This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository lldb.
from fcb35af Another little example use of scripted thread plans.
new 1d5d1b4 Add a read_full_buffer argument to ConnectionFileDescriptor::Read
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
include/lldb/Core/Connection.h | 15 ++-
include/lldb/Core/ConnectionSharedMemory.h | 7 +-
.../Host/posix/ConnectionFileDescriptorPosix.h | 4 +-
source/Core/Communication.cpp | 6 +-
source/Core/ConnectionSharedMemory.cpp | 7 +-
source/Host/common/Editline.cpp | 5 +-
.../Host/posix/ConnectionFileDescriptorPosix.cpp | 49 +++++---
source/Plugins/Platform/Android/AdbClient.cpp | 24 ++--
unittests/Host/CMakeLists.txt | 1 +
.../Host/ConnectionFileDescriptorPosixTest.cpp | 135 +++++++++++++++++++++
unittests/Host/SocketTest.cpp | 80 +-----------
unittests/Host/SocketUtil.h | 66 ++++++++++
12 files changed, 272 insertions(+), 127 deletions(-)
create mode 100644 unittests/Host/ConnectionFileDescriptorPosixTest.cpp
create mode 100644 unittests/Host/SocketUtil.h
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository compiler-rt.
from 3bd85d7 Relax stack check as on some platforms demanglers fail
new be916c3 tsan: update tsan_analyze to what tip clang generates
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
lib/tsan/check_analyze.sh | 16 ++--------------
1 file changed, 2 insertions(+), 14 deletions(-)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository lld.
from de69b4a ELF: Remove the function SymbolTable<ELFT>::findFile.
new 0a72dce Remove unused includes.
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
ELF/Driver.cpp | 1 -
ELF/SymbolListFile.cpp | 1 -
2 files changed, 2 deletions(-)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository gcc.
from 5ab325d * gcc.dg/ipa/inline-8.c: Require c99_runtime.
new 60b27d5 S/390: Add patterns for r<nox>sbg instructions.
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
gcc/ChangeLog | 7 +
gcc/config/s390/s390.md | 34 ++++-
gcc/testsuite/ChangeLog | 6 +
gcc/testsuite/gcc.target/s390/md/rXsbg_mode_sXl.c | 151 ++++++++++++++++++++++
gcc/testsuite/gcc.target/s390/s390.exp | 11 ++
5 files changed, 207 insertions(+), 2 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/s390/md/rXsbg_mode_sXl.c
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository llvm.
from d56fee3 [X86][SSSE3] Missing combine opportunity to simplify to a MOV [...]
new c8b0a11 [mips] Use MipsMCExpr instead of MCSymbolRefExpr for all relo [...]
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
include/llvm/MC/MCExpr.h | 30 +--
lib/MC/ELFObjectWriter.cpp | 1 -
lib/MC/MCELFStreamer.cpp | 4 -
lib/MC/MCExpr.cpp | 31 +--
lib/Target/Mips/AsmParser/MipsAsmParser.cpp | 179 +++++--------
lib/Target/Mips/InstPrinter/MipsInstPrinter.cpp | 37 +--
lib/Target/Mips/MCTargetDesc/MipsAsmBackend.cpp | 21 +-
lib/Target/Mips/MCTargetDesc/MipsBaseInfo.h | 3 +-
.../Mips/MCTargetDesc/MipsELFObjectWriter.cpp | 5 +-
lib/Target/Mips/MCTargetDesc/MipsFixupKinds.h | 7 +-
lib/Target/Mips/MCTargetDesc/MipsMCCodeEmitter.cpp | 173 ++++++------
lib/Target/Mips/MCTargetDesc/MipsMCExpr.cpp | 295 +++++++++++++++++----
lib/Target/Mips/MCTargetDesc/MipsMCExpr.h | 66 +++--
.../Mips/MCTargetDesc/MipsTargetStreamer.cpp | 25 +-
lib/Target/Mips/MipsISelLowering.cpp | 2 +-
lib/Target/Mips/MipsMCInstLower.cpp | 155 +++++++----
lib/Target/Mips/MipsMCInstLower.h | 7 +-
test/MC/Mips/cprestore-noreorder.s | 10 +-
test/MC/Mips/cprestore-reorder.s | 10 +-
test/MC/Mips/expansion-jal-sym-pic.s | 72 ++---
test/MC/Mips/expr1.s | 44 +--
test/MC/Mips/macro-la.s | 8 +-
test/MC/Mips/micromips-expansions.s | 8 +-
test/MC/Mips/micromips-relocations.s | 28 +-
test/MC/Mips/mips-expansions.s | 32 +--
test/MC/Mips/mips32r6/relocations.s | 4 +-
test/MC/Mips/mips64r6/relocations.s | 4 +-
test/MC/Mips/mips_directives.s | 2 +-
test/MC/Mips/relocation.s | 46 ++--
29 files changed, 710 insertions(+), 599 deletions(-)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
christophe.lyon pushed a change to branch abe-july-2015
in repository toolchain/abe.
from 4393004 bug #2011: fix path to Foundation Model.
new 78f978e Add the right configure options for armv8l.
new ad8197c config/gcc.conf: Default armv8l to Thumb.
The 2 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
config/gcc.conf | 6 ++++++
1 file changed, 6 insertions(+)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository gcc.
from c548a0a [RS6000] Fix ICE caused by rs6000_savres_strategy thinko
new 5ab325d * gcc.dg/ipa/inline-8.c: Require c99_runtime.
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
gcc/testsuite/ChangeLog | 4 ++++
gcc/testsuite/gcc.dg/ipa/inline-8.c | 2 ++
2 files changed, 6 insertions(+)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository llvm.
from b663ea6 [AVX512] Add support for commutative MAX/MIN . In general VMA [...]
new d56fee3 [X86][SSSE3] Missing combine opportunity to simplify to a MOV [...]
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
test/CodeGen/X86/vector-shuffle-combining-ssse3.ll | 15 +++++++++++++++
1 file changed, 15 insertions(+)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
charles.baylis pushed a change to branch aarch64-v2.1
in repository toolchain/luajit-aarch64.
at 677781f WIP
This branch includes the following new commits:
new 55b1695 RELEASE LuaJIT-2.0.0-beta1
new 1d1fed4 RELEASE LuaJIT-2.0.0-beta2
new 2d0ef45 LuaJIT 2.0.0-beta2 hotfix #1
new 5287b93 LuaJIT-2.0.0-beta2 hotfix #2
new 3f1f9e1 Fast forward to sync public repo.
new 699232f Add DynASM-built files to public repo.
new 6163a90 Remove asm auto-detection due to cross-compilation issues.
new ec24428 Reorg Makefile to allow more overrides on the make command line.
new 4523c46 Fix text about x64 port in status page.
new fe36e4a Add DynASM x64 module (non-functional). Add extra x64 registers.
new 0acb234 Add REX encoding to DynASM x64.
new 8d564c7 Show memory operand size in cvtsi2sd/cvtsi2ss disassembly.
new 33171e8 Add 64 bit instruction templates to DynASM x64 and fix REX encoding.
new d12d036 Pick unused X char to force REX.W in DynASM x64.
new ef885f4 Add support for 64 bit immediates/displacements to DynASM x64.
new 547508f Fix off-by-one error in err_chunkid().
new b2b8577 Add usage message for mov64 to DynASM x64.
new 31f0074 Make DynASM conditionals available before .arch statement.
new 267d715 Allow 32 bit regs for indexing even for DynASM x64.
new 1eedc6d First bunch of register definitions for x64 interpreter.
new 8df9603 Unify interpreter reg saves/restores for WIN64 prolog/epilog req.
new 4cb357d Define x64 interpreter frame and cleanup use of stack temps.
new d2156c7 Warn about use of esp with DynASM x64.
new 85d32ae Change some misuses of esp in x64 interpreter.
new 3a18dba Add support for rip-relative displacements to DynASM x64.
new eb8ee66 Fix handling of byte-addressable registers in DynASM x64.
new 1e7951e Cleanup DynASM action list flushes.
new 1353aca Allow quadword lea for DynASM x64.
new 6adab43 Adapt primary inbound calls in x64 interpreter.
new db29a8c Change KBASE/PC regs for x64/POSIX to get shorter encodings.
new ed8d86b Better change the saved regs for the x64 interpreter, too.
new a1d4d05 Adapt most outbound calls in interpreter to x64 calling conventions.
new 34d7169 Properly compile on Debian kFreeBSD.
new 64a4528 Fix memory operand sizes for some SSE ops in DynASM.
new ab02f06 Add SSE2 variants of basic arithmetic ops in interpreter.
new 298e3f5 Add SSE2 variants for all FP ops (except vm_pow*) in interpreter.
new 44a9d7b Use SSE variants for IRFPM_FLOOR/CEIL/TRUNC unless SSE4.1 available.
new c4e9dc0 Miscellaneous cleanups for x64 interpreter.
new a5faa29 Fix last commit.
new 6ce0c90 Add build infrastructure for the SSE2-enabled interpreter.
new 690760a Add SSE variant of pow/powi to interpreter.
new bc47063 Use fastcall for remaining 1-arg/2-arg calls from interpreter.
new e5438da Ignore lea operand size in DynASM x86/x64.
new 3a15e46 More calling convention cleanups for x64 interpreter.
new 8bb38bd Final calling convention cleanup for x64 interpreter.
new da65e0d Fix DynASM x64 encoding for qword-only instructions.
new 7b7c6ed Change callee-save regs for x64 interpreter to shorten code.
new 07e1a0d Fix POSIX/x64 call argument order.
new 9c8f425 Fix size calculation for closure structs.
new 61abf34 Linux/x64 mremap() does not obey MAP_32BIT, so make it non-moving.
new 8a9cfa4 Define CFRAME structure for x64 interpreter.
new 374f534 Logical 'not' must be sign-extended for address operands.
new d18d157 Save all callee-saved x64 integer regs for unwinding.
new 52f310b Fix x64 lj_vm_pow_sse().
new 241e8db Fix bad stack setup in collectgarbage().
new 81c9f5d Fix various 32/64 bit issues in interpreter.
new 24429ed Fix alloc/free sizes of internal GCRef arrays.
new 4212422 Add DWARF2 unwind info for x64 interpreter.
new 82f7d09 Fix PE object build on x64 and with disabled interpreter.
