Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(643)

Side by Side Diff: src/IceTargetLoweringX8632.cpp

Issue 567553003: Mark setjmp as "returns twice" and turn off SimpleCoalescing when called. (Closed) Base URL: https://chromium.googlesource.com/native_client/pnacl-subzero.git@master
Patch Set: line up slash again Created 6 years, 3 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « src/IceTargetLowering.cpp ('k') | tests_lit/llvm2ice_tests/returns_twice_no_coalesce.ll » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 //===- subzero/src/IceTargetLoweringX8632.cpp - x86-32 lowering -----------===// 1 //===- subzero/src/IceTargetLoweringX8632.cpp - x86-32 lowering -----------===//
2 // 2 //
3 // The Subzero Code Generator 3 // The Subzero Code Generator
4 // 4 //
5 // This file is distributed under the University of Illinois Open Source 5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details. 6 // License. See LICENSE.TXT for details.
7 // 7 //
8 //===----------------------------------------------------------------------===// 8 //===----------------------------------------------------------------------===//
9 // 9 //
10 // This file implements the TargetLoweringX8632 class, which 10 // This file implements the TargetLoweringX8632 class, which
(...skipping 629 matching lines...) Expand 10 before | Expand all | Expand 10 after
640 // * GlobalsSize: area 4 640 // * GlobalsSize: area 4
641 // * GlobalsAndSubsequentPaddingSize: areas 4 - 5 641 // * GlobalsAndSubsequentPaddingSize: areas 4 - 5
642 // * LocalsSpillAreaSize: area 6 642 // * LocalsSpillAreaSize: area 6
643 // * SpillAreaSizeBytes: areas 3 - 7 643 // * SpillAreaSizeBytes: areas 3 - 7
644 644
645 // If SimpleCoalescing is false, each variable without a register 645 // If SimpleCoalescing is false, each variable without a register
646 // gets its own unique stack slot, which leads to large stack 646 // gets its own unique stack slot, which leads to large stack
647 // frames. If SimpleCoalescing is true, then each "global" variable 647 // frames. If SimpleCoalescing is true, then each "global" variable
648 // without a register gets its own slot, but "local" variable slots 648 // without a register gets its own slot, but "local" variable slots
649 // are reused across basic blocks. E.g., if A and B are local to 649 // are reused across basic blocks. E.g., if A and B are local to
650 // block 1 and C is local to block 2, then C may share a slot with A 650 // block 1 and C is local to block 2, then C may share a slot with A or B.
651 // or B. 651 //
652 const bool SimpleCoalescing = true; 652 // We cannot coalesce stack slots if this function calls a "returns twice"
653 // function. In that case, basic blocks may be revisited, and variables
654 // local to those basic blocks are actually live until after the
655 // called function returns a second time.
656 const bool SimpleCoalescing = !callsReturnsTwice();
653 size_t InArgsSizeBytes = 0; 657 size_t InArgsSizeBytes = 0;
654 size_t PreservedRegsSizeBytes = 0; 658 size_t PreservedRegsSizeBytes = 0;
655 SpillAreaSizeBytes = 0; 659 SpillAreaSizeBytes = 0;
656 Context.init(Node); 660 Context.init(Node);
657 Context.setInsertPoint(Context.getCur()); 661 Context.setInsertPoint(Context.getCur());
658 662
659 // Determine stack frame offsets for each Variable without a 663 // Determine stack frame offsets for each Variable without a
660 // register assignment. This can be done as one variable per stack 664 // register assignment. This can be done as one variable per stack
661 // slot. Or, do coalescing by running the register allocator again 665 // slot. Or, do coalescing by running the register allocator again
662 // with an infinite set of registers (as a side effect, this gives 666 // with an infinite set of registers (as a side effect, this gives
(...skipping 3794 matching lines...) Expand 10 before | Expand all | Expand 10 after
4457 Str << "\t.align\t" << Align << "\n"; 4461 Str << "\t.align\t" << Align << "\n";
4458 Str << MangledName << ":\n"; 4462 Str << MangledName << ":\n";
4459 for (SizeT i = 0; i < Size; ++i) { 4463 for (SizeT i = 0; i < Size; ++i) {
4460 Str << "\t.byte\t" << (((unsigned)Data[i]) & 0xff) << "\n"; 4464 Str << "\t.byte\t" << (((unsigned)Data[i]) & 0xff) << "\n";
4461 } 4465 }
4462 Str << "\t.size\t" << MangledName << ", " << Size << "\n"; 4466 Str << "\t.size\t" << MangledName << ", " << Size << "\n";
4463 } 4467 }
4464 } 4468 }
4465 4469
4466 } // end of namespace Ice 4470 } // end of namespace Ice
OLDNEW
« no previous file with comments | « src/IceTargetLowering.cpp ('k') | tests_lit/llvm2ice_tests/returns_twice_no_coalesce.ll » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698