1 //===-- X86FloatingPoint.cpp - Floating point Reg -> Stack converter ------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file defines the pass which converts floating point instructions from
10 // pseudo registers into register stack instructions. This pass uses live
11 // variable information to indicate where the FPn registers are used and their
12 // lifetimes.
13 //
14 // The x87 hardware tracks liveness of the stack registers, so it is necessary
15 // to implement exact liveness tracking between basic blocks. The CFG edges are
16 // partitioned into bundles where the same FP registers must be live in
17 // identical stack positions. Instructions are inserted at the end of each basic
18 // block to rearrange the live registers to match the outgoing bundle.
19 //
20 // This approach avoids splitting critical edges at the potential cost of more
21 // live register shuffling instructions when critical edges are present.
22 //
23 //===----------------------------------------------------------------------===//
24
25 #include "X86.h"
26 #include "X86InstrInfo.h"
27 #include "llvm/ADT/DepthFirstIterator.h"
28 #include "llvm/ADT/STLExtras.h"
29 #include "llvm/ADT/SmallPtrSet.h"
30 #include "llvm/ADT/SmallSet.h"
31 #include "llvm/ADT/SmallVector.h"
32 #include "llvm/ADT/Statistic.h"
33 #include "llvm/CodeGen/EdgeBundles.h"
34 #include "llvm/CodeGen/LivePhysRegs.h"
35 #include "llvm/CodeGen/MachineFunctionPass.h"
36 #include "llvm/CodeGen/MachineInstrBuilder.h"
37 #include "llvm/CodeGen/MachineRegisterInfo.h"
38 #include "llvm/CodeGen/Passes.h"
39 #include "llvm/CodeGen/TargetInstrInfo.h"
40 #include "llvm/CodeGen/TargetSubtargetInfo.h"
41 #include "llvm/Config/llvm-config.h"
42 #include "llvm/IR/InlineAsm.h"
43 #include "llvm/InitializePasses.h"
44 #include "llvm/Support/Debug.h"
45 #include "llvm/Support/ErrorHandling.h"
46 #include "llvm/Support/raw_ostream.h"
47 #include "llvm/Target/TargetMachine.h"
48 #include <algorithm>
49 #include <bitset>
50 using namespace llvm;
51
52 #define DEBUG_TYPE "x86-codegen"
53
54 STATISTIC(NumFXCH, "Number of fxch instructions inserted");
55 STATISTIC(NumFP , "Number of floating point instructions");
56
57 namespace {
58 const unsigned ScratchFPReg = 7;
59
60 struct FPS : public MachineFunctionPass {
61 static char ID;
FPS__anon3deeeb160111::FPS62 FPS() : MachineFunctionPass(ID) {
63 // This is really only to keep valgrind quiet.
64 // The logic in isLive() is too much for it.
65 memset(Stack, 0, sizeof(Stack));
66 memset(RegMap, 0, sizeof(RegMap));
67 }
68
getAnalysisUsage__anon3deeeb160111::FPS69 void getAnalysisUsage(AnalysisUsage &AU) const override {
70 AU.setPreservesCFG();
71 AU.addRequired<EdgeBundles>();
72 AU.addPreservedID(MachineLoopInfoID);
73 AU.addPreservedID(MachineDominatorsID);
74 MachineFunctionPass::getAnalysisUsage(AU);
75 }
76
77 bool runOnMachineFunction(MachineFunction &MF) override;
78
getRequiredProperties__anon3deeeb160111::FPS79 MachineFunctionProperties getRequiredProperties() const override {
80 return MachineFunctionProperties().set(
81 MachineFunctionProperties::Property::NoVRegs);
82 }
83
getPassName__anon3deeeb160111::FPS84 StringRef getPassName() const override { return "X86 FP Stackifier"; }
85
86 private:
87 const TargetInstrInfo *TII = nullptr; // Machine instruction info.
88
89 // Two CFG edges are related if they leave the same block, or enter the same
90 // block. The transitive closure of an edge under this relation is a
91 // LiveBundle. It represents a set of CFG edges where the live FP stack
92 // registers must be allocated identically in the x87 stack.
93 //
94 // A LiveBundle is usually all the edges leaving a block, or all the edges
95 // entering a block, but it can contain more edges if critical edges are
96 // present.
97 //
98 // The set of live FP registers in a LiveBundle is calculated by bundleCFG,
99 // but the exact mapping of FP registers to stack slots is fixed later.
100 struct LiveBundle {
101 // Bit mask of live FP registers. Bit 0 = FP0, bit 1 = FP1, &c.
102 unsigned Mask;
103
104 // Number of pre-assigned live registers in FixStack. This is 0 when the
105 // stack order has not yet been fixed.
106 unsigned FixCount;
107
108 // Assigned stack order for live-in registers.
109 // FixStack[i] == getStackEntry(i) for all i < FixCount.
110 unsigned char FixStack[8];
111
LiveBundle__anon3deeeb160111::FPS::LiveBundle112 LiveBundle() : Mask(0), FixCount(0) {}
113
114 // Have the live registers been assigned a stack order yet?
isFixed__anon3deeeb160111::FPS::LiveBundle115 bool isFixed() const { return !Mask || FixCount; }
116 };
117
118 // Numbered LiveBundle structs. LiveBundles[0] is used for all CFG edges
119 // with no live FP registers.
120 SmallVector<LiveBundle, 8> LiveBundles;
121
122 // The edge bundle analysis provides indices into the LiveBundles vector.
123 EdgeBundles *Bundles = nullptr;
124
125 // Return a bitmask of FP registers in block's live-in list.
calcLiveInMask__anon3deeeb160111::FPS126 static unsigned calcLiveInMask(MachineBasicBlock *MBB, bool RemoveFPs) {
127 unsigned Mask = 0;
128 for (MachineBasicBlock::livein_iterator I = MBB->livein_begin();
129 I != MBB->livein_end(); ) {
130 MCPhysReg Reg = I->PhysReg;
131 static_assert(X86::FP6 - X86::FP0 == 6, "sequential regnums");
132 if (Reg >= X86::FP0 && Reg <= X86::FP6) {
133 Mask |= 1 << (Reg - X86::FP0);
134 if (RemoveFPs) {
135 I = MBB->removeLiveIn(I);
136 continue;
137 }
138 }
139 ++I;
140 }
141 return Mask;
142 }
143
144 // Partition all the CFG edges into LiveBundles.
145 void bundleCFGRecomputeKillFlags(MachineFunction &MF);
146
147 MachineBasicBlock *MBB = nullptr; // Current basic block
148
149 // The hardware keeps track of how many FP registers are live, so we have
150 // to model that exactly. Usually, each live register corresponds to an
151 // FP<n> register, but when dealing with calls, returns, and inline
152 // assembly, it is sometimes necessary to have live scratch registers.
153 unsigned Stack[8]; // FP<n> Registers in each stack slot...
154 unsigned StackTop = 0; // The current top of the FP stack.
155
156 enum {
157 NumFPRegs = 8 // Including scratch pseudo-registers.
158 };
159
160 // For each live FP<n> register, point to its Stack[] entry.
161 // The first entries correspond to FP0-FP6, the rest are scratch registers
162 // used when we need slightly different live registers than what the
163 // register allocator thinks.
164 unsigned RegMap[NumFPRegs];
165
166 // Set up our stack model to match the incoming registers to MBB.
167 void setupBlockStack();
168
169 // Shuffle live registers to match the expectations of successor blocks.
170 void finishBlockStack();
171
172 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
dumpStack__anon3deeeb160111::FPS173 void dumpStack() const {
174 dbgs() << "Stack contents:";
175 for (unsigned i = 0; i != StackTop; ++i) {
176 dbgs() << " FP" << Stack[i];
177 assert(RegMap[Stack[i]] == i && "Stack[] doesn't match RegMap[]!");
178 }
179 }
180 #endif
181
182 /// getSlot - Return the stack slot number a particular register number is
183 /// in.
getSlot__anon3deeeb160111::FPS184 unsigned getSlot(unsigned RegNo) const {
185 assert(RegNo < NumFPRegs && "Regno out of range!");
186 return RegMap[RegNo];
187 }
188
189 /// isLive - Is RegNo currently live in the stack?
isLive__anon3deeeb160111::FPS190 bool isLive(unsigned RegNo) const {
191 unsigned Slot = getSlot(RegNo);
192 return Slot < StackTop && Stack[Slot] == RegNo;
193 }
194
195 /// getStackEntry - Return the X86::FP<n> register in register ST(i).
getStackEntry__anon3deeeb160111::FPS196 unsigned getStackEntry(unsigned STi) const {
197 if (STi >= StackTop)
198 report_fatal_error("Access past stack top!");
199 return Stack[StackTop-1-STi];
200 }
201
202 /// getSTReg - Return the X86::ST(i) register which contains the specified
203 /// FP<RegNo> register.
getSTReg__anon3deeeb160111::FPS204 unsigned getSTReg(unsigned RegNo) const {
205 return StackTop - 1 - getSlot(RegNo) + X86::ST0;
206 }
207
208 // pushReg - Push the specified FP<n> register onto the stack.
pushReg__anon3deeeb160111::FPS209 void pushReg(unsigned Reg) {
210 assert(Reg < NumFPRegs && "Register number out of range!");
211 if (StackTop >= 8)
212 report_fatal_error("Stack overflow!");
213 Stack[StackTop] = Reg;
214 RegMap[Reg] = StackTop++;
215 }
216
217 // popReg - Pop a register from the stack.
popReg__anon3deeeb160111::FPS218 void popReg() {
219 if (StackTop == 0)
220 report_fatal_error("Cannot pop empty stack!");
221 RegMap[Stack[--StackTop]] = ~0; // Update state
222 }
223
isAtTop__anon3deeeb160111::FPS224 bool isAtTop(unsigned RegNo) const { return getSlot(RegNo) == StackTop-1; }
moveToTop__anon3deeeb160111::FPS225 void moveToTop(unsigned RegNo, MachineBasicBlock::iterator I) {
226 DebugLoc dl = I == MBB->end() ? DebugLoc() : I->getDebugLoc();
227 if (isAtTop(RegNo)) return;
228
229 unsigned STReg = getSTReg(RegNo);
230 unsigned RegOnTop = getStackEntry(0);
231
232 // Swap the slots the regs are in.
233 std::swap(RegMap[RegNo], RegMap[RegOnTop]);
234
235 // Swap stack slot contents.
236 if (RegMap[RegOnTop] >= StackTop)
237 report_fatal_error("Access past stack top!");
238 std::swap(Stack[RegMap[RegOnTop]], Stack[StackTop-1]);
239
240 // Emit an fxch to update the runtime processors version of the state.
