullptr),
- offset_low_(INT_MAX),
- offset_high_(INT_MIN) {
+ offset_low_(std::numeric_limits<int32_t>::max()),
+ offset_high_(std::numeric_limits<int32_t>::min()) {
Run();
}
@@ -492,7 +495,7 @@ class MonotonicValueRange : public ValueRange {
HInstruction* initial,
int32_t increment,
ValueBound bound)
- // To be conservative, give it full range [INT_MIN, INT_MAX] in case it's
+ // To be conservative, give it full range [Min(), Max()] in case it's
// used as a regular value range, due to possible overflow/underflow.
: ValueRange(allocator, ValueBound::Min(), ValueBound::Max()),
induction_variable_(induction_variable),
@@ -554,19 +557,19 @@ class MonotonicValueRange : public ValueRange {
if (increment_ > 0) {
// Monotonically increasing.
ValueBound lower = ValueBound::NarrowLowerBound(bound_, range->GetLower());
- if (!lower.IsConstant() || lower.GetConstant() == INT_MIN) {
+ if (!lower.IsConstant() || lower.GetConstant() == std::numeric_limits<int32_t>::min()) {
// Lower bound isn't useful. Leave it to deoptimization.
return this;
}
- // We currently conservatively assume max array length is INT_MAX. If we can
- // make assumptions about the max array length, e.g. due to the max heap size,
+ // We currently conservatively assume max array length is Max().
+ // If we can make assumptions about the max array length, e.g. due to the max heap size,
// divided by the element size (such as 4 bytes for each integer array), we can
// lower this number and rule out some possible overflows.
- int32_t max_array_len = INT_MAX;
+ int32_t max_array_len = std::numeric_limits<int32_t>::max();
// max possible integer value of range's upper value.
- int32_t upper = INT_MAX;
+ int32_t upper = std::numeric_limits<int32_t>::max();
// Try to lower upper.
ValueBound upper_bound = range->GetUpper();
if (upper_bound.IsConstant()) {
@@ -593,7 +596,7 @@ class MonotonicValueRange : public ValueRange {
((int64_t)upper - (int64_t)initial_constant) / increment_ * increment_;
}
}
- if (last_num_in_sequence <= INT_MAX - increment_) {
+ if (last_num_in_sequence <= (std::numeric_limits<int32_t>::max() - increment_)) {
// No overflow. The sequence will be stopped by the upper bound test as expected.
return new (GetAllocator()) ValueRange(GetAllocator(), lower, range->GetUpper());
}
@@ -604,7 +607,7 @@ class MonotonicValueRange : public ValueRange {
DCHECK_NE(increment_, 0);
// Monotonically decreasing.
ValueBound upper = ValueBound::NarrowUpperBound(bound_, range->GetUpper());
- if ((!upper.IsConstant() || upper.GetConstant() == INT_MAX) &&
+ if ((!upper.IsConstant() || upper.GetConstant() == std::numeric_limits<int32_t>::max()) &&
!upper.IsRelatedToArrayLength()) {
// Upper bound isn't useful. Leave it to deoptimization.
return this;
@@ -614,7 +617,7 @@ class MonotonicValueRange : public ValueRange {
// for common cases.
if (range->GetLower().IsConstant()) {
int32_t constant = range->GetLower().GetConstant();
- if (constant >= INT_MIN - increment_) {
+ if (constant >= (std::numeric_limits<int32_t>::min() - increment_)) {
return new (GetAllocator()) ValueRange(GetAllocator(), range->GetLower(), upper);
}
}
@@ -1099,7 +1102,8 @@ class BCEVisitor : public HGraphVisitor {
// Very large constant index is considered as an anomaly. This is a threshold
// beyond which we don't bother to apply the deoptimization technique since
// it's likely some AIOOBE will be thrown.
- static constexpr int32_t kMaxConstantForAddingDeoptimize = INT_MAX - 1024 * 1024;
+ static constexpr int32_t kMaxConstantForAddingDeoptimize =
+ std::numeric_limits<int32_t>::max() - 1024 * 1024;
// Added blocks for loop body entry test.
bool IsAddedBlock(HBasicBlock* block) const {
@@ -1165,8 +1169,8 @@ class BCEVisitor : public HGraphVisitor {
ValueRange* LookupInductionRange(HInstruction* context, HInstruction* instruction) {
InductionVarRange::Value v1 = induction_range_.GetMinInduction(context, instruction);
InductionVarRange::Value v2 = induction_range_.GetMaxInduction(context, instruction);
- if ((v1.a_constant == 0 || v1.a_constant == 1) && v1.b_constant != INT_MIN &&
- (v2.a_constant == 0 || v2.a_constant == 1) && v2.b_constant != INT_MAX) {
+ if (v1.is_known && (v1.a_constant == 0 || v1.a_constant == 1) &&
+ v2.is_known && (v2.a_constant == 0 || v2.a_constant == 1)) {
DCHECK(v1.a_constant == 1 || v1.instruction == nullptr);
DCHECK(v2.a_constant == 1 || v2.instruction == nullptr);
ValueBound low = ValueBound(v1.instruction, v1.b_constant);
@@ -1467,8 +1471,8 @@ class BCEVisitor : public HGraphVisitor {
// Once we have an array access like 'array[5] = 1', we record array.length >= 6.
// We currently don't do it for non-constant index since a valid array[i] can't prove
// a valid array[i-1] yet due to the lower bound side.
- if (constant == INT_MAX) {
- // INT_MAX as an index will definitely throw AIOOBE.
+ if (constant == std::numeric_limits<int32_t>::max()) {
+ // Max() as an index will definitely throw AIOOBE.
return;
}
ValueBound lower = ValueBound(nullptr, constant + 1);
@@ -1690,8 +1694,8 @@ class BCEVisitor : public HGraphVisitor {
// The value of left input of instruction equals (left + c).
// (array_length + 1) or smaller divided by two or more
- // always generate a value in [INT_MIN, array_length].
- // This is true even if array_length is INT_MAX.
+ // always generate a value in [Min(), array_length].
+ // This is true even if array_length is Max().
if (left->IsArrayLength() && c <= 1) {
if (instruction->IsUShr() && c < 0) {
// Make sure for unsigned shift, left side is not negative.
@@ -1701,7 +1705,7 @@ class BCEVisitor : public HGraphVisitor {
}
ValueRange* range = new (GetGraph()->GetArena()) ValueRange(
GetGraph()->GetArena(),
- ValueBound(nullptr, INT_MIN),
+ ValueBound(nullptr, std::numeric_limits<int32_t>::min()),
ValueBound(left, 0));
GetValueRangeMap(instruction->GetBlock())->Overwrite(instruction->GetId(), range);
}
@@ -1811,7 +1815,7 @@ class BCEVisitor : public HGraphVisitor {
continue;
}
HIntConstant* lower_bound_const_instr = nullptr;
- int32_t lower_bound_const = INT_MIN;
+ int32_t lower_bound_const = std::numeric_limits<int32_t>::min();
size_t counter = 0;
// Count the constant indexing for which bounds checks haven't
// been removed yet.
diff --git a/compiler/optimizing/builder.cc b/compiler/optimizing/builder.cc
index f2bde899d2..e19e74f37a 100644
--- a/compiler/optimizing/builder.cc
+++ b/compiler/optimizing/builder.cc
@@ -262,22 +262,6 @@ bool HGraphBuilder::SkipCompilation(const DexFile::CodeItem& code_item,
return false;
}
-static const DexFile::TryItem* GetTryItem(HBasicBlock* block,
- const DexFile::CodeItem& code_item,
- const ArenaBitVector& can_block_throw) {
- DCHECK(!block->IsSingleTryBoundary());
-
- // Block does not contain throwing instructions. Even if it is covered by
- // a TryItem, we will consider it not in a try block.
- if (!can_block_throw.IsBitSet(block->GetBlockId())) {
- return nullptr;
- }
-
- // Instructions in the block may throw. Find a TryItem covering this block.
- int32_t try_item_idx = DexFile::FindTryItem(code_item, block->GetDexPc());
- return (try_item_idx == -1) ? nullptr : DexFile::GetTryItems(code_item, try_item_idx);
-}
-
void HGraphBuilder::CreateBlocksForTryCatch(const DexFile::CodeItem& code_item) {
if (code_item.tries_size_ == 0) {
return;
@@ -316,18 +300,18 @@ void HGraphBuilder::CreateBlocksForTryCatch(const DexFile::CodeItem& code_item)
}
}
-void HGraphBuilder::SplitTryBoundaryEdge(HBasicBlock* predecessor,
- HBasicBlock* successor,
- HTryBoundary::BoundaryKind kind,
- const DexFile::CodeItem& code_item,
- const DexFile::TryItem& try_item) {
- // Split the edge with a single TryBoundary instruction.
- HTryBoundary* try_boundary = new (arena_) HTryBoundary(kind, successor->GetDexPc());
- HBasicBlock* try_entry_block = graph_->SplitEdge(predecessor, successor);
- try_entry_block->AddInstruction(try_boundary);
-
- // Link the TryBoundary to the handlers of `try_item`.
- for (CatchHandlerIterator it(code_item, try_item); it.HasNext(); it.Next()) {
+// Returns the TryItem stored for `block` or nullptr if there is no info for it.
+static const DexFile::TryItem* GetTryItem(
+ HBasicBlock* block,
+ const ArenaSafeMap<uint32_t, const DexFile::TryItem*>& try_block_info) {
+ auto iterator = try_block_info.find(block->GetBlockId());
+ return (iterator == try_block_info.end()) ? nullptr : iterator->second;
+}
+
+void HGraphBuilder::LinkToCatchBlocks(HTryBoundary* try_boundary,
+ const DexFile::CodeItem& code_item,
+ const DexFile::TryItem* try_item) {
+ for (CatchHandlerIterator it(code_item, *try_item); it.HasNext(); it.Next()) {
try_boundary->AddExceptionHandler(FindBlockStartingAt(it.GetHandlerAddress()));
}
}
@@ -337,132 +321,103 @@ void HGraphBuilder::InsertTryBoundaryBlocks(const DexFile::CodeItem& code_item)
return;
}
- // Bit vector stores information on which blocks contain throwing instructions.
- // Must be expandable because catch blocks may be split into two.
- ArenaBitVector can_block_throw(arena_, graph_->GetBlocks().size(), /* expandable */ true);
+ // Keep a map of all try blocks and their respective TryItems. We do not use
+ // the block's pointer but rather its id to ensure deterministic iteration.
+ ArenaSafeMap<uint32_t, const DexFile::TryItem*> try_block_info(
+ std::less<uint32_t>(), arena_->Adapter());
+
+ // Obtain TryItem information for blocks with throwing instructions, and split
+ // blocks which are both try & catch to simplify the graph.
+ // NOTE: We are appending new blocks inside the loop, so we need to use index
+ // because iterators can be invalidated. We remember the initial size to avoid
+ // iterating over the new blocks which cannot throw.
+ for (size_t i = 0, e = graph_->GetBlocks().size(); i < e; ++i) {
+ HBasicBlock* block = graph_->GetBlocks()[i];
+
+ // Do not bother creating exceptional edges for try blocks which have no
+ // throwing instructions. In that case we simply assume that the block is
+ // not covered by a TryItem. This prevents us from creating a throw-catch
+ // loop for synchronized blocks.
+ if (block->HasThrowingInstructions()) {
+ // Try to find a TryItem covering the block.
+ DCHECK_NE(block->GetDexPc(), kNoDexPc) << "Block must have a dec_pc to find its TryItem.";
+ const int32_t try_item_idx = DexFile::FindTryItem(code_item, block->GetDexPc());
+ if (try_item_idx != -1) {
+ // Block throwing and in a TryItem. Store the try block information.
+ HBasicBlock* throwing_block = block;
+ if (block->IsCatchBlock()) {
+ // Simplify blocks which are both try and catch, otherwise we would
+ // need a strategy for splitting exceptional edges. We split the block
+ // after the move-exception (if present) and mark the first part not
+ // throwing. The normal-flow edge between them will be split later.
+ HInstruction* first_insn = block->GetFirstInstruction();
+ if (first_insn->IsLoadException()) {
+ // Catch block starts with a LoadException. Split the block after
+ // the StoreLocal and ClearException which must come after the load.
+ DCHECK(first_insn->GetNext()->IsStoreLocal());
+ DCHECK(first_insn->GetNext()->GetNext()->IsClearException());
+ throwing_block = block->SplitBefore(first_insn->GetNext()->GetNext()->GetNext());
+ } else {
+ // Catch block does not load the exception. Split at the beginning
+ // to create an empty catch block.
+ throwing_block = block->SplitBefore(first_insn);
+ }
+ }
- // Scan blocks and mark those which contain throwing instructions.
- // NOTE: We're appending new blocks inside the loop, so we need to use index because iterators
- // can be invalidated. We remember the initial size to avoid iterating over the new blocks.
- for (size_t block_id = 0u, end = graph_->GetBlocks().size(); block_id != end; ++block_id) {
- HBasicBlock* block = graph_->GetBlocks()[block_id];
- bool can_throw = false;
- for (HInstructionIterator insn(block->GetInstructions()); !insn.Done(); insn.Advance()) {
- if (insn.Current()->CanThrow()) {
- can_throw = true;
- break;
+ try_block_info.Put(throwing_block->GetBlockId(),
+ DexFile::GetTryItems(code_item, try_item_idx));
}
}
+ }
- if (can_throw) {
- if (block->IsCatchBlock()) {
- // Catch blocks are always considered an entry point into the TryItem in
- // order to avoid splitting exceptional edges. We split the block after
- // the move-exception (if present) and mark the first part non-throwing.
- // Later on, a TryBoundary will be inserted between the two blocks.
- HInstruction* first_insn = block->GetFirstInstruction();
- if (first_insn->IsLoadException()) {
- // Catch block starts with a LoadException. Split the block after the
- // StoreLocal and ClearException which must come after the load.
- DCHECK(first_insn->GetNext()->IsStoreLocal());
- DCHECK(first_insn->GetNext()->GetNext()->IsClearException());
- block = block->SplitBefore(first_insn->GetNext()->GetNext()->GetNext());
- } else {
- // Catch block does not load the exception. Split at the beginning to
- // create an empty catch block.
- block = block->SplitBefore(first_insn);
- }
+ // Do a pass over the try blocks and insert entering TryBoundaries where at
+ // least one predecessor is not covered by the same TryItem as the try block.
+ // We do not split each edge separately, but rather create one boundary block
+ // that all predecessors are relinked to. This preserves loop headers (b/23895756).
+ for (auto entry : try_block_info) {
+ HBasicBlock* try_block = graph_->GetBlock(entry.first);
+ for (HBasicBlock* predecessor : try_block->GetPredecessors()) {
+ if (GetTryItem(predecessor, try_block_info) != entry.second) {
+ // Found a predecessor not covered by the same TryItem. Insert entering
+ // boundary block.
+ HTryBoundary* try_entry =
+ new (arena_) HTryBoundary(HTryBoundary::kEntry, try_block->GetDexPc());
+ try_block->CreateImmediateDominator()->AddInstruction(try_entry);
+ LinkToCatchBlocks(try_entry, code_item, entry.second);
+ break;
}
- can_block_throw.SetBit(block->GetBlockId());
- }
- }
-
- // Iterate over all blocks, find those covered by some TryItem and:
- // (a) split edges which enter/exit the try range,
- // (b) create TryBoundary instructions in the new blocks,
- // (c) link the new blocks to corresponding exception handlers.
- // We cannot iterate only over blocks in `branch_targets_` because switch-case
- // blocks share the same dex_pc.
- // NOTE: We're appending new blocks inside the loop, so we need to use index because iterators
- // can be invalidated. We remember the initial size to avoid iterating over the new blocks.
- for (size_t block_id = 0u, end = graph_->GetBlocks().size(); block_id != end; ++block_id) {
- HBasicBlock* try_block = graph_->GetBlocks()[block_id];
- // TryBoundary blocks are added at the end of the list and not iterated over.
- DCHECK(!try_block->IsSingleTryBoundary());
-
- // Find the TryItem for this block.
- const DexFile::TryItem* try_item = GetTryItem(try_block, code_item, can_block_throw);
- if (try_item == nullptr) {
- continue;
- }
-
- // Catch blocks were split earlier and cannot throw.
