blob: 63cef7e3482dc16cb0649422307e66ccc2a7fb6e [file] [log] [blame]
/*
* Copyright (C) 2015 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/* This file contains codegen for the Mips64 ISA */
#include "codegen_mips64.h"
#include "base/logging.h"
#include "dex/mir_graph.h"
#include "dex/quick/mir_to_lir-inl.h"
#include "entrypoints/quick/quick_entrypoints.h"
#include "gc/accounting/card_table.h"
#include "mips64_lir.h"
#include "mirror/art_method.h"
#include "mirror/object_array-inl.h"
namespace art {
bool Mips64Mir2Lir::GenSpecialCase(BasicBlock* bb, MIR* mir, const InlineMethod& special) {
// TODO
UNUSED(bb, mir, special);
return false;
}
/*
* The lack of pc-relative loads on Mips64 presents somewhat of a challenge
* for our PIC switch table strategy. To materialize the current location
* we'll do a dummy JAL and reference our tables using rRA as the
* base register. Note that rRA will be used both as the base to
* locate the switch table data and as the reference base for the switch
* target offsets stored in the table. We'll use a special pseudo-instruction
* to represent the jal and trigger the construction of the
* switch table offsets (which will happen after final assembly and all
* labels are fixed).
*
* The test loop will look something like:
*
* ori r_end, rZERO, #table_size ; size in bytes
* jal BaseLabel ; stores "return address" (BaseLabel) in rRA
* nop ; opportunistically fill
* BaseLabel:
* addiu r_base, rRA, <table> - <BaseLabel> ; table relative to BaseLabel
addu r_end, r_end, r_base ; end of table
* lw r_val, [rSP, v_reg_off] ; Test Value
* loop:
* beq r_base, r_end, done
* lw r_key, 0(r_base)
* addu r_base, 8
* bne r_val, r_key, loop
* lw r_disp, -4(r_base)
* addu rRA, r_disp
* jalr rZERO, rRA
* done:
*
*/
void Mips64Mir2Lir::GenLargeSparseSwitch(MIR* mir, DexOffset table_offset, RegLocation rl_src) {
const uint16_t* table = mir_graph_->GetTable(mir, table_offset);
// Add the table to the list - we'll process it later.
SwitchTable* tab_rec = static_cast<SwitchTable*>(arena_->Alloc(sizeof(SwitchTable),
kArenaAllocData));
tab_rec->switch_mir = mir;
tab_rec->table = table;
tab_rec->vaddr = current_dalvik_offset_;
int elements = table[1];
switch_tables_.push_back(tab_rec);
// The table is composed of 8-byte key/disp pairs.
int byte_size = elements * 8;
int size_hi = byte_size >> 16;
int size_lo = byte_size & 0xffff;
RegStorage r_end = AllocTempWide();
if (size_hi) {
NewLIR2(kMips64Lui, r_end.GetReg(), size_hi);
}
// Must prevent code motion for the curr pc pair.
GenBarrier(); // Scheduling barrier.
NewLIR0(kMips64CurrPC); // Really a jal to .+8.
// Now, fill the branch delay slot.
if (size_hi) {
NewLIR3(kMips64Ori, r_end.GetReg(), r_end.GetReg(), size_lo);
} else {
NewLIR3(kMips64Ori, r_end.GetReg(), rZERO, size_lo);
}
GenBarrier(); // Scheduling barrier.
// Construct BaseLabel and set up table base register.
LIR* base_label = NewLIR0(kPseudoTargetLabel);
// Remember base label so offsets can be computed later.
tab_rec->anchor = base_label;
RegStorage r_base = AllocTempWide();
NewLIR4(kMips64Delta, r_base.GetReg(), 0, WrapPointer(base_label), WrapPointer(tab_rec));
OpRegRegReg(kOpAdd, r_end, r_end, r_base);
// Grab switch test value.
rl_src = LoadValue(rl_src, kCoreReg);
// Test loop.
RegStorage r_key = AllocTemp();
LIR* loop_label = NewLIR0(kPseudoTargetLabel);
LIR* exit_branch = OpCmpBranch(kCondEq, r_base, r_end, NULL);
Load32Disp(r_base, 0, r_key);
OpRegImm(kOpAdd, r_base, 8);
OpCmpBranch(kCondNe, rl_src.reg, r_key, loop_label);
RegStorage r_disp = AllocTemp();
Load32Disp(r_base, -4, r_disp);
OpRegRegReg(kOpAdd, TargetReg(kLr, kWide), TargetReg(kLr, kWide), r_disp);
OpReg(kOpBx, TargetReg(kLr, kWide));
// Loop exit.