241 BuildMI(*MBB, I, dl, TII->get(X86::XCH_F)).addReg(STReg);
242 ++NumFXCH;
243 }
244
duplicateToTop__anon3deeeb160111::FPS245 void duplicateToTop(unsigned RegNo, unsigned AsReg,
246 MachineBasicBlock::iterator I) {
247 DebugLoc dl = I == MBB->end() ? DebugLoc() : I->getDebugLoc();
248 unsigned STReg = getSTReg(RegNo);
249 pushReg(AsReg); // New register on top of stack
250
251 BuildMI(*MBB, I, dl, TII->get(X86::LD_Frr)).addReg(STReg);
252 }
253
254 /// popStackAfter - Pop the current value off of the top of the FP stack
255 /// after the specified instruction.
256 void popStackAfter(MachineBasicBlock::iterator &I);
257
258 /// freeStackSlotAfter - Free the specified register from the register
259 /// stack, so that it is no longer in a register. If the register is
260 /// currently at the top of the stack, we just pop the current instruction,
261 /// otherwise we store the current top-of-stack into the specified slot,
262 /// then pop the top of stack.
263 void freeStackSlotAfter(MachineBasicBlock::iterator &I, unsigned Reg);
264
265 /// freeStackSlotBefore - Just the pop, no folding. Return the inserted
266 /// instruction.
267 MachineBasicBlock::iterator
268 freeStackSlotBefore(MachineBasicBlock::iterator I, unsigned FPRegNo);
269
270 /// Adjust the live registers to be the set in Mask.
271 void adjustLiveRegs(unsigned Mask, MachineBasicBlock::iterator I);
272
273 /// Shuffle the top FixCount stack entries such that FP reg FixStack[0] is
274 /// st(0), FP reg FixStack[1] is st(1) etc.
275 void shuffleStackTop(const unsigned char *FixStack, unsigned FixCount,
276 MachineBasicBlock::iterator I);
277
278 bool processBasicBlock(MachineFunction &MF, MachineBasicBlock &MBB);
279
280 void handleCall(MachineBasicBlock::iterator &I);
281 void handleReturn(MachineBasicBlock::iterator &I);
282 void handleZeroArgFP(MachineBasicBlock::iterator &I);
283 void handleOneArgFP(MachineBasicBlock::iterator &I);
284 void handleOneArgFPRW(MachineBasicBlock::iterator &I);
285 void handleTwoArgFP(MachineBasicBlock::iterator &I);
286 void handleCompareFP(MachineBasicBlock::iterator &I);
287 void handleCondMovFP(MachineBasicBlock::iterator &I);
288 void handleSpecialFP(MachineBasicBlock::iterator &I);
289
290 // Check if a COPY instruction is using FP registers.
isFPCopy__anon3deeeb160111::FPS291 static bool isFPCopy(MachineInstr &MI) {
292 Register DstReg = MI.getOperand(0).getReg();
293 Register SrcReg = MI.getOperand(1).getReg();
294
295 return X86::RFP80RegClass.contains(DstReg) ||
296 X86::RFP80RegClass.contains(SrcReg);
297 }
298
299 void setKillFlags(MachineBasicBlock &MBB) const;
300 };
301 }
302
303 char FPS::ID = 0;
304
305 INITIALIZE_PASS_BEGIN(FPS, DEBUG_TYPE, "X86 FP Stackifier",
306 false, false)
INITIALIZE_PASS_DEPENDENCY(EdgeBundles)307 INITIALIZE_PASS_DEPENDENCY(EdgeBundles)
308 INITIALIZE_PASS_END(FPS, DEBUG_TYPE, "X86 FP Stackifier",
309 false, false)
310
311 FunctionPass *llvm::createX86FloatingPointStackifierPass() { return new FPS(); }
312
313 /// getFPReg - Return the X86::FPx register number for the specified operand.
314 /// For example, this returns 3 for X86::FP3.
getFPReg(const MachineOperand & MO)315 static unsigned getFPReg(const MachineOperand &MO) {
316 assert(MO.isReg() && "Expected an FP register!");
317 Register Reg = MO.getReg();
318 assert(Reg >= X86::FP0 && Reg <= X86::FP6 && "Expected FP register!");
319 return Reg - X86::FP0;
320 }
321
322 /// runOnMachineFunction - Loop over all of the basic blocks, transforming FP
323 /// register references into FP stack references.
324 ///
runOnMachineFunction(MachineFunction & MF)325 bool FPS::runOnMachineFunction(MachineFunction &MF) {
326 // We only need to run this pass if there are any FP registers used in this
327 // function. If it is all integer, there is nothing for us to do!
328 bool FPIsUsed = false;
329
330 static_assert(X86::FP6 == X86::FP0+6, "Register enums aren't sorted right!");
331 const MachineRegisterInfo &MRI = MF.getRegInfo();
332 for (unsigned i = 0; i <= 6; ++i)
333 if (!MRI.reg_nodbg_empty(X86::FP0 + i)) {
334 FPIsUsed = true;
335 break;
336 }
337
338 // Early exit.
339 if (!FPIsUsed) return false;
340
341 Bundles = &getAnalysis<EdgeBundles>();
342 TII = MF.getSubtarget().getInstrInfo();
343
344 // Prepare cross-MBB liveness.
345 bundleCFGRecomputeKillFlags(MF);
346
347 StackTop = 0;
348
349 // Process the function in depth first order so that we process at least one
350 // of the predecessors for every reachable block in the function.
351 df_iterator_default_set<MachineBasicBlock*> Processed;
352 MachineBasicBlock *Entry = &MF.front();
353
354 LiveBundle &Bundle =
355 LiveBundles[Bundles->getBundle(Entry->getNumber(), false)];
356
357 // In regcall convention, some FP registers may not be passed through
358 // the stack, so they will need to be assigned to the stack first
359 if ((Entry->getParent()->getFunction().getCallingConv() ==
360 CallingConv::X86_RegCall) && (Bundle.Mask && !Bundle.FixCount)) {
361 // In the register calling convention, up to one FP argument could be
362 // saved in the first FP register.
363 // If bundle.mask is non-zero and Bundle.FixCount is zero, it means
364 // that the FP registers contain arguments.
365 // The actual value is passed in FP0.
366 // Here we fix the stack and mark FP0 as pre-assigned register.
367 assert((Bundle.Mask & 0xFE) == 0 &&
368 "Only FP0 could be passed as an argument");
369 Bundle.FixCount = 1;
370 Bundle.FixStack[0] = 0;
371 }
372
373 bool Changed = false;
374 for (MachineBasicBlock *BB : depth_first_ext(Entry, Processed))
375 Changed |= processBasicBlock(MF, *BB);
376
377 // Process any unreachable blocks in arbitrary order now.
378 if (MF.size() != Processed.size())
379 for (MachineBasicBlock &BB : MF)
380 if (Processed.insert(&BB).second)
381 Changed |= processBasicBlock(MF, BB);
382
383 LiveBundles.clear();
384
385 return Changed;
386 }
387
388 /// bundleCFG - Scan all the basic blocks to determine consistent live-in and
389 /// live-out sets for the FP registers. Consistent means that the set of
390 /// registers live-out from a block is identical to the live-in set of all
391 /// successors. This is not enforced by the normal live-in lists since
392 /// registers may be implicitly defined, or not used by all successors.
bundleCFGRecomputeKillFlags(MachineFunction & MF)393 void FPS::bundleCFGRecomputeKillFlags(MachineFunction &MF) {
394 assert(LiveBundles.empty() && "Stale data in LiveBundles");
395 LiveBundles.resize(Bundles->getNumBundles());
396
397 // Gather the actual live-in masks for all MBBs.
398 for (MachineBasicBlock &MBB : MF) {
399 setKillFlags(MBB);
400
401 const unsigned Mask = calcLiveInMask(&MBB, false);
402 if (!Mask)
403 continue;
404 // Update MBB ingoing bundle mask.
405 LiveBundles[Bundles->getBundle(MBB.getNumber(), false)].Mask |= Mask;
406 }
407 }
408
409 /// processBasicBlock - Loop over all of the instructions in the basic block,
410 /// transforming FP instructions into their stack form.
411 ///
processBasicBlock(MachineFunction & MF,MachineBasicBlock & BB)412 bool FPS::processBasicBlock(MachineFunction &MF, MachineBasicBlock &BB) {
413 bool Changed = false;
414 MBB = &BB;
415
416 setupBlockStack();
417
418 for (MachineBasicBlock::iterator I = BB.begin(); I != BB.end(); ++I) {
419 MachineInstr &MI = *I;
420 uint64_t Flags = MI.getDesc().TSFlags;
421
422 unsigned FPInstClass = Flags & X86II::FPTypeMask;
423 if (MI.isInlineAsm())
424 FPInstClass = X86II::SpecialFP;
425
426 if (MI.isCopy() && isFPCopy(MI))
427 FPInstClass = X86II::SpecialFP;
428
429 if (MI.isImplicitDef() &&
430 X86::RFP80RegClass.contains(MI.getOperand(0).getReg()))
431 FPInstClass = X86II::SpecialFP;
432
433 if (MI.isCall())
434 FPInstClass = X86II::SpecialFP;
435
436 if (FPInstClass == X86II::NotFP)
437 continue; // Efficiently ignore non-fp insts!
438
439 MachineInstr *PrevMI = nullptr;
440 if (I != BB.begin())
441 PrevMI = &*std::prev(I);
442
443 ++NumFP; // Keep track of # of pseudo instrs
444 LLVM_DEBUG(dbgs() << "\nFPInst:\t" << MI);
445
446 // Get dead variables list now because the MI pointer may be deleted as part
447 // of processing!
448 SmallVector<unsigned, 8> DeadRegs;
449 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
450 const MachineOperand &MO = MI.getOperand(i);
451 if (MO.isReg() && MO.isDead())
452 DeadRegs.push_back(MO.getReg());
453 }
454
455 switch (FPInstClass) {
456 case X86II::ZeroArgFP: handleZeroArgFP(I); break;
457 case X86II::OneArgFP: handleOneArgFP(I); break; // fstp ST(0)
458 case X86II::OneArgFPRW: handleOneArgFPRW(I); break; // ST(0) = fsqrt(ST(0))
459 case X86II::TwoArgFP: handleTwoArgFP(I); break;
460 case X86II::CompareFP: handleCompareFP(I); break;
461 case X86II::CondMovFP: handleCondMovFP(I); break;
462 case X86II::SpecialFP: handleSpecialFP(I); break;
463 default: llvm_unreachable("Unknown FP Type!");
464 }
465
466 // Check to see if any of the values defined by this instruction are dead
467 // after definition. If so, pop them.