- DCHECK(!try_block->IsCatchBlock());
-
- // Find predecessors which are not covered by the same TryItem range. Such
- // edges enter the try block and will have a TryBoundary inserted.
- for (size_t i = 0; i < try_block->GetPredecessors().size(); ++i) {
- HBasicBlock* predecessor = try_block->GetPredecessor(i);
- if (predecessor->IsSingleTryBoundary()) {
- // The edge was already split because of an exit from a neighbouring
- // TryItem. We split it again and insert an entry point.
- if (kIsDebugBuild) {
- HTryBoundary* last_insn = predecessor->GetLastInstruction()->AsTryBoundary();
- const DexFile::TryItem* predecessor_try_item =
- GetTryItem(predecessor->GetSinglePredecessor(), code_item, can_block_throw);
- DCHECK(!last_insn->IsEntry());
- DCHECK_EQ(last_insn->GetNormalFlowSuccessor(), try_block);
- DCHECK(try_block->IsFirstIndexOfPredecessor(predecessor, i));
- DCHECK_NE(try_item, predecessor_try_item);
- }
- } else if (GetTryItem(predecessor, code_item, can_block_throw) != try_item) {
- // This is an entry point into the TryItem and the edge has not been
- // split yet. That means that `predecessor` is not in a TryItem, or
- // it is in a different TryItem and we happened to iterate over this
- // block first. We split the edge and insert an entry point.
- } else {
- // Not an edge on the boundary of the try block.
+ }
+ }
+
+ // Do a second pass over the try blocks and insert exit TryBoundaries where
+ // the successor is not in the same TryItem.
+ for (auto entry : try_block_info) {
+ HBasicBlock* try_block = graph_->GetBlock(entry.first);
+ // NOTE: Do not use iterators because SplitEdge would invalidate them.
+ for (size_t i = 0, e = try_block->GetSuccessors().size(); i < e; ++i) {
+ HBasicBlock* successor = try_block->GetSuccessor(i);
+
+ // If the successor is a try block, all of its predecessors must be
+ // covered by the same TryItem. Otherwise the previous pass would have
+ // created a non-throwing boundary block.
+ if (GetTryItem(successor, try_block_info) != nullptr) {
+ DCHECK_EQ(entry.second, GetTryItem(successor, try_block_info));
continue;
}
- SplitTryBoundaryEdge(predecessor, try_block, HTryBoundary::kEntry, code_item, *try_item);
- }
-
- // Find successors which are not covered by the same TryItem range. Such
- // edges exit the try block and will have a TryBoundary inserted.
- for (HBasicBlock* successor : try_block->GetSuccessors()) {
- if (successor->IsCatchBlock()) {
- // A catch block is always considered an entry point into its TryItem.
- // We therefore assume this is an exit point, regardless of whether
- // the catch block is in a different TryItem or not.
- } else if (successor->IsSingleTryBoundary()) {
- // The edge was already split because of an entry into a neighbouring
- // TryItem. We split it again and insert an exit.
- if (kIsDebugBuild) {
- HTryBoundary* last_insn = successor->GetLastInstruction()->AsTryBoundary();
- const DexFile::TryItem* successor_try_item =
- GetTryItem(last_insn->GetNormalFlowSuccessor(), code_item, can_block_throw);
- DCHECK_EQ(try_block, successor->GetSinglePredecessor());
- DCHECK(last_insn->IsEntry());
- DCHECK_NE(try_item, successor_try_item);
- }
- } else if (GetTryItem(successor, code_item, can_block_throw) != try_item) {
- // This is an exit out of the TryItem and the edge has not been split
- // yet. That means that either `successor` is not in a TryItem, or it
- // is in a different TryItem and we happened to iterate over this
- // block first. We split the edge and insert an exit.
- HInstruction* last_instruction = try_block->GetLastInstruction();
- if (last_instruction->IsReturn() || last_instruction->IsReturnVoid()) {
- DCHECK_EQ(successor, exit_block_);
- // Control flow exits the try block with a Return(Void). Because
- // splitting the edge would invalidate the invariant that Return
- // always jumps to Exit, we move the Return outside the try block.
- successor = try_block->SplitBefore(last_instruction);
- }
- } else {
- // Not an edge on the boundary of the try block.
- continue;
+
+ // Preserve the invariant that Return(Void) always jumps to Exit by moving
+ // it outside the try block if necessary.
+ HInstruction* last_instruction = try_block->GetLastInstruction();
+ if (last_instruction->IsReturn() || last_instruction->IsReturnVoid()) {
+ DCHECK_EQ(successor, exit_block_);
+ successor = try_block->SplitBefore(last_instruction);
}
- SplitTryBoundaryEdge(try_block, successor, HTryBoundary::kExit, code_item, *try_item);
+
+ // Insert TryBoundary and link to catch blocks.
+ HTryBoundary* try_exit =
+ new (arena_) HTryBoundary(HTryBoundary::kExit, successor->GetDexPc());
+ graph_->SplitEdge(try_block, successor)->AddInstruction(try_exit);
+ LinkToCatchBlocks(try_exit, code_item, entry.second);
}
}
}
@@ -1685,6 +1640,34 @@ bool HGraphBuilder::NeedsAccessCheck(uint32_t type_index) const {
dex_compilation_unit_->GetDexMethodIndex(), *dex_file_, type_index);
}
+void HGraphBuilder::BuildSwitchJumpTable(const SwitchTable& table,
+ const Instruction& instruction,
+ HInstruction* value,
+ uint32_t dex_pc) {
+ // Add the successor blocks to the current block.
+ uint16_t num_entries = table.GetNumEntries();
+ for (size_t i = 1; i <= num_entries; i++) {
+ int32_t target_offset = table.GetEntryAt(i);
+ HBasicBlock* case_target = FindBlockStartingAt(dex_pc + target_offset);
+ DCHECK(case_target != nullptr);
+
+ // Add the target block as a successor.
+ current_block_->AddSuccessor(case_target);
+ }
+
+ // Add the default target block as the last successor.
+ HBasicBlock* default_target = FindBlockStartingAt(dex_pc + instruction.SizeInCodeUnits());
+ DCHECK(default_target != nullptr);
+ current_block_->AddSuccessor(default_target);
+
+ // Now add the Switch instruction.
+ int32_t starting_key = table.GetEntryAt(0);
+ current_block_->AddInstruction(
+ new (arena_) HPackedSwitch(starting_key, num_entries, value, dex_pc));
+ // This block ends with control flow.
+ current_block_ = nullptr;
+}
+
void HGraphBuilder::BuildPackedSwitch(const Instruction& instruction, uint32_t dex_pc) {
// Verifier guarantees that the payload for PackedSwitch contains:
// (a) number of entries (may be zero)
@@ -1695,18 +1678,24 @@ void HGraphBuilder::BuildPackedSwitch(const Instruction& instruction, uint32_t d
// Value to test against.
HInstruction* value = LoadLocal(instruction.VRegA(), Primitive::kPrimInt, dex_pc);
+ // Starting key value.
+ int32_t starting_key = table.GetEntryAt(0);
+
// Retrieve number of entries.
uint16_t num_entries = table.GetNumEntries();
if (num_entries == 0) {
return;
}
- // Chained cmp-and-branch, starting from starting_key.
- int32_t starting_key = table.GetEntryAt(0);
-
- for (size_t i = 1; i <= num_entries; i++) {
- BuildSwitchCaseHelper(instruction, i, i == num_entries, table, value, starting_key + i - 1,
- table.GetEntryAt(i), dex_pc);
+ // Don't use a packed switch if there are very few entries.
+ if (num_entries > kSmallSwitchThreshold) {
+ BuildSwitchJumpTable(table, instruction, value, dex_pc);
+ } else {
+ // Chained cmp-and-branch, starting from starting_key.
+ for (size_t i = 1; i <= num_entries; i++) {
+ BuildSwitchCaseHelper(instruction, i, i == num_entries, table, value,
+ starting_key + i - 1, table.GetEntryAt(i), dex_pc);
+ }
}
}
diff --git a/compiler/optimizing/builder.h b/compiler/optimizing/builder.h
index 8d40d69a32..4c8e3d0442 100644
--- a/compiler/optimizing/builder.h
+++ b/compiler/optimizing/builder.h
@@ -90,6 +90,9 @@ class HGraphBuilder : public ValueObject {
static constexpr const char* kBuilderPassName = "builder";
+ // The number of entries in a packed switch before we use a jump table.
+ static constexpr uint16_t kSmallSwitchThreshold = 5;
+
private:
// Analyzes the dex instruction and adds HInstruction to the graph
// to execute that instruction. Returns whether the instruction can
@@ -118,13 +121,13 @@ class HGraphBuilder : public ValueObject {
// instructions and links them to the corresponding catch blocks.
void InsertTryBoundaryBlocks(const DexFile::CodeItem& code_item);
- // Splits a single edge, inserting a TryBoundary of given `kind` and linking
- // it to exception handlers of `try_item`.
- void SplitTryBoundaryEdge(HBasicBlock* predecessor,
- HBasicBlock* successor,
- HTryBoundary::BoundaryKind kind,
- const DexFile::CodeItem& code_item,
- const DexFile::TryItem& try_item);
+ // Iterates over the exception handlers of `try_item`, finds the corresponding
+ // catch blocks and makes them successors of `try_boundary`. The order of
+ // successors matches the order in which runtime exception delivery searches
+ // for a handler.
+ void LinkToCatchBlocks(HTryBoundary* try_boundary,
+ const DexFile::CodeItem& code_item,
+ const DexFile::TryItem* try_item);
bool CanDecodeQuickenedInfo() const;
uint16_t LookupQuickenedInfo(uint32_t dex_pc);
@@ -239,6 +242,12 @@ class HGraphBuilder : public ValueObject {
// Builds an instruction sequence for a packed switch statement.
void BuildPackedSwitch(const Instruction& instruction, uint32_t dex_pc);
+ // Build a switch instruction from a packed switch statement.
+ void BuildSwitchJumpTable(const SwitchTable& table,
+ const Instruction& instruction,
+ HInstruction* value,
+ uint32_t dex_pc);
+
// Builds an instruction sequence for a sparse switch statement.
void BuildSparseSwitch(const Instruction& instruction, uint32_t dex_pc);
diff --git a/compiler/optimizing/code_generator_arm.cc b/compiler/optimizing/code_generator_arm.cc
index 411e05f443..d7b1d24887 100644
--- a/compiler/optimizing/code_generator_arm.cc
+++ b/compiler/optimizing/code_generator_arm.cc
@@ -4953,6 +4953,33 @@ void InstructionCodeGeneratorARM::VisitFakeString(HFakeString* instruction ATTRI
// Will be generated at use site.
}
+// Simple implementation of packed switch - generate cascaded compare/jumps.
+void LocationsBuilderARM::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ LocationSummary* locations =
+ new (GetGraph()->GetArena()) LocationSummary(switch_instr, LocationSummary::kNoCall);
+ locations->SetInAt(0, Location::RequiresRegister());
+}
+
+void InstructionCodeGeneratorARM::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ int32_t lower_bound = switch_instr->GetStartValue();
+ int32_t num_entries = switch_instr->GetNumEntries();
+ LocationSummary* locations = switch_instr->GetLocations();
+ Register value_reg = locations->InAt(0).AsRegister<Register>();
+ HBasicBlock* default_block = switch_instr->GetDefaultBlock();
+
+ // Create a series of compare/jumps.
+ const ArenaVector<HBasicBlock*>& successors = switch_instr->GetBlock()->GetSuccessors();
+ for (int32_t i = 0; i < num_entries; i++) {
+ GenerateCompareWithImmediate(value_reg, lower_bound + i);
+ __ b(codegen_->GetLabelOf(successors.at(i)), EQ);
+ }
+
+ // And the default for any other value.
+ if (!codegen_->GoesToNextBlock(switch_instr->GetBlock(), default_block)) {
+ __ b(codegen_->GetLabelOf(default_block));
+ }
+}
+
void CodeGeneratorARM::MoveFromReturnRegister(Location trg, Primitive::Type type) {
if (!trg.IsValid()) {
DCHECK(type == Primitive::kPrimVoid);
diff --git a/compiler/optimizing/code_generator_arm64.cc b/compiler/optimizing/code_generator_arm64.cc
index 8e1260e821..d175532f4c 100644
--- a/compiler/optimizing/code_generator_arm64.cc
+++ b/compiler/optimizing/code_generator_arm64.cc
@@ -3540,6 +3540,38 @@ void InstructionCodeGeneratorARM64::VisitFakeString(HFakeString* instruction ATT
// Will be generated at use site.
}
+// Simple implementation of packed switch - generate cascaded compare/jumps.
+void LocationsBuilderARM64::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ LocationSummary* locations =
+ new (GetGraph()->GetArena()) LocationSummary(switch_instr, LocationSummary::kNoCall);
+ locations->SetInAt(0, Location::RequiresRegister());
+}
+
+void InstructionCodeGeneratorARM64::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ int32_t lower_bound = switch_instr->GetStartValue();
+ int32_t num_entries = switch_instr->GetNumEntries();
+ Register value_reg = InputRegisterAt(switch_instr, 0);
+ HBasicBlock* default_block = switch_instr->GetDefaultBlock();
+
+ // Create a series of compare/jumps.
+ const ArenaVector<HBasicBlock*>& successors = switch_instr->GetBlock()->GetSuccessors();
+ for (int32_t i = 0; i < num_entries; i++) {
+ int32_t case_value = lower_bound + i;
+ vixl::Label* succ = codegen_->GetLabelOf(successors.at(i));
+ if (case_value == 0) {
+ __ Cbz(value_reg, succ);
+ } else {
+ __ Cmp(value_reg, vixl::Operand(case_value));
+ __ B(eq, succ);
+ }
+ }
+
+ // And the default for any other value.
+ if (!codegen_->GoesToNextBlock(switch_instr->GetBlock(), default_block)) {
+ __ B(codegen_->GetLabelOf(default_block));
+ }
+}
+
#undef __
#undef QUICK_ENTRY_POINT
diff --git a/compiler/optimizing/code_generator_mips64.cc b/compiler/optimizing/code_generator_mips64.cc
index f4f53d5f32..8fdd56e0bc 100644
--- a/compiler/optimizing/code_generator_mips64.cc
+++ b/compiler/optimizing/code_generator_mips64.cc
@@ -3364,5 +3364,38 @@ void InstructionCodeGeneratorMIPS64::VisitFakeString(HFakeString* instruction AT
// Will be generated at use site.
}
+// Simple implementation of packed switch - generate cascaded compare/jumps.
+void LocationsBuilderMIPS64::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ LocationSummary* locations =
+ new (GetGraph()->GetArena()) LocationSummary(switch_instr, LocationSummary::kNoCall);
+ locations->SetInAt(0, Location::RequiresRegister());
+}
+
+void InstructionCodeGeneratorMIPS64::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ int32_t lower_bound = switch_instr->GetStartValue();
+ int32_t num_entries = switch_instr->GetNumEntries();
+ LocationSummary* locations = switch_instr->GetLocations();
+ GpuRegister value_reg = locations->InAt(0).AsRegister<GpuRegister>();
+ HBasicBlock* default_block = switch_instr->GetDefaultBlock();
+
+ // Create a series of compare/jumps.
+ const ArenaVector<HBasicBlock*>& successors = switch_instr->GetBlock()->GetSuccessors();
+ for (int32_t i = 0; i < num_entries; i++) {
+ int32_t case_value = lower_bound + i;
+ Label* succ = codegen_->GetLabelOf(successors.at(i));
+ if (case_value == 0) {
+ __ Beqzc(value_reg, succ);
+ } else {
+ __ LoadConst32(TMP, case_value);
+ __ Beqc(value_reg, TMP, succ);
+ }
+ }
+
+ // And the default for any other value.
+ if (!codegen_->GoesToNextBlock(switch_instr->GetBlock(), default_block)) {
+ __ B(codegen_->GetLabelOf(default_block));
+ }
+}
+
} // namespace mips64
} // namespace art
diff --git a/compiler/optimizing/code_generator_x86.cc b/compiler/optimizing/code_generator_x86.cc
index 86cdbdde8c..ab3d1d1924 100644
--- a/compiler/optimizing/code_generator_x86.cc
+++ b/compiler/optimizing/code_generator_x86.cc
@@ -5488,6 +5488,38 @@ void InstructionCodeGeneratorX86::VisitFakeString(HFakeString* instruction ATTRI
// Will be generated at use site.