new d64b031 Minor fixes for x64 interpreter.
new 52eb887 Fix narrowing casts of pointer differences for x64.
new 9de0f53 Implement yield from C hooks.
new 8fc9430 Rename NRESULTS to MULTRES on the assembler side, too.
new 93ee106 Major rewrite of error handling to allow external/internal un [...]
new 9df01f9 Allow @ in DynASM globals.
new b3e3bad Fix PE object build for fastcall entry points into interpreter.
new c31ac26 Add support for WIN64 exception handling to external unwinder.
new 4d9be5b Bump all copyright dates to 2010.
new ece20f8 Fix x64 PE object emitter.
new a33204a Fix 32/64 bit portability issue with upval->v.
new 2cc554d Avoid int16_t widening for pt->uv elements.
new 99d153b Improve alias analysis of upvalues using a disambiguation has [...]
new 767035f Strip '@' suffix from external symbols for MACH-O, too.
new a431d6f Shorten %p format for pure 32 bit pointers on x64.
new 82417ee Add build infrastructure for x64 interpreter.
new 3100026 Error for blacklisted loop bytecodes has no info argument.
new 62fafb5 Fix off-by-one error in x64 PE object symbol mangling.
new 49e3bdf Allocate 32 bit memory on WIN64 using NtAllocateVirtualMemory.
new 09ca54d Enable build for x64 interpreter on WIN64.
new 0954802 Silence 64 bit conversion warning.
new 4b0725d Fix more 64 bit conversion warnings.
new 379b8c5 Update docs with x64 build instructions.
new 32969ab Reduce non-numeric tag range by bumping up 64 bit lightud tag.
new 4e39597 Add some sanity checks for allocator in 64 bit mode.
new 36769c2 Fix broken intarith + testop optimization.
new a61df8e Decouple guard vs. INT check vs. TYPECHECK semantics for SLOAD.
new d62dcd4 Update docs about exception handling.
new 419cbb3 Fix ordered string comparisons. Unsigned arithmetic is evil.
new c56811b Fix undefined behavior in table resizing calculation.
new 43f1e13 Integrate MinGW build with DWARF2 exception handling.
new 055396a Force error if lua_newstate() is used in 64 bit mode.
new 47f1bc8 Use dedicated type for snapshot map entry.
new 34d84f8 Fill gaps in frames (caused by metamethod calls) with nil.
new e058714 Add missing eviction in asm_obar().
new 67ca399 Compress snapshots using a simple, extensible 1D-compression.
new c8a80fa Cleanup types for narrowing stack machine.
new 05d67cf Add missing check for return to lower frame.
new d216cdb Add frame and continuation flags to TRef and SnapEntry.
new 3aeb892 Eliminate IR_FRAME. Replace with KGC and TRef/SnapEntry flags.
new 09b8f67 Expose compressed snapshot map to reflection API.
new 21d6709 Avoid reuse of PHI registers, even for duplicate right PHIs.
new fb2c583 Fix snapshot dumps.
new 725da92 Don't modify jit_State and exit counters while in vmevent.
new a6565d2 Fix recording of metamethod result adjustment.
new 2ed0bb7 Drop obsolete frame shrinking after continuation return.
new 272b2f7 Followup fix: set maxslot for continuation return.
new 8681b73 Compile return to lower frame. Only for Lua frames right now.
new 96e8a56 Loop formation test must check for return to lower frame.
new 02e58f5 Add support for weak IR references to register allocator.
new 9a682f3 Revise hardcoded inlining in lj_asm.c. Saves 1-2K.
new 61c5a60 Improve coalescing of BASE register in side traces.
new 305ecbb Another loop formation test must check for return to lower frame.
new c1c9abe Add extra check to suppress hotcall event during recording.
new 318e86c Clean up frame depth checks and loop detection.
new 78f5e2f Fix unroll limit checks and frame depth adjustment for pcall.
new 7256690 Add shadow frame link stack for trace recorder.
new 5d2690c Reset the hotcount table after a JIT off to on transition.
new 370c868 RETF modifies BASE. Treat it like a store or it gets CSEd.
new c8d55e8 32/64 bit memory ref cleanup, part 1: GCproto ->bc and ->k.
new d778982 32/64 bit memory ref cleanup, part 2: GCproto ->uvname and -> [...]
new 8dcc436 32/64 bit memory ref cleanup, part 3: GCproto ->lineinfo.
new a0914c4 32/64 bit memory ref cleanup, part 4: GCproto ->varinfo.
new c4dadf1 Move bytecode offsets from lj_vm.* to generated header.
new 1f39cc5 Update .gitignore.
new 1307f49 Major cleanup of bytecode parser.
new 60b5af4 Redesign of prototype generation, part 1: varinfo and uvname.
new 48d93d8 Redesign of prototype generation, part 2: late init of chunkname.
new 4424027 Redesign of prototype generation, part 3: bc and lineinfo.
new f275a9d Redesign of prototype generation, part 4: late creation of pr [...]
new 6194b1c Redesign of prototype generation, part 5: colocation of proto [...]
new f86f9e8 Reduce whitespace in lj_bcdef.h.
new 2ddea38 Fix constructor bytecode generation for conditional values.
new d5a4662 Drop bc field in GCproto since the bytecode is colocated.
new bb0384c Replace GCproto reference with bytecode PC in GCfuncL.
new ab90b8f Switch to pre-initialized stacks. Drop frame clearing in inte [...]
new 565eb39 Drop frame clearing in exit handling and JIT compiled code.
new b206723 Move dispatch tables out of GG_State struct.
new 2e5d374 Fix dump of bytecode instructions with missing operands.
new 4f8d7be Extend scope of local vars in debug info beyond final return.
new c93138b Major redesign of function call handling.
new 8e38231 Implement call hooks (zero-cost if disabled).
new 2a2f8ed Implement return hooks for Lua functions (zero-cost if disabled).
new b6e4fde Improve FOR loop const specialization and integerness checks.
new 1288e04 Add missing FORI coercions in recorder.
new b838cd8 Minor cleanup of trace event handling.
new 3452bfc Add generic function handling for debug modules.
new e1905f4 Use a different marker for hot calls.
new e664945 Change record vmevent parameters. Dump function id on FUNC*.
new afa07e0 Improve error reporting for traces aborted in non-Lua functions.
new c225ee8 Split CALL/FUNC recording.
new 8060f5b Just disable JIT compiler for non-SSE2 CPUs instead of aborting.
new 4a7ee2c Update docs and changelog.
new 00ad80c Clear frame gaps in recorder to avoid resurrecting previous refs.
new 265a24b Add more assertions to compare the recorder state and the VM state.
new 2b22b9e Drop obsolete shadow frame link stack.
new 2f300fb Fix some uses of tref_isint().
new 59f54be Update trace recorder infrastructure for hot calls.
new 65586ca Adapt bytecode patching/unpatching for hot calls.
new b11eeab Use a limited history buffer for tailcall counts while recording.
new bbe7d81 Allow linking to already compiled functions.
new c52da1f Rethrow errors from trace exit handling from the right C frame.
new c1658dd Eliminate redundant stack checks. Add checks for growing root [...]
new e76bcd2 Refactor Lua stack handling in lj_asm.c.
new 19af483 Add region selection for up-recursion and tail-recursion.
new 86494c7 Back out history buffer for tailcall counts.
new 3336434 Fix TRef for (dummy) 2nd arg of __len metamethod.
new 659ea9d Ensure function and all args have a reference for call recording.
new f751cd1 Do not patch parent exit for a stack check.
new c1362dc Don't eliminate SLOAD restores across RETF.
new d5c8fe4 Mark renamed registers as modified.
new 8ae2f9f Randomize penalties for aborts and add blacklisting.
new 857f538 Fix exit state for 64 bit mode.
new 47da0d8 Fix type() implementation for x64 lightuserdata.
new 2f3a917 Check relative jump distances for x64.
new 13a9d7e Add trace exit helper for x64. Fix trace entry for x64.
new e46f4c8 Fix 64 bit portability problem in rec_ret().
new 4c9f71b Major 32/64 bit cleanups in assembler and exit handling.
new 4e73488 Fix 64 bit conversion warnings.
new b32f4f4 Handle negative continuation offsets (WIN64 with debug).
new b952945 Move SIMD constants to jit_State to keep them in the low 4GB.
new 3c6cec0 Add x64 call argument setup. More 32/64 bit cleanups in assembler.
new a0fbb05 Make penalty PRNG available for general use by compiler.
new 52b7651 Place dynamically generated code near static code on x64.
new ec78d31 Correctly align and free allocated machine code areas.
new 2144d55 Save/restore remaining callee-save regs on trace<->interp tra [...]
new 956065f Add frame info for x64 GDB JIT API.
new 2e22d33 Enable JIT compiler for x64.
new 8cc50cf Conditionally compile functions that are unused with JIT disabled.
new 89c90f0 Fix error message generation for OOM error.
new 7e69612 Resize stack up to the true limit.
new 69ea553 Fix pc to line number translation for GDB JIT API.
new e7b737a Implement down-recursion.
new 5d0b655 Enable tracing of recursion.
new a1bacd8 Fix amalgamated build.
new eed5585 Add support for recursion to changelog.
new 1cda54a Add compile errors for 64 bit OS without an equivalent of MAP_32BIT.
new 20fa30b Fix comments about 64 bit FreeBSD mmap() behavior.
new 3230958 Improve register allocation on x64.
new 5fdb6e2 Fix 64 bit conversion warning.
new 15c3bd7 Improve performance of HREF/HREFK on x64.
new 3d2abf3 Build as a native 32 or 64 bit binary by default.
new f76e5a3 Allocate 32 bit memory on OSX/x64 with mmap() hinting.
new af92a14 Update docs: native build default, cross-compilation, embedding.
new f8f6d02 Document jit.status().
new 5594a0c Fix generated 64 bit Mach-O assembler output.
new 88244c6 Fix disassembly of call/jmp ModRM.
new b429925 Fix unwind info for assembler part for OSX.
new e807d46 Generate indirect calls for out-of-range distances on x64.
new e38fbb3 Fix assertion in rec_check_slots.
new e40b001 Improve placement of dynamically generated code on x64.
new 6769397 Rebase dynamic library on OSX/x64.
new 9fd1c6c Add OS/CPU/CC compatibility matrix to docs. Fix spelling.