468 for (unsigned i = 0, e = DeadRegs.size(); i != e; ++i) {
469 unsigned Reg = DeadRegs[i];
470 // Check if Reg is live on the stack. An inline-asm register operand that
471 // is in the clobber list and marked dead might not be live on the stack.
472 static_assert(X86::FP7 - X86::FP0 == 7, "sequential FP regnumbers");
473 if (Reg >= X86::FP0 && Reg <= X86::FP6 && isLive(Reg-X86::FP0)) {
474 LLVM_DEBUG(dbgs() << "Register FP#" << Reg - X86::FP0 << " is dead!\n");
475 freeStackSlotAfter(I, Reg-X86::FP0);
476 }
477 }
478
479 // Print out all of the instructions expanded to if -debug
480 LLVM_DEBUG({
481 MachineBasicBlock::iterator PrevI = PrevMI;
482 if (I == PrevI) {
483 dbgs() << "Just deleted pseudo instruction\n";
484 } else {
485 MachineBasicBlock::iterator Start = I;
486 // Rewind to first instruction newly inserted.
487 while (Start != BB.begin() && std::prev(Start) != PrevI)
488 --Start;
489 dbgs() << "Inserted instructions:\n\t";
490 Start->print(dbgs());
491 while (++Start != std::next(I)) {
492 }
493 }
494 dumpStack();
495 });
496 (void)PrevMI;
497
498 Changed = true;
499 }
500
501 finishBlockStack();
502
503 return Changed;
504 }
505
506 /// setupBlockStack - Use the live bundles to set up our model of the stack
507 /// to match predecessors' live out stack.
setupBlockStack()508 void FPS::setupBlockStack() {
509 LLVM_DEBUG(dbgs() << "\nSetting up live-ins for " << printMBBReference(*MBB)
510 << " derived from " << MBB->getName() << ".\n");
511 StackTop = 0;
512 // Get the live-in bundle for MBB.
513 const LiveBundle &Bundle =
514 LiveBundles[Bundles->getBundle(MBB->getNumber(), false)];
515
516 if (!Bundle.Mask) {
517 LLVM_DEBUG(dbgs() << "Block has no FP live-ins.\n");
518 return;
519 }
520
521 // Depth-first iteration should ensure that we always have an assigned stack.
522 assert(Bundle.isFixed() && "Reached block before any predecessors");
523
524 // Push the fixed live-in registers.
525 for (unsigned i = Bundle.FixCount; i > 0; --i) {
526 LLVM_DEBUG(dbgs() << "Live-in st(" << (i - 1) << "): %fp"
527 << unsigned(Bundle.FixStack[i - 1]) << '\n');
528 pushReg(Bundle.FixStack[i-1]);
529 }
530
531 // Kill off unwanted live-ins. This can happen with a critical edge.
532 // FIXME: We could keep these live registers around as zombies. They may need
533 // to be revived at the end of a short block. It might save a few instrs.
534 unsigned Mask = calcLiveInMask(MBB, /*RemoveFPs=*/true);
535 adjustLiveRegs(Mask, MBB->begin());
536 LLVM_DEBUG(MBB->dump());
537 }
538
539 /// finishBlockStack - Revive live-outs that are implicitly defined out of
540 /// MBB. Shuffle live registers to match the expected fixed stack of any
541 /// predecessors, and ensure that all predecessors are expecting the same
542 /// stack.
finishBlockStack()543 void FPS::finishBlockStack() {
544 // The RET handling below takes care of return blocks for us.
545 if (MBB->succ_empty())
546 return;
547
548 LLVM_DEBUG(dbgs() << "Setting up live-outs for " << printMBBReference(*MBB)
549 << " derived from " << MBB->getName() << ".\n");
550
551 // Get MBB's live-out bundle.
552 unsigned BundleIdx = Bundles->getBundle(MBB->getNumber(), true);
553 LiveBundle &Bundle = LiveBundles[BundleIdx];
554
555 // We may need to kill and define some registers to match successors.
556 // FIXME: This can probably be combined with the shuffle below.
557 MachineBasicBlock::iterator Term = MBB->getFirstTerminator();
558 adjustLiveRegs(Bundle.Mask, Term);
559
560 if (!Bundle.Mask) {
561 LLVM_DEBUG(dbgs() << "No live-outs.\n");
562 return;
563 }
564
565 // Has the stack order been fixed yet?
566 LLVM_DEBUG(dbgs() << "LB#" << BundleIdx << ": ");
567 if (Bundle.isFixed()) {
568 LLVM_DEBUG(dbgs() << "Shuffling stack to match.\n");
569 shuffleStackTop(Bundle.FixStack, Bundle.FixCount, Term);
570 } else {
571 // Not fixed yet, we get to choose.
572 LLVM_DEBUG(dbgs() << "Fixing stack order now.\n");
573 Bundle.FixCount = StackTop;
574 for (unsigned i = 0; i < StackTop; ++i)
575 Bundle.FixStack[i] = getStackEntry(i);
576 }
577 }
578
579
580 //===----------------------------------------------------------------------===//
581 // Efficient Lookup Table Support
582 //===----------------------------------------------------------------------===//
583
584 namespace {
585 struct TableEntry {
586 uint16_t from;
587 uint16_t to;
operator <__anon3deeeb160311::TableEntry588 bool operator<(const TableEntry &TE) const { return from < TE.from; }
operator <(const TableEntry & TE,unsigned V)589 friend bool operator<(const TableEntry &TE, unsigned V) {
590 return TE.from < V;
591 }
operator <(unsigned V,const TableEntry & TE)592 friend bool LLVM_ATTRIBUTE_UNUSED operator<(unsigned V,
593 const TableEntry &TE) {
594 return V < TE.from;
595 }
596 };
597 }
598
Lookup(ArrayRef<TableEntry> Table,unsigned Opcode)599 static int Lookup(ArrayRef<TableEntry> Table, unsigned Opcode) {
600 const TableEntry *I = llvm::lower_bound(Table, Opcode);
601 if (I != Table.end() && I->from == Opcode)
602 return I->to;
603 return -1;
604 }
605
606 #ifdef NDEBUG
607 #define ASSERT_SORTED(TABLE)
608 #else
609 #define ASSERT_SORTED(TABLE) \
610 { \
611 static std::atomic<bool> TABLE##Checked(false); \
612 if (!TABLE##Checked.load(std::memory_order_relaxed)) { \
613 assert(std::is_sorted(std::begin(TABLE), std::end(TABLE)) && \
614 "All lookup tables must be sorted for efficient access!"); \
615 TABLE##Checked.store(true, std::memory_order_relaxed); \
616 } \
617 }
618 #endif
619
620 //===----------------------------------------------------------------------===//
621 // Register File -> Register Stack Mapping Methods
622 //===----------------------------------------------------------------------===//
623
624 // OpcodeTable - Sorted map of register instructions to their stack version.
625 // The first element is an register file pseudo instruction, the second is the
626 // concrete X86 instruction which uses the register stack.
627 //
628 static const TableEntry OpcodeTable[] = {
629 { X86::ABS_Fp32 , X86::ABS_F },
630 { X86::ABS_Fp64 , X86::ABS_F },
631 { X86::ABS_Fp80 , X86::ABS_F },
632 { X86::ADD_Fp32m , X86::ADD_F32m },
633 { X86::ADD_Fp64m , X86::ADD_F64m },
634 { X86::ADD_Fp64m32 , X86::ADD_F32m },
635 { X86::ADD_Fp80m32 , X86::ADD_F32m },
636 { X86::ADD_Fp80m64 , X86::ADD_F64m },
637 { X86::ADD_FpI16m32 , X86::ADD_FI16m },
638 { X86::ADD_FpI16m64 , X86::ADD_FI16m },
639 { X86::ADD_FpI16m80 , X86::ADD_FI16m },
640 { X86::ADD_FpI32m32 , X86::ADD_FI32m },
641 { X86::ADD_FpI32m64 , X86::ADD_FI32m },
642 { X86::ADD_FpI32m80 , X86::ADD_FI32m },
643 { X86::CHS_Fp32 , X86::CHS_F },
644 { X86::CHS_Fp64 , X86::CHS_F },
645 { X86::CHS_Fp80 , X86::CHS_F },
646 { X86::CMOVBE_Fp32 , X86::CMOVBE_F },
647 { X86::CMOVBE_Fp64 , X86::CMOVBE_F },
648 { X86::CMOVBE_Fp80 , X86::CMOVBE_F },
649 { X86::CMOVB_Fp32 , X86::CMOVB_F },
650 { X86::CMOVB_Fp64 , X86::CMOVB_F },
651 { X86::CMOVB_Fp80 , X86::CMOVB_F },
652 { X86::CMOVE_Fp32 , X86::CMOVE_F },
653 { X86::CMOVE_Fp64 , X86::CMOVE_F },
654 { X86::CMOVE_Fp80 , X86::CMOVE_F },
655 { X86::CMOVNBE_Fp32 , X86::CMOVNBE_F },
656 { X86::CMOVNBE_Fp64 , X86::CMOVNBE_F },
657 { X86::CMOVNBE_Fp80 , X86::CMOVNBE_F },
658 { X86::CMOVNB_Fp32 , X86::CMOVNB_F },
659 { X86::CMOVNB_Fp64 , X86::CMOVNB_F },
660 { X86::CMOVNB_Fp80 , X86::CMOVNB_F },
661 { X86::CMOVNE_Fp32 , X86::CMOVNE_F },