}
+// Simple implementation of packed switch - generate cascaded compare/jumps.
+void LocationsBuilderX86::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ LocationSummary* locations =
+ new (GetGraph()->GetArena()) LocationSummary(switch_instr, LocationSummary::kNoCall);
+ locations->SetInAt(0, Location::RequiresRegister());
+}
+
+void InstructionCodeGeneratorX86::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ int32_t lower_bound = switch_instr->GetStartValue();
+ int32_t num_entries = switch_instr->GetNumEntries();
+ LocationSummary* locations = switch_instr->GetLocations();
+ Register value_reg = locations->InAt(0).AsRegister<Register>();
+ HBasicBlock* default_block = switch_instr->GetDefaultBlock();
+
+ // Create a series of compare/jumps.
+ const ArenaVector<HBasicBlock*>& successors = switch_instr->GetBlock()->GetSuccessors();
+ for (int i = 0; i < num_entries; i++) {
+ int32_t case_value = lower_bound + i;
+ if (case_value == 0) {
+ __ testl(value_reg, value_reg);
+ } else {
+ __ cmpl(value_reg, Immediate(case_value));
+ }
+ __ j(kEqual, codegen_->GetLabelOf(successors.at(i)));
+ }
+
+ // And the default for any other value.
+ if (!codegen_->GoesToNextBlock(switch_instr->GetBlock(), default_block)) {
+ __ jmp(codegen_->GetLabelOf(default_block));
+ }
+}
+
void LocationsBuilderX86::VisitX86ComputeBaseMethodAddress(
HX86ComputeBaseMethodAddress* insn) {
LocationSummary* locations =
diff --git a/compiler/optimizing/code_generator_x86_64.cc b/compiler/optimizing/code_generator_x86_64.cc
index b78b017e23..cfce7a0faa 100644
--- a/compiler/optimizing/code_generator_x86_64.cc
+++ b/compiler/optimizing/code_generator_x86_64.cc
@@ -5203,6 +5203,38 @@ void InstructionCodeGeneratorX86_64::VisitFakeString(HFakeString* instruction AT
// Will be generated at use site.
}
+// Simple implementation of packed switch - generate cascaded compare/jumps.
+void LocationsBuilderX86_64::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ LocationSummary* locations =
+ new (GetGraph()->GetArena()) LocationSummary(switch_instr, LocationSummary::kNoCall);
+ locations->SetInAt(0, Location::RequiresRegister());
+}
+
+void InstructionCodeGeneratorX86_64::VisitPackedSwitch(HPackedSwitch* switch_instr) {
+ int32_t lower_bound = switch_instr->GetStartValue();
+ int32_t num_entries = switch_instr->GetNumEntries();
+ LocationSummary* locations = switch_instr->GetLocations();
+ CpuRegister value_reg = locations->InAt(0).AsRegister<CpuRegister>();
+ HBasicBlock* default_block = switch_instr->GetDefaultBlock();
+
+ // Create a series of compare/jumps.
+ const ArenaVector<HBasicBlock*>& successors = switch_instr->GetBlock()->GetSuccessors();
+ for (int i = 0; i < num_entries; i++) {
+ int32_t case_value = lower_bound + i;
+ if (case_value == 0) {
+ __ testl(value_reg, value_reg);
+ } else {
+ __ cmpl(value_reg, Immediate(case_value));
+ }
+ __ j(kEqual, codegen_->GetLabelOf(successors.at(i)));
+ }
+
+ // And the default for any other value.
+ if (!codegen_->GoesToNextBlock(switch_instr->GetBlock(), default_block)) {
+ __ jmp(codegen_->GetLabelOf(default_block));
+ }
+}
+
void CodeGeneratorX86_64::Load64BitValue(CpuRegister dest, int64_t value) {
if (value == 0) {
__ xorl(dest, dest);
diff --git a/compiler/optimizing/dead_code_elimination.cc b/compiler/optimizing/dead_code_elimination.cc
index 7d509a22a6..b322759a6c 100644
--- a/compiler/optimizing/dead_code_elimination.cc
+++ b/compiler/optimizing/dead_code_elimination.cc
@@ -16,34 +16,63 @@
#include "dead_code_elimination.h"
+#include "utils/array_ref.h"
#include "base/bit_vector-inl.h"
#include "ssa_phi_elimination.h"
namespace art {
-static void MarkReachableBlocks(HBasicBlock* block, ArenaBitVector* visited) {
- int block_id = block->GetBlockId();
- if (visited->IsBitSet(block_id)) {
- return;
- }
- visited->SetBit(block_id);
-
- HInstruction* last_instruction = block->GetLastInstruction();
- if (last_instruction->IsIf()) {
- HIf* if_instruction = last_instruction->AsIf();
- HInstruction* condition = if_instruction->InputAt(0);
- if (!condition->IsIntConstant()) {
- MarkReachableBlocks(if_instruction->IfTrueSuccessor(), visited);
- MarkReachableBlocks(if_instruction->IfFalseSuccessor(), visited);
- } else if (condition->AsIntConstant()->IsOne()) {
- MarkReachableBlocks(if_instruction->IfTrueSuccessor(), visited);
- } else {
- DCHECK(condition->AsIntConstant()->IsZero());
- MarkReachableBlocks(if_instruction->IfFalseSuccessor(), visited);
+static void MarkReachableBlocks(HGraph* graph, ArenaBitVector* visited) {
+ ArenaVector<HBasicBlock*> worklist(graph->GetArena()->Adapter());
+ constexpr size_t kDefaultWorlistSize = 8;
+ worklist.reserve(kDefaultWorlistSize);
+ visited->SetBit(graph->GetEntryBlock()->GetBlockId());
+ worklist.push_back(graph->GetEntryBlock());
+
+ while (!worklist.empty()) {
+ HBasicBlock* block = worklist.back();
+ worklist.pop_back();
+ int block_id = block->GetBlockId();
+ DCHECK(visited->IsBitSet(block_id));
+
+ ArrayRef<HBasicBlock* const> live_successors(block->GetSuccessors());
+ HInstruction* last_instruction = block->GetLastInstruction();
+ if (last_instruction->IsIf()) {
+ HIf* if_instruction = last_instruction->AsIf();
+ HInstruction* condition = if_instruction->InputAt(0);
+ if (condition->IsIntConstant()) {
+ if (condition->AsIntConstant()->IsOne()) {
+ live_successors = live_successors.SubArray(0u, 1u);
+ DCHECK_EQ(live_successors[0], if_instruction->IfTrueSuccessor());
+ } else {
+ DCHECK(condition->AsIntConstant()->IsZero());
+ live_successors = live_successors.SubArray(1u, 1u);
+ DCHECK_EQ(live_successors[0], if_instruction->IfFalseSuccessor());
+ }
+ }
+ } else if (last_instruction->IsPackedSwitch()) {
+ HPackedSwitch* switch_instruction = last_instruction->AsPackedSwitch();
+ HInstruction* switch_input = switch_instruction->InputAt(0);
+ if (switch_input->IsIntConstant()) {
+ int32_t switch_value = switch_input->AsIntConstant()->GetValue();
+ int32_t start_value = switch_instruction->GetStartValue();
+ uint32_t switch_index = static_cast<uint32_t>(switch_value - start_value);
+ if (switch_index < switch_instruction->GetNumEntries()) {
+ live_successors = live_successors.SubArray(switch_index, 1u);
+ DCHECK_EQ(live_successors[0], block->GetSuccessor(switch_index));
+ } else {
+ live_successors = live_successors.SubArray(switch_instruction->GetNumEntries(), 1u);
+ DCHECK_EQ(live_successors[0], switch_instruction->GetDefaultBlock());
+ }
+ }
}
- } else {
- for (HBasicBlock* successor : block->GetSuccessors()) {
- MarkReachableBlocks(successor, visited);
+
+ for (HBasicBlock* successor : live_successors) {
+ // Add only those successors that have not been visited yet.
+ if (!visited->IsBitSet(successor->GetBlockId())) {
+ visited->SetBit(successor->GetBlockId());
+ worklist.push_back(successor);
+ }
}
}
}
@@ -67,7 +96,7 @@ void HDeadCodeElimination::RemoveDeadBlocks() {
ArenaBitVector live_blocks(allocator, graph_->GetBlocks().size(), false);
ArenaBitVector affected_loops(allocator, graph_->GetBlocks().size(), false);
- MarkReachableBlocks(graph_->GetEntryBlock(), &live_blocks);
+ MarkReachableBlocks(graph_, &live_blocks);
bool removed_one_or_more_blocks = false;
// Remove all dead blocks. Iterate in post order because removal needs the
diff --git a/compiler/optimizing/graph_checker.cc b/compiler/optimizing/graph_checker.cc
index 583da30438..4e1cafee66 100644
--- a/compiler/optimizing/graph_checker.cc
+++ b/compiler/optimizing/graph_checker.cc
@@ -743,6 +743,22 @@ void SSAChecker::HandleBooleanInput(HInstruction* instruction, size_t input_inde
}
}
+void SSAChecker::VisitPackedSwitch(HPackedSwitch* instruction) {
+ VisitInstruction(instruction);
+ // Check that the number of block successors matches the switch count plus
+ // one for the default block.
+ HBasicBlock* block = instruction->GetBlock();
+ if (instruction->GetNumEntries() + 1u != block->GetSuccessors().size()) {
+ AddError(StringPrintf(
+ "%s instruction %d in block %d expects %u successors to the block, but found: %zu.",
+ instruction->DebugName(),
+ instruction->GetId(),
+ block->GetBlockId(),
+ instruction->GetNumEntries() + 1u,
+ block->GetSuccessors().size()));
+ }
+}
+
void SSAChecker::VisitIf(HIf* instruction) {
VisitInstruction(instruction);
HandleBooleanInput(instruction, 0);
diff --git a/compiler/optimizing/graph_checker.h b/compiler/optimizing/graph_checker.h
index 0e270dbe18..7ddffc136a 100644
--- a/compiler/optimizing/graph_checker.h
+++ b/compiler/optimizing/graph_checker.h
@@ -125,6 +125,7 @@ class SSAChecker : public GraphChecker {
void VisitBinaryOperation(HBinaryOperation* op) OVERRIDE;
void VisitCondition(HCondition* op) OVERRIDE;
void VisitIf(HIf* instruction) OVERRIDE;
+ void VisitPackedSwitch(HPackedSwitch* instruction) OVERRIDE;
void VisitBooleanNot(HBooleanNot* instruction) OVERRIDE;
void VisitConstant(HConstant* instruction) OVERRIDE;
diff --git a/compiler/optimizing/induction_var_analysis.cc b/compiler/optimizing/induction_var_analysis.cc
index 92c732c0c3..9fb4304450 100644
--- a/compiler/optimizing/induction_var_analysis.cc
+++ b/compiler/optimizing/induction_var_analysis.cc
@@ -33,17 +33,6 @@ static bool IsLoopInvariant(HLoopInformation* loop, HInstruction* instruction) {
}
/**
- * Returns true if instruction is proper entry-phi-operation for given loop
- * (referred to as mu-operation in Gerlek's paper).
- */
-static bool IsEntryPhi(HLoopInformation* loop, HInstruction* instruction) {
- return
- instruction->IsPhi() &&
- instruction->InputCount() == 2 &&
- instruction->GetBlock() == loop->GetHeader();
-}
-
-/**
* Since graph traversal may enter a SCC at any position, an initial representation may be rotated,
* along dependences, viz. any of (a, b, c, d), (d, a, b, c) (c, d, a, b), (b, c, d, a) assuming
* a chain of dependences (mutual independent items may occur in arbitrary order). For proper
@@ -58,8 +47,9 @@ static void RotateEntryPhiFirst(HLoopInformation* loop,
size_t phi_pos = -1;
const size_t size = scc->size();
for (size_t i = 0; i < size; i++) {
- if (IsEntryPhi(loop, scc->at(i)) && (phi == nullptr || phis.FoundBefore(scc->at(i), phi))) {
- phi = scc->at(i);
+ HInstruction* other = scc->at(i);
+ if (other->IsLoopHeaderPhi() && (phi == nullptr || phis.FoundBefore(other, phi))) {
+ phi = other;
phi_pos = i;
}
}
@@ -168,7 +158,7 @@ void HInductionVarAnalysis::VisitNode(HLoopInformation* loop, HInstruction* inst
}
// Classify the SCC.
- if (scc_.size() == 1 && !IsEntryPhi(loop, scc_[0])) {
+ if (scc_.size() == 1 && !scc_[0]->IsLoopHeaderPhi()) {
ClassifyTrivial(loop, scc_[0]);
} else {
ClassifyNonTrivial(loop);
@@ -200,10 +190,7 @@ uint32_t HInductionVarAnalysis::VisitDescendant(HLoopInformation* loop, HInstruc
void HInductionVarAnalysis::ClassifyTrivial(HLoopInformation* loop, HInstruction* instruction) {
InductionInfo* info = nullptr;
if (instruction->IsPhi()) {
- for (size_t i = 1, count = instruction->InputCount(); i < count; i++) {
- info = TransferPhi(LookupInfo(loop, instruction->InputAt(0)),
- LookupInfo(loop, instruction->InputAt(i)));
- }
+ info = TransferPhi(loop, instruction, /* input_index */ 0);
} else if (instruction->IsAdd()) {
info = TransferAddSub(LookupInfo(loop, instruction->InputAt(0)),
LookupInfo(loop, instruction->InputAt(1)), kAdd);
@@ -245,21 +232,21 @@ void HInductionVarAnalysis::ClassifyNonTrivial(HLoopInformation* loop) {
RotateEntryPhiFirst(loop, &scc_, &other);
}
- // Analyze from phi onwards.
+ // Analyze from entry-phi onwards.
HInstruction* phi = scc_[0];
- if (!IsEntryPhi(loop, phi)) {
+ if (!phi->IsLoopHeaderPhi()) {
return;
}
- HInstruction* external = phi->InputAt(0);
- HInstruction* internal = phi->InputAt(1);
- InductionInfo* initial = LookupInfo(loop, external);
+
+ // External link should be loop invariant.
+ InductionInfo* initial = LookupInfo(loop, phi->InputAt(0));
if (initial == nullptr || initial->induction_class != kInvariant) {
return;
}
- // Singleton entry-phi-operation may be a wrap-around induction.
+ // Singleton is wrap-around induction if all internal links have the same meaning.
if (size == 1) {
- InductionInfo* update = LookupInfo(loop, internal);
+ InductionInfo* update = TransferPhi(loop, phi, /* input_index */ 1);
if (update != nullptr) {
AssignInfo(loop, phi, CreateInduction(kWrapAround, initial, update));
}
@@ -272,7 +259,7 @@ void HInductionVarAnalysis::ClassifyNonTrivial(HLoopInformation* loop) {
HInstruction* instruction = scc_[i];
InductionInfo* update = nullptr;
if (instruction->IsPhi()) {
- update = SolvePhi(loop, phi, instruction);
+ update = SolvePhiAllInputs(loop, phi, instruction);
} else if (instruction->IsAdd()) {
update = SolveAddSub(
loop, phi, instruction, instruction->InputAt(0), instruction->InputAt(1), kAdd, true);
@@ -286,10 +273,9 @@ void HInductionVarAnalysis::ClassifyNonTrivial(HLoopInformation* loop) {
cycle_.Put(instruction, update);
}
- // Success if the internal link received a meaning.
- auto it = cycle_.find(internal);
- if (it != cycle_.end()) {
- InductionInfo* induction = it->second;
+ // Success if all internal links received the same temporary meaning.
+ InductionInfo* induction = SolvePhi(phi, /* input_index */ 1);
+ if (induction != nullptr) {
switch (induction->induction_class) {
case kInvariant:
// Classify first phi and then the rest of the cycle "on-demand".
@@ -329,13 +315,20 @@ HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::RotatePeriodicInduc
return CreateInduction(kPeriodic, induction->op_a, RotatePeriodicInduction(induction->op_b, last));
}
-HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::TransferPhi(InductionInfo* a,
- InductionInfo* b) {
- // Transfer over a phi: if both inputs are identical, result is input.
- if (InductionEqual(a, b)) {
- return a;
- }
- return nullptr;
+HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::TransferPhi(HLoopInformation* loop,
+ HInstruction* phi,
+ size_t input_index) {
+ // Match all phi inputs from input_index onwards exactly.