new f354b46 Fix tracebacks for failed coroutines.
new 09e8755 RELEASE LuaJIT-2.0.0-beta3
new 2e614ad Do not fuse SLOAD across RETF.
new d87cb5b Add WinSDK v7.0 requirement for building on Windows/x64.
new 6e6034e Avoid tracing the nil return case of tonumber().
new f49649d Fix recording of getmetatable() for non-tables.
new c56d791 Fix handling of bad argument types in recorder.
new 4f19a15 Fix potential hang in UCLO redirection handling in hooks.
new 16f3342 Restore MULTRES for snapshots pointing to CALLM etc. bytecodes.
new a25a717 Fix folding of (comparison x x).
new 24402ed Reorganize scalar evolution analysis.
new c472722 Add array bounds check elimination (-Oabc, on by default).
new e57390f Update changelog.
new 871941a Add static target to msvcbuild.bat.
new 1fea5cb Fix TSETM on x64/SSE builds when table is resized.
new 37a3ca3 Reorder various structs to reduce padding (thanks to /usr/bin [...]
new 5c8f1f2 Increase max. fusion distance.
new 8e20b03 Add links to interactive performance comparison to offline docs.
new 51c14bf Avoid snapshots for returns to known callers.
new 3612665 Move free node pos to t->node[0].freetop. Saves 4 bytes in GCtab.
new 097db73 Move colocated array part after GCtab (now properly aligned).
new 6038866 Pass MULTRES or negated error code in RD to lj_vm_exit_interp.
new 69df01e Fix precondition check for NEWREF.
new 44756de Only fuse with spill slots for cross-section references.
new 1130ade Also merge EQ(HREF, niltv) in backend.
new e4bca95 Generate EQ(HREF, niltv) for load path, too (better CSE).
new 1d1d922 Fold HREF of TNEW/TDUP to niltv. Fold HLOAD of niltv to nil.
new 3a8b9ba Update changelog.
new 23189fa RELEASE LuaJIT-2.0.0-beta4
new db75643 Don't copy SNAP_NORESTORE mark into loops and suppress restor [...]
new fbe092c Fix setup of RD when dispatching to function headers after exit.
new ff82df7 Refactor buildvm symbol generation.
new 932cda0 Replace on-trace GC frame syncing with interpreter exit.
new d8cb69e Simplify GC step calls from on-trace code.
new ab45481 No longer let the GC replace dead keys with the LJ_TDEADKEY tag.
new a6c52d8 Avoid compiler warning.
new 28a6284 Add assertions to guard against using lua_*call on dead coroutines.
new 3a32bbc Simplify lexer a bit.
new 721b73f Turn traces into true GC objects (GCtrace).
new 2e24770 Simplify management of current trace. Drop lazy save.
new 87e4daf Reorganize and fix placement of generated machine code on x64.
new ef0904e Make metamethod names proper GC roots.
new 2ccb24f Fix TNEW in x64 interpreter. Do not force a full GC (ouch).
new f396f3d Avoid starting a GC cycle immediately after library init.
new 4137912 Treat the tag of a TValue as unsigned everywhere.
new 28cfccf Turn TValue setter macros into inline functions.
new f28a42c Fix PHI elimination: mark right PHI as used, too.
new 8fa1db8 Create symbol table of JIT-compiled code for use with Linux p [...]
new 5ff994f Add weak guards. Emit TNEW/TDUP with a guard bit.
new 41ec9a9 Avoid string allocation in GDB JIT API.
new 655401f Avoid name clash with standard POSIX function.
new c7a3103 Fix handling of failed fp:read(..., "*n").
new 926f688 Canonicalize string conversion of nan, inf, -inf.
new fa673eb Fix string.char() range check on x64.
new cdc37a3 Fallback to metamethod resolving for tostring in print().
new 9f6707d Fix frame setup for error handling in next().
new bbd1584 Do not eliminate PHIs for values only referenced from side exits.
new 7d5a3cb Rechain pseudo-resurrected string keys with colliding hashes.
new 0d3f190 Fix passing of constant args in FPRs on x64.
new 2ca58aa Fix arg check for two-arg SSE math functions in interpreter.
new 6667ab0 Use independent hash for VM event dispatch.
new d05873e Abstract out pointer hash to hashrot(). Tune hash constants.
new 4201243 Switch to fast string hash.
new c5feda2 Speed up string compares in string interning.
new a6b7952 Use FuncState typedef in favor of struct.
new 33ef1fa Untangle some target dependencies.
new 653327b Remove bogus global in -jdump.
new 309d8f5 Cleanup library function caching in dasm_x86.lua.
new 0a3329c Fix some compiler warnings.
new 3d81368 Fix lua_cpcall(). Duplicate save corrupted C frame chain.
new 44de7eb Keep framesize in RA of FUNCC/FUNCCW instructions, too
new be19218 Fix return from pcall within active hook.
new 0d6f6f3 Clean up some more DynASM target dependencies.
new 65b8c36 Fix encoding of doubles for mixed-endian cross builds.
new 55017d4 Fix formatting of canonicalized number to string conversions.
new d9cdefe Update changelog.
new d668373 RELEASE LuaJIT-2.0.0-beta5
new 06ae534 Fix string.format() for non-finite numbers.
new 5526fa2 PPC: Add PowerPC target architecture selection.
new 4f47d31 PPC: Add DynASM PowerPC module. Standard instructions only.
new 4aa8757 PPC: Add DynASM PowerPC encoding engine.
new 1b57266 PPC: Add SPE instructions to DynASM PowerPC module.
new a0e3c9b PPC: Add support for word-wise relocations to buildvm asm targets.
new 877ff1e Always enable check mode for DynASM (only used at build-time).
new 96b60d5 PPC: Add skeleton for PowerPC interpreter.
new a8d23e8 PPC: Add build rules and preprocessed skeleton for PPCSPE target.
new b3bd9b5 Improve target architecture detection in src/Makefile.
new 9cb5046 PPC: Add stack frame layout for PPCSPE target.
new 9c16a9f PPC: Amend DynASM PowerPC conditional branch instructions.
new 844e40d PPC: Clean up masked shift/rotate target settings.
new 7313a08 PPC: Add missing PowerPC subtract mnemonics.
new e1efd0d Minor tweaks to integration of assembler part.
new c7f91f8 PPC: Add register assignments and type definitions.
new 4ef6564 PPC: Add instruction/call decode + dispatch macros.
new 690b1f3 PPC: Add entry and exit points into VM. Add type check macros.
new f4d4af1 PPC: Add support to call C functions and to return back to C.
new 96c1d8a PPC: Clean up JIT/non-JIT symbol declarations in assembler part.
new 233dc76 PPC: Add support to call Lua functions.
new fc1866f PPC: Handle stack growth during function calls.
new 00d1071 PPC: Add return instructions.
new 8876704 Improve coalescing of multiple KPRI instructions to KNIL.
new d71bd78 PPC: Add move and constant instructions.
new d67eb4d PPC: Handle stack growth on return to C.
new 7ccdc80 PPC: Unwind the stack after an error is thrown.
new 222e01f PPC: Add loop instructions.
new 7457ee8 PPC: Add comparison instructions.
new 44fb3eb PPC: Add unary test and copy instructions.
new f708d31 Explicitly indicate tailcall from fast function fallback.
new 5043efa PPC: Add table indexing with string keys. Add get/set global.
new b4c8b05 PPC: Add table indexing with variable keys or uint8_t.
new 684acbe PPC: Add call and iterator call instructions.
new 711e55b PPC: Add tailcall instructions.
new 3418209 PPC: Add table constructors.
new 15834c3 PPC: Add unary instructions.
new db735e0 PPC: Add binary arithmetic instructions (except MOD).
new e9e7df5 PPC: Add heavily optimized floor/ceil/trunc functions.
new 932193b PPC: Add modulo instructions.
new b2aa08c PPC: Add helper function for folding arithmetic ops.
new ec8f291 PPC: Add closure-related instructions.
new 9fd1096 PPC: Add upvalue get/set instructions.
new 8a1af8c PPC: Use unsigned compares for pointer comparisons.
new fd63b05 Use biased integer constant for TSETM array index.
new 96957a4 Turn some lua_State fields into 32 bit pointers.
new 6fd3155 Optimize BC_VARG: use RC for numparams.
new daafc0b Reduce minimum Lua frame size to 1.
new e15e4c4 PPC: Add metamethod handlers for call instructions.
new cabcce1 PPC: Add metamethod handlers for arithmetic instructions.
new c330af2 PPC: Add continuation dispatch and RA/nop continuations.
new 90f2276 PPC: Add argument coercion for BC_FORI.
new b72ae54 Fix broken MSVC build.
new c2c08ba Record calls to vararg functions.
new 4da0396 Record vararg expressions with known fixed number of results.
new 52f75e9 Fix off-by-one errors in maxslot calculation of trace recorder.
new e32f7d9 Fix tailcalls from vararg functions.
new 847b9cf Record vararg expressions with varargs defined off-trace.
new 4ba0eb5 Record select().
new c88169d Record y = select(x, ...) idiom.
new 32bfece Release all memory when using the builtin allocator.
new cd6a457 PPC: Teach DynASM to reject r0 for some operands.
new dcfc385 PPC: Fix write barrier and linkage for BC_USETV/BC_USETS.
new 7a766c7 PPC: Add missing write barrier to BC_TSETV.
new 0001916 PPC: Add metamethod handlers for indexing instructions.
new 8a87395 PPC: Add support for shifted MULTRES.
new 3d17f71 x86/x64: Avoid unnecessary resize in BC_TSETM.
new a36287b PPC: Add better diagnostics to DynASM for reg/imm mismatches.
new fe159ea PPC: Keep MULTRES in a register.
new 2bdf3fc PPC: Fix KBASE restore in BC_CALLT.
new f8b7fb7 PPC: Fix write barrier in BC_TSETV.
new 5b74004 PPC: Fix line number in stack overflow error.
new b69c02e Avoid unnecessary vararg loads.
new 23655bd Improve alias analysis for vararg loads.
new 8dc76ee Add IR_VLOAD for vararg loads.
new 77267bc Show names of IR calls in disassembler output.
new 4470272 Don't traverse inactive GCtrace objects.
new 028a65a PPC: Add BC_VARG.