662 { X86::CMOVNE_Fp64 , X86::CMOVNE_F },
663 { X86::CMOVNE_Fp80 , X86::CMOVNE_F },
664 { X86::CMOVNP_Fp32 , X86::CMOVNP_F },
665 { X86::CMOVNP_Fp64 , X86::CMOVNP_F },
666 { X86::CMOVNP_Fp80 , X86::CMOVNP_F },
667 { X86::CMOVP_Fp32 , X86::CMOVP_F },
668 { X86::CMOVP_Fp64 , X86::CMOVP_F },
669 { X86::CMOVP_Fp80 , X86::CMOVP_F },
670 { X86::COM_FpIr32 , X86::COM_FIr },
671 { X86::COM_FpIr64 , X86::COM_FIr },
672 { X86::COM_FpIr80 , X86::COM_FIr },
673 { X86::COM_Fpr32 , X86::COM_FST0r },
674 { X86::COM_Fpr64 , X86::COM_FST0r },
675 { X86::COM_Fpr80 , X86::COM_FST0r },
676 { X86::DIVR_Fp32m , X86::DIVR_F32m },
677 { X86::DIVR_Fp64m , X86::DIVR_F64m },
678 { X86::DIVR_Fp64m32 , X86::DIVR_F32m },
679 { X86::DIVR_Fp80m32 , X86::DIVR_F32m },
680 { X86::DIVR_Fp80m64 , X86::DIVR_F64m },
681 { X86::DIVR_FpI16m32, X86::DIVR_FI16m},
682 { X86::DIVR_FpI16m64, X86::DIVR_FI16m},
683 { X86::DIVR_FpI16m80, X86::DIVR_FI16m},
684 { X86::DIVR_FpI32m32, X86::DIVR_FI32m},
685 { X86::DIVR_FpI32m64, X86::DIVR_FI32m},
686 { X86::DIVR_FpI32m80, X86::DIVR_FI32m},
687 { X86::DIV_Fp32m , X86::DIV_F32m },
688 { X86::DIV_Fp64m , X86::DIV_F64m },
689 { X86::DIV_Fp64m32 , X86::DIV_F32m },
690 { X86::DIV_Fp80m32 , X86::DIV_F32m },
691 { X86::DIV_Fp80m64 , X86::DIV_F64m },
692 { X86::DIV_FpI16m32 , X86::DIV_FI16m },
693 { X86::DIV_FpI16m64 , X86::DIV_FI16m },
694 { X86::DIV_FpI16m80 , X86::DIV_FI16m },
695 { X86::DIV_FpI32m32 , X86::DIV_FI32m },
696 { X86::DIV_FpI32m64 , X86::DIV_FI32m },
697 { X86::DIV_FpI32m80 , X86::DIV_FI32m },
698 { X86::ILD_Fp16m32 , X86::ILD_F16m },
699 { X86::ILD_Fp16m64 , X86::ILD_F16m },
700 { X86::ILD_Fp16m80 , X86::ILD_F16m },
701 { X86::ILD_Fp32m32 , X86::ILD_F32m },
702 { X86::ILD_Fp32m64 , X86::ILD_F32m },
703 { X86::ILD_Fp32m80 , X86::ILD_F32m },
704 { X86::ILD_Fp64m32 , X86::ILD_F64m },
705 { X86::ILD_Fp64m64 , X86::ILD_F64m },
706 { X86::ILD_Fp64m80 , X86::ILD_F64m },
707 { X86::ISTT_Fp16m32 , X86::ISTT_FP16m},
708 { X86::ISTT_Fp16m64 , X86::ISTT_FP16m},
709 { X86::ISTT_Fp16m80 , X86::ISTT_FP16m},
710 { X86::ISTT_Fp32m32 , X86::ISTT_FP32m},
711 { X86::ISTT_Fp32m64 , X86::ISTT_FP32m},
712 { X86::ISTT_Fp32m80 , X86::ISTT_FP32m},
713 { X86::ISTT_Fp64m32 , X86::ISTT_FP64m},
714 { X86::ISTT_Fp64m64 , X86::ISTT_FP64m},
715 { X86::ISTT_Fp64m80 , X86::ISTT_FP64m},
716 { X86::IST_Fp16m32 , X86::IST_F16m },
717 { X86::IST_Fp16m64 , X86::IST_F16m },
718 { X86::IST_Fp16m80 , X86::IST_F16m },
719 { X86::IST_Fp32m32 , X86::IST_F32m },
720 { X86::IST_Fp32m64 , X86::IST_F32m },
721 { X86::IST_Fp32m80 , X86::IST_F32m },
722 { X86::IST_Fp64m32 , X86::IST_FP64m },
723 { X86::IST_Fp64m64 , X86::IST_FP64m },
724 { X86::IST_Fp64m80 , X86::IST_FP64m },
725 { X86::LD_Fp032 , X86::LD_F0 },
726 { X86::LD_Fp064 , X86::LD_F0 },
727 { X86::LD_Fp080 , X86::LD_F0 },
728 { X86::LD_Fp132 , X86::LD_F1 },
729 { X86::LD_Fp164 , X86::LD_F1 },
730 { X86::LD_Fp180 , X86::LD_F1 },
731 { X86::LD_Fp32m , X86::LD_F32m },
732 { X86::LD_Fp32m64 , X86::LD_F32m },
733 { X86::LD_Fp32m80 , X86::LD_F32m },
734 { X86::LD_Fp64m , X86::LD_F64m },
735 { X86::LD_Fp64m80 , X86::LD_F64m },
736 { X86::LD_Fp80m , X86::LD_F80m },
737 { X86::MUL_Fp32m , X86::MUL_F32m },
738 { X86::MUL_Fp64m , X86::MUL_F64m },
739 { X86::MUL_Fp64m32 , X86::MUL_F32m },
740 { X86::MUL_Fp80m32 , X86::MUL_F32m },
741 { X86::MUL_Fp80m64 , X86::MUL_F64m },
742 { X86::MUL_FpI16m32 , X86::MUL_FI16m },
743 { X86::MUL_FpI16m64 , X86::MUL_FI16m },
744 { X86::MUL_FpI16m80 , X86::MUL_FI16m },
745 { X86::MUL_FpI32m32 , X86::MUL_FI32m },
746 { X86::MUL_FpI32m64 , X86::MUL_FI32m },
747 { X86::MUL_FpI32m80 , X86::MUL_FI32m },
748 { X86::SQRT_Fp32 , X86::SQRT_F },
749 { X86::SQRT_Fp64 , X86::SQRT_F },
750 { X86::SQRT_Fp80 , X86::SQRT_F },
751 { X86::ST_Fp32m , X86::ST_F32m },
752 { X86::ST_Fp64m , X86::ST_F64m },
753 { X86::ST_Fp64m32 , X86::ST_F32m },
754 { X86::ST_Fp80m32 , X86::ST_F32m },
755 { X86::ST_Fp80m64 , X86::ST_F64m },
756 { X86::ST_FpP80m , X86::ST_FP80m },
757 { X86::SUBR_Fp32m , X86::SUBR_F32m },
758 { X86::SUBR_Fp64m , X86::SUBR_F64m },
759 { X86::SUBR_Fp64m32 , X86::SUBR_F32m },
760 { X86::SUBR_Fp80m32 , X86::SUBR_F32m },
761 { X86::SUBR_Fp80m64 , X86::SUBR_F64m },
762 { X86::SUBR_FpI16m32, X86::SUBR_FI16m},
763 { X86::SUBR_FpI16m64, X86::SUBR_FI16m},
764 { X86::SUBR_FpI16m80, X86::SUBR_FI16m},
765 { X86::SUBR_FpI32m32, X86::SUBR_FI32m},
766 { X86::SUBR_FpI32m64, X86::SUBR_FI32m},
767 { X86::SUBR_FpI32m80, X86::SUBR_FI32m},
768 { X86::SUB_Fp32m , X86::SUB_F32m },
769 { X86::SUB_Fp64m , X86::SUB_F64m },
770 { X86::SUB_Fp64m32 , X86::SUB_F32m },
771 { X86::SUB_Fp80m32 , X86::SUB_F32m },
772 { X86::SUB_Fp80m64 , X86::SUB_F64m },
773 { X86::SUB_FpI16m32 , X86::SUB_FI16m },
774 { X86::SUB_FpI16m64 , X86::SUB_FI16m },
775 { X86::SUB_FpI16m80 , X86::SUB_FI16m },
776 { X86::SUB_FpI32m32 , X86::SUB_FI32m },
777 { X86::SUB_FpI32m64 , X86::SUB_FI32m },
778 { X86::SUB_FpI32m80 , X86::SUB_FI32m },
779 { X86::TST_Fp32 , X86::TST_F },
780 { X86::TST_Fp64 , X86::TST_F },
781 { X86::TST_Fp80 , X86::TST_F },
782 { X86::UCOM_FpIr32 , X86::UCOM_FIr },
783 { X86::UCOM_FpIr64 , X86::UCOM_FIr },
784 { X86::UCOM_FpIr80 , X86::UCOM_FIr },
785 { X86::UCOM_Fpr32 , X86::UCOM_Fr },
786 { X86::UCOM_Fpr64 , X86::UCOM_Fr },
787 { X86::UCOM_Fpr80 , X86::UCOM_Fr },
788 };
789
getConcreteOpcode(unsigned Opcode)790 static unsigned getConcreteOpcode(unsigned Opcode) {
791 ASSERT_SORTED(OpcodeTable);
792 int Opc = Lookup(OpcodeTable, Opcode);
793 assert(Opc != -1 && "FP Stack instruction not in OpcodeTable!");
794 return Opc;
795 }
796
797 //===----------------------------------------------------------------------===//
798 // Helper Methods
799 //===----------------------------------------------------------------------===//
800
801 // PopTable - Sorted map of instructions to their popping version. The first
802 // element is an instruction, the second is the version which pops.
803 //
804 static const TableEntry PopTable[] = {
805 { X86::ADD_FrST0 , X86::ADD_FPrST0 },
806
807 { X86::COMP_FST0r, X86::FCOMPP },
808 { X86::COM_FIr , X86::COM_FIPr },
809 { X86::COM_FST0r , X86::COMP_FST0r },
810
811 { X86::DIVR_FrST0, X86::DIVR_FPrST0 },
812 { X86::DIV_FrST0 , X86::DIV_FPrST0 },
813
814 { X86::IST_F16m , X86::IST_FP16m },
815 { X86::IST_F32m , X86::IST_FP32m },
816
817 { X86::MUL_FrST0 , X86::MUL_FPrST0 },
818
819 { X86::ST_F32m , X86::ST_FP32m },
820 { X86::ST_F64m , X86::ST_FP64m },
821 { X86::ST_Frr , X86::ST_FPrr },
822
823 { X86::SUBR_FrST0, X86::SUBR_FPrST0 },
824 { X86::SUB_FrST0 , X86::SUB_FPrST0 },
825
826 { X86::UCOM_FIr , X86::UCOM_FIPr },
827
828 { X86::UCOM_FPr , X86::UCOM_FPPr },
829 { X86::UCOM_Fr , X86::UCOM_FPr },
830 };
831
832 /// popStackAfter - Pop the current value off of the top of the FP stack after
833 /// the specified instruction. This attempts to be sneaky and combine the pop
834 /// into the instruction itself if possible. The iterator is left pointing to
835 /// the last instruction, be it a new pop instruction inserted, or the old
836 /// instruction if it was modified in place.
837 ///
popStackAfter(MachineBasicBlock::iterator & I)838 void FPS::popStackAfter(MachineBasicBlock::iterator &I) {
839 MachineInstr &MI = *I;
840 const DebugLoc &dl = MI.getDebugLoc();
841 ASSERT_SORTED(PopTable);
842
843 popReg();
844
845 // Check to see if there is a popping version of this instruction...