+ const size_t count = phi->InputCount();
+ DCHECK_LT(input_index, count);
+ InductionInfo* a = LookupInfo(loop, phi->InputAt(input_index));
+ for (size_t i = input_index + 1; i < count; i++) {
+ InductionInfo* b = LookupInfo(loop, phi->InputAt(i));
+ if (!InductionEqual(a, b)) {
+ return nullptr;
+ }
+ }
+ return a;
}
HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::TransferAddSub(InductionInfo* a,
@@ -421,47 +414,56 @@ HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::TransferNeg(Inducti
return nullptr;
}
-HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::SolvePhi(HLoopInformation* loop,
- HInstruction* phi,
- HInstruction* instruction) {
- // Solve within a cycle over a phi: identical inputs are combined into that input as result.
- const size_t count = instruction->InputCount();
- DCHECK_GT(count, 0u);
- auto ita = cycle_.find(instruction->InputAt(0));
+HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::SolvePhi(HInstruction* phi,
+ size_t input_index) {
+ // Match all phi inputs from input_index onwards exactly.
+ const size_t count = phi->InputCount();
+ DCHECK_LT(input_index, count);
+ auto ita = cycle_.find(phi->InputAt(input_index));
if (ita != cycle_.end()) {
- InductionInfo* a = ita->second;
- for (size_t i = 1; i < count; i++) {
- auto itb = cycle_.find(instruction->InputAt(i));
- if (itb == cycle_.end() || !HInductionVarAnalysis::InductionEqual(a, itb->second)) {
+ for (size_t i = input_index + 1; i < count; i++) {
+ auto itb = cycle_.find(phi->InputAt(i));
+ if (itb == cycle_.end() ||
+ !HInductionVarAnalysis::InductionEqual(ita->second, itb->second)) {
return nullptr;
}
}
- return a;
+ return ita->second;
}
+ return nullptr;
+}
- // Solve within a cycle over another entry-phi: add invariants into a periodic.
- if (IsEntryPhi(loop, instruction)) {
- InductionInfo* a = LookupInfo(loop, instruction->InputAt(0));
+HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::SolvePhiAllInputs(
+ HLoopInformation* loop,
+ HInstruction* entry_phi,
+ HInstruction* phi) {
+ // Match all phi inputs.
+ InductionInfo* match = SolvePhi(phi, /* input_index */ 0);
+ if (match != nullptr) {
+ return match;
+ }
+
+ // Otherwise, try to solve for a periodic seeded from phi onward.
+ // Only tight multi-statement cycles are considered in order to
+ // simplify rotating the periodic during the final classification.
+ if (phi->IsLoopHeaderPhi() && phi->InputCount() == 2) {
+ InductionInfo* a = LookupInfo(loop, phi->InputAt(0));
if (a != nullptr && a->induction_class == kInvariant) {
- if (instruction->InputAt(1) == phi) {
- InductionInfo* initial = LookupInfo(loop, phi->InputAt(0));
+ if (phi->InputAt(1) == entry_phi) {
+ InductionInfo* initial = LookupInfo(loop, entry_phi->InputAt(0));
return CreateInduction(kPeriodic, a, initial);
}
- auto it = cycle_.find(instruction->InputAt(1));
- if (it != cycle_.end()) {
- InductionInfo* b = it->second;
- if (b->induction_class == kPeriodic) {
- return CreateInduction(kPeriodic, a, b);
- }
+ InductionInfo* b = SolvePhi(phi, /* input_index */ 1);
+ if (b != nullptr && b->induction_class == kPeriodic) {
+ return CreateInduction(kPeriodic, a, b);
}
}
}
-
return nullptr;
}
HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::SolveAddSub(HLoopInformation* loop,
- HInstruction* phi,
+ HInstruction* entry_phi,
HInstruction* instruction,
HInstruction* x,
HInstruction* y,
@@ -471,7 +473,7 @@ HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::SolveAddSub(HLoopIn
// invariant value, seeded from phi, keeps adding to the stride of the induction.
InductionInfo* b = LookupInfo(loop, y);
if (b != nullptr && b->induction_class == kInvariant) {
- if (x == phi) {
+ if (x == entry_phi) {
return (op == kAdd) ? b : CreateInvariantOp(kNeg, nullptr, b);
}
auto it = cycle_.find(x);
@@ -487,14 +489,15 @@ HInductionVarAnalysis::InductionInfo* HInductionVarAnalysis::SolveAddSub(HLoopIn
if (op == kAdd) {
// Try the other way around for an addition if considered for first time.
if (is_first_call) {
- return SolveAddSub(loop, phi, instruction, y, x, op, false);
+ return SolveAddSub(loop, entry_phi, instruction, y, x, op, false);
}
} else if (op == kSub) {
- // Solve within a tight cycle for a periodic idiom k = c - k;
- if (y == phi && instruction == phi->InputAt(1)) {
+ // Solve within a tight cycle that is formed by exactly two instructions,
+ // one phi and one update, for a periodic idiom of the form k = c - k;
+ if (y == entry_phi && entry_phi->InputCount() == 2 && instruction == entry_phi->InputAt(1)) {
InductionInfo* a = LookupInfo(loop, x);
if (a != nullptr && a->induction_class == kInvariant) {
- InductionInfo* initial = LookupInfo(loop, phi->InputAt(0));
+ InductionInfo* initial = LookupInfo(loop, entry_phi->InputAt(0));
return CreateInduction(kPeriodic, CreateInvariantOp(kSub, a, initial), initial);
}
}
@@ -539,32 +542,47 @@ void HInductionVarAnalysis::VisitCondition(HLoopInformation* loop,
Primitive::Type type,
IfCondition cmp) {
if (a->induction_class == kInvariant && b->induction_class == kLinear) {
- // Swap conditions (e.g. U > i is same as i < U).
+ // Swap condition if induction is at right-hand-side (e.g. U > i is same as i < U).
switch (cmp) {
case kCondLT: VisitCondition(loop, b, a, type, kCondGT); break;
case kCondLE: VisitCondition(loop, b, a, type, kCondGE); break;
case kCondGT: VisitCondition(loop, b, a, type, kCondLT); break;
case kCondGE: VisitCondition(loop, b, a, type, kCondLE); break;
+ case kCondNE: VisitCondition(loop, b, a, type, kCondNE); break;
default: break;
}
} else if (a->induction_class == kLinear && b->induction_class == kInvariant) {
- // Normalize a linear loop control with a constant, nonzero stride:
- // stride > 0, either i < U or i <= U
- // stride < 0, either i > U or i >= U
+ // Analyze condition with induction at left-hand-side (e.g. i < U).
InductionInfo* stride = a->op_a;
InductionInfo* lo_val = a->op_b;
InductionInfo* hi_val = b;
- // Analyze the stride thoroughly, since its representation may be compound at this point.
- InductionVarRange::Value v1 = InductionVarRange::GetMin(stride, nullptr);
- InductionVarRange::Value v2 = InductionVarRange::GetMax(stride, nullptr);
- if (v1.a_constant == 0 && v2.a_constant == 0 && v1.b_constant == v2.b_constant) {
- const int32_t stride_value = v1.b_constant;
- if ((stride_value > 0 && (cmp == kCondLT || cmp == kCondLE)) ||
- (stride_value < 0 && (cmp == kCondGT || cmp == kCondGE))) {
- bool is_strict = cmp == kCondLT || cmp == kCondGT;
- VisitTripCount(loop, lo_val, hi_val, stride, stride_value, type, is_strict);
+ // Analyze stride (may be compound).
+ InductionVarRange::Value v1 = InductionVarRange::GetVal(stride, nullptr, /* is_min */ true);
+ InductionVarRange::Value v2 = InductionVarRange::GetVal(stride, nullptr, /* is_min */ false);
+ if (v1.a_constant != 0 || v2.a_constant != 0 || v1.b_constant != v2.b_constant) {
+ return;
+ }
+ // Rewrite safe condition i != U with unit stride into i < U or i > U
+ // (unit stride guarantees that the end condition is always reached).
+ const int32_t stride_value = v1.b_constant;
+ int64_t lo_value = 0;
+ int64_t hi_value = 0;
+ if (cmp == kCondNE && IsIntAndGet(lo_val, &lo_value) && IsIntAndGet(hi_val, &hi_value)) {
+ if ((stride_value == +1 && lo_value < hi_value) ||
+ (stride_value == -1 && lo_value > hi_value)) {
+ cmp = stride_value > 0 ? kCondLT : kCondGT;
}
}
+ // Normalize a linear loop control with a nonzero stride:
+ // stride > 0, either i < U or i <= U
+ // stride < 0, either i > U or i >= U
+ //
+ // TODO: construct conditions for constant/symbolic safety of trip-count
+ //
+ if ((stride_value > 0 && (cmp == kCondLT || cmp == kCondLE)) ||
+ (stride_value < 0 && (cmp == kCondGT || cmp == kCondGE))) {
+ VisitTripCount(loop, lo_val, hi_val, stride, stride_value, type, cmp);
+ }
}
}
@@ -574,7 +592,7 @@ void HInductionVarAnalysis::VisitTripCount(HLoopInformation* loop,
InductionInfo* stride,
int32_t stride_value,
Primitive::Type type,
- bool is_strict) {
+ IfCondition cmp) {
// Any loop of the general form:
//
// for (i = L; i <= U; i += S) // S > 0
@@ -586,26 +604,27 @@ void HInductionVarAnalysis::VisitTripCount(HLoopInformation* loop,
// for (n = 0; n < TC; n++) // where TC = (U + S - L) / S
// .. L + S * n ..
//
- // NOTE: The TC (trip-count) expression is only valid if the top-test path is taken at
- // least once. Otherwise TC is 0. Also, the expression assumes the loop does not
- // have any early-exits. Otherwise, TC is an upper bound.
+ // NOTE: The TC (trip-count) expression is only valid when safe. Otherwise TC is 0
+ // (or possibly infinite). Also, the expression assumes the loop does not have
+ // early-exits. Otherwise, TC is an upper bound.
//
- bool cancels = is_strict && std::abs(stride_value) == 1; // compensation cancels conversion?
+ bool cancels = (cmp == kCondLT || cmp == kCondGT) && std::abs(stride_value) == 1;
if (!cancels) {
// Convert exclusive integral inequality into inclusive integral inequality,
// viz. condition i < U is i <= U - 1 and condition i > U is i >= U + 1.
- if (is_strict) {
- const InductionOp op = stride_value > 0 ? kSub : kAdd;
- hi_val = CreateInvariantOp(op, hi_val, CreateConstant(1, type));
+ if (cmp == kCondLT) {
+ hi_val = CreateInvariantOp(kSub, hi_val, CreateConstant(1, type));
+ } else if (cmp == kCondGT) {
+ hi_val = CreateInvariantOp(kAdd, hi_val, CreateConstant(1, type));
}
// Compensate for stride.
hi_val = CreateInvariantOp(kAdd, hi_val, stride);
}
// Assign the trip-count expression to the loop control. Clients that use the information
- // should be aware that due to the top-test assumption, the expression is only valid in the
- // loop-body proper, and not yet in the loop-header. If the loop has any early exits, the
- // trip-count forms a conservative upper bound on the number of loop iterations.
+ // should be aware that the expression is only valid in the loop-body proper (when symbolically
+ // safe), and not yet in the loop-header (unless constant safe). If the loop has any early exits,
+ // the trip-count forms a conservative upper bound on the number of loop iterations.
InductionInfo* trip_count =
CreateInvariantOp(kDiv, CreateInvariantOp(kSub, hi_val, lo_val), stride);
AssignInfo(loop, loop->GetHeader()->GetLastInstruction(), trip_count);
diff --git a/compiler/optimizing/induction_var_analysis.h b/compiler/optimizing/induction_var_analysis.h
index 8eccf925c1..190a0db65c 100644
--- a/compiler/optimizing/induction_var_analysis.h
+++ b/compiler/optimizing/induction_var_analysis.h
@@ -121,26 +121,27 @@ class HInductionVarAnalysis : public HOptimization {
uint32_t VisitDescendant(HLoopInformation* loop, HInstruction* instruction);
void ClassifyTrivial(HLoopInformation* loop, HInstruction* instruction);
void ClassifyNonTrivial(HLoopInformation* loop);
+ InductionInfo* RotatePeriodicInduction(InductionInfo* induction, InductionInfo* last);
// Transfer operations.
- InductionInfo* TransferPhi(InductionInfo* a, InductionInfo* b);
+ InductionInfo* TransferPhi(HLoopInformation* loop, HInstruction* phi, size_t input_index);
InductionInfo* TransferAddSub(InductionInfo* a, InductionInfo* b, InductionOp op);
InductionInfo* TransferMul(InductionInfo* a, InductionInfo* b);
InductionInfo* TransferShl(InductionInfo* a, InductionInfo* b, Primitive::Type type);
InductionInfo* TransferNeg(InductionInfo* a);
// Solvers.
- InductionInfo* SolvePhi(HLoopInformation* loop,
- HInstruction* phi,
- HInstruction* instruction);
+ InductionInfo* SolvePhi(HInstruction* phi, size_t input_index);
+ InductionInfo* SolvePhiAllInputs(HLoopInformation* loop,
+ HInstruction* entry_phi,
+ HInstruction* phi);
InductionInfo* SolveAddSub(HLoopInformation* loop,
- HInstruction* phi,
+ HInstruction* entry_phi,
HInstruction* instruction,
HInstruction* x,
HInstruction* y,
InductionOp op,
bool is_first_call);
- InductionInfo* RotatePeriodicInduction(InductionInfo* induction, InductionInfo* last);
// Trip count information.
void VisitControl(HLoopInformation* loop);
@@ -155,7 +156,7 @@ class HInductionVarAnalysis : public HOptimization {
InductionInfo* stride,
int32_t stride_value,
Primitive::Type type,
- bool is_strict);
+ IfCondition cmp);
// Assign and lookup.
void AssignInfo(HLoopInformation* loop, HInstruction* instruction, InductionInfo* info);
diff --git a/compiler/optimizing/induction_var_analysis_test.cc b/compiler/optimizing/induction_var_analysis_test.cc
index fca1ca55e5..e519e77f60 100644
--- a/compiler/optimizing/induction_var_analysis_test.cc
+++ b/compiler/optimizing/induction_var_analysis_test.cc
@@ -20,6 +20,7 @@
#include "builder.h"
#include "gtest/gtest.h"
#include "induction_var_analysis.h"
+#include "induction_var_range.h"
#include "nodes.h"
#include "optimizing_unit_test.h"
@@ -388,7 +389,7 @@ TEST_F(InductionVarAnalysisTest, FindSecondOrderWrapAroundInduction) {
HInstruction* store = InsertArrayStore(induc_, 0);
InsertLocalStore(induc_, InsertLocalLoad(tmp_, 0), 0);
HInstruction *sub = InsertInstruction(
- new (&allocator_) HSub(Primitive::kPrimInt, constant100_, InsertLocalLoad(basic_[0], 0)), 0);
+ new (&allocator_) HSub(Primitive::kPrimInt, constant100_, InsertLocalLoad(basic_[0], 0)), 0);
InsertLocalStore(tmp_, sub, 0);
PerformInductionVarAnalysis();
@@ -412,16 +413,16 @@ TEST_F(InductionVarAnalysisTest, FindWrapAroundDerivedInduction) {
new (&allocator_) HAdd(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
InsertLocalStore(tmp_, add, 0);
HInstruction *sub = InsertInstruction(
- new (&allocator_) HSub(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
+ new (&allocator_) HSub(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
InsertLocalStore(tmp_, sub, 0);
HInstruction *mul = InsertInstruction(
- new (&allocator_) HMul(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
+ new (&allocator_) HMul(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
InsertLocalStore(tmp_, mul, 0);
HInstruction *shl = InsertInstruction(
- new (&allocator_) HShl(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant1_), 0);
+ new (&allocator_) HShl(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant1_), 0);
InsertLocalStore(tmp_, shl, 0);
HInstruction *neg = InsertInstruction(
- new (&allocator_) HNeg(Primitive::kPrimInt, InsertLocalLoad(induc_, 0)), 0);
+ new (&allocator_) HNeg(Primitive::kPrimInt, InsertLocalLoad(induc_, 0)), 0);
InsertLocalStore(tmp_, neg, 0);
InsertLocalStore(
induc_,
@@ -471,7 +472,7 @@ TEST_F(InductionVarAnalysisTest, FindIdiomaticPeriodicInduction) {
BuildLoopNest(1);
HInstruction* store = InsertArrayStore(induc_, 0);
HInstruction *sub = InsertInstruction(
- new (&allocator_) HSub(Primitive::kPrimInt, constant1_, InsertLocalLoad(induc_, 0)), 0);
+ new (&allocator_) HSub(Primitive::kPrimInt, constant1_, InsertLocalLoad(induc_, 0)), 0);
InsertLocalStore(induc_, sub, 0);
PerformInductionVarAnalysis();
@@ -497,19 +498,19 @@ TEST_F(InductionVarAnalysisTest, FindDerivedPeriodicInduction) {
HSub(Primitive::kPrimInt, constant1_, InsertLocalLoad(induc_, 0)), 0), 0);
// Derived expressions.
