I talked to Kevin about this and he agreed to take a look at your proposed change.
On Thu, Dec 9, 2010 at 3:09 PM, <[email protected]> wrote: > Reviewers: Kasper Lund, > > Description: > RFC: Switch to ast ids (instead of positions) for type feedback. > > The goal here is to avoid having several pieces of type feedback tied > to a single position. Since everything else already relied on ast ids, > it seemed natural to use them for type feedback as well[*]. The > implementation squeezes in a new reloc info mode (CODE_TARGET_WITH_ID) > which is used for calls to IC stubs from the code generated by the > full codegen. The new mode should (in most cases) use just one extra > byte per call in the encoded form. The type feedback oracle can simply > focus on this reloc info mode. > > [*]: I did play with disambiguating positions, but because of the > implicit way they are set, it's hard to be sure all the cases are > fixed. > > Please review this at http://codereview.chromium.org/5699002/ > > Affected files: > M src/accessors.cc > M src/arm/lithium-arm.h > M src/arm/lithium-arm.cc > M src/assembler.h > M src/assembler.cc > M src/ast.h > M src/ast.cc > M src/compiler.h > M src/compiler.cc > M src/deoptimizer.h > M src/deoptimizer.cc > M src/disassembler.cc > M src/frames.h > M src/frames.cc > M src/full-codegen.h > M src/full-codegen.cc > M src/handles.h > M src/handles.cc > M src/hydrogen-instructions.h > M src/hydrogen.h > M src/hydrogen.cc > M src/ia32/assembler-ia32-inl.h > M src/ia32/assembler-ia32.h > M src/ia32/assembler-ia32.cc > M src/ia32/deoptimizer-ia32.cc > M src/ia32/full-codegen-ia32.cc > M src/ia32/lithium-ia32.h > M src/ia32/lithium-ia32.cc > M src/ia32/macro-assembler-ia32.h > M src/ia32/macro-assembler-ia32.cc > M src/objects.h > M src/objects.cc > M src/runtime.cc > M src/safepoint-table.h > M src/type-info.h > M src/type-info.cc > M src/v8globals.h > M test/cctest/test-debug.cc > > > -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