LIR* exit_label = NewLIR0(kPseudoTargetLabel);
exit_branch->target = exit_label;
}
/*
* Code pattern will look something like:
*
* lw r_val
* jal BaseLabel ; stores "return address" (BaseLabel) in rRA
* nop ; opportunistically fill
* [subiu r_val, bias] ; Remove bias if low_val != 0
* bound check -> done
* lw r_disp, [rRA, r_val]
* addu rRA, r_disp
* jalr rZERO, rRA
* done:
*/
void Mips64Mir2Lir::GenLargePackedSwitch(MIR* mir, DexOffset table_offset, RegLocation rl_src) {
const uint16_t* table = mir_graph_->GetTable(mir, table_offset);
// Add the table to the list - we'll process it later.
SwitchTable* tab_rec =
static_cast<SwitchTable*>(arena_->Alloc(sizeof(SwitchTable), kArenaAllocData));
tab_rec->switch_mir = mir;
tab_rec->table = table;
tab_rec->vaddr = current_dalvik_offset_;
int size = table[1];
switch_tables_.push_back(tab_rec);
// Get the switch value.
rl_src = LoadValue(rl_src, kCoreReg);
// Prepare the bias. If too big, handle 1st stage here.
int low_key = s4FromSwitchData(&table[2]);
bool large_bias = false;
RegStorage r_key;
if (low_key == 0) {
r_key = rl_src.reg;
} else if ((low_key & 0xffff) != low_key) {
r_key = AllocTemp();
LoadConstant(r_key, low_key);
large_bias = true;
} else {
r_key = AllocTemp();
}
// Must prevent code motion for the curr pc pair.
GenBarrier();
NewLIR0(kMips64CurrPC); // Really a jal to .+8.
// Now, fill the branch delay slot with bias strip.
if (low_key == 0) {
NewLIR0(kMips64Nop);
} else {
if (large_bias) {
OpRegRegReg(kOpSub, r_key, rl_src.reg, r_key);
} else {
OpRegRegImm(kOpSub, r_key, rl_src.reg, low_key);
}
}
GenBarrier(); // Scheduling barrier.
// Construct BaseLabel and set up table base register.
LIR* base_label = NewLIR0(kPseudoTargetLabel);
// Remember base label so offsets can be computed later.
tab_rec->anchor = base_label;
// Bounds check - if < 0 or >= size continue following switch.
LIR* branch_over = OpCmpImmBranch(kCondHi, r_key, size-1, NULL);
// Materialize the table base pointer.
RegStorage r_base = AllocTempWide();
NewLIR4(kMips64Delta, r_base.GetReg(), 0, WrapPointer(base_label), WrapPointer(tab_rec));
// Load the displacement from the switch table.
RegStorage r_disp = AllocTemp();
LoadBaseIndexed(r_base, r_key, r_disp, 2, k32);
// Add to rAP and go.
OpRegRegReg(kOpAdd, TargetReg(kLr, kWide), TargetReg(kLr, kWide), r_disp);
OpReg(kOpBx, TargetReg(kLr, kWide));
// Branch_over target here.
LIR* target = NewLIR0(kPseudoTargetLabel);
branch_over->target = target;
}
void Mips64Mir2Lir::GenMoveException(RegLocation rl_dest) {
int ex_offset = Thread::ExceptionOffset<8>().Int32Value();
RegLocation rl_result = EvalLoc(rl_dest, kRefReg, true);
RegStorage reset_reg = AllocTempRef();
LoadRefDisp(rs_rMIPS64_SELF, ex_offset, rl_result.reg, kNotVolatile);
LoadConstant(reset_reg, 0);
StoreRefDisp(rs_rMIPS64_SELF, ex_offset, reset_reg, kNotVolatile);
FreeTemp(reset_reg);
StoreValue(rl_dest, rl_result);
}
void Mips64Mir2Lir::UnconditionallyMarkGCCard(RegStorage tgt_addr_reg) {
RegStorage reg_card_base = AllocTempWide();
RegStorage reg_card_no = AllocTempWide();
// NOTE: native pointer.