HInstruction *add = InsertInstruction(
- new (&allocator_) HAdd(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
+ new (&allocator_) HAdd(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
InsertLocalStore(tmp_, add, 0);
HInstruction *sub = InsertInstruction(
- new (&allocator_) HSub(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
+ new (&allocator_) HSub(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
InsertLocalStore(tmp_, sub, 0);
HInstruction *mul = InsertInstruction(
- new (&allocator_) HMul(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
+ new (&allocator_) HMul(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant100_), 0);
InsertLocalStore(tmp_, mul, 0);
HInstruction *shl = InsertInstruction(
- new (&allocator_) HShl(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant1_), 0);
+ new (&allocator_) HShl(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant1_), 0);
InsertLocalStore(tmp_, shl, 0);
HInstruction *neg = InsertInstruction(
- new (&allocator_) HNeg(Primitive::kPrimInt, InsertLocalLoad(induc_, 0)), 0);
+ new (&allocator_) HNeg(Primitive::kPrimInt, InsertLocalLoad(induc_, 0)), 0);
InsertLocalStore(tmp_, neg, 0);
PerformInductionVarAnalysis();
@@ -520,6 +521,34 @@ TEST_F(InductionVarAnalysisTest, FindDerivedPeriodicInduction) {
EXPECT_STREQ("periodic(( - (1)), (0))", GetInductionInfo(neg, 0).c_str());
}
+TEST_F(InductionVarAnalysisTest, FindRange) {
+ // Setup:
+ // for (int i = 0; i < 100; i++) {
+ // k = i << 1;
+ // k = k + 1;
+ // a[k] = 0;
+ // }
+ BuildLoopNest(1);
+ HInstruction *shl = InsertInstruction(
+ new (&allocator_) HShl(Primitive::kPrimInt, InsertLocalLoad(basic_[0], 0), constant1_), 0);
+ InsertLocalStore(induc_, shl, 0);
+ HInstruction *add = InsertInstruction(
+ new (&allocator_) HAdd(Primitive::kPrimInt, InsertLocalLoad(induc_, 0), constant1_), 0);
+ InsertLocalStore(induc_, add, 0);
+ HInstruction* store = InsertArrayStore(induc_, 0);
+ PerformInductionVarAnalysis();
+
+ EXPECT_STREQ("((2) * i + (1))", GetInductionInfo(store->InputAt(1), 0).c_str());
+
+ InductionVarRange range(iva_);
+ InductionVarRange::Value v_min = range.GetMinInduction(store, store->InputAt(1));
+ InductionVarRange::Value v_max = range.GetMaxInduction(store, store->InputAt(1));
+ EXPECT_EQ(0, v_min.a_constant);
+ EXPECT_EQ(1, v_min.b_constant);
+ EXPECT_EQ(0, v_max.a_constant);
+ EXPECT_EQ(199, v_max.b_constant);
+}
+
TEST_F(InductionVarAnalysisTest, FindDeepLoopInduction) {
// Setup:
// k = 0;
diff --git a/compiler/optimizing/induction_var_range.cc b/compiler/optimizing/induction_var_range.cc
index 486e904bd1..119a80b6f4 100644
--- a/compiler/optimizing/induction_var_range.cc
+++ b/compiler/optimizing/induction_var_range.cc
@@ -14,94 +14,95 @@
* limitations under the License.
*/
-#include <limits.h>
-
#include "induction_var_range.h"
-namespace art {
+#include <limits>
-static bool IsValidConstant32(int32_t c) {
- return INT_MIN < c && c < INT_MAX;
-}
+namespace art {
-static bool IsValidConstant64(int64_t c) {
- return INT_MIN < c && c < INT_MAX;
+/** Returns true if 64-bit constant fits in 32-bit constant. */
+static bool CanLongValueFitIntoInt(int64_t c) {
+ return std::numeric_limits<int32_t>::min() <= c && c <= std::numeric_limits<int32_t>::max();
}
-/** Returns true if 32-bit addition can be done safely (and is not an unknown range). */
+/** Returns true if 32-bit addition can be done safely. */
static bool IsSafeAdd(int32_t c1, int32_t c2) {
- if (IsValidConstant32(c1) && IsValidConstant32(c2)) {
- return IsValidConstant64(static_cast<int64_t>(c1) + static_cast<int64_t>(c2));
- }
- return false;
+ return CanLongValueFitIntoInt(static_cast<int64_t>(c1) + static_cast<int64_t>(c2));
}
-/** Returns true if 32-bit subtraction can be done safely (and is not an unknown range). */
+/** Returns true if 32-bit subtraction can be done safely. */
static bool IsSafeSub(int32_t c1, int32_t c2) {
- if (IsValidConstant32(c1) && IsValidConstant32(c2)) {
- return IsValidConstant64(static_cast<int64_t>(c1) - static_cast<int64_t>(c2));
- }
- return false;
+ return CanLongValueFitIntoInt(static_cast<int64_t>(c1) - static_cast<int64_t>(c2));
}
-/** Returns true if 32-bit multiplication can be done safely (and is not an unknown range). */
+/** Returns true if 32-bit multiplication can be done safely. */
static bool IsSafeMul(int32_t c1, int32_t c2) {
- if (IsValidConstant32(c1) && IsValidConstant32(c2)) {
- return IsValidConstant64(static_cast<int64_t>(c1) * static_cast<int64_t>(c2));
- }
- return false;
+ return CanLongValueFitIntoInt(static_cast<int64_t>(c1) * static_cast<int64_t>(c2));
}
-/** Returns true if 32-bit division can be done safely (and is not an unknown range). */
+/** Returns true if 32-bit division can be done safely. */
static bool IsSafeDiv(int32_t c1, int32_t c2) {
- if (IsValidConstant32(c1) && IsValidConstant32(c2) && c2 != 0) {
- return IsValidConstant64(static_cast<int64_t>(c1) / static_cast<int64_t>(c2));
- }
- return false;
+ return c2 != 0 && CanLongValueFitIntoInt(static_cast<int64_t>(c1) / static_cast<int64_t>(c2));
}
-/** Returns true for 32/64-bit integral constant within known range. */
+/** Returns true for 32/64-bit integral constant. */
static bool IsIntAndGet(HInstruction* instruction, int32_t* value) {
if (instruction->IsIntConstant()) {
- const int32_t c = instruction->AsIntConstant()->GetValue();
- if (IsValidConstant32(c)) {
- *value = c;
- return true;
- }
+ *value = instruction->AsIntConstant()->GetValue();
+ return true;
} else if (instruction->IsLongConstant()) {
const int64_t c = instruction->AsLongConstant()->GetValue();
- if (IsValidConstant64(c)) {
- *value = c;
+ if (CanLongValueFitIntoInt(c)) {
+ *value = static_cast<int32_t>(c);
return true;
}
}
return false;
}
+/**
+ * An upper bound a * (length / a) + b, where a > 0, can be conservatively rewritten as length + b
+ * because length >= 0 is true. This makes it more likely the bound is useful to clients.
+ */
+static InductionVarRange::Value SimplifyMax(InductionVarRange::Value v) {
+ int32_t value;
+ if (v.a_constant > 1 &&
+ v.instruction->IsDiv() &&
+ v.instruction->InputAt(0)->IsArrayLength() &&
+ IsIntAndGet(v.instruction->InputAt(1), &value) && v.a_constant == value) {
+ return InductionVarRange::Value(v.instruction->InputAt(0), 1, v.b_constant);
+ }
+ return v;
+}
+
//
// Public class methods.
//
InductionVarRange::InductionVarRange(HInductionVarAnalysis* induction_analysis)
: induction_analysis_(induction_analysis) {
+ DCHECK(induction_analysis != nullptr);
}
InductionVarRange::Value InductionVarRange::GetMinInduction(HInstruction* context,
HInstruction* instruction) {
HLoopInformation* loop = context->GetBlock()->GetLoopInformation();
- if (loop != nullptr && induction_analysis_ != nullptr) {
- return GetMin(induction_analysis_->LookupInfo(loop, instruction), GetTripCount(loop, context));
+ if (loop != nullptr) {
+ return GetVal(induction_analysis_->LookupInfo(loop, instruction),
+ GetTripCount(loop, context), /* is_min */ true);
}
- return Value(INT_MIN);
+ return Value();
}
InductionVarRange::Value InductionVarRange::GetMaxInduction(HInstruction* context,
HInstruction* instruction) {
HLoopInformation* loop = context->GetBlock()->GetLoopInformation();
- if (loop != nullptr && induction_analysis_ != nullptr) {
- return GetMax(induction_analysis_->LookupInfo(loop, instruction), GetTripCount(loop, context));
+ if (loop != nullptr) {
+ return SimplifyMax(
+ GetVal(induction_analysis_->LookupInfo(loop, instruction),
+ GetTripCount(loop, context), /* is_min */ false));
}
- return Value(INT_MAX);
+ return Value();
}
//
@@ -113,6 +114,9 @@ HInductionVarAnalysis::InductionInfo* InductionVarRange::GetTripCount(HLoopInfor
// The trip-count expression is only valid when the top-test is taken at least once,
// that means, when the analyzed context appears outside the loop header itself.
// Early-exit loops are okay, since in those cases, the trip-count is conservative.
+ //
+ // TODO: deal with runtime safety issues on TCs
+ //
if (context->GetBlock() != loop->GetHeader()) {
HInductionVarAnalysis::InductionInfo* trip =
induction_analysis_->LookupInfo(loop, loop->GetHeader()->GetLastInstruction());
@@ -127,7 +131,7 @@ HInductionVarAnalysis::InductionInfo* InductionVarRange::GetTripCount(HLoopInfor
InductionVarRange::Value InductionVarRange::GetFetch(HInstruction* instruction,
HInductionVarAnalysis::InductionInfo* trip,
- int32_t fail_value) {
+ bool is_min) {
// Detect constants and chase the fetch a bit deeper into the HIR tree, so that it becomes
// more likely range analysis will compare the same instructions as terminal nodes.
int32_t value;
@@ -135,14 +139,12 @@ InductionVarRange::Value InductionVarRange::GetFetch(HInstruction* instruction,
return Value(value);
} else if (instruction->IsAdd()) {
if (IsIntAndGet(instruction->InputAt(0), &value)) {
- return AddValue(Value(value),
- GetFetch(instruction->InputAt(1), trip, fail_value), fail_value);
+ return AddValue(Value(value), GetFetch(instruction->InputAt(1), trip, is_min));
} else if (IsIntAndGet(instruction->InputAt(1), &value)) {
- return AddValue(GetFetch(instruction->InputAt(0), trip, fail_value),
- Value(value), fail_value);
+ return AddValue(GetFetch(instruction->InputAt(0), trip, is_min), Value(value));
}
- } else if (fail_value < 0) {
- // Special case: within the loop-body, minimum of trip-count is 1.
+ } else if (is_min) {
+ // Special case for finding minimum: minimum of trip-count is 1.
if (trip != nullptr && instruction == trip->op_b->fetch) {
return Value(1);
}
@@ -150,142 +152,111 @@ InductionVarRange::Value InductionVarRange::GetFetch(HInstruction* instruction,
return Value(instruction, 1, 0);
}
-InductionVarRange::Value InductionVarRange::GetMin(HInductionVarAnalysis::InductionInfo* info,
- HInductionVarAnalysis::InductionInfo* trip) {
- if (info != nullptr) {
- switch (info->induction_class) {
- case HInductionVarAnalysis::kInvariant:
- // Invariants.
- switch (info->operation) {
- case HInductionVarAnalysis::kNop: // normalized: 0
- DCHECK_EQ(info->op_a, info->op_b);
- return Value(0);
- case HInductionVarAnalysis::kAdd:
- return AddValue(GetMin(info->op_a, trip), GetMin(info->op_b, trip), INT_MIN);
- case HInductionVarAnalysis::kSub: // second max!
- return SubValue(GetMin(info->op_a, trip), GetMax(info->op_b, trip), INT_MIN);
- case HInductionVarAnalysis::kNeg: // second max!
- return SubValue(Value(0), GetMax(info->op_b, trip), INT_MIN);
- case HInductionVarAnalysis::kMul:
- return GetMul(info->op_a, info->op_b, trip, INT_MIN);
- case HInductionVarAnalysis::kDiv:
- return GetDiv(info->op_a, info->op_b, trip, INT_MIN);
- case HInductionVarAnalysis::kFetch:
- return GetFetch(info->fetch, trip, INT_MIN);
- }
- break;
- case HInductionVarAnalysis::kLinear:
- // Minimum over linear induction a * i + b, for normalized 0 <= i < TC.
- return AddValue(GetMul(info->op_a, trip, trip, INT_MIN),
- GetMin(info->op_b, trip), INT_MIN);
- case HInductionVarAnalysis::kWrapAround:
- case HInductionVarAnalysis::kPeriodic:
- // Minimum over all values in the wrap-around/periodic.
- return MinValue(GetMin(info->op_a, trip), GetMin(info->op_b, trip));
- }
- }
- return Value(INT_MIN);
-}
-
-InductionVarRange::Value InductionVarRange::GetMax(HInductionVarAnalysis::InductionInfo* info,
- HInductionVarAnalysis::InductionInfo* trip) {
+InductionVarRange::Value InductionVarRange::GetVal(HInductionVarAnalysis::InductionInfo* info,
+ HInductionVarAnalysis::InductionInfo* trip,
+ bool is_min) {
if (info != nullptr) {
switch (info->induction_class) {
case HInductionVarAnalysis::kInvariant:
// Invariants.
switch (info->operation) {
- case HInductionVarAnalysis::kNop: // normalized: TC - 1
+ case HInductionVarAnalysis::kNop: // normalized: 0 or TC-1
DCHECK_EQ(info->op_a, info->op_b);
- return SubValue(GetMax(info->op_b, trip), Value(1), INT_MAX);
+ return is_min ? Value(0)
+ : SubValue(GetVal(info->op_b, trip, is_min), Value(1));
case HInductionVarAnalysis::kAdd:
- return AddValue(GetMax(info->op_a, trip), GetMax(info->op_b, trip), INT_MAX);
- case HInductionVarAnalysis::kSub: // second min!
- return SubValue(GetMax(info->op_a, trip), GetMin(info->op_b, trip), INT_MAX);
- case HInductionVarAnalysis::kNeg: // second min!
- return SubValue(Value(0), GetMin(info->op_b, trip), INT_MAX);
+ return AddValue(GetVal(info->op_a, trip, is_min),
+ GetVal(info->op_b, trip, is_min));
+ case HInductionVarAnalysis::kSub: // second reversed!
+ return SubValue(GetVal(info->op_a, trip, is_min),
+ GetVal(info->op_b, trip, !is_min));
+ case HInductionVarAnalysis::kNeg: // second reversed!
+ return SubValue(Value(0),
+ GetVal(info->op_b, trip, !is_min));
case HInductionVarAnalysis::kMul:
- return GetMul(info->op_a, info->op_b, trip, INT_MAX);
+ return GetMul(info->op_a, info->op_b, trip, is_min);
case HInductionVarAnalysis::kDiv:
- return GetDiv(info->op_a, info->op_b, trip, INT_MAX);
+ return GetDiv(info->op_a, info->op_b, trip, is_min);
case HInductionVarAnalysis::kFetch:
- return GetFetch(info->fetch, trip, INT_MAX);
+ return GetFetch(info->fetch, trip, is_min);
}
break;
case HInductionVarAnalysis::kLinear:
- // Maximum over linear induction a * i + b, for normalized 0 <= i < TC.