846 int Opcode = Lookup(PopTable, I->getOpcode());
847 if (Opcode != -1) {
848 I->setDesc(TII->get(Opcode));
849 if (Opcode == X86::FCOMPP || Opcode == X86::UCOM_FPPr)
850 I->RemoveOperand(0);
851 } else { // Insert an explicit pop
852 I = BuildMI(*MBB, ++I, dl, TII->get(X86::ST_FPrr)).addReg(X86::ST0);
853 }
854 }
855
856 /// freeStackSlotAfter - Free the specified register from the register stack, so
857 /// that it is no longer in a register. If the register is currently at the top
858 /// of the stack, we just pop the current instruction, otherwise we store the
859 /// current top-of-stack into the specified slot, then pop the top of stack.
freeStackSlotAfter(MachineBasicBlock::iterator & I,unsigned FPRegNo)860 void FPS::freeStackSlotAfter(MachineBasicBlock::iterator &I, unsigned FPRegNo) {
861 if (getStackEntry(0) == FPRegNo) { // already at the top of stack? easy.
862 popStackAfter(I);
863 return;
864 }
865
866 // Otherwise, store the top of stack into the dead slot, killing the operand
867 // without having to add in an explicit xchg then pop.
868 //
869 I = freeStackSlotBefore(++I, FPRegNo);
870 }
871
872 /// freeStackSlotBefore - Free the specified register without trying any
873 /// folding.
874 MachineBasicBlock::iterator
freeStackSlotBefore(MachineBasicBlock::iterator I,unsigned FPRegNo)875 FPS::freeStackSlotBefore(MachineBasicBlock::iterator I, unsigned FPRegNo) {
876 unsigned STReg = getSTReg(FPRegNo);
877 unsigned OldSlot = getSlot(FPRegNo);
878 unsigned TopReg = Stack[StackTop-1];
879 Stack[OldSlot] = TopReg;
880 RegMap[TopReg] = OldSlot;
881 RegMap[FPRegNo] = ~0;
882 Stack[--StackTop] = ~0;
883 return BuildMI(*MBB, I, DebugLoc(), TII->get(X86::ST_FPrr))
884 .addReg(STReg)
885 .getInstr();
886 }
887
888 /// adjustLiveRegs - Kill and revive registers such that exactly the FP
889 /// registers with a bit in Mask are live.
adjustLiveRegs(unsigned Mask,MachineBasicBlock::iterator I)890 void FPS::adjustLiveRegs(unsigned Mask, MachineBasicBlock::iterator I) {
891 unsigned Defs = Mask;
892 unsigned Kills = 0;
893 for (unsigned i = 0; i < StackTop; ++i) {
894 unsigned RegNo = Stack[i];
895 if (!(Defs & (1 << RegNo)))
896 // This register is live, but we don't want it.
897 Kills |= (1 << RegNo);
898 else
899 // We don't need to imp-def this live register.
900 Defs &= ~(1 << RegNo);
901 }
902 assert((Kills & Defs) == 0 && "Register needs killing and def'ing?");
903
904 // Produce implicit-defs for free by using killed registers.
905 while (Kills && Defs) {
906 unsigned KReg = countTrailingZeros(Kills);
907 unsigned DReg = countTrailingZeros(Defs);
908 LLVM_DEBUG(dbgs() << "Renaming %fp" << KReg << " as imp %fp" << DReg
909 << "\n");
910 std::swap(Stack[getSlot(KReg)], Stack[getSlot(DReg)]);
911 std::swap(RegMap[KReg], RegMap[DReg]);
912 Kills &= ~(1 << KReg);
913 Defs &= ~(1 << DReg);
914 }
915
916 // Kill registers by popping.
917 if (Kills && I != MBB->begin()) {
918 MachineBasicBlock::iterator I2 = std::prev(I);
919 while (StackTop) {
920 unsigned KReg = getStackEntry(0);
921 if (!(Kills & (1 << KReg)))
922 break;
923 LLVM_DEBUG(dbgs() << "Popping %fp" << KReg << "\n");
924 popStackAfter(I2);
925 Kills &= ~(1 << KReg);
926 }
927 }
928
929 // Manually kill the rest.
930 while (Kills) {
931 unsigned KReg = countTrailingZeros(Kills);
932 LLVM_DEBUG(dbgs() << "Killing %fp" << KReg << "\n");
933 freeStackSlotBefore(I, KReg);
934 Kills &= ~(1 << KReg);
935 }
936
937 // Load zeros for all the imp-defs.
938 while(Defs) {
939 unsigned DReg = countTrailingZeros(Defs);
940 LLVM_DEBUG(dbgs() << "Defining %fp" << DReg << " as 0\n");
941 BuildMI(*MBB, I, DebugLoc(), TII->get(X86::LD_F0));
942 pushReg(DReg);
943 Defs &= ~(1 << DReg);
944 }
945
946 // Now we should have the correct registers live.
947 LLVM_DEBUG(dumpStack());
948 assert(StackTop == countPopulation(Mask) && "Live count mismatch");
949 }
950
951 /// shuffleStackTop - emit fxch instructions before I to shuffle the top
952 /// FixCount entries into the order given by FixStack.
953 /// FIXME: Is there a better algorithm than insertion sort?
shuffleStackTop(const unsigned char * FixStack,unsigned FixCount,MachineBasicBlock::iterator I)954 void FPS::shuffleStackTop(const unsigned char *FixStack,
955 unsigned FixCount,
956 MachineBasicBlock::iterator I) {
957 // Move items into place, starting from the desired stack bottom.
958 while (FixCount--) {
959 // Old register at position FixCount.
960 unsigned OldReg = getStackEntry(FixCount);
961 // Desired register at position FixCount.
962 unsigned Reg = FixStack[FixCount];
963 if (Reg == OldReg)
964 continue;
965 // (Reg st0) (OldReg st0) = (Reg OldReg st0)
966 moveToTop(Reg, I);
967 if (FixCount > 0)
968 moveToTop(OldReg, I);
969 }
970 LLVM_DEBUG(dumpStack());
971 }
972
973
974 //===----------------------------------------------------------------------===//
975 // Instruction transformation implementation
976 //===----------------------------------------------------------------------===//
977
handleCall(MachineBasicBlock::iterator & I)978 void FPS::handleCall(MachineBasicBlock::iterator &I) {
979 MachineInstr &MI = *I;
980 unsigned STReturns = 0;
981
982 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
983 MachineOperand &Op = MI.getOperand(i);
984 if (!Op.isReg() || Op.getReg() < X86::FP0 || Op.getReg() > X86::FP6)
985 continue;
986
987 assert(Op.isImplicit() && "Expected implicit def/use");
988
989 if (Op.isDef())
990 STReturns |= 1 << getFPReg(Op);
991
992 // Remove the operand so that later passes don't see it.
993 MI.RemoveOperand(i);
994 --i;
995 --e;
996 }
997
998 unsigned N = countTrailingOnes(STReturns);
999
1000 // FP registers used for function return must be consecutive starting at
1001 // FP0
1002 assert(STReturns == 0 || (isMask_32(STReturns) && N <= 2));
1003
1004 // Reset the FP Stack - It is required because of possible leftovers from
1005 // passed arguments. The caller should assume that the FP stack is
1006 // returned empty (unless the callee returns values on FP stack).
1007 while (StackTop > 0)
1008 popReg();
1009
1010 for (unsigned I = 0; I < N; ++I)
1011 pushReg(N - I - 1);
1012 }
1013
1014 /// If RET has an FP register use operand, pass the first one in ST(0) and
1015 /// the second one in ST(1).
handleReturn(MachineBasicBlock::iterator & I)1016 void FPS::handleReturn(MachineBasicBlock::iterator &I) {
1017 MachineInstr &MI = *I;
1018
1019 // Find the register operands.
1020 unsigned FirstFPRegOp = ~0U, SecondFPRegOp = ~0U;
1021 unsigned LiveMask = 0;
1022
1023 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
1024 MachineOperand &Op = MI.getOperand(i);
1025 if (!Op.isReg() || Op.getReg() < X86::FP0 || Op.getReg() > X86::FP6)
1026 continue;
1027 // FP Register uses must be kills unless there are two uses of the same
1028 // register, in which case only one will be a kill.
1029 assert(Op.isUse() &&
1030 (Op.isKill() || // Marked kill.
1031 getFPReg(Op) == FirstFPRegOp || // Second instance.
1032 MI.killsRegister(Op.getReg())) && // Later use is marked kill.
1033 "Ret only defs operands, and values aren't live beyond it");
1034
1035 if (FirstFPRegOp == ~0U)
1036 FirstFPRegOp = getFPReg(Op);
1037 else {
1038 assert(SecondFPRegOp == ~0U && "More than two fp operands!");
1039 SecondFPRegOp = getFPReg(Op);
1040 }
1041 LiveMask |= (1 << getFPReg(Op));
1042
1043 // Remove the operand so that later passes don't see it.
1044 MI.RemoveOperand(i);
1045 --i;
1046 --e;
1047 }
1048
1049 // We may have been carrying spurious live-ins, so make sure only the
1050 // returned registers are left live.
1051 adjustLiveRegs(LiveMask, MI);
1052 if (!LiveMask) return; // Quick check to see if any are possible.
1053
1054 // There are only four possibilities here:
1055 // 1) we are returning a single FP value. In this case, it has to be in
1056 // ST(0) already, so just declare success by removing the value from the
1057 // FP Stack.
1058 if (SecondFPRegOp == ~0U) {
1059 // Assert that the top of stack contains the right FP register.
1060 assert(StackTop == 1 && FirstFPRegOp == getStackEntry(0) &&
1061 "Top of stack not the right register for RET!");
1062
1063 // Ok, everything is good, mark the value as not being on the stack
1064 // anymore so that our assertion about the stack being empty at end of
1065 // block doesn't fire.
1066 StackTop = 0;
1067 return;
1068 }
1069
1070 // Otherwise, we are returning two values:
1071 // 2) If returning the same value for both, we only have one thing in the FP
1072 // stack. Consider: RET FP1, FP1
1073 if (StackTop == 1) {
1074 assert(FirstFPRegOp == SecondFPRegOp && FirstFPRegOp == getStackEntry(0)&&
1075 "Stack misconfiguration for RET!");
1076
1077 // Duplicate the TOS so that we return it twice. Just pick some other FPx
1078 // register to hold it.