new 3f32f4a PPC: Add BC_TSETM.
new 028f80c PPC: Avoid comparison with NaN in BC_ISEQN/BC_ISNEN.
new 4f1d43d PPC: Fix ordered comparisons for NaN operands.
new 66f0c43 PPC: Add BC_CAT and lj_cont_cat.
new 9dc9828 PPC: Reschedule bytecode instruction fetch.
new 9203d27 PPC: Add metamethod handlers and continuations for comparisons.
new 52b922c Avoid fusing potentially negative indexes into AREF on x64.
new 23b5c56 Improve alias analysis: disambiguate new allocations.
new 38628d9 Improve FOLD/CSE of field loads and array/hash refs across NEWREF.
new 3ae27ff Drop redundant (UGE any +0).
new 4cd2299 Improve AA for field accesses: disambiguate new allocations.
new bc146dd Consistently round int args of string functions on x64.
new 1c3b8c5 PPC: Fix constant folding of modulo operator.
new 7336305 PPC: Fix result copy for return to Lua.
new 0c0047e PPC: Add missing register setup after unwind to ff pcall.
new 201eec4 PPC: Add fixed register for TOBIT bias.
new aff6b30 Whitespace.
new d270e89 PPC: Reschedule continuation dispatch.
new 5baf0f9 Fix stack growth for coroutine.wrap().
new d351c2a x86/x64: Fix stack overflow handling for coroutine results.
new 5beedd6 Workaround for missing fseeko/ftello prototypes on ancient glibc.
new 6aedfa7 PPC: Add lj_vm_resume().
new 493d0aa PPC: Add dispatch targets for instruction and return hooks.
new f635e3e PPC: Add dispatch target for call hook.
new a500752 Fix checks for minimum required compiler versions.
new dc4fdec PPC: Don't use callee-save registers when folding '%' operator.
new a471360 PPC: Add frame unwind info for assembler part.
new 408cd05 PPC: Add basic infrastructure for fast functions.
new 4afff97 PPC: Add bit.* fast functions.
new c0a8f5f PPC: Add dispatch to fast function fallback handlers.
new fce05fd x64: Use 64 bit slot copies in the interpreter.
new ac76493 Specialize bytecode for pairs()/next() iterator. Speedup: 3.5x.
new dcf006b PPC: Add BC_ITERN and BC_ISNEXT.
new 61d498b PPC: Add next() and pairs() fast functions.
new 05099f0 PPC: Add ipairs() fast function and its iterator.
new e396bef PPC: Add pcall() and xpcall() fast functions.
new ddae887 Update docs for PPC port. Clarify and extend install instructions.
new 8b096c5 PPC: Add assert() and type() fast functions.
new 08ac8b0 PPC: Add getmetatable(), setmetatable() and rawget() fast functions.
new 2f5f629 PPC: Add tonumber() and tostring() fast functions.
new 14cd10e PPC: Add table.getn() fast function.
new 3418512 PPC: Remove pointless optimization in BC_UNM.
new 52b67fd PPC: Add math.abs/deg/rad/min/max() fast functions.
new 793fb4a PPC: Add math.ldexp/frexp/modf() fast functions.
new 8f5efb2 PPC: Use own implementation for math.floor/ceil() fast functions.
new 443d28b PPC: Fix math.deg/rad() fast functions.
new 82b2f61 PPC: Add string.len/byte/char() fast functions.
new b4f3d45 PPC: Add string.rep/reverse/lower/upper() fast functions.
new 9f14e03 PPC: Add string.sub() fast function.
new 9d201c9 PPC: Add coroutine.resume/wrap_aux/yield() fast functions.
new a7caa48 PPC: Use ZERO register instead of extra load immediate.
new a3eeee5 PPC: Update DynASM-generated file.
new 89fdee9 Fix external link in docs.
new 5dcf630 Reorganize compatibility table in install docs.
new 5140b40 x64: Optimize internal/external tag conversion in lua_type().
new cc62ede x64: Fix type check for numbers in compiled code.
new b3cf2c7 Decouple SLOAD type and optional conversion.
new 5391a5f Fix recording of y = select(n, ...) for non-int indexes.
new 65dec38 PPC: Preserve RD during array resizing in BC_TSETM.
new 02dc9d1 Avoid compiler warning.
new 08b0ede Reduce 32GB RLIMIT_DATA on FreeBSD/x64 to allocate low-2GB memory.
new ab1b67f Fix collectgarbage("count") result if more than 2GB is in use.
new 7e5cb31 Fix anchors in API docs.
new 44372a4 Fix parsing of hex floats.
new 188f0b0 Number parser shouldn't accept '0x' without hex digits.
new fe21a42 Fix conflict between loop branch inversion and HREF+NE/EQ merging.
new ad29c1f Rename character type handling from lj_ctype* to lj_char*.
new b45e324 Split up extension/API docs into sub-pages.
new 1de05d1 Avoid using negative hex numbers with DynASM.
new 24baf77 Cleanup architecture, ABI and OS definitions.
new 751ff5b Add jit.os string.
new 3754a8f Allow running C functions with coroutine.create(), too.
new ba602c9 Add support for __pairs and __ipairs metamethods (from Lua 5.2).
new 57cd502 Disable Lua 5.2 features by default. See -DLUAJIT_ENABLE_LUA5 [...]
new 29b8959 Parse hexadecimal escapes in strings (from Lua 5.2).
new 7cc981c string.format("%q", str) is now fully reversible (from Lua 5.2).
new 52fd87b Add support for "%g" character class in patterns (from Lua 5.2).
new b776bf9 Tighter check on table.sort function compliance (from Lua 5.2).
new 367ff0c Parse '\*' escape in strings (from Lua 5.2).
new 7338456 Support os.exit(status|true|false [,close]) (from Lua 5.2).
new 18fbf86 Improve unrolling heuristics for non-looping inner loops.
new 7840d0b Shrink slots for 'break' statement to help data-flow analysis.
new 6290d6f Display caller location in errors from metamethods.
new 2fd1292 FFI: Add cdata object type.
new ef8c264 FFI: Add macros for consistent number to integer truncation.
new 35890f6 FFI: Don't propagate cdata objects.
new 24186e0 Remove metamethod frame for errors from internal metamethods.
new 2033d63 Add shorter defines for endianess checks.
new 592cf3f Don't mark REF_BASE in DCE.
new b3ef604 Update copyright for Lua 5.2-derived features.
new cd19013 Rearrange src/Makefile. Split features and debugging support.
new 443c542 Fix check for enabled JIT compiler in lj_trace.h.
new 4ed8a19 Add support for deferred library loads to lib_init.c.
new 0ffd06a FFI: Add define to disable the FFI library.
new d98cd19 Allow access to raw error messages from outside lj_err.c.
new 100a00d Undo ef8c2648.
new 8679ca5 POSIX is an acronym.
new 05973ee FFI: Add C type management.
new 526e087 FFI: Add C data handling and C type conversions.
new 233076d FFI: Add C declaration parser.
new 513b0ba FFI: Add support for converting cdata to tonumber().
new 7cb250c FFI: Add ffi.* library.
new ffa5a2b Don't export internal symbol.
new 559545e Whitespace.
new e7f8cc9 Split off fast function recording to lj_ffrecord.c.
new 26830d0 Make sure to use irt_toitype() macro everywhere.
new 6a7605e Rename IRT_PTR to IRT_P32.
new b1fb71f Cleanup 64 bit IR type handling.
new 5a13fa6 Add IR_KINT64.
new c5f4f60 Fix xmm spill/restore broken by b1fb71fb.
new 588fa04 Add support for 64 bit integer arithmetic to x64 backend.
new 281f426 Add minimal set of fold rules for KINT64.
new db3d16b Add IR_XSTORE.
new 4755765 Avoid warning.
new 73bc83c Simplify lj_ir_kvalue(). Fix IR_KNULL + IRT_P64 case.
new e137408 Use native IRT_PTR instead of IRT_LIGHTUD workaround for FILE*.
new a850b27 Avoid compiler warnings.
new b56b834 Support all kinds of XLOAD/XSTORE references in backend.
new 86fd228 Add support for integer IR_MUL.
new 5270586 Fix XLOAD/XSTORE references.
new a5bdae8 Fix 64 bit shifts in backend. Fix shift by 0.
new f6c4e8d Add FOLD rule to turn i << 1 into i + i.
new 159f51d Add IR_TOI64.
new b3c4c08 Add IRT_NUM support to XLOAD/XSTORE.
new e6093b1 Fix register allocation for 8 bit stores in x86 backend.
new d4d95ee Fix x64 code generation for A16+REX prefixed instructions.
new 2a917a9 FFI: Treat cdata metatable+methods as immutable in trace recorder.
new d778680 FFI: Record cdata index operations (preliminary, lots of NYI cases).
new 3ef6a53 FFI: Turn cdata indexing into x86/x64 [base+idx*sz+ofs] addressing.
new 72744da Fix IRT_NUM support for XLOAD/XSTORE.
new 3c78a7f Avoid fusing loads if there are multiple references.
new 6a04591 FOLD (base+(idx+k)*sz)+ofs ==> (base+idx*sz)+(ofs+k*sz).
new d0b283e Avoid stack resizes while recording calls to vararg functions.
new 7007124 Fix type-based colorization of snapshot slots in debug output.
new f308e00 Add FOLD rule to reassociate 64 bit (x+k1)+k2.
new 4f0d3e8 Apply narrowing optimization to IR_TOI64, too.
new e9cab5a x86/x64 backend: keep invariants on the right rather than fus [...]
new 36fed9f Allow integer literals in FOLD rules.
new 1d86090 Strength-reduce 32 to 64 bit widening using scalar evolution [...]
new cc0efd8 FFI: Fix alignment of scalars/vectors modified with mode attribute.
new 870bcd7 Add missing guard for check against "#" when recording select().
new 09e2bdb FFI: Always specialize to the field name for struct access.
new 6947493 Turn lj_mem_newgco() into a fastcall.
new cdac0fa FFI: Add IR_CNEW/IR_CNEWI to allocate/init cdata objects.
new 44101c6 FFI: Don't use KNULL for arbitrary NULL pointers.
new 5f7e917 Regroup FOLD rules for constant folding.
new 476259e Add 64 bit bswap and rotates.
new 42f9b38 Extend all FOLD rules to work on 64 bit integers.