- return AddValue(GetMul(info->op_a, trip, trip, INT_MAX),
- GetMax(info->op_b, trip), INT_MAX);
+ // Linear induction a * i + b, for normalized 0 <= i < TC.
+ return AddValue(GetMul(info->op_a, trip, trip, is_min),
+ GetVal(info->op_b, trip, is_min));
case HInductionVarAnalysis::kWrapAround:
case HInductionVarAnalysis::kPeriodic:
- // Maximum over all values in the wrap-around/periodic.
- return MaxValue(GetMax(info->op_a, trip), GetMax(info->op_b, trip));
+ // Merge values in the wrap-around/periodic.
+ return MergeVal(GetVal(info->op_a, trip, is_min),
+ GetVal(info->op_b, trip, is_min), is_min);
}
}
- return Value(INT_MAX);
+ return Value();
}
InductionVarRange::Value InductionVarRange::GetMul(HInductionVarAnalysis::InductionInfo* info1,
HInductionVarAnalysis::InductionInfo* info2,
HInductionVarAnalysis::InductionInfo* trip,
- int32_t fail_value) {
- Value v1_min = GetMin(info1, trip);
- Value v1_max = GetMax(info1, trip);
- Value v2_min = GetMin(info2, trip);
- Value v2_max = GetMax(info2, trip);
- if (v1_min.a_constant == 0 && v1_min.b_constant >= 0) {
+ bool is_min) {
+ Value v1_min = GetVal(info1, trip, /* is_min */ true);
+ Value v1_max = GetVal(info1, trip, /* is_min */ false);
+ Value v2_min = GetVal(info2, trip, /* is_min */ true);
+ Value v2_max = GetVal(info2, trip, /* is_min */ false);
+ if (v1_min.is_known && v1_min.a_constant == 0 && v1_min.b_constant >= 0) {
// Positive range vs. positive or negative range.
- if (v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
- return (fail_value < 0) ? MulValue(v1_min, v2_min, fail_value)
- : MulValue(v1_max, v2_max, fail_value);
- } else if (v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
- return (fail_value < 0) ? MulValue(v1_max, v2_min, fail_value)
- : MulValue(v1_min, v2_max, fail_value);
+ if (v2_min.is_known && v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
+ return is_min ? MulValue(v1_min, v2_min)
+ : MulValue(v1_max, v2_max);
+ } else if (v2_max.is_known && v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
+ return is_min ? MulValue(v1_max, v2_min)
+ : MulValue(v1_min, v2_max);
}
- } else if (v1_min.a_constant == 0 && v1_min.b_constant <= 0) {
+ } else if (v1_min.is_known && v1_min.a_constant == 0 && v1_min.b_constant <= 0) {
// Negative range vs. positive or negative range.
- if (v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
- return (fail_value < 0) ? MulValue(v1_min, v2_max, fail_value)
- : MulValue(v1_max, v2_min, fail_value);
- } else if (v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
- return (fail_value < 0) ? MulValue(v1_max, v2_max, fail_value)
- : MulValue(v1_min, v2_min, fail_value);
+ if (v2_min.is_known && v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
+ return is_min ? MulValue(v1_min, v2_max)
+ : MulValue(v1_max, v2_min);
+ } else if (v2_max.is_known && v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
+ return is_min ? MulValue(v1_max, v2_max)
+ : MulValue(v1_min, v2_min);
}
}
- return Value(fail_value);
+ return Value();
}
InductionVarRange::Value InductionVarRange::GetDiv(HInductionVarAnalysis::InductionInfo* info1,
HInductionVarAnalysis::InductionInfo* info2,
HInductionVarAnalysis::InductionInfo* trip,
- int32_t fail_value) {
- Value v1_min = GetMin(info1, trip);
- Value v1_max = GetMax(info1, trip);
- Value v2_min = GetMin(info2, trip);
- Value v2_max = GetMax(info2, trip);
- if (v1_min.a_constant == 0 && v1_min.b_constant >= 0) {
+ bool is_min) {
+ Value v1_min = GetVal(info1, trip, /* is_min */ true);
+ Value v1_max = GetVal(info1, trip, /* is_min */ false);
+ Value v2_min = GetVal(info2, trip, /* is_min */ true);
+ Value v2_max = GetVal(info2, trip, /* is_min */ false);
+ if (v1_min.is_known && v1_min.a_constant == 0 && v1_min.b_constant >= 0) {
// Positive range vs. positive or negative range.
- if (v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
- return (fail_value < 0) ? DivValue(v1_min, v2_max, fail_value)
- : DivValue(v1_max, v2_min, fail_value);
- } else if (v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
- return (fail_value < 0) ? DivValue(v1_max, v2_max, fail_value)
- : DivValue(v1_min, v2_min, fail_value);
+ if (v2_min.is_known && v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
+ return is_min ? DivValue(v1_min, v2_max)
+ : DivValue(v1_max, v2_min);
+ } else if (v2_max.is_known && v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
+ return is_min ? DivValue(v1_max, v2_max)
+ : DivValue(v1_min, v2_min);
}
- } else if (v1_min.a_constant == 0 && v1_min.b_constant <= 0) {
+ } else if (v1_min.is_known && v1_min.a_constant == 0 && v1_min.b_constant <= 0) {
// Negative range vs. positive or negative range.
- if (v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
- return (fail_value < 0) ? DivValue(v1_min, v2_min, fail_value)
- : DivValue(v1_max, v2_max, fail_value);
- } else if (v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
- return (fail_value < 0) ? DivValue(v1_max, v2_min, fail_value)
- : DivValue(v1_min, v2_max, fail_value);
+ if (v2_min.is_known && v2_min.a_constant == 0 && v2_min.b_constant >= 0) {
+ return is_min ? DivValue(v1_min, v2_min)
+ : DivValue(v1_max, v2_max);
+ } else if (v2_max.is_known && v2_max.a_constant == 0 && v2_max.b_constant <= 0) {
+ return is_min ? DivValue(v1_max, v2_min)
+ : DivValue(v1_min, v2_max);
}
}
- return Value(fail_value);
+ return Value();
}
-InductionVarRange::Value InductionVarRange::AddValue(Value v1, Value v2, int32_t fail_value) {
- if (IsSafeAdd(v1.b_constant, v2.b_constant)) {
+InductionVarRange::Value InductionVarRange::AddValue(Value v1, Value v2) {
+ if (v1.is_known && v2.is_known && IsSafeAdd(v1.b_constant, v2.b_constant)) {
const int32_t b = v1.b_constant + v2.b_constant;
if (v1.a_constant == 0) {
return Value(v2.instruction, v2.a_constant, b);
@@ -295,11 +266,11 @@ InductionVarRange::Value InductionVarRange::AddValue(Value v1, Value v2, int32_t
return Value(v1.instruction, v1.a_constant + v2.a_constant, b);
}
}
- return Value(fail_value);
+ return Value();
}
-InductionVarRange::Value InductionVarRange::SubValue(Value v1, Value v2, int32_t fail_value) {
- if (IsSafeSub(v1.b_constant, v2.b_constant)) {
+InductionVarRange::Value InductionVarRange::SubValue(Value v1, Value v2) {
+ if (v1.is_known && v2.is_known && IsSafeSub(v1.b_constant, v2.b_constant)) {
const int32_t b = v1.b_constant - v2.b_constant;
if (v1.a_constant == 0 && IsSafeSub(0, v2.a_constant)) {
return Value(v2.instruction, -v2.a_constant, b);
@@ -309,43 +280,42 @@ InductionVarRange::Value InductionVarRange::SubValue(Value v1, Value v2, int32_t
return Value(v1.instruction, v1.a_constant - v2.a_constant, b);
}
}
- return Value(fail_value);
+ return Value();
}
-InductionVarRange::Value InductionVarRange::MulValue(Value v1, Value v2, int32_t fail_value) {
- if (v1.a_constant == 0) {
- if (IsSafeMul(v1.b_constant, v2.a_constant) && IsSafeMul(v1.b_constant, v2.b_constant)) {
- return Value(v2.instruction, v1.b_constant * v2.a_constant, v1.b_constant * v2.b_constant);
- }
- } else if (v2.a_constant == 0) {
- if (IsSafeMul(v1.a_constant, v2.b_constant) && IsSafeMul(v1.b_constant, v2.b_constant)) {
- return Value(v1.instruction, v1.a_constant * v2.b_constant, v1.b_constant * v2.b_constant);
+InductionVarRange::Value InductionVarRange::MulValue(Value v1, Value v2) {
+ if (v1.is_known && v2.is_known) {
+ if (v1.a_constant == 0) {
+ if (IsSafeMul(v1.b_constant, v2.a_constant) && IsSafeMul(v1.b_constant, v2.b_constant)) {
+ return Value(v2.instruction, v1.b_constant * v2.a_constant, v1.b_constant * v2.b_constant);
+ }
+ } else if (v2.a_constant == 0) {
+ if (IsSafeMul(v1.a_constant, v2.b_constant) && IsSafeMul(v1.b_constant, v2.b_constant)) {
+ return Value(v1.instruction, v1.a_constant * v2.b_constant, v1.b_constant * v2.b_constant);
+ }
}
}
- return Value(fail_value);
+ return Value();
}
-InductionVarRange::Value InductionVarRange::DivValue(Value v1, Value v2, int32_t fail_value) {
- if (v1.a_constant == 0 && v2.a_constant == 0) {
+InductionVarRange::Value InductionVarRange::DivValue(Value v1, Value v2) {
+ if (v1.is_known && v2.is_known && v1.a_constant == 0 && v2.a_constant == 0) {
if (IsSafeDiv(v1.b_constant, v2.b_constant)) {
return Value(v1.b_constant / v2.b_constant);
}
}
- return Value(fail_value);
+ return Value();
}
-InductionVarRange::Value InductionVarRange::MinValue(Value v1, Value v2) {
- if (v1.instruction == v2.instruction && v1.a_constant == v2.a_constant) {
- return Value(v1.instruction, v1.a_constant, std::min(v1.b_constant, v2.b_constant));
- }
- return Value(INT_MIN);
-}
-
-InductionVarRange::Value InductionVarRange::MaxValue(Value v1, Value v2) {
- if (v1.instruction == v2.instruction && v1.a_constant == v2.a_constant) {
- return Value(v1.instruction, v1.a_constant, std::max(v1.b_constant, v2.b_constant));
+InductionVarRange::Value InductionVarRange::MergeVal(Value v1, Value v2, bool is_min) {
+ if (v1.is_known && v2.is_known) {
+ if (v1.instruction == v2.instruction && v1.a_constant == v2.a_constant) {
+ return Value(v1.instruction, v1.a_constant,
+ is_min ? std::min(v1.b_constant, v2.b_constant)
+ : std::max(v1.b_constant, v2.b_constant));
+ }
}
- return Value(INT_MAX);
+ return Value();
}
} // namespace art
diff --git a/compiler/optimizing/induction_var_range.h b/compiler/optimizing/induction_var_range.h
index e002e5ff6c..8280c8bedc 100644
--- a/compiler/optimizing/induction_var_range.h
+++ b/compiler/optimizing/induction_var_range.h
@@ -22,30 +22,36 @@
namespace art {
/**
- * This class implements induction variable based range analysis on expressions within loops.
- * It takes the results of induction variable analysis in the constructor and provides a public
- * API to obtain a conservative lower and upper bound value on each instruction in the HIR.
+ * This class implements range analysis on expressions within loops. It takes the results
+ * of induction variable analysis in the constructor and provides a public API to obtain
+ * a conservative lower and upper bound value on each instruction in the HIR.
*
- * For example, given a linear induction 2 * i + x where 0 <= i <= 10, range analysis yields lower
- * bound value x and upper bound value x + 20 for the expression, thus, the range [x, x + 20].
+ * The range analysis is done with a combination of symbolic and partial integral evaluation
+ * of expressions. The analysis avoids complications with wrap-around arithmetic on the integral
+ * parts but all clients should be aware that wrap-around may occur on any of the symbolic parts.
+ * For example, given a known range for [0,100] for i, the evaluation yields range [-100,100]
+ * for expression -2*i+100, which is exact, and range [x,x+100] for expression i+x, which may
+ * wrap-around anywhere in the range depending on the actual value of x.
*/
class InductionVarRange {
public:
/*
* A value that can be represented as "a * instruction + b" for 32-bit constants, where
- * Value(INT_MIN) and Value(INT_MAX) denote an unknown lower and upper bound, respectively.
- * Although range analysis could yield more complex values, the format is sufficiently powerful
- * to represent useful cases and feeds directly into optimizations like bounds check elimination.
+ * Value() denotes an unknown lower and upper bound. Although range analysis could yield
+ * more complex values, the format is sufficiently powerful to represent useful cases
+ * and feeds directly into optimizations like bounds check elimination.
*/
struct Value {
+ Value() : instruction(nullptr), a_constant(0), b_constant(0), is_known(false) {}
Value(HInstruction* i, int32_t a, int32_t b)
- : instruction(a != 0 ? i : nullptr),
- a_constant(a),
- b_constant(b) {}
+ : instruction(a != 0 ? i : nullptr), a_constant(a), b_constant(b), is_known(true) {}
explicit Value(int32_t b) : Value(nullptr, 0, b) {}
+ // Representation as: a_constant x instruction + b_constant.
HInstruction* instruction;
int32_t a_constant;
int32_t b_constant;
+ // If true, represented by prior fields. Otherwise unknown value.
+ bool is_known;
};
explicit InductionVarRange(HInductionVarAnalysis* induction);
@@ -67,32 +73,29 @@ class InductionVarRange {
// Private helper methods.
//
- HInductionVarAnalysis::InductionInfo* GetTripCount(HLoopInformation* loop,
- HInstruction* context);
+ HInductionVarAnalysis::InductionInfo* GetTripCount(HLoopInformation* loop, HInstruction* context);
static Value GetFetch(HInstruction* instruction,
HInductionVarAnalysis::InductionInfo* trip,
- int32_t fail_value);
+ bool is_min);
- static Value GetMin(HInductionVarAnalysis::InductionInfo* info,
- HInductionVarAnalysis::InductionInfo* trip);
- static Value GetMax(HInductionVarAnalysis::InductionInfo* info,
- HInductionVarAnalysis::InductionInfo* trip);
+ static Value GetVal(HInductionVarAnalysis::InductionInfo* info,
+ HInductionVarAnalysis::InductionInfo* trip,
+ bool is_min);
static Value GetMul(HInductionVarAnalysis::InductionInfo* info1,
HInductionVarAnalysis::InductionInfo* info2,
HInductionVarAnalysis::InductionInfo* trip,
- int32_t fail_value);
+ bool is_min);
static Value GetDiv(HInductionVarAnalysis::InductionInfo* info1,
HInductionVarAnalysis::InductionInfo* info2,
HInductionVarAnalysis::InductionInfo* trip,
- int32_t fail_value);
-
- static Value AddValue(Value v1, Value v2, int32_t fail_value);
- static Value SubValue(Value v1, Value v2, int32_t fail_value);
- static Value MulValue(Value v1, Value v2, int32_t fail_value);
- static Value DivValue(Value v1, Value v2, int32_t fail_value);
- static Value MinValue(Value v1, Value v2);
- static Value MaxValue(Value v1, Value v2);
+ bool is_min);
+
+ static Value AddValue(Value v1, Value v2);
+ static Value SubValue(Value v1, Value v2);
+ static Value MulValue(Value v1, Value v2);
+ static Value DivValue(Value v1, Value v2);
+ static Value MergeVal(Value v1, Value v2, bool is_min);
/** Results of prior induction variable analysis. */
HInductionVarAnalysis *induction_analysis_;
diff --git a/compiler/optimizing/induction_var_range_test.cc b/compiler/optimizing/induction_var_range_test.cc
index d3c3518193..5d9a075aef 100644
--- a/compiler/optimizing/induction_var_range_test.cc
+++ b/compiler/optimizing/induction_var_range_test.cc
@@ -14,8 +14,6 @@
* limitations under the License.