1079 unsigned NewReg = ScratchFPReg;
1080 duplicateToTop(FirstFPRegOp, NewReg, MI);
1081 FirstFPRegOp = NewReg;
1082 }
1083
1084 /// Okay we know we have two different FPx operands now:
1085 assert(StackTop == 2 && "Must have two values live!");
1086
1087 /// 3) If SecondFPRegOp is currently in ST(0) and FirstFPRegOp is currently
1088 /// in ST(1). In this case, emit an fxch.
1089 if (getStackEntry(0) == SecondFPRegOp) {
1090 assert(getStackEntry(1) == FirstFPRegOp && "Unknown regs live");
1091 moveToTop(FirstFPRegOp, MI);
1092 }
1093
1094 /// 4) Finally, FirstFPRegOp must be in ST(0) and SecondFPRegOp must be in
1095 /// ST(1). Just remove both from our understanding of the stack and return.
1096 assert(getStackEntry(0) == FirstFPRegOp && "Unknown regs live");
1097 assert(getStackEntry(1) == SecondFPRegOp && "Unknown regs live");
1098 StackTop = 0;
1099 }
1100
1101 /// handleZeroArgFP - ST(0) = fld0 ST(0) = flds <mem>
1102 ///
handleZeroArgFP(MachineBasicBlock::iterator & I)1103 void FPS::handleZeroArgFP(MachineBasicBlock::iterator &I) {
1104 MachineInstr &MI = *I;
1105 unsigned DestReg = getFPReg(MI.getOperand(0));
1106
1107 // Change from the pseudo instruction to the concrete instruction.
1108 MI.RemoveOperand(0); // Remove the explicit ST(0) operand
1109 MI.setDesc(TII->get(getConcreteOpcode(MI.getOpcode())));
1110 MI.addOperand(
1111 MachineOperand::CreateReg(X86::ST0, /*isDef*/ true, /*isImp*/ true));
1112
1113 // Result gets pushed on the stack.
1114 pushReg(DestReg);
1115 }
1116
1117 /// handleOneArgFP - fst <mem>, ST(0)
1118 ///
handleOneArgFP(MachineBasicBlock::iterator & I)1119 void FPS::handleOneArgFP(MachineBasicBlock::iterator &I) {
1120 MachineInstr &MI = *I;
1121 unsigned NumOps = MI.getDesc().getNumOperands();
1122 assert((NumOps == X86::AddrNumOperands + 1 || NumOps == 1) &&
1123 "Can only handle fst* & ftst instructions!");
1124
1125 // Is this the last use of the source register?
1126 unsigned Reg = getFPReg(MI.getOperand(NumOps - 1));
1127 bool KillsSrc = MI.killsRegister(X86::FP0 + Reg);
1128
1129 // FISTP64m is strange because there isn't a non-popping versions.
1130 // If we have one _and_ we don't want to pop the operand, duplicate the value
1131 // on the stack instead of moving it. This ensure that popping the value is
1132 // always ok.
1133 // Ditto FISTTP16m, FISTTP32m, FISTTP64m, ST_FpP80m.
1134 //
1135 if (!KillsSrc && (MI.getOpcode() == X86::IST_Fp64m32 ||
1136 MI.getOpcode() == X86::ISTT_Fp16m32 ||
1137 MI.getOpcode() == X86::ISTT_Fp32m32 ||
1138 MI.getOpcode() == X86::ISTT_Fp64m32 ||
1139 MI.getOpcode() == X86::IST_Fp64m64 ||
1140 MI.getOpcode() == X86::ISTT_Fp16m64 ||
1141 MI.getOpcode() == X86::ISTT_Fp32m64 ||
1142 MI.getOpcode() == X86::ISTT_Fp64m64 ||
1143 MI.getOpcode() == X86::IST_Fp64m80 ||
1144 MI.getOpcode() == X86::ISTT_Fp16m80 ||
1145 MI.getOpcode() == X86::ISTT_Fp32m80 ||
1146 MI.getOpcode() == X86::ISTT_Fp64m80 ||
1147 MI.getOpcode() == X86::ST_FpP80m)) {
1148 duplicateToTop(Reg, ScratchFPReg, I);
1149 } else {
1150 moveToTop(Reg, I); // Move to the top of the stack...
1151 }
1152
1153 // Convert from the pseudo instruction to the concrete instruction.
1154 MI.RemoveOperand(NumOps - 1); // Remove explicit ST(0) operand
1155 MI.setDesc(TII->get(getConcreteOpcode(MI.getOpcode())));
1156 MI.addOperand(
1157 MachineOperand::CreateReg(X86::ST0, /*isDef*/ false, /*isImp*/ true));
1158
1159 if (MI.getOpcode() == X86::IST_FP64m || MI.getOpcode() == X86::ISTT_FP16m ||
1160 MI.getOpcode() == X86::ISTT_FP32m || MI.getOpcode() == X86::ISTT_FP64m ||
1161 MI.getOpcode() == X86::ST_FP80m) {
1162 if (StackTop == 0)
1163 report_fatal_error("Stack empty??");
1164 --StackTop;
1165 } else if (KillsSrc) { // Last use of operand?
1166 popStackAfter(I);
1167 }
1168 }
1169
1170
1171 /// handleOneArgFPRW: Handle instructions that read from the top of stack and
1172 /// replace the value with a newly computed value. These instructions may have
1173 /// non-fp operands after their FP operands.
1174 ///
1175 /// Examples:
1176 /// R1 = fchs R2
1177 /// R1 = fadd R2, [mem]
1178 ///
handleOneArgFPRW(MachineBasicBlock::iterator & I)1179 void FPS::handleOneArgFPRW(MachineBasicBlock::iterator &I) {
1180 MachineInstr &MI = *I;
1181 #ifndef NDEBUG
1182 unsigned NumOps = MI.getDesc().getNumOperands();
1183 assert(NumOps >= 2 && "FPRW instructions must have 2 ops!!");
1184 #endif
1185
1186 // Is this the last use of the source register?
1187 unsigned Reg = getFPReg(MI.getOperand(1));
1188 bool KillsSrc = MI.killsRegister(X86::FP0 + Reg);
1189
1190 if (KillsSrc) {
1191 // If this is the last use of the source register, just make sure it's on
1192 // the top of the stack.
1193 moveToTop(Reg, I);
1194 if (StackTop == 0)
1195 report_fatal_error("Stack cannot be empty!");
1196 --StackTop;
1197 pushReg(getFPReg(MI.getOperand(0)));
1198 } else {
1199 // If this is not the last use of the source register, _copy_ it to the top
1200 // of the stack.
1201 duplicateToTop(Reg, getFPReg(MI.getOperand(0)), I);
1202 }
1203
1204 // Change from the pseudo instruction to the concrete instruction.
1205 MI.RemoveOperand(1); // Drop the source operand.
1206 MI.RemoveOperand(0); // Drop the destination operand.
1207 MI.setDesc(TII->get(getConcreteOpcode(MI.getOpcode())));
1208 }
1209
1210
1211 //===----------------------------------------------------------------------===//
1212 // Define tables of various ways to map pseudo instructions
1213 //
1214
1215 // ForwardST0Table - Map: A = B op C into: ST(0) = ST(0) op ST(i)
1216 static const TableEntry ForwardST0Table[] = {
1217 { X86::ADD_Fp32 , X86::ADD_FST0r },
1218 { X86::ADD_Fp64 , X86::ADD_FST0r },
1219 { X86::ADD_Fp80 , X86::ADD_FST0r },
1220 { X86::DIV_Fp32 , X86::DIV_FST0r },
1221 { X86::DIV_Fp64 , X86::DIV_FST0r },
1222 { X86::DIV_Fp80 , X86::DIV_FST0r },
1223 { X86::MUL_Fp32 , X86::MUL_FST0r },
1224 { X86::MUL_Fp64 , X86::MUL_FST0r },
1225 { X86::MUL_Fp80 , X86::MUL_FST0r },
1226 { X86::SUB_Fp32 , X86::SUB_FST0r },
1227 { X86::SUB_Fp64 , X86::SUB_FST0r },
1228 { X86::SUB_Fp80 , X86::SUB_FST0r },
1229 };
1230
1231 // ReverseST0Table - Map: A = B op C into: ST(0) = ST(i) op ST(0)
1232 static const TableEntry ReverseST0Table[] = {
1233 { X86::ADD_Fp32 , X86::ADD_FST0r }, // commutative
1234 { X86::ADD_Fp64 , X86::ADD_FST0r }, // commutative
1235 { X86::ADD_Fp80 , X86::ADD_FST0r }, // commutative
1236 { X86::DIV_Fp32 , X86::DIVR_FST0r },
1237 { X86::DIV_Fp64 , X86::DIVR_FST0r },
1238 { X86::DIV_Fp80 , X86::DIVR_FST0r },
1239 { X86::MUL_Fp32 , X86::MUL_FST0r }, // commutative
1240 { X86::MUL_Fp64 , X86::MUL_FST0r }, // commutative
1241 { X86::MUL_Fp80 , X86::MUL_FST0r }, // commutative
1242 { X86::SUB_Fp32 , X86::SUBR_FST0r },
1243 { X86::SUB_Fp64 , X86::SUBR_FST0r },
1244 { X86::SUB_Fp80 , X86::SUBR_FST0r },
1245 };
1246
1247 // ForwardSTiTable - Map: A = B op C into: ST(i) = ST(0) op ST(i)
1248 static const TableEntry ForwardSTiTable[] = {
1249 { X86::ADD_Fp32 , X86::ADD_FrST0 }, // commutative
1250 { X86::ADD_Fp64 , X86::ADD_FrST0 }, // commutative
1251 { X86::ADD_Fp80 , X86::ADD_FrST0 }, // commutative
1252 { X86::DIV_Fp32 , X86::DIVR_FrST0 },
1253 { X86::DIV_Fp64 , X86::DIVR_FrST0 },
1254 { X86::DIV_Fp80 , X86::DIVR_FrST0 },
1255 { X86::MUL_Fp32 , X86::MUL_FrST0 }, // commutative
1256 { X86::MUL_Fp64 , X86::MUL_FrST0 }, // commutative
1257 { X86::MUL_Fp80 , X86::MUL_FrST0 }, // commutative
1258 { X86::SUB_Fp32 , X86::SUBR_FrST0 },
1259 { X86::SUB_Fp64 , X86::SUBR_FrST0 },
1260 { X86::SUB_Fp80 , X86::SUBR_FrST0 },
1261 };
1262
1263 // ReverseSTiTable - Map: A = B op C into: ST(i) = ST(i) op ST(0)
1264 static const TableEntry ReverseSTiTable[] = {
1265 { X86::ADD_Fp32 , X86::ADD_FrST0 },
1266 { X86::ADD_Fp64 , X86::ADD_FrST0 },
1267 { X86::ADD_Fp80 , X86::ADD_FrST0 },
1268 { X86::DIV_Fp32 , X86::DIV_FrST0 },
1269 { X86::DIV_Fp64 , X86::DIV_FrST0 },
1270 { X86::DIV_Fp80 , X86::DIV_FrST0 },
1271 { X86::MUL_Fp32 , X86::MUL_FrST0 },
1272 { X86::MUL_Fp64 , X86::MUL_FrST0 },
1273 { X86::MUL_Fp80 , X86::MUL_FrST0 },
1274 { X86::SUB_Fp32 , X86::SUB_FrST0 },
1275 { X86::SUB_Fp64 , X86::SUB_FrST0 },
1276 { X86::SUB_Fp80 , X86::SUB_FrST0 },
1277 };
1278
1279
1280 /// handleTwoArgFP - Handle instructions like FADD and friends which are virtual
1281 /// instructions which need to be simplified and possibly transformed.