new a403cb1 Add missing PHI barrier to strength reduction of widening.
new a22ea88 FOLD (base+k)+(idx*sz)+ofs ==> (base+idx*sz)+(ofs+k).
new 74317fa Add alias analysis for XLOAD/XSTORE. Add DSE for XSTORE.
new 9666cf5 Fix type-based alias analysis for XLOAD/XSTORE.
new badff4e Fix error location while recording metamethods.
new fbcc925 FFI: Fix auto-deref of pointers to structs.
new 09ef7eb FFI: Simplify copy-by-value. Strip attributes, keep qualifiers.
new c4a0fb4 FFI: Record copy-by-value for pointer and complex C types.
new c98132e FFI: FOLD load of initializers even across PHIs.
new 255c6e8 FFI: Drop IR_CNEWI. Add IR_CNEWP only for pointers/refs.
new c9a0b9e Improve alias analysis for cdata allocations.
new 1eb6e23 Fix dump of pointer constants.
new c8d6f07 Fix XSTORE for IRT_NUM constants.
new dbab6cf Reassociate XLOAD across PHIs to handle a[i-1] forwarding case.
new 4f2f44d Avoid compiler warnings.
new c0ffe7d FFI: Record cdata allocations.
new 6082f93 Improve register allocation for loops with variable shifts.
new ba20871 Fix display of branches to exit stubs in 32 bit disassembly.
new 79ab1a4 FFI: Add pointer arithmetic.
new aa705f7 FFI: Fix C parser for hex literals.
new 4850865 Fix misnamed macro in Lua parser.
new 461bf77 FFI: Parse complex and 64 bit integer literals.
new dd65e00 FFI: Fix C type representation for valarrays.
new 44935da FFI: Add 64 bit integer arithmetic.
new a70fb94 Make recording of type() independent of LJ_T*/IRT_* mapping.
new d176433 FFI: Catch various NYI cases while recording.
new 06c61d9 Use macro-map for IRT_* enum. Add IRT_FLOAT.
new 2ca2de7 Refactoring of conversion ops, part 1: add IR_CONV.
new 8bcc5ed Add support for IRT_FLOAT to XLOAD/XSTORE.
new 158de60 Refactoring of conversion ops, part 2: cleanup cdata conversions.
new 4668b22 FFI: Add missing GC steps for implicit allocations.
new dc4b82c Add lj_num2u64 for number to uint64_t conversion.
new 65b194a Copy destination type for CONV from ir->t to op2, too.
new 1f26961 Refactoring of conversion ops, part 3: add FOLD rules for IR_CONV.
new 1716540 Refactoring of conversion ops, part 4: use CONV instead of TO [...]
new 783dbd3 Refactoring of conversion ops, part 5: drop TOINT/TOI64/TONUM.
new 007067b Fix register allocation for FP <- int conversion.
new 3b47eba Add 8/16 to 32 bit sign/zero-extension variants to CONV.
new ddfa7fc Add conversions for type mismatches in XSTORE forwarding.
new 0005b2c Add support for uint64_t <-> FP conversions to x64 backend.
new 55a9be7 FFI: Record tonumber() for boxed cdata.
new e66b5b6 Improve uint64_t <-> FP conversions in x64 backend.
new 331b148 Use cdata to pass IR_KINT64 to -jdump.
new b2c74c5 FFI: Add support for cdata constants to IR.
new 3291bb6 FFI: Record pointer arithmetic.
new 593821f Add support for integer IR_NEG to backend.
new 44a22b9 Add support for non-constant integer IR_MUL to backend.
new eaf88df FFI: Record 64 bit integer arithmetic.
new 3b468ce FFI: Force snapshot after store to cdata.
new bc37edc Fix handling of floats in x86/x64 backend.
new a148a6a FFI: Add missing link to parameters for C function declaration.
new 97875c4 Avoid compiler warnings.
new fb0a18c FFI: Add support for calling C functions.
new aca1e5b Fix Windows/x86 build and update pregenerated files.
new bfc638c FFI: Add missing GC steps for C function calls.
new 4c0330a FFI: Handle NYI cases for cdata call metamethod.
new 06f99fc Bump copyright date to 2011.
new 9502afb FFI: Preserve stack top across implicit load of FFI library i [...]
new d215747 FFI: Fix function pointer lookup for calls.
new ddf6596 FFI: Add ffi.load() and ffi.C default namespace.
new 87a6c8a FFI: Auto-detect __stdcall and fix up C function declarations.
new a54843b FFI: Lookup decorated name for fastcall/stdcall functions.
new 8d858bf FFI: Fix C calls with complex values on x64.
new 1548383 Avoid compiler warnings.
new 2363399 FFI: Untangle and fix calling convention definitions.
new a5e4828 FFI: Rearrange code in lib_ffi.c.
new e7b08b2 FFI: Simplify logic for pointer arithmetic.
new 0ec7f5e FFI: Add 64 bit integer comparisons and pointer comparisons.
new 6e702d7 FFI: Cleanup some type conversions.
new 706bbe3 Fix recording of select() with multi-char string as 1st argument.
new 59643c8 Add FOLD rule for CONV.num.u32 KINT.
new 0fa32e5 FFI: Record conversions from strings to enums or pointers.
new 6f74657 FFI: Allow indexing a struct constructor to get constants.
new 2940ab0 Fix assertion.
new 335232b FFI: Record conversions to bool ctype.
new b2ff889 FFI: Don't swap operands for cdata __eq metamethod call.
new 66ba1e6 Add trace recorder infrastructure for instruction post-processing.
new 83d8c86 FFI: Record conversions from bool ctype.
new f385af7 FFI: Record 64 bit integer comparisons and pointer comparisons.
new 751eff9 Cleanup and fix trace flush logic.
new 685dfc3 Fix for the fix for the trace flush logic.
new 925050f Differentiate between IR_KPTR and IR_KKPTR.
new 8407ac0 Add volatile XLOADs.
new d20e53c Combine i8/u8 XLOAD with BAND+comparison into test byte mrm, imm8.
new 6bce6b1 Add compile-time option LUAJIT_ENABLE_CHECKHOOK. Disabled by [...]
new 8396c3c PPC: Fix ipairs() for keys in the hash part.
new e985aed FFI: Add preliminary FFI documentation (still incomplete).
new f529d22 Another fix for the trace flush logic. I'll get this right someday.
new 72b3fff FFI: Simplify initializer rules. Clarify docs.
new ef3b209 FFI: Fix ffi.load() error message on Windows.
new 96ef87b Clear bytecode penalty cache in jit.flush().
new 705f593 FFI: Don't force zero terminator on string to array conversion.
new ed6c895 ARM: Add ARM target architecture selection (disabled).
new 1b75ec2 FFI: Add symbol name redirection.
new a285415 One more fix for the trace flush logic. Sigh.
new 30db71e FFI: Convert enum return value of C function to underlying type.
new 0622ad9 Fix comments for BC_ITERN.
new de0781d FFI: Allow cdata indexing with integer cdata.
new 2526498 FFI: Record cdata indexing with integer cdata.
new 3184f17 FFI: Move code for cdata arithmetic to lj_carith.c.
new 279b6ec FFI: Fix symbol name redirection.
new cd9b8f9 FFI: Fix various issues with C type table reallocations.
new 07d8a53 FFI: Split up 64 bit x^k helper into signed/unsigned.
new 6fee000 Fix 64 bit case of (SUB x x) and (BXOR x x) FOLD rules.
new 6fd721c FFI: Implement POSIX/x64 struct-by-value calling conventions.
new 992bc2c FFI: Limit index range for complex numbers.
new c539c0c Add SSE3 CPU feature detection.
new b613216 Add SPLIT pass to split 64 bit IR instructions for 32 bit CPUs.
new 1505d6f Fix OSX build to work with newer ld64 versions.
new 2c8945d Use names defined in lualib.h for library registration.
new a1e0f99 FFI: Record 64 bit integer divide and modulo.
new 433fd18 FFI: Add basic FOLD rules for 64 bit integer DIV, MOD and POWI.
new 1027018 Rename IR_POWI to IR_POW.
new df65b8b FFI: Rename IR_CNEWP to IR_CNEWI and use it to box 64 bit integers.
new 1abf5d4 FFI: Disable MUL => BSHL FOLD rule on 32 bit.
new 618b451 FFI: Avoid intermediate boxes for tonumber(), too.
new c29ed4d FFI: Record ffi.string().
new 223d85a Treat metatables of special userdata objects as immutable.
new 14f5103 FFI: Record C library namespace lookups.
new da3f256 Fix metamethod comparisons triggered by BC_ISEQP/BC_ISNEP.
new bf05209 FFI: Optimize snapshots for cdata comparisons.
new 26fadcd FFI: Record simple C function calls.
new 26bb0fe FFI: Limit number of arguments for recorded calls.
new 4c74f70 Strength-reduce 32 to 64 bit widening for XLOAD U8/U16 inputs.
new bf58160 FFI: Simplify and fix tonumber() for cdata objects.
new ceece4d FFI: Perform stricter checks in ffi.cast(). Record ffi.cast().
new 8da287c FFI: Fix handling of enum arguments to C calls.
new 51e8fe9 FFI: Allow cdata types for integer arguments of ffi.* functions.
new 0a50356 Improve static assertion macro.
new 97d0874 Add IR_XBAR, a barrier against XLOAD/XSTORE optimizations.
new fb53d4a FFI: Fix recording of pointer arithmetic.
new 7a37b93 FFI: Record ffi.copy() and ffi.fill().
new c39ca7c FFI: Record ffi.abi().
new 9c81c81 Fix bytecode optimization of and/or operators.
new 2388a7f FFI: Document current FFI implementation status.
new 24c314e FFI: Add more docs on FFI semantics.
new a7ab6b2 Fix various HTML errors in the docs.
new a2f9f1f FFI: Finish docs on FFI semantics. Phew.
new a5aade2 FFI: Finish FFI docs.
new 1f0006a Cleanup of docs.
new a8114d5 Update changelog.
new 7c61073 FFI: Disable unused code for some build modes.
new 97d8411 RELEASE LuaJIT-2.0.0-beta6
new 19b8400 Fix type mismatch in XSTORE forwarding.
new 4117a61 PPC: Disable FFI due to NYI: comparisons and calls.