LoadWordDisp(rs_rMIPS64_SELF, Thread::CardTableOffset<8>().Int32Value(), reg_card_base);
OpRegRegImm(kOpLsr, reg_card_no, tgt_addr_reg, gc::accounting::CardTable::kCardShift);
StoreBaseIndexed(reg_card_base, reg_card_no, As32BitReg(reg_card_base), 0, kUnsignedByte);
FreeTemp(reg_card_base);
FreeTemp(reg_card_no);
}
void Mips64Mir2Lir::GenEntrySequence(RegLocation* ArgLocs, RegLocation rl_method) {
int spill_count = num_core_spills_ + num_fp_spills_;
/*
* On entry, rMIPS64_ARG0, rMIPS64_ARG1, rMIPS64_ARG2, rMIPS64_ARG3,
* rMIPS64_ARG4, rMIPS64_ARG5, rMIPS64_ARG6 & rMIPS64_ARG7 are live.
* Let the register allocation mechanism know so it doesn't try to
* use any of them when expanding the frame or flushing.
*/
LockTemp(rs_rMIPS64_ARG0);
LockTemp(rs_rMIPS64_ARG1);
LockTemp(rs_rMIPS64_ARG2);
LockTemp(rs_rMIPS64_ARG3);
LockTemp(rs_rMIPS64_ARG4);
LockTemp(rs_rMIPS64_ARG5);
LockTemp(rs_rMIPS64_ARG6);
LockTemp(rs_rMIPS64_ARG7);
/*
* We can safely skip the stack overflow check if we're
* a leaf *and* our frame size < fudge factor.
*/
bool skip_overflow_check = mir_graph_->MethodIsLeaf() && !FrameNeedsStackCheck(frame_size_,
kMips64);
NewLIR0(kPseudoMethodEntry);
RegStorage check_reg = AllocTempWide();
RegStorage new_sp = AllocTempWide();
if (!skip_overflow_check) {
// Load stack limit.
LoadWordDisp(rs_rMIPS64_SELF, Thread::StackEndOffset<8>().Int32Value(), check_reg);
}
// Spill core callee saves.
SpillCoreRegs();
// NOTE: promotion of FP regs currently unsupported, thus no FP spill.
DCHECK_EQ(num_fp_spills_, 0);
const int frame_sub = frame_size_ - spill_count * 8;
if (!skip_overflow_check) {
class StackOverflowSlowPath : public LIRSlowPath {
public:
StackOverflowSlowPath(Mir2Lir* m2l, LIR* branch, size_t sp_displace)
: LIRSlowPath(m2l, m2l->GetCurrentDexPc(), branch, nullptr), sp_displace_(sp_displace) {
}
void Compile() OVERRIDE {
m2l_->ResetRegPool();
m2l_->ResetDefTracking();
GenerateTargetLabel(kPseudoThrowTarget);
// Load RA from the top of the frame.
m2l_->LoadWordDisp(rs_rMIPS64_SP, sp_displace_ - 8, rs_rRAd);
m2l_->OpRegImm(kOpAdd, rs_rMIPS64_SP, sp_displace_);
m2l_->ClobberCallerSave();
RegStorage r_tgt = m2l_->CallHelperSetup(kQuickThrowStackOverflow); // Doesn't clobber LR.
m2l_->CallHelper(r_tgt, kQuickThrowStackOverflow, false /* MarkSafepointPC */,
false /* UseLink */);
}
private:
const size_t sp_displace_;
};
OpRegRegImm(kOpSub, new_sp, rs_rMIPS64_SP, frame_sub);
LIR* branch = OpCmpBranch(kCondUlt, new_sp, check_reg, nullptr);
AddSlowPath(new(arena_)StackOverflowSlowPath(this, branch, spill_count * 8));
// TODO: avoid copy for small frame sizes.
OpRegCopy(rs_rMIPS64_SP, new_sp); // Establish stack.
} else {
OpRegImm(kOpSub, rs_rMIPS64_SP, frame_sub);
}
FlushIns(ArgLocs, rl_method);
FreeTemp(rs_rMIPS64_ARG0);
FreeTemp(rs_rMIPS64_ARG1);
FreeTemp(rs_rMIPS64_ARG2);
FreeTemp(rs_rMIPS64_ARG3);
FreeTemp(rs_rMIPS64_ARG4);
FreeTemp(rs_rMIPS64_ARG5);
FreeTemp(rs_rMIPS64_ARG6);
FreeTemp(rs_rMIPS64_ARG7);
}
void Mips64Mir2Lir::GenExitSequence() {
/*
* In the exit path, rMIPS64_RET0/rMIPS64_RET1 are live - make sure they aren't
* allocated by the register utilities as temps.