1282 ///
1283 /// Result: ST(0) = fsub ST(0), ST(i)
1284 /// ST(i) = fsub ST(0), ST(i)
1285 /// ST(0) = fsubr ST(0), ST(i)
1286 /// ST(i) = fsubr ST(0), ST(i)
1287 ///
handleTwoArgFP(MachineBasicBlock::iterator & I)1288 void FPS::handleTwoArgFP(MachineBasicBlock::iterator &I) {
1289 ASSERT_SORTED(ForwardST0Table); ASSERT_SORTED(ReverseST0Table);
1290 ASSERT_SORTED(ForwardSTiTable); ASSERT_SORTED(ReverseSTiTable);
1291 MachineInstr &MI = *I;
1292
1293 unsigned NumOperands = MI.getDesc().getNumOperands();
1294 assert(NumOperands == 3 && "Illegal TwoArgFP instruction!");
1295 unsigned Dest = getFPReg(MI.getOperand(0));
1296 unsigned Op0 = getFPReg(MI.getOperand(NumOperands - 2));
1297 unsigned Op1 = getFPReg(MI.getOperand(NumOperands - 1));
1298 bool KillsOp0 = MI.killsRegister(X86::FP0 + Op0);
1299 bool KillsOp1 = MI.killsRegister(X86::FP0 + Op1);
1300 DebugLoc dl = MI.getDebugLoc();
1301
1302 unsigned TOS = getStackEntry(0);
1303
1304 // One of our operands must be on the top of the stack. If neither is yet, we
1305 // need to move one.
1306 if (Op0 != TOS && Op1 != TOS) { // No operand at TOS?
1307 // We can choose to move either operand to the top of the stack. If one of
1308 // the operands is killed by this instruction, we want that one so that we
1309 // can update right on top of the old version.
1310 if (KillsOp0) {
1311 moveToTop(Op0, I); // Move dead operand to TOS.
1312 TOS = Op0;
1313 } else if (KillsOp1) {
1314 moveToTop(Op1, I);
1315 TOS = Op1;
1316 } else {
1317 // All of the operands are live after this instruction executes, so we
1318 // cannot update on top of any operand. Because of this, we must
1319 // duplicate one of the stack elements to the top. It doesn't matter
1320 // which one we pick.
1321 //
1322 duplicateToTop(Op0, Dest, I);
1323 Op0 = TOS = Dest;
1324 KillsOp0 = true;
1325 }
1326 } else if (!KillsOp0 && !KillsOp1) {
1327 // If we DO have one of our operands at the top of the stack, but we don't
1328 // have a dead operand, we must duplicate one of the operands to a new slot
1329 // on the stack.
1330 duplicateToTop(Op0, Dest, I);
1331 Op0 = TOS = Dest;
1332 KillsOp0 = true;
1333 }
1334
1335 // Now we know that one of our operands is on the top of the stack, and at
1336 // least one of our operands is killed by this instruction.
1337 assert((TOS == Op0 || TOS == Op1) && (KillsOp0 || KillsOp1) &&
1338 "Stack conditions not set up right!");
1339
1340 // We decide which form to use based on what is on the top of the stack, and
1341 // which operand is killed by this instruction.
1342 ArrayRef<TableEntry> InstTable;
1343 bool isForward = TOS == Op0;
1344 bool updateST0 = (TOS == Op0 && !KillsOp1) || (TOS == Op1 && !KillsOp0);
1345 if (updateST0) {
1346 if (isForward)
1347 InstTable = ForwardST0Table;
1348 else
1349 InstTable = ReverseST0Table;
1350 } else {
1351 if (isForward)
1352 InstTable = ForwardSTiTable;
1353 else
1354 InstTable = ReverseSTiTable;
1355 }
1356
1357 int Opcode = Lookup(InstTable, MI.getOpcode());
1358 assert(Opcode != -1 && "Unknown TwoArgFP pseudo instruction!");
1359
1360 // NotTOS - The register which is not on the top of stack...
1361 unsigned NotTOS = (TOS == Op0) ? Op1 : Op0;
1362
1363 // Replace the old instruction with a new instruction
1364 MBB->remove(&*I++);
1365 I = BuildMI(*MBB, I, dl, TII->get(Opcode)).addReg(getSTReg(NotTOS));
1366
1367 // If both operands are killed, pop one off of the stack in addition to
1368 // overwriting the other one.
1369 if (KillsOp0 && KillsOp1 && Op0 != Op1) {
1370 assert(!updateST0 && "Should have updated other operand!");
1371 popStackAfter(I); // Pop the top of stack
1372 }
1373
1374 // Update stack information so that we know the destination register is now on
1375 // the stack.
1376 unsigned UpdatedSlot = getSlot(updateST0 ? TOS : NotTOS);
1377 assert(UpdatedSlot < StackTop && Dest < 7);
1378 Stack[UpdatedSlot] = Dest;
1379 RegMap[Dest] = UpdatedSlot;
1380 MBB->getParent()->DeleteMachineInstr(&MI); // Remove the old instruction
1381 }
1382
1383 /// handleCompareFP - Handle FUCOM and FUCOMI instructions, which have two FP
1384 /// register arguments and no explicit destinations.
1385 ///
handleCompareFP(MachineBasicBlock::iterator & I)1386 void FPS::handleCompareFP(MachineBasicBlock::iterator &I) {
1387 MachineInstr &MI = *I;
1388
1389 unsigned NumOperands = MI.getDesc().getNumOperands();
1390 assert(NumOperands == 2 && "Illegal FUCOM* instruction!");
1391 unsigned Op0 = getFPReg(MI.getOperand(NumOperands - 2));
1392 unsigned Op1 = getFPReg(MI.getOperand(NumOperands - 1));
1393 bool KillsOp0 = MI.killsRegister(X86::FP0 + Op0);
1394 bool KillsOp1 = MI.killsRegister(X86::FP0 + Op1);
1395
1396 // Make sure the first operand is on the top of stack, the other one can be
1397 // anywhere.
1398 moveToTop(Op0, I);
1399
1400 // Change from the pseudo instruction to the concrete instruction.
1401 MI.getOperand(0).setReg(getSTReg(Op1));
1402 MI.RemoveOperand(1);
1403 MI.setDesc(TII->get(getConcreteOpcode(MI.getOpcode())));
1404
1405 // If any of the operands are killed by this instruction, free them.
1406 if (KillsOp0) freeStackSlotAfter(I, Op0);
1407 if (KillsOp1 && Op0 != Op1) freeStackSlotAfter(I, Op1);
1408 }
1409
1410 /// handleCondMovFP - Handle two address conditional move instructions. These
1411 /// instructions move a st(i) register to st(0) iff a condition is true. These
1412 /// instructions require that the first operand is at the top of the stack, but
1413 /// otherwise don't modify the stack at all.
handleCondMovFP(MachineBasicBlock::iterator & I)1414 void FPS::handleCondMovFP(MachineBasicBlock::iterator &I) {
1415 MachineInstr &MI = *I;
1416
1417 unsigned Op0 = getFPReg(MI.getOperand(0));
1418 unsigned Op1 = getFPReg(MI.getOperand(2));
1419 bool KillsOp1 = MI.killsRegister(X86::FP0 + Op1);
1420
1421 // The first operand *must* be on the top of the stack.
1422 moveToTop(Op0, I);
1423
1424 // Change the second operand to the stack register that the operand is in.
1425 // Change from the pseudo instruction to the concrete instruction.
1426 MI.RemoveOperand(0);
1427 MI.RemoveOperand(1);
1428 MI.getOperand(0).setReg(getSTReg(Op1));
1429 MI.setDesc(TII->get(getConcreteOpcode(MI.getOpcode())));
1430
1431 // If we kill the second operand, make sure to pop it from the stack.
1432 if (Op0 != Op1 && KillsOp1) {
1433 // Get this value off of the register stack.
1434 freeStackSlotAfter(I, Op1);
1435 }
1436 }
1437
1438
1439 /// handleSpecialFP - Handle special instructions which behave unlike other
1440 /// floating point instructions. This is primarily intended for use by pseudo
1441 /// instructions.
1442 ///
handleSpecialFP(MachineBasicBlock::iterator & Inst)1443 void FPS::handleSpecialFP(MachineBasicBlock::iterator &Inst) {
1444 MachineInstr &MI = *Inst;
1445
1446 if (MI.isCall()) {
1447 handleCall(Inst);
1448 return;
1449 }
1450
1451 if (MI.isReturn()) {
1452 handleReturn(Inst);
1453 return;
1454 }
1455
1456 switch (MI.getOpcode()) {
1457 default: llvm_unreachable("Unknown SpecialFP instruction!");
1458 case TargetOpcode::COPY: {
1459 // We handle three kinds of copies: FP <- FP, FP <- ST, and ST <- FP.
1460 const MachineOperand &MO1 = MI.getOperand(1);
1461 const MachineOperand &MO0 = MI.getOperand(0);
1462 bool KillsSrc = MI.killsRegister(MO1.getReg());
1463
1464 // FP <- FP copy.
1465 unsigned DstFP = getFPReg(MO0);
1466 unsigned SrcFP = getFPReg(MO1);
1467 assert(isLive(SrcFP) && "Cannot copy dead register");
1468 if (KillsSrc) {
1469 // If the input operand is killed, we can just change the owner of the
1470 // incoming stack slot into the result.
1471 unsigned Slot = getSlot(SrcFP);
1472 Stack[Slot] = DstFP;
1473 RegMap[DstFP] = Slot;
1474 } else {
1475 // For COPY we just duplicate the specified value to a new stack slot.
1476 // This could be made better, but would require substantial changes.