new 4dca589 Add workaround for lj_meta_tset() newkey inconsistency.
new 5fe7282 Fix error handling within metamethods. Special-case FFI mm.
new 54b6054 FFI: Fix management of C library handles for default namespace.
new 963f05c DUALNUM: Add build infrastructure.
new 03946ac DUALNUM: Add integer type to core VM.
new 3e2a7a2 FFI: Fix docs about C bool -> Lua boolean conversion.
new bd29d16 FFI: Improve conversion error messages.
new 32eaa48 FFI: Fix recording of ffi.copy() and ffi.fill().
new cf88c43 FFI: Fix 64 bit to 32 bit truncations on x64.
new 34bcf63 FFI: Fix recording of userdata conversions.
new d82022a From Lua 5.2: coroutine.running(). Needs -DLUAJIT_ENABLE_LUA5 [...]
new f83d58d From Lua 5.2: Empty statement. Needs -DLUAJIT_ENABLE_LUA52COMPAT.
new baef199 From Lua 5.2: Add package.searchpath().
new c823e26 From Lua 5.2: table.unpack(). Needs -DLUAJIT_ENABLE_LUA52COMPAT.
new 2bae11a From Lua 5.2: fp:read("*L").
new 8d0b073 From Lua 5.2: Support load(string).
new 4c97cc7 Eliminate dead slots in snapshots using bytecode data-flow analysis.
new a720db4 FFI: Record calls to functions with void results.
new b90d80d Fix table.maxn().
new d437086 x64: Improve accuracy of x^-k.
new c031d4b DUALNUM: Handle integer type in FFI.
new cead25f DUALNUM: Handle integer type in x86/x64 interpreter and libraries.
new 83a37ae FFI: Add ffi.gc() function for finalization of cdata objects.
new e1aa8d0 FFI: Fix compiled ffi.string() semantics.
new 6c43767 FFI: Resolve ld script redirection in ffi.load().
new 1cb58fb Fix code generation for Intel Atom in x64 mode.
new f29a9d4 Fix narrowing of POW.
new 8eb09fe Suppress recording of retried fast functions.
new a30f58b DUALNUM: Fix x87 code for BC_ISEQN/BC_ISNEN.
new 54978e4 DUALNUM: Make overflow guards weak. Add IR_USE and IR_MULOV.
new 449cee4 Fix code generation for bit.bnot().
new 0125ee8 DUALNUM: Add extra assertions to interpreter.
new 30cc4d3 Fix handling of err_loc() within cpcall frames.
new a127eaf FFI: Fix x86 code generation for ffi.string().
new 1c9981a x86/x64: Fix code generation for fused IR_MUL/IR_MULOV.
new 3f26e3a FFI: Fix and optimize recording of cdata[cdata].
new bfce3c1 DUALNUM: Handle integer type in JIT compiler.
new 889368e Get rid of the remaining silly cast macros from Lua.
new 7fb1075 DUALNUM: Add integer variant of MIN/MAX.
new a226fb4 FFI: Force snapshot after C call to preserve semantics.
new dcbae09 DUALNUM: Narrow unary minus.
new 6299485 DUALNUM: Narrow result of math.floor() and math.ceil().
new ac3b1dc x64: Workaround for libgcc unwind bug (still present in RHEL 5.5).
new 063182d FFI: Fix marking of ffi.gc() finalizer table.
new 642ae06 x64: Use external unwinding for lua_yield().
new a1c887e Fix data-flow analysis for BC_ITERC.
new 7088abc DUALNUM: Fix narrowing of unary minus.
new 156bf15 ARM: Add DynASM ARM module and encoding engine.
new b6ed984 ARM: Add support for ARM relocations to buildvm.
new 32aef8a ARM: Add skeleton for ARM interpreter.
new 492efb7 Clean up DynASM glue macros. Thanks to Josh Haberman.
new 4c9a10f Avoid compiler warning.
new 98e3c8a ARM: Add register assignments, type definitions and stack layout.
new 71f976b ARM: Add instruction/call decode + dispatch macros.
new 3bcb48c ARM: Add frame unwind info for assembler part.
new cff08b2 ARM: Add entry and exit points into VM. Add type check macros.
new c04a252 ARM: Add support to call C functions.
new ca494b7 ARM: Add support to call Lua functions and return from them.
new 7b0a125 ARM: Add basic unary bytecode instructions.
new a0e4711 ARM: Add basic loop and branch instructions.
new 5b06b29 ARM: Add binary arithmetic instructions.
new 2138df2 ARM: Fix escaping of opcodes in DynASM.
new 33bab1f ARM: Don't conditionalize non-opcodes in DynASM.
new f1c79f8 ARM: Minor fixes and cleanups for type checks.
new 85fff38 ARM: Add missing parts of return handling.
new 81fa9e3 ARM: Add call and iterator call instructions.
new 0e7009d ARM: Add table constructors.
new dd4a043 ARM: Add table indexing with string keys. Add get global.
new 6110faa ARM: Add basic infrastructure for fast functions.
new 5b146f6 ARM: Add tostring() fast function.
new 6511941 ARM: Add fast functions pairs()/next() and ipairs() + iterator.
new 959683c PPC: Minor cleanups of interpreter.
new 43d7db6 ARM: Add pcall() and xpcall() fast functions and unwinders.
new a7874cb ARM: Add assert() and type() fast functions.
new 314995a ARM: Finish basic table indexing and add metamethod handlers.
new 5b7a8ba x64: Fix getmetatable(lightuserdata).
new 48438b9 Fix code generation for PHIs with type IRT_FLOAT.
new 3acd489 Add FOLD rule for FLOAT->NUM->FLOAT conversions.
new 66a1006 Fuse XLOAD/XSTORE operands more aggressively.
new 734e96f Remove links to shootout.
new ee64487 FFI: Register the FFI library as a loaded module in luaopen_ffi().
new 1a56dac ARM: Add build rules (still disabled).
new 3f8fed5 ARM: Add pc-relative loads to DynASM.
new 0046f95 ARM: Handle stack growth during function calls.
new 1f4f805 ARM: Add getmetatable(), setmetatable() and tonumber() fast f [...]
new 83e3029 ARM: Add math.* fast functions.
new 3af4106 ARM: Add test/copy and comparison instructions.
new 9e18260 ARM: Add upvalue get/set and closure-related instructions.
new 16e0eca ARM: Add table.getn() fast function.
new aee129a ARM: Add tailcall instructions.
new b1912f3 ARM: Add missing metamethod handlers and continuations.
new b56784a ARM: Add BC_CAT and lj_cont_cat.
new 27bf481 ARM: Misc. fixes for interpreter.
new 161bc37 ARM: Add GPR variants for shift aliases to DynASM.
new f089f39 ARM: Add bit.* fast functions and lj_vm_tobit().
new 89022b4 Use lj_vm_tobit() on targets without FPU.
new da1ef8c ARM: Add rawget() fast function.
new 040c8c5 ARM: Add string.* fast functions and GC steps.
new 54e986a ARM: Misc. fixes for interpreter.
new f69524f PPC: Misc. fixes for interpreter.
new d85b7e6 Fix recording of JLOOP with higher maxslot.
new efcaef2 Inline lj_gc_barrierback().
new a1f66ab Avoid phantom stores to proxy tables.
new 23f847f Optimize lookups in empty proxy tables.
new fa5cd01 Add support for tailcalls from internal C functions.
new 3b6f37d FFI: Add ctype metamethods and ffi.metatype().
new 46d97d2 ARM: Add modulo subroutine and temp. math.floor()/math.ceil().
new 7aa7069 ARM: Add dispatch targets for hooks.
new 0001a7f ARM: Add BC_VARG and BC_TSETM.
new 1fa0cc0 PPC: Fix L->status for resume after yield.
new 31e0774 FFI: Fix cdata finalization.
new 5f08a50 FFI: Fix arithmetic ctype metamethods for references.
new fc40214 ARM: Add coroutine.resume/wrap_aux/yield() fast functions.
new d636a3d ARM: Add partial support for FFI.
new add553e ARM: Add integer variant of modulo operator.
new 32db452 ARM: Misc. fixes for interpreter.
new 7bb4b0f ARM: Normalize results of math.floor()/ceil() fast functions.
new 54994ef ARM: Enable ARM port.
new 38a842a Add common TARGET_FLAGS option to Makefile.
new b53ca06 ARM: Add install docs.
new ca18293 Add remark about target OS to cross-compilation instructions.
new a1ffb61 ARM: Add missing FFI functionality. Enable FFI for ARM.
new 4fbacaf ARM: Allow compilation in thumb mode (CCOPT_ARM=-mthumb).
new b3d10cd ARM: Workaround for unaligned accesses.
new e94a12f Fix compiler warning.
new 9ea6794 Workaround to compile with Clang. Fix Clang warnings.
new 5d096dc FFI: Add ffi.istype() function.
new 0b60606 ARM: Use own lj_bswap(). Reduce min. req. version of GCC to 4.2.
new 7ff8409 DynASM: x86/x64: Fix vreg in opcode. x64: Fix imm "I" after a [...]
new 0ba34ff ARM: Fix ABI and build issues for iOS. Now works on iOS 3.0+.
new b4cfaa6 Improve target OS selection for buildvm.
new eb7b452 x86: Pop unused FP result from x87 stack.
new c77ca54 FFI: Compile C function calls with 64 bit args/results in 32 [...]
new 139175f Tune loop unrolling heuristics. Increase trace recorder limits.
new b4006ba Fix stack slot updates for down-recursion.
new fb8b87e More tuning of loop unrolling heuristics.
new 704752c FFI: Fix field alignment for iOS.
new 6534152 Fix compiler warning.
new b760add Misc. updates to docs.
new 693b629 Improve bytecode optimization of and/or operators.
new 998061f Update changelog.
new cfdc356 RELEASE LuaJIT-2.0.0-beta7
new 16369d5 x86/x64: Fix bit.* argument checking.
new 87553d9 Avoid name clash with Windows MM_MAX define.
new 77ba772 FFI: Save errno/GetLastError() around allocations, hooks etc.
new 868ecad FFI: Add ffi.errno().
new d9c1f77 FFI: Allow setting errno with ffi.errno(), too.
new 28e87d3 Use common helper functions for unaligned loads.
new 67d3ac9 Fix some portability issues with the JIT compiler.