*/
-#include <limits.h>
-
#include "base/arena_allocator.h"
#include "builder.h"
#include "gtest/gtest.h"
@@ -45,6 +43,7 @@ class InductionVarRangeTest : public testing::Test {
EXPECT_EQ(v1.instruction, v2.instruction);
EXPECT_EQ(v1.a_constant, v2.a_constant);
EXPECT_EQ(v1.b_constant, v2.b_constant);
+ EXPECT_EQ(v1.is_known, v2.is_known);
}
/** Constructs bare minimum graph. */
@@ -113,30 +112,32 @@ class InductionVarRangeTest : public testing::Test {
Value GetMin(HInductionVarAnalysis::InductionInfo* info,
HInductionVarAnalysis::InductionInfo* induc) {
- return InductionVarRange::GetMin(info, induc);
+ return InductionVarRange::GetVal(info, induc, /* is_min */ true);
}
Value GetMax(HInductionVarAnalysis::InductionInfo* info,
HInductionVarAnalysis::InductionInfo* induc) {
- return InductionVarRange::GetMax(info, induc);
+ return InductionVarRange::GetVal(info, induc, /* is_min */ false);
}
Value GetMul(HInductionVarAnalysis::InductionInfo* info1,
- HInductionVarAnalysis::InductionInfo* info2, int32_t fail_value) {
- return InductionVarRange::GetMul(info1, info2, nullptr, fail_value);
+ HInductionVarAnalysis::InductionInfo* info2,
+ bool is_min) {
+ return InductionVarRange::GetMul(info1, info2, nullptr, is_min);
}
Value GetDiv(HInductionVarAnalysis::InductionInfo* info1,
- HInductionVarAnalysis::InductionInfo* info2, int32_t fail_value) {
- return InductionVarRange::GetDiv(info1, info2, nullptr, fail_value);
+ HInductionVarAnalysis::InductionInfo* info2,
+ bool is_min) {
+ return InductionVarRange::GetDiv(info1, info2, nullptr, is_min);
}
- Value AddValue(Value v1, Value v2) { return InductionVarRange::AddValue(v1, v2, INT_MIN); }
- Value SubValue(Value v1, Value v2) { return InductionVarRange::SubValue(v1, v2, INT_MIN); }
- Value MulValue(Value v1, Value v2) { return InductionVarRange::MulValue(v1, v2, INT_MIN); }
- Value DivValue(Value v1, Value v2) { return InductionVarRange::DivValue(v1, v2, INT_MIN); }
- Value MinValue(Value v1, Value v2) { return InductionVarRange::MinValue(v1, v2); }
- Value MaxValue(Value v1, Value v2) { return InductionVarRange::MaxValue(v1, v2); }
+ Value AddValue(Value v1, Value v2) { return InductionVarRange::AddValue(v1, v2); }
+ Value SubValue(Value v1, Value v2) { return InductionVarRange::SubValue(v1, v2); }
+ Value MulValue(Value v1, Value v2) { return InductionVarRange::MulValue(v1, v2); }
+ Value DivValue(Value v1, Value v2) { return InductionVarRange::DivValue(v1, v2); }
+ Value MinValue(Value v1, Value v2) { return InductionVarRange::MergeVal(v1, v2, true); }
+ Value MaxValue(Value v1, Value v2) { return InductionVarRange::MergeVal(v1, v2, false); }
// General building fields.
ArenaPool pool_;
@@ -154,8 +155,8 @@ class InductionVarRangeTest : public testing::Test {
//
TEST_F(InductionVarRangeTest, GetMinMaxNull) {
- ExpectEqual(Value(INT_MIN), GetMin(nullptr, nullptr));
- ExpectEqual(Value(INT_MAX), GetMax(nullptr, nullptr));
+ ExpectEqual(Value(), GetMin(nullptr, nullptr));
+ ExpectEqual(Value(), GetMax(nullptr, nullptr));
}
TEST_F(InductionVarRangeTest, GetMinMaxAdd) {
@@ -251,91 +252,91 @@ TEST_F(InductionVarRangeTest, GetMinMaxPeriodic) {
}
TEST_F(InductionVarRangeTest, GetMulMin) {
- ExpectEqual(Value(6), GetMul(CreateRange(2, 10), CreateRange(3, 5), INT_MIN));
- ExpectEqual(Value(-50), GetMul(CreateRange(2, 10), CreateRange(-5, -3), INT_MIN));
- ExpectEqual(Value(-50), GetMul(CreateRange(-10, -2), CreateRange(3, 5), INT_MIN));
- ExpectEqual(Value(6), GetMul(CreateRange(-10, -2), CreateRange(-5, -3), INT_MIN));
+ ExpectEqual(Value(6), GetMul(CreateRange(2, 10), CreateRange(3, 5), true));
+ ExpectEqual(Value(-50), GetMul(CreateRange(2, 10), CreateRange(-5, -3), true));
+ ExpectEqual(Value(-50), GetMul(CreateRange(-10, -2), CreateRange(3, 5), true));
+ ExpectEqual(Value(6), GetMul(CreateRange(-10, -2), CreateRange(-5, -3), true));
}
TEST_F(InductionVarRangeTest, GetMulMax) {
- ExpectEqual(Value(50), GetMul(CreateRange(2, 10), CreateRange(3, 5), INT_MAX));
- ExpectEqual(Value(-6), GetMul(CreateRange(2, 10), CreateRange(-5, -3), INT_MAX));
- ExpectEqual(Value(-6), GetMul(CreateRange(-10, -2), CreateRange(3, 5), INT_MAX));
- ExpectEqual(Value(50), GetMul(CreateRange(-10, -2), CreateRange(-5, -3), INT_MAX));
+ ExpectEqual(Value(50), GetMul(CreateRange(2, 10), CreateRange(3, 5), false));
+ ExpectEqual(Value(-6), GetMul(CreateRange(2, 10), CreateRange(-5, -3), false));
+ ExpectEqual(Value(-6), GetMul(CreateRange(-10, -2), CreateRange(3, 5), false));
+ ExpectEqual(Value(50), GetMul(CreateRange(-10, -2), CreateRange(-5, -3), false));
}
TEST_F(InductionVarRangeTest, GetDivMin) {
- ExpectEqual(Value(10), GetDiv(CreateRange(40, 1000), CreateRange(2, 4), INT_MIN));
- ExpectEqual(Value(-500), GetDiv(CreateRange(40, 1000), CreateRange(-4, -2), INT_MIN));
- ExpectEqual(Value(-500), GetDiv(CreateRange(-1000, -40), CreateRange(2, 4), INT_MIN));
- ExpectEqual(Value(10), GetDiv(CreateRange(-1000, -40), CreateRange(-4, -2), INT_MIN));
+ ExpectEqual(Value(10), GetDiv(CreateRange(40, 1000), CreateRange(2, 4), true));
+ ExpectEqual(Value(-500), GetDiv(CreateRange(40, 1000), CreateRange(-4, -2), true));
+ ExpectEqual(Value(-500), GetDiv(CreateRange(-1000, -40), CreateRange(2, 4), true));
+ ExpectEqual(Value(10), GetDiv(CreateRange(-1000, -40), CreateRange(-4, -2), true));
}
TEST_F(InductionVarRangeTest, GetDivMax) {
- ExpectEqual(Value(500), GetDiv(CreateRange(40, 1000), CreateRange(2, 4), INT_MAX));
- ExpectEqual(Value(-10), GetDiv(CreateRange(40, 1000), CreateRange(-4, -2), INT_MAX));
- ExpectEqual(Value(-10), GetDiv(CreateRange(-1000, -40), CreateRange(2, 4), INT_MAX));
- ExpectEqual(Value(500), GetDiv(CreateRange(-1000, -40), CreateRange(-4, -2), INT_MAX));
+ ExpectEqual(Value(500), GetDiv(CreateRange(40, 1000), CreateRange(2, 4), false));
+ ExpectEqual(Value(-10), GetDiv(CreateRange(40, 1000), CreateRange(-4, -2), false));
+ ExpectEqual(Value(-10), GetDiv(CreateRange(-1000, -40), CreateRange(2, 4), false));
+ ExpectEqual(Value(500), GetDiv(CreateRange(-1000, -40), CreateRange(-4, -2), false));
}
TEST_F(InductionVarRangeTest, AddValue) {
ExpectEqual(Value(110), AddValue(Value(10), Value(100)));
ExpectEqual(Value(-5), AddValue(Value(&x_, 1, -4), Value(&x_, -1, -1)));
ExpectEqual(Value(&x_, 3, -5), AddValue(Value(&x_, 2, -4), Value(&x_, 1, -1)));
- ExpectEqual(Value(INT_MIN), AddValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
+ ExpectEqual(Value(), AddValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
ExpectEqual(Value(&x_, 1, 23), AddValue(Value(&x_, 1, 20), Value(3)));
ExpectEqual(Value(&y_, 1, 5), AddValue(Value(55), Value(&y_, 1, -50)));
- // Unsafe.
- ExpectEqual(Value(INT_MIN), AddValue(Value(INT_MAX - 5), Value(6)));
+ const int32_t max_value = std::numeric_limits<int32_t>::max();
+ ExpectEqual(Value(max_value), AddValue(Value(max_value - 5), Value(5)));
+ ExpectEqual(Value(), AddValue(Value(max_value - 5), Value(6))); // unsafe
}
TEST_F(InductionVarRangeTest, SubValue) {
ExpectEqual(Value(-90), SubValue(Value(10), Value(100)));
ExpectEqual(Value(-3), SubValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
ExpectEqual(Value(&x_, 2, -3), SubValue(Value(&x_, 3, -4), Value(&x_, 1, -1)));
- ExpectEqual(Value(INT_MIN), SubValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
+ ExpectEqual(Value(), SubValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
ExpectEqual(Value(&x_, 1, 17), SubValue(Value(&x_, 1, 20), Value(3)));
ExpectEqual(Value(&y_, -4, 105), SubValue(Value(55), Value(&y_, 4, -50)));
- // Unsafe.
- ExpectEqual(Value(INT_MIN), SubValue(Value(INT_MIN + 5), Value(6)));
+ const int32_t min_value = std::numeric_limits<int32_t>::min();
+ ExpectEqual(Value(min_value), SubValue(Value(min_value + 5), Value(5)));
+ ExpectEqual(Value(), SubValue(Value(min_value + 5), Value(6))); // unsafe
}
TEST_F(InductionVarRangeTest, MulValue) {
ExpectEqual(Value(1000), MulValue(Value(10), Value(100)));
- ExpectEqual(Value(INT_MIN), MulValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
- ExpectEqual(Value(INT_MIN), MulValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
+ ExpectEqual(Value(), MulValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
+ ExpectEqual(Value(), MulValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
ExpectEqual(Value(&x_, 9, 60), MulValue(Value(&x_, 3, 20), Value(3)));
ExpectEqual(Value(&y_, 55, -110), MulValue(Value(55), Value(&y_, 1, -2)));
- // Unsafe.
- ExpectEqual(Value(INT_MIN), MulValue(Value(90000), Value(-90000)));
+ ExpectEqual(Value(), MulValue(Value(90000), Value(-90000))); // unsafe
}
TEST_F(InductionVarRangeTest, DivValue) {
ExpectEqual(Value(25), DivValue(Value(100), Value(4)));
- ExpectEqual(Value(INT_MIN), DivValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
- ExpectEqual(Value(INT_MIN), DivValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
- ExpectEqual(Value(INT_MIN), DivValue(Value(&x_, 12, 24), Value(3)));
- ExpectEqual(Value(INT_MIN), DivValue(Value(55), Value(&y_, 1, -50)));
- // Unsafe.
- ExpectEqual(Value(INT_MIN), DivValue(Value(1), Value(0)));
+ ExpectEqual(Value(), DivValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
+ ExpectEqual(Value(), DivValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
+ ExpectEqual(Value(), DivValue(Value(&x_, 12, 24), Value(3)));
+ ExpectEqual(Value(), DivValue(Value(55), Value(&y_, 1, -50)));
+ ExpectEqual(Value(), DivValue(Value(1), Value(0))); // unsafe
}
TEST_F(InductionVarRangeTest, MinValue) {
ExpectEqual(Value(10), MinValue(Value(10), Value(100)));
ExpectEqual(Value(&x_, 1, -4), MinValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
ExpectEqual(Value(&x_, 4, -4), MinValue(Value(&x_, 4, -4), Value(&x_, 4, -1)));
- ExpectEqual(Value(INT_MIN), MinValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
- ExpectEqual(Value(INT_MIN), MinValue(Value(&x_, 1, 20), Value(3)));
- ExpectEqual(Value(INT_MIN), MinValue(Value(55), Value(&y_, 1, -50)));
+ ExpectEqual(Value(), MinValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
+ ExpectEqual(Value(), MinValue(Value(&x_, 1, 20), Value(3)));
+ ExpectEqual(Value(), MinValue(Value(55), Value(&y_, 1, -50)));
}
TEST_F(InductionVarRangeTest, MaxValue) {
ExpectEqual(Value(100), MaxValue(Value(10), Value(100)));
ExpectEqual(Value(&x_, 1, -1), MaxValue(Value(&x_, 1, -4), Value(&x_, 1, -1)));
ExpectEqual(Value(&x_, 4, -1), MaxValue(Value(&x_, 4, -4), Value(&x_, 4, -1)));
- ExpectEqual(Value(INT_MAX), MaxValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
- ExpectEqual(Value(INT_MAX), MaxValue(Value(&x_, 1, 20), Value(3)));
- ExpectEqual(Value(INT_MAX), MaxValue(Value(55), Value(&y_, 1, -50)));
+ ExpectEqual(Value(), MaxValue(Value(&x_, 1, 5), Value(&y_, 1, -7)));
+ ExpectEqual(Value(), MaxValue(Value(&x_, 1, 20), Value(3)));
+ ExpectEqual(Value(), MaxValue(Value(55), Value(&y_, 1, -50)));
}
} // namespace art
diff --git a/compiler/optimizing/nodes.cc b/compiler/optimizing/nodes.cc
index b2407c520c..ef89932e3b 100644
--- a/compiler/optimizing/nodes.cc
+++ b/compiler/optimizing/nodes.cc
@@ -20,6 +20,7 @@
#include "ssa_builder.h"
#include "base/bit_vector-inl.h"
#include "base/bit_utils.h"
+#include "base/stl_util.h"
#include "mirror/class-inl.h"
#include "utils/growable_array.h"
#include "scoped_thread_state_change.h"
@@ -32,8 +33,41 @@ void HGraph::AddBlock(HBasicBlock* block) {
}
void HGraph::FindBackEdges(ArenaBitVector* visited) {
+ // "visited" must be empty on entry, it's an output argument for all visited (i.e. live) blocks.
+ DCHECK_EQ(visited->GetHighestBitSet(), -1);
+
+ // Nodes that we're currently visiting, indexed by block id.
ArenaBitVector visiting(arena_, blocks_.size(), false);
- VisitBlockForBackEdges(entry_block_, visited, &visiting);
+ // Number of successors visited from a given node, indexed by block id.
+ ArenaVector<size_t> successors_visited(blocks_.size(), 0u, arena_->Adapter());
+ // Stack of nodes that we're currently visiting (same as marked in "visiting" above).