1477 duplicateToTop(SrcFP, DstFP, Inst);
1478 }
1479 break;
1480 }
1481
1482 case TargetOpcode::IMPLICIT_DEF: {
1483 // All FP registers must be explicitly defined, so load a 0 instead.
1484 unsigned Reg = MI.getOperand(0).getReg() - X86::FP0;
1485 LLVM_DEBUG(dbgs() << "Emitting LD_F0 for implicit FP" << Reg << '\n');
1486 BuildMI(*MBB, Inst, MI.getDebugLoc(), TII->get(X86::LD_F0));
1487 pushReg(Reg);
1488 break;
1489 }
1490
1491 case TargetOpcode::INLINEASM:
1492 case TargetOpcode::INLINEASM_BR: {
1493 // The inline asm MachineInstr currently only *uses* FP registers for the
1494 // 'f' constraint. These should be turned into the current ST(x) register
1495 // in the machine instr.
1496 //
1497 // There are special rules for x87 inline assembly. The compiler must know
1498 // exactly how many registers are popped and pushed implicitly by the asm.
1499 // Otherwise it is not possible to restore the stack state after the inline
1500 // asm.
1501 //
1502 // There are 3 kinds of input operands:
1503 //
1504 // 1. Popped inputs. These must appear at the stack top in ST0-STn. A
1505 // popped input operand must be in a fixed stack slot, and it is either
1506 // tied to an output operand, or in the clobber list. The MI has ST use
1507 // and def operands for these inputs.
1508 //
1509 // 2. Fixed inputs. These inputs appear in fixed stack slots, but are
1510 // preserved by the inline asm. The fixed stack slots must be STn-STm
1511 // following the popped inputs. A fixed input operand cannot be tied to
1512 // an output or appear in the clobber list. The MI has ST use operands
1513 // and no defs for these inputs.
1514 //
1515 // 3. Preserved inputs. These inputs use the "f" constraint which is
1516 // represented as an FP register. The inline asm won't change these
1517 // stack slots.
1518 //
1519 // Outputs must be in ST registers, FP outputs are not allowed. Clobbered
1520 // registers do not count as output operands. The inline asm changes the
1521 // stack as if it popped all the popped inputs and then pushed all the
1522 // output operands.
1523
1524 // Scan the assembly for ST registers used, defined and clobbered. We can
1525 // only tell clobbers from defs by looking at the asm descriptor.
1526 unsigned STUses = 0, STDefs = 0, STClobbers = 0;
1527 unsigned NumOps = 0;
1528 SmallSet<unsigned, 1> FRegIdx;
1529 unsigned RCID;
1530
1531 for (unsigned i = InlineAsm::MIOp_FirstOperand, e = MI.getNumOperands();
1532 i != e && MI.getOperand(i).isImm(); i += 1 + NumOps) {
1533 unsigned Flags = MI.getOperand(i).getImm();
1534
1535 NumOps = InlineAsm::getNumOperandRegisters(Flags);
1536 if (NumOps != 1)
1537 continue;
1538 const MachineOperand &MO = MI.getOperand(i + 1);
1539 if (!MO.isReg())
1540 continue;
1541 unsigned STReg = MO.getReg() - X86::FP0;
1542 if (STReg >= 8)
1543 continue;
1544
1545 // If the flag has a register class constraint, this must be an operand
1546 // with constraint "f". Record its index and continue.
1547 if (InlineAsm::hasRegClassConstraint(Flags, RCID)) {
1548 FRegIdx.insert(i + 1);
1549 continue;
1550 }
1551
1552 switch (InlineAsm::getKind(Flags)) {
1553 case InlineAsm::Kind_RegUse:
1554 STUses |= (1u << STReg);
1555 break;
1556 case InlineAsm::Kind_RegDef:
1557 case InlineAsm::Kind_RegDefEarlyClobber:
1558 STDefs |= (1u << STReg);
1559 break;
1560 case InlineAsm::Kind_Clobber:
1561 STClobbers |= (1u << STReg);
1562 break;
1563 default:
1564 break;
1565 }
1566 }
1567
1568 if (STUses && !isMask_32(STUses))
1569 MI.emitError("fixed input regs must be last on the x87 stack");
1570 unsigned NumSTUses = countTrailingOnes(STUses);
1571
1572 // Defs must be contiguous from the stack top. ST0-STn.
1573 if (STDefs && !isMask_32(STDefs)) {
1574 MI.emitError("output regs must be last on the x87 stack");
1575 STDefs = NextPowerOf2(STDefs) - 1;
1576 }
1577 unsigned NumSTDefs = countTrailingOnes(STDefs);
1578
1579 // So must the clobbered stack slots. ST0-STm, m >= n.
1580 if (STClobbers && !isMask_32(STDefs | STClobbers))
1581 MI.emitError("clobbers must be last on the x87 stack");
1582
1583 // Popped inputs are the ones that are also clobbered or defined.
1584 unsigned STPopped = STUses & (STDefs | STClobbers);
1585 if (STPopped && !isMask_32(STPopped))
1586 MI.emitError("implicitly popped regs must be last on the x87 stack");
1587 unsigned NumSTPopped = countTrailingOnes(STPopped);
1588
1589 LLVM_DEBUG(dbgs() << "Asm uses " << NumSTUses << " fixed regs, pops "
1590 << NumSTPopped << ", and defines " << NumSTDefs
1591 << " regs.\n");
1592
1593 #ifndef NDEBUG
1594 // If any input operand uses constraint "f", all output register
1595 // constraints must be early-clobber defs.
1596 for (unsigned I = 0, E = MI.getNumOperands(); I < E; ++I)
1597 if (FRegIdx.count(I)) {
1598 assert((1 << getFPReg(MI.getOperand(I)) & STDefs) == 0 &&
1599 "Operands with constraint \"f\" cannot overlap with defs");
1600 }
1601 #endif
1602
1603 // Collect all FP registers (register operands with constraints "t", "u",
1604 // and "f") to kill afer the instruction.
1605 unsigned FPKills = ((1u << NumFPRegs) - 1) & ~0xff;
1606 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
1607 MachineOperand &Op = MI.getOperand(i);
1608 if (!Op.isReg() || Op.getReg() < X86::FP0 || Op.getReg() > X86::FP6)
1609 continue;
1610 unsigned FPReg = getFPReg(Op);
1611
1612 // If we kill this operand, make sure to pop it from the stack after the
1613 // asm. We just remember it for now, and pop them all off at the end in
1614 // a batch.
1615 if (Op.isUse() && Op.isKill())
1616 FPKills |= 1U << FPReg;
1617 }
1618
1619 // Do not include registers that are implicitly popped by defs/clobbers.
1620 FPKills &= ~(STDefs | STClobbers);
1621
1622 // Now we can rearrange the live registers to match what was requested.
1623 unsigned char STUsesArray[8];
1624
1625 for (unsigned I = 0; I < NumSTUses; ++I)
1626 STUsesArray[I] = I;
1627
1628 shuffleStackTop(STUsesArray, NumSTUses, Inst);
1629 LLVM_DEBUG({
1630 dbgs() << "Before asm: ";
1631 dumpStack();
1632 });
1633
1634 // With the stack layout fixed, rewrite the FP registers.
1635 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
1636 MachineOperand &Op = MI.getOperand(i);
1637 if (!Op.isReg() || Op.getReg() < X86::FP0 || Op.getReg() > X86::FP6)
1638 continue;
1639
1640 unsigned FPReg = getFPReg(Op);
1641
1642 if (FRegIdx.count(i))
1643 // Operand with constraint "f".
1644 Op.setReg(getSTReg(FPReg));
1645 else
1646 // Operand with a single register class constraint ("t" or "u").
1647 Op.setReg(X86::ST0 + FPReg);
1648 }
1649
1650 // Simulate the inline asm popping its inputs and pushing its outputs.
1651 StackTop -= NumSTPopped;
1652
1653 for (unsigned i = 0; i < NumSTDefs; ++i)
1654 pushReg(NumSTDefs - i - 1);
1655
1656 // If this asm kills any FP registers (is the last use of them) we must
1657 // explicitly emit pop instructions for them. Do this now after the asm has
1658 // executed so that the ST(x) numbers are not off (which would happen if we
1659 // did this inline with operand rewriting).
1660 //
1661 // Note: this might be a non-optimal pop sequence. We might be able to do
1662 // better by trying to pop in stack order or something.
1663 while (FPKills) {
1664 unsigned FPReg = countTrailingZeros(FPKills);
1665 if (isLive(FPReg))
1666 freeStackSlotAfter(Inst, FPReg);
1667 FPKills &= ~(1U << FPReg);
1668 }
1669
1670 // Don't delete the inline asm!
1671 return;
1672 }
1673 }
1674
1675 Inst = MBB->erase(Inst); // Remove the pseudo instruction
1676
1677 // We want to leave I pointing to the previous instruction, but what if we
1678 // just erased the first instruction?
1679 if (Inst == MBB->begin()) {
1680 LLVM_DEBUG(dbgs() << "Inserting dummy KILL\n");
1681 Inst = BuildMI(*MBB, Inst, DebugLoc(), TII->get(TargetOpcode::KILL));
1682 } else
1683 --Inst;
1684 }
1685
setKillFlags(MachineBasicBlock & MBB) const1686 void FPS::setKillFlags(MachineBasicBlock &MBB) const {
1687 const TargetRegisterInfo &TRI =
1688 *MBB.getParent()->getSubtarget().getRegisterInfo();
1689 LivePhysRegs LPR(TRI);
1690
1691 LPR.addLiveOuts(MBB);
1692
1693 for (MachineBasicBlock::reverse_iterator I = MBB.rbegin(), E = MBB.rend();
1694 I != E; ++I) {
1695 if (I->isDebugInstr())
1696 continue;
1697
1698 std::bitset<8> Defs;
1699 SmallVector<MachineOperand *, 2> Uses;
1700 MachineInstr &MI = *I;
1701
1702 for (auto &MO : I->operands()) {
1703 if (!MO.isReg())
1704 continue;
1705
1706 unsigned Reg = MO.getReg() - X86::FP0;
1707
1708 if (Reg >= 8)
1709 continue;
1710
1711 if (MO.isDef()) {
1712 Defs.set(Reg);
1713 if (!LPR.contains(MO.getReg()))
1714 MO.setIsDead();
1715 } else
1716 Uses.push_back(&MO);
1717 }
1718
1719 for (auto *MO : Uses)
1720 if (Defs.test(getFPReg(*MO)) || !LPR.contains(MO->getReg()))
1721 MO->setIsKill();
1722
1723 LPR.stepBackward(MI);
1724 }
1725 }
1726