new c8308e2 ARM: Interpreter/JIT integration, part #1.
new 65d9c7c ARM: Fix table/userdata equality comparisons in interpreter.
new b79cdba Tuning loop unrolling heuristics again.
new 8003493 Disentangle target-specific parts of JIT assembler backend.
new 58f38c2 Move x86/x64 parts of JIT assembler backend to extra files.
new be73a96 ARM: Generalize machine code co-location. ARM has a +-32MB range.
new 1ac71f1 ARM: Add LJ_SOFTFP define. Add support for soft-float slot handling.
new f0a7293 ARM: Remove redundant parse_gpr() in DynASM.
new a251a28 ARM: Add support for GDB JIT API.
new 6cb123b Fix compiler warning.
new dab4bca Add target-specific macro for stack alignment in assembler backend.
new fb46370 Cleanup of target dependencies.
new 3168d9a Clarify docs wrt. table initializer handling for VLA/VLS.
new 3a942d6 Portability cleanups for assembler backend.
new effdfda Move IR_CALL* definitions to lj_ircall.h.
new 919ba5f Add missing guard to for loop argument conversion.
new d0115c6 More portability cleanups for assembler backend.
new 138f543 Split up FP IR instructions with SPLIT pass for soft-float targets.
new aaba681 ARM: Add rematerialization scheduler for constants to backend.
new afad72a ARM: Add ARM-specific tuning to generic assembler backend.
new c156c8a ARM: Add ARM disassembler.
new 288085a Get register maps for jit.dump from target disassembler.
new 7b21a66 FFI: Fix __gc for VLA/VLS cdata objects.
new 185554b ARM: Unify rotates to simplify assembler backend.
new ae31799 ARM: Add CPU detection.
new 625ef86 Simplify helper routines for soft-float targets. Add POW rejoin.
new 4057620 Treat extra exit from stack check like a parent exit.
new 8d44003 ARM: Flush instruction cache in assembler backend.
new 840a067 ARM: Fix rematerialization scheduler.
new 46955be Fix handling of number constants in snapshots in SPLIT pass.
new 7bad42a Fix compiler warning.
new 7683f39 ARM: Fix math.ldexp() recording and folding.
new 12dc697 ARM: Interpreter/JIT integration, part #2.
new 638f968 FFI: Avoid unused functions if FFI is disabled.
new 03b5c8c Clean up memory allocator initialization and catch early OOM.
new 0764c61 ARM: Fix string.sub() argument checks.
new 514ccfc ARM: Fix __newindex metamethod handling for BC_TSETB.
new 496d375 ARM: Optimize hash algorithm for 3-operand CPUs.
new a522087 x86/x64: Minor cleanup in assembler backend.
new 865ec11 FFI: Fix recording of casts from 32 bit pointers to integers.
new 2d0b325 ARM: Tune rematerialization scheduler.
new 8dd09d5 DUALNUM: Fix missing type check for loop index slot.
new 9e15300 ARM: Catch C++ exceptions in interpreter frames.
new a770bf3 Many soft-float-related fixes to SPLIT pass and assembler backend.
new fff2fb3 ARM: Don't generate unaligned XLOADs.
new 5d82cfd ARM: Add ARM-specific assembler backend.
new 84e3333 ARM: Enable JIT compiler (except for iOS).
new 77bb8de ARM: Disable C++ exception catching on Symbian.
new eec28aa Move math helpers to lj_vmmath.c. Add missing log2/exp2 for Symbian.
new 8c59350 Add narrowing of modulo operator.
new b88eaa1 Add stdint.h replacement for Symbian. Sigh.
new 7ad4908 ARM: Fuse XLOAD/XSTORE operands.
new a175754 PPC: Add lj_vm_floor which is callable from C.
new a48058a ARM: Add fast assembler implementation of floor/ceil/trunc.
new 22946f2 FFI: Change tonumber(cdata) semantics. Return nil for non-numbers.
new f50d898 Fix data-flow analysis for BC_UCLO.
new 51ee3d9 Minor change to bytecode penalty code for consistency.
new a7ca722 FFI: Show address pointed to for tostring(pointer).
new a0840d2 FFI: Print NULL pointers as "cdata<... *>: NULL".
new b44ccd4 Fix compiler warning.
new 585cf05 Move debugging/introspection functionality to lj_debug.c.
new 8c32b38 Flatten and compress in-memory debug info (saves ~70%).
new 96cc99b Remove bogus assertion for GC estimate vs. threshold.
new b6a7fc5 Cleanup prototype flags.
new 27d8d3b Fix check for missing arguments in string.format().
new a58eb60 Make .eh_frame section read-write on Solaris.
new 287de61 FFI: Support __tostring metamethod for pointers to structs, too.
new 0c8696d No need for L argument to lj_str_initbuf().
new 9da94d1 Add recursive dump option to jit.bc.dump().
new 4994fcc Add support for bytecode loading/saving.
new 0eee70c Add -b command line option to save/list bytecode.
new bcc196e Fix dumping of already stripped functions with debug info.
new ecab831 Add more docs for bytecode load/save feature.
new 9cf2cd2 Fix bytecode loading on Windows.
new 3d15635 More fixes for the Solaris/x86 build. Thanks to Will Metcalf.
new 0f55b46 Update man page for -b command line option.
new 31d5664 From Lua 5.2: __len for tables. Needs -DLUAJIT_ENABLE_LUA52COMPAT.
new 4df40a2 Fix recording of loops with instable directions in side traces.
new b6e4ffd Avoid pointless calls for std file handles on __gc.
new e2246db x86/x64: Fix fusion of comparisons with u8/u16 XLOAD.
new d4da9ff x86/x64: Fix register allocation for variable shifts.
new 513bd6e x86/x64: Don't evict BASE in ra_evictk().
new 6691e72 Fix assertion.
new f182559 Misc. updates to docs.
new 2513ddc Update changelog.
new 9b0c641 RELEASE LuaJIT-2.0.0-beta8
new f50c962 Fix documentation for -b command line option.
new 29e89ad RELEASE LuaJIT-2.0.0-beta8 (fixed)
new e613cb7 FFI: Clarify docs wrt. string initializers for byte arrays.
new 113f8e9 Fix data-flow analysis for BC_ITERL.
new db22ce2 Fix iOS build.
new 2dd5250 Use thread-safe exceptions for external unwinding on x64/GCC.
new 999d952 Fix forced unwinding triggered by external unwinder.
new 3dbae4f Clean up temporary bytecode patching.
new deeb819 Reorganize trace linking and track link types.
new 82eca89 Prefer recording loops over calls.
new 64dbe7d Return to lower frame via interpreter for unhandled cases.
new 0af502f Don't use thread-safe exceptions on OSX. Complain to Apple, not me.
new 8ccf897 Add debug option to msvcbuild.bat and improve error handling.
new fe1f913 PPC: Fix range check for scaled immediates in DynASM.
new 8addfef PPC: Untangle PPC vs. PPCSPE target defines.
new 161f1a5 PPC: Separate PPC vs. PPCSPE interpreter source files.
new 0239a56 PPC: Add build rules for PPC interpreter (non-functional, yet).
new 89201f2 Fix install_name and versioning on OSX.
new ee4cffb ARM: Install dis_arm.lua.
new 29ac908 FFI: Fix comment parsing.
new fe6ded9 Fix doc link.
new eb92daa ARM: Don't evict ASMREF_L in ra_evictk().
new 9687472 Cleanup single-number vs. dual-number mode selection.
new 6599bc7 PPCSPE: Fix dispatch for binop metamethods.
new ca8dadb PPC: Add dual-number mode interpreter.
new 9c58bd6 PPC: Update install docs.
new b261d0e From Lua 5.2: Change \* to \z. Reject undefined escape sequences.
new d001d7b FFI: Treat function ctypes like pointers in comparisons.
new ee2bb5e Use different workaround for alignment of jit_State.
new 551f09c PPC: Add dual-number mode interpreter. Enable by default.
new 2610675 PPC: Fix floor/ceil for dual-number mode.
new 4008b35 PPC: Add machine-specific part of FFI.
new d9b5188 FFI: Limit snapshot optimization for cdata comparisons.
new bed0f18 Remove temporary mcode limit error from application stack.
new 4c50e36 Fix stack check when L->top is above L->maxstack.
new f333dfd Fix compiler warning.
new 28b98ac PPC: Tune and reschedule interpreter for PPC/e300.
new 5385809 PPC: Add mtcrf instruction to DynASM.
new d84207b PPC: Save/restore cr2, cr3 and cr4 when entering/leaving VM.
new e8fa76f PPCSPE: Save/restore cr2, cr3 and cr4 when entering/leaving VM.
new 05ef03e Don't bother to chain NOPs.
new 10ffd05 x64: Fix result type of cdata index conversions.
new 4a42933 Replace some _Unwind_* types. Fixes build issues on OSX 10.7.
new b500b50 Clarify name of MIT license.
new aad7ea3 Load embedded bytecode with require().
new a9dd47b Extend -b to generate c/h/obj/o files with embedded bytecode.
new 607bd88 Add potential PHI for number conversions due to type instabil [...]
new ec1b70a PPC: Add PPC disassembler.
new 78cc9c8 Record missing FORI slot loads (e.g. after return to lower frame).
new 6627744 FFI/x64: Fix struct-by-value calling conventions.
new 32c1d47 Avoid conversion narrowing of arithmetic with large constants.
new 3e904ae Mark installed shared library as executable.
new 25a0e66 Always use ANSI variants of Windows system functions.
new e2b89de PPC: Add y-bit variants of conditional branches to DynASM.
new a5b5f41 PPC: Disassemble y-bit instead of at-bits for conditional branches.
new d9fe562 PPC: Export register names from disassembler.
new 690ff90 PPC: More interpreter tuning. Use y-bit for branch predictions.
new bab2f0e PPC: Fix or remove assembler helper functions.
new ae3317b PPC: Interpreter/JIT integration.
new 1b949dc Use some register allocator helpers for multiple architectures.
new c56ae9f Increase max. number of exit stub groups to correspond with - [...]
new 1c5fbd4 ARM: Disable CPU detection for interpreter-only builds.
new c3116b4 Fix typo.
new 6d854c0 ARM: Fix GC barrier for TSETM.
new 1a899ff PPC: Fix GC barrier for TSETM.