+ ArenaVector<HBasicBlock*> worklist(arena_->Adapter());
+ constexpr size_t kDefaultWorklistSize = 8;
+ worklist.reserve(kDefaultWorklistSize);
+ visited->SetBit(entry_block_->GetBlockId());
+ visiting.SetBit(entry_block_->GetBlockId());
+ worklist.push_back(entry_block_);
+
+ while (!worklist.empty()) {
+ HBasicBlock* current = worklist.back();
+ uint32_t current_id = current->GetBlockId();
+ if (successors_visited[current_id] == current->GetSuccessors().size()) {
+ visiting.ClearBit(current_id);
+ worklist.pop_back();
+ } else {
+ DCHECK_LT(successors_visited[current_id], current->GetSuccessors().size());
+ HBasicBlock* successor = current->GetSuccessors()[successors_visited[current_id]++];
+ uint32_t successor_id = successor->GetBlockId();
+ if (visiting.IsBitSet(successor_id)) {
+ DCHECK(ContainsElement(worklist, successor));
+ successor->AddBackEdge(current);
+ } else if (!visited->IsBitSet(successor_id)) {
+ visited->SetBit(successor_id);
+ visiting.SetBit(successor_id);
+ worklist.push_back(successor);
+ }
+ }
+ }
}
static void RemoveAsUser(HInstruction* instruction) {
@@ -79,24 +113,6 @@ void HGraph::RemoveDeadBlocks(const ArenaBitVector& visited) {
}
}
-void HGraph::VisitBlockForBackEdges(HBasicBlock* block,
- ArenaBitVector* visited,
- ArenaBitVector* visiting) {
- int id = block->GetBlockId();
- if (visited->IsBitSet(id)) return;
-
- visited->SetBit(id);
- visiting->SetBit(id);
- for (HBasicBlock* successor : block->GetSuccessors()) {
- if (visiting->IsBitSet(successor->GetBlockId())) {
- successor->AddBackEdge(block);
- } else {
- VisitBlockForBackEdges(successor, visited, visiting);
- }
- }
- visiting->ClearBit(id);
-}
-
void HGraph::BuildDominatorTree() {
// (1) Simplify the CFG so that catch blocks have only exceptional incoming
// edges. This invariant simplifies building SSA form because Phis cannot
@@ -141,10 +157,43 @@ void HBasicBlock::ClearDominanceInformation() {
void HGraph::ComputeDominanceInformation() {
DCHECK(reverse_post_order_.empty());
reverse_post_order_.reserve(blocks_.size());
- ArenaVector<size_t> visits(blocks_.size(), 0u, arena_->Adapter());
reverse_post_order_.push_back(entry_block_);
- for (HBasicBlock* successor : entry_block_->GetSuccessors()) {
- VisitBlockForDominatorTree(successor, entry_block_, &visits);
+
+ // Number of visits of a given node, indexed by block id.
+ ArenaVector<size_t> visits(blocks_.size(), 0u, arena_->Adapter());
+ // Number of successors visited from a given node, indexed by block id.
+ ArenaVector<size_t> successors_visited(blocks_.size(), 0u, arena_->Adapter());
+ // Nodes for which we need to visit successors.
+ ArenaVector<HBasicBlock*> worklist(arena_->Adapter());
+ constexpr size_t kDefaultWorklistSize = 8;
+ worklist.reserve(kDefaultWorklistSize);
+ worklist.push_back(entry_block_);
+
+ while (!worklist.empty()) {
+ HBasicBlock* current = worklist.back();
+ uint32_t current_id = current->GetBlockId();
+ if (successors_visited[current_id] == current->GetSuccessors().size()) {
+ worklist.pop_back();
+ } else {
+ DCHECK_LT(successors_visited[current_id], current->GetSuccessors().size());
+ HBasicBlock* successor = current->GetSuccessors()[successors_visited[current_id]++];
+
+ if (successor->GetDominator() == nullptr) {
+ successor->SetDominator(current);
+ } else {
+ successor->SetDominator(FindCommonDominator(successor->GetDominator(), current));
+ }
+
+ // Once all the forward edges have been visited, we know the immediate
+ // dominator of the block. We can then start visiting its successors.
+ DCHECK_LT(successor->GetBlockId(), visits.size());
+ if (++visits[successor->GetBlockId()] ==
+ successor->GetPredecessors().size() - successor->NumberOfBackEdges()) {
+ successor->GetDominator()->AddDominatedBlock(successor);
+ reverse_post_order_.push_back(successor);
+ worklist.push_back(successor);
+ }
+ }
}
}
@@ -166,28 +215,6 @@ HBasicBlock* HGraph::FindCommonDominator(HBasicBlock* first, HBasicBlock* second
return nullptr;
}
-void HGraph::VisitBlockForDominatorTree(HBasicBlock* block,
- HBasicBlock* predecessor,
- ArenaVector<size_t>* visits) {
- if (block->GetDominator() == nullptr) {
- block->SetDominator(predecessor);
- } else {
- block->SetDominator(FindCommonDominator(block->GetDominator(), predecessor));
- }
-
- // Once all the forward edges have been visited, we know the immediate
- // dominator of the block. We can then start visiting its successors.
- DCHECK_LT(block->GetBlockId(), visits->size());
- if (++(*visits)[block->GetBlockId()] ==
- block->GetPredecessors().size() - block->NumberOfBackEdges()) {
- block->GetDominator()->AddDominatedBlock(block);
- reverse_post_order_.push_back(block);
- for (HBasicBlock* successor : block->GetSuccessors()) {
- VisitBlockForDominatorTree(successor, block, visits);
- }
- }
-}
-
void HGraph::TransformToSsa() {
DCHECK(!reverse_post_order_.empty());
SsaBuilder ssa_builder(this);
@@ -1143,6 +1170,23 @@ HBasicBlock* HBasicBlock::SplitBefore(HInstruction* cursor) {
return new_block;
}
+HBasicBlock* HBasicBlock::CreateImmediateDominator() {
+ DCHECK(!graph_->IsInSsaForm()) << "Support for SSA form not implemented";
+ DCHECK(!IsCatchBlock()) << "Support for updating try/catch information not implemented.";
+
+ HBasicBlock* new_block = new (GetGraph()->GetArena()) HBasicBlock(GetGraph(), GetDexPc());
+
+ for (HBasicBlock* predecessor : GetPredecessors()) {
+ new_block->predecessors_.push_back(predecessor);
+ predecessor->successors_[predecessor->GetSuccessorIndexOf(this)] = new_block;
+ }
+ predecessors_.clear();
+ AddPredecessor(new_block);
+
+ GetGraph()->AddBlock(new_block);
+ return new_block;
+}
+
HBasicBlock* HBasicBlock::SplitAfter(HInstruction* cursor) {
DCHECK(!cursor->IsControlFlow());
DCHECK_NE(instructions_.last_instruction_, cursor);
@@ -1188,6 +1232,15 @@ const HTryBoundary* HBasicBlock::ComputeTryEntryOfSuccessors() const {
}
}
+bool HBasicBlock::HasThrowingInstructions() const {
+ for (HInstructionIterator it(GetInstructions()); !it.Done(); it.Advance()) {
+ if (it.Current()->CanThrow()) {
+ return true;
+ }
+ }
+ return false;
+}
+
static bool HasOnlyOneInstruction(const HBasicBlock& block) {
return block.GetPhis().IsEmpty()
&& !block.GetInstructions().IsEmpty()
@@ -1297,16 +1350,25 @@ void HBasicBlock::DisconnectAndDelete() {
// instructions.
for (HBasicBlock* predecessor : predecessors_) {
HInstruction* last_instruction = predecessor->GetLastInstruction();
- predecessor->RemoveInstruction(last_instruction);
predecessor->RemoveSuccessor(this);
- if (predecessor->GetSuccessors().size() == 1u) {
- DCHECK(last_instruction->IsIf());
+ uint32_t num_pred_successors = predecessor->GetSuccessors().size();
+ if (num_pred_successors == 1u) {
+ // If we have one successor after removing one, then we must have
+ // had an HIf or HPackedSwitch, as they have more than one successor.
+ // Replace those with a HGoto.
+ DCHECK(last_instruction->IsIf() || last_instruction->IsPackedSwitch());
+ predecessor->RemoveInstruction(last_instruction);
predecessor->AddInstruction(new (graph_->GetArena()) HGoto(last_instruction->GetDexPc()));
- } else {
+ } else if (num_pred_successors == 0u) {
// The predecessor has no remaining successors and therefore must be dead.
// We deliberately leave it without a control-flow instruction so that the
// SSAChecker fails unless it is not removed during the pass too.
- DCHECK_EQ(predecessor->GetSuccessors().size(), 0u);
+ predecessor->RemoveInstruction(last_instruction);
+ } else {
+ // There are multiple successors left. This must come from a HPackedSwitch
+ // and we are in the middle of removing the HPackedSwitch. Like above, leave
+ // this alone, and the SSAChecker will fail if it is not removed as well.
+ DCHECK(last_instruction->IsPackedSwitch());
}
}
predecessors_.clear();
diff --git a/compiler/optimizing/nodes.h b/compiler/optimizing/nodes.h
index 8dd31bef86..6b0ccf8690 100644
--- a/compiler/optimizing/nodes.h
+++ b/compiler/optimizing/nodes.h
@@ -370,13 +370,7 @@ class HGraph : public ArenaObject<kArenaAllocGraph> {
void SetHasTryCatch(bool value) { has_try_catch_ = value; }
private:
- void VisitBlockForDominatorTree(HBasicBlock* block,
- HBasicBlock* predecessor,
- ArenaVector<size_t>* visits);
void FindBackEdges(ArenaBitVector* visited);
- void VisitBlockForBackEdges(HBasicBlock* block,
- ArenaBitVector* visited,
- ArenaBitVector* visiting);
void RemoveInstructionsAsUsersFromDeadBlocks(const ArenaBitVector& visited) const;
void RemoveDeadBlocks(const ArenaBitVector& visited);
@@ -825,11 +819,17 @@ class HBasicBlock : public ArenaObject<kArenaAllocBasicBlock> {
return EndsWithTryBoundary() ? 1 : GetSuccessors().size();
}
+ // Create a new block between this block and its predecessors. The new block
+ // is added to the graph, all predecessor edges are relinked to it and an edge
+ // is created to `this`. Returns the new empty block. Reverse post order or
+ // loop and try/catch information are not updated.
+ HBasicBlock* CreateImmediateDominator();
+
// Split the block into two blocks just before `cursor`. Returns the newly
// created, latter block. Note that this method will add the block to the
// graph, create a Goto at the end of the former block and will create an edge
// between the blocks. It will not, however, update the reverse post order or
- // loop information.
+ // loop and try/catch information.
HBasicBlock* SplitBefore(HInstruction* cursor);
// Split the block into two blocks just after `cursor`. Returns the newly
@@ -940,6 +940,8 @@ class HBasicBlock : public ArenaObject<kArenaAllocBasicBlock> {
// the appropriate try entry will be returned.
const HTryBoundary* ComputeTryEntryOfSuccessors() const;
+ bool HasThrowingInstructions() const;
+
// Returns whether this block dominates the blocked passed as parameter.
bool Dominates(HBasicBlock* block) const;
@@ -949,7 +951,6 @@ class HBasicBlock : public ArenaObject<kArenaAllocBasicBlock> {
void SetLifetimeStart(size_t start) { lifetime_start_ = start; }
void SetLifetimeEnd(size_t end) { lifetime_end_ = end; }
-
bool EndsWithControlFlowInstruction() const;
bool EndsWithIf() const;
bool EndsWithTryBoundary() const;
@@ -1056,6 +1057,7 @@ class HLoopInformationOutwardIterator : public ValueObject {
M(NullConstant, Instruction) \
M(NullCheck, Instruction) \
M(Or, BinaryOperation) \
+ M(PackedSwitch, Instruction) \
M(ParallelMove, Instruction) \
M(ParameterValue, Instruction) \
M(Phi, Instruction) \
@@ -2402,6 +2404,38 @@ class HCurrentMethod : public HExpression<0> {
DISALLOW_COPY_AND_ASSIGN(HCurrentMethod);
};
+// PackedSwitch (jump table). A block ending with a PackedSwitch instruction will
+// have one successor for each entry in the switch table, and the final successor
+// will be the block containing the next Dex opcode.
+class HPackedSwitch : public HTemplateInstruction<1> {
+ public:
+ HPackedSwitch(int32_t start_value, uint32_t num_entries, HInstruction* input,
+ uint32_t dex_pc = kNoDexPc)
+ : HTemplateInstruction(SideEffects::None(), dex_pc),
+ start_value_(start_value),
+ num_entries_(num_entries) {
+ SetRawInputAt(0, input);
+ }
+
+ bool IsControlFlow() const OVERRIDE { return true; }
+
+ int32_t GetStartValue() const { return start_value_; }
+
+ uint32_t GetNumEntries() const { return num_entries_; }
+
+ HBasicBlock* GetDefaultBlock() const {
+ // Last entry is the default block.
+ return GetBlock()->GetSuccessor(num_entries_);
+ }
+ DECLARE_INSTRUCTION(PackedSwitch);
+
+ private:
+ int32_t start_value_;
+ uint32_t num_entries_;
+
+ DISALLOW_COPY_AND_ASSIGN(HPackedSwitch);
+};
+
class HUnaryOperation : public HExpression<1> {
public:
HUnaryOperation(Primitive::Type result_type, HInstruction* input, uint32_t dex_pc = kNoDexPc)
diff --git a/compiler/optimizing/register_allocator.cc b/compiler/optimizing/register_allocator.cc
index e6b2a8bb3e..c43e58ffc4 100644
--- a/compiler/optimizing/register_allocator.cc
+++ b/compiler/optimizing/register_allocator.cc
@@ -1527,10 +1527,10 @@ void RegisterAllocator::InsertParallelMoveAtExitOf(HBasicBlock* block,
DCHECK_EQ(block->NumberOfNormalSuccessors(), 1u);
HInstruction* last = block->GetLastInstruction();
// We insert moves at exit for phi predecessors and connecting blocks.
- // A block ending with an if cannot branch to a block with phis because
- // we do not allow critical edges. It can also not connect
+ // A block ending with an if or a packed switch cannot branch to a block
+ // with phis because we do not allow critical edges. It can also not connect
// a split interval between two blocks: the move has to happen in the successor.
- DCHECK(!last->IsIf());
+ DCHECK(!last->IsIf() && !last->IsPackedSwitch());
HInstruction* previous = last->GetPrevious();
HParallelMove* move;
// This is a parallel move for connecting blocks. We need to differentiate
diff --git a/compiler/optimizing/ssa_builder.cc b/compiler/optimizing/ssa_builder.cc
index 0ef86d80ed..ad8c682b3a 100644
--- a/compiler/optimizing/ssa_builder.cc
+++ b/compiler/optimizing/ssa_builder.cc
@@ -191,12 +191,6 @@ void DeadPhiHandling::Run() {
ProcessWorklist();
}
-static bool IsPhiEquivalentOf(HInstruction* instruction, HPhi* phi) {
- return instruction != nullptr
- && instruction->IsPhi()
- && instruction->AsPhi()->GetRegNumber() == phi->GetRegNumber();
-}
-
void SsaBuilder::FixNullConstantType() {
// The order doesn't matter here.
for (HReversePostOrderIterator itb(*GetGraph()); !itb.Done(); itb.Advance()) {
@@ -324,13 +318,13 @@ void SsaBuilder::BuildSsa() {
// If the phi is not dead, or has no environment uses, there is nothing to do.
if (!phi->IsDead() || !phi->HasEnvironmentUses()) continue;
HInstruction* next = phi->GetNext();
- if (!IsPhiEquivalentOf(next, phi)) continue;
+ if (!phi->IsVRegEquivalentOf(next)) continue;
if (next->AsPhi()->IsDead()) {
// If the phi equivalent is dead, check if there is another one.
next = next->GetNext();
- if (!IsPhiEquivalentOf(next, phi)) continue;
+ if (!phi->IsVRegEquivalentOf(next)) continue;
// There can be at most two phi equivalents.
- DCHECK(!IsPhiEquivalentOf(next->GetNext(), phi));
+ DCHECK(!phi->IsVRegEquivalentOf(next->GetNext()));
if (next->AsPhi()->IsDead()) continue;
}
// We found a live phi equivalent. Update the environment uses of `phi` with it.
@@ -403,6 +397,24 @@ void SsaBuilder::VisitBasicBlock(HBasicBlock* block) {
if (block->IsCatchBlock()) {
// Catch phis were already created and inputs collected from throwing sites.
+ if (kIsDebugBuild) {
+ // Make sure there was at least one throwing instruction which initialized
+ // locals (guaranteed by HGraphBuilder) and that all try blocks have been
+ // visited already (from HTryBoundary scoping and reverse post order).
+ bool throwing_instruction_found = false;
+ bool catch_block_visited = false;
+ for (HReversePostOrderIterator it(*GetGraph()); !it.Done(); it.Advance()) {
+ HBasicBlock* current = it.Current();
+ if (current == block) {
+ catch_block_visited = true;
+ } else if (current->IsTryBlock() &&
+ current->GetTryCatchInformation()->GetTryEntry().HasExceptionHandler(*block)) {
+ DCHECK(!catch_block_visited) << "Catch block visited before its try block.";
+ throwing_instruction_found |= current->HasThrowingInstructions();
+ }
+ }
+ DCHECK(throwing_instruction_found) << "No instructions throwing into a live catch block.";
+ }
} else if (block->IsLoopHeader()) {
// If the block is a loop header, we know we only have visited the pre header
// because we are visiting in reverse post order. We create phis for all initialized