*/
LockTemp(rs_rMIPS64_RET0);
LockTemp(rs_rMIPS64_RET1);
NewLIR0(kPseudoMethodExit);
UnSpillCoreRegs();
OpReg(kOpBx, rs_rRAd);
}
void Mips64Mir2Lir::GenSpecialExitSequence() {
OpReg(kOpBx, rs_rRAd);
}
void Mips64Mir2Lir::GenSpecialEntryForSuspend() {
// Keep 16-byte stack alignment - push A0, i.e. ArtMethod* and RA.
core_spill_mask_ = (1u << rs_rRAd.GetRegNum());
num_core_spills_ = 1u;
fp_spill_mask_ = 0u;
num_fp_spills_ = 0u;
frame_size_ = 16u;
core_vmap_table_.clear();
fp_vmap_table_.clear();
OpRegImm(kOpSub, rs_rMIPS64_SP, frame_size_);
StoreWordDisp(rs_rMIPS64_SP, frame_size_ - 8, rs_rRAd);
StoreWordDisp(rs_rMIPS64_SP, 0, rs_rA0d);
}
void Mips64Mir2Lir::GenSpecialExitForSuspend() {
// Pop the frame. Don't pop ArtMethod*, it's no longer needed.
LoadWordDisp(rs_rMIPS64_SP, frame_size_ - 8, rs_rRAd);
OpRegImm(kOpAdd, rs_rMIPS64_SP, frame_size_);
}
/*
* Bit of a hack here - in the absence of a real scheduling pass,
* emit the next instruction in static & direct invoke sequences.
*/
static int Mips64NextSDCallInsn(CompilationUnit* cu, CallInfo* info ATTRIBUTE_UNUSED, int state,
const MethodReference& target_method, uint32_t,
uintptr_t direct_code, uintptr_t direct_method, InvokeType type) {
Mir2Lir* cg = static_cast<Mir2Lir*>(cu->cg.get());
if (direct_code != 0 && direct_method != 0) {
switch (state) {
case 0: // Get the current Method* [sets kArg0]
if (direct_code != static_cast<uintptr_t>(-1)) {
cg->LoadConstant(cg->TargetPtrReg(kInvokeTgt), direct_code);
} else {
cg->LoadCodeAddress(target_method, type, kInvokeTgt);
}
if (direct_method != static_cast<uintptr_t>(-1)) {
cg->LoadConstant(cg->TargetReg(kArg0, kRef), direct_method);
} else {
cg->LoadMethodAddress(target_method, type, kArg0);
}
break;
default:
return -1;
}
} else {
RegStorage arg0_ref = cg->TargetReg(kArg0, kRef);
switch (state) {
case 0: // Get the current Method* [sets kArg0]
// TUNING: we can save a reg copy if Method* has been promoted.
cg->LoadCurrMethodDirect(arg0_ref);
break;
case 1: // Get method->dex_cache_resolved_methods_
cg->LoadRefDisp(arg0_ref, mirror::ArtMethod::DexCacheResolvedMethodsOffset().Int32Value(),
arg0_ref, kNotVolatile);
// Set up direct code if known.
if (direct_code != 0) {
if (direct_code != static_cast<uintptr_t>(-1)) {
cg->LoadConstant(cg->TargetPtrReg(kInvokeTgt), direct_code);
} else {
CHECK_LT(target_method.dex_method_index, target_method.dex_file->NumMethodIds());
cg->LoadCodeAddress(target_method, type, kInvokeTgt);
}
}
break;
case 2: // Grab target method*
CHECK_EQ(cu->dex_file, target_method.dex_file);
cg->LoadRefDisp(arg0_ref, mirror::ObjectArray<mirror::Object>::
OffsetOfElement(target_method.dex_method_index).Int32Value(), arg0_ref,
kNotVolatile);
break;
case 3: // Grab the code from the method*
if (direct_code == 0) {
int32_t offset = mirror::ArtMethod::EntryPointFromQuickCompiledCodeOffset(
InstructionSetPointerSize(cu->instruction_set)).Int32Value();
// Get the compiled code address [use *alt_from or kArg0, set kInvokeTgt]
cg->LoadWordDisp(arg0_ref, offset, cg->TargetPtrReg(kInvokeTgt));
}
break;
default:
return -1;
}
}
return state + 1;
}
NextCallInsn Mips64Mir2Lir::GetNextSDCallInsn() {
return Mips64NextSDCallInsn;
}
LIR* Mips64Mir2Lir::GenCallInsn(const MirMethodLoweringInfo& method_info ATTRIBUTE_UNUSED) {
return OpReg(kOpBlx, TargetPtrReg(kInvokeTgt));
}
} // namespace art