new 5fedf72 PPC/e500: Fix GC barrier for TSETM.
new c5e75e7 Generate a portable vmdef.lua.
new da9c74f ARM: Fix HREFK code generation.
new 3c998f6 Fix (harmless) typo.
new 76a2a7c Avoid overwriting argument with error from metacall recording.
new f34ac81 Remove unneeded PHI barrier for reassociation of duplicate ops.
new 4a14462 FFI: Fix recording of const/enum lookups in namespaces.
new 120c3ad FFI: Record loads/stores to external variables in namespaces.
new bc95d54 FFI: Fix call argument and return handling for I8/U8/I16/U16 types.
new 882537a FFI: Fix unfused loads of float fields.
new 4c93187 Save tmptv state for trace recorder across RECORD vmevent.
new c5164b1 FFI/ARM: Ensure invocation of SPLIT pass for float conversions.
new 270de2d Add missing FOLD rule for integer NEG of constants.
new e5f310e PPC: Distinguish slwi, srwi and rlwinm in disassembly.
new f50075a Fixup PC in tracebacks after exits from down-recursive traces.
new 2fcd562 Don't fuse colocated array offsets without -Ofuse.
new 45df079 FFI/x86: Fix CONV.u32.num rounding.
new a39aac0 Fix FOLD rule for CONV.flt.num(CONV.num.flt(x)) => x.
new 00591a2 Fix PE/COFF bytecode file format for big-endian targets.
new e80478c Rearrange defines for workarounds to embedded operating systems.
new ff7f7a7 Cleanup various endianess issues in assembler backend.
new c805a49 Separate/rename NUM/FLOAT <-> I64/U64 conversion functions.
new c01aaad Fix register hints for arithmetic helper functions.
new 2e5c087 x64: Fix math.random() code generation.
new fee957b FFI: Cleanup/fix 64 bit arithmetic in assembler backend.
new aa50220 Add missing FOLD rule for CONV.u64.int + sign-extension.
new 7aa2c99 Whitespace.
new 86a5cc4 Fix compiler warning.
new 8468340 Don't save/restore tmptv with copyTV.
new a0d7827 Generalize handling of stack checks indicated by highest exit + 1.
new cb1dd15 PPC: Add support for per-trace exit stubs.
new 3209892 Fix handling of alignment arguments (nil) to CALLX.
new 8fc9bc1 Make sure to flush/sync the D/I-cache for all generated code.
new 3591609 PPC: Add Linux/PPC cache flush mechanism.
new 826fb37 PPC: Add support for GDB JIT API.
new 167f638 PPC: Generalize rematerialization handling.
new 96d8611 PPC: Fix/improve interpreter/JIT integration.
new fa9ade3 Allow selection of single-number vs. dual-number mode in Makefile.
new 0cf8c20 PPC: Integrate and enable JIT compiler.
new bec4d20 PPC: Fix write barrier in USETV/USETS.
new 9df9c7e PPC: Fix externally provided relocation offsets in DynASM.
new ca2b923 PPC/e500: Fix write barrier in USETS.
new 179fe9e PPC: Fuse BSWAP with XLOAD/XSTORE to lwbrx/stwbrx.
new 17cb366 x64: Fix code generation for BSWAP.
new 455917f Add more FOLD rules for shift + BAND simplifications.
new 1d190c9 PPC: Fuse BAND with shift to rlwinm.
new 2dc574d FFI: Compile calls to stdcall, fastcall and vararg functions.
new ec56c25 PPC: Install dis_ppc.lua.
new 7b0157a Spelling.
new f1849fa x64: Fix lightuserdata comparisons.
new 4bcf2b4 ARM: Fix code generation for type-check only softfp [AHUV]LOAD.
new fa1675b ARM: Fix IR splitting for softfp XLOAD.
new 0123e4f FFI: Extend metamethod tutorial.
new 635371c FFI: Add unwind definitions for lj_vm_ffi_call.
new 4d59752 Avoid dependence on lj_jit.h if only including lj_target.h.
new ae79ff4 FFI: Resolve __call metamethod for pointers, too.
new e9eb4fd PPC: Clarify docs.
new 71d00a5 FFI: Add callback support (for x86/x64).
new 87456f7 Fix hint-based register allocation in ra_dest().
new 8437d0c Eliminate some rare occurrences of redundant PHIs.
new d4df8d7 FFI: Save GetLastError() around ffi.load() and symbol resolvi [...]
new b1efcef FFI: Add missing GC barrier for callback registration.
new e549ffc FFI: Fix EnumWindows example in docs.
new 726dc42 Fix debug option for msvcbuild.bat.
new dc2a39e Get rid of snap->depth.
new f8a4769 Keep maximum frame extent in snap->topslot.
new cc7a12b Count frames with same prototype in call unroll limit check.
new cecbe3c Specialize to prototype for non-monomorphic functions.
new bd758df Replace stack slot for implicit number->string conv. in Lua/C API.
new 9237384 Do not eliminate PHIs only referenced from other PHIs.
new c2dcf39 FFI: Improve ld script detection in ffi.load().
new 90c4458 FFI: Fix GetLastError() save/restore around ffi.load().
new b3f16cb FFI: Fix line info for result conversion errors in callbacks.
new c142b6c FFI: Record C function calls with bool return values.
new 1a42c03 Fix stack check in side exit.
new e3c7476 FFI: Record ffi.errno().
new 0a31c63 Don't increment module reference counts due to symbol resolving.
new d1569a5 Fix docs. The code base doesn't work on Windows 98 anymore.
new ff7e9bc FFI: Need to treat strlen() as a load.
new e162d8c FFI: Fix stack adjustment for calls to stdcall/fastcall functions.
new 6398143 Fix compiler warning.
new 2591b45 x64: Always extend stack arguments to pointer size.
new 60b8e9e ARM: Fix stack check for spilled BASE in parent.
new 2d8f3d7 ARM: Error out for attempt to compile with hard-float ABI.
new 1047498 Move helper for syncing data/instruction cache to lj_mcode.c.
new 1b0d646 FFI: Add callback support for PPC.
new 03c51fc FFI: Add callback support for ARM.
new a064156 PPC: Avoid undefined operand dereference in BAND/shift fusion.
new a302468 Update changelog.
new 6ace80c RELEASE LuaJIT-2.0.0-beta9
new b61be29 Auto-detect target arch via cross-compiler. Drop TARGET=arch.
new b330b46 MIPS: Add build rules (non-functional, yet).
new ba4917b MIPS: Add DynASM MIPS module and encoding engine.
new d050b2f FFI: Ignore empty statements in ffi.cdef().
new 248cf2f FFI: Ignore number parsing errors while skipping definitions.
new 2f396c6 FFI: Don't touch frame in callbacks with tailcalls to fast fu [...]
new a72134e Fix parsing of hex literals with exponents.
new b9651b4 MIPS: Add missing opcodes to the DynASM MIPS module plus mino [...]
new 93a71f6 ARM: Avoid pointless arg clearing in BC_IFUNCF. Fix comments.
new 0a8fd0a PPC: Fix resume after yield from hook.
new 9f48511 PPC: Fix argument checking for rawget.
new 7d2774e PPC: Avoid pointless arg clearing in BC_IFUNCF. Remove dead code.
new 5bed11e MIPS: Add interpreter. Enable MIPS build rules.
new 20813bc MIPS: Update install docs.
new 10ef109 Bump copyright date to 2012.
new d0b1646 MIPS: Add MIPS disassembler.
new d72d758 FFI: Fix library unloading on POSIX systems.
new 8e524d4 Fix bytecode dump for certain number constants.
new ff71394 From Lua 5.2: Try __tostring metamethod on non-string error m [...]
new 6c05739 DynASM: Lua 5.2 compatibility fixes.
new 5e113d0 Fix argument type in error message for relative arguments.
new 7498ee2 Fix argument error handling on Lua stacks without a frame.
new 8ae38dc FFI: Finalize cdata before userdata when closing the state.
new 6670a13 Add missing mcode limit check in assembler backend.
new 5dbb667 Fix compilation on OpenBSD.
new 9589e11 Avoid recursive GC steps after GC-triggered trace exit.
new 339142c FFI: Change ffi.load() library name resolution for Cygwin.
new 0187924 Replace unwind.h definitions with our own.
new 6777316 Fix OSX build issues. Bump minimum required OSX version to 10.4
new 509ca0f PPC: Fix fusion of floating-point XLOAD/XSTORE.
new a53a549 PPC: Fix HREFK code generation for huge tables.
new 1980ee9 PPC: Use builtin D-Cache/I-Cache sync code.
new aaaf0e0 MIPS: Interpreter/JIT integration
new 2225c9a MIPS: Add MIPS32R2 compile-time/runtime CPU detection.
new bcd459a MIPS: Integrate and enable JIT compiler.
new 03275ba FFI: Fix resolving of function name redirects on Windows/x86.
new 407cc5e FFI: Fix symbol resolving error messages on Windows.
new 5a83601 MIPS: Add support for GDB JIT API.
new c3e1d72 MIPS: Update docs.
new d621a6c Fix discharge order of comparisons in Lua parser.
new e14b634 Ensure running __gc of userdata created in __gc at state close.
new 44876ec Limit number of userdata __gc separations at state close.
new 39112fd Fix bytecode JMP slot range after const + and/or optimization.
new 3ed5172 Fix DSE of USTORE. Have to remove OBAR, too.
new 68ca796 Make lua_concat() work from C hook with partial frame.
new cf5a54f FFI: Fix blacklisting of C functions calling callbacks.
new 7aa34aa Fix dependencies.
new ac6b678 FFI: Fix result type of pointer difference.
new ba3cad0 Add required PHIs for implicit conversions (via XREF fwd).
new 633b248 Update changelog.
new 5439631 FFI: Clarify docs wrt. 'local ffi = ...' and the interactive prompt.
new 904cc1f ARM: Reorder type/value tests to silence Valgrind.
new 0522f74 ARM: Fix register allocation for ldrd-optimized HREFK.
new 325a182 ARM: Fix conditional branch fixup for OBAR.
new d4a0943 ARM/FFI: Invoke SPLIT pass for double args in FFI call.
new e099059 Add more comparison variants to Valgrind suppressions file.
new 9a2ae72 ARM: Handle all CALL* ops with double results in SPLIT pass.
new 8d31b5c ARM: Fix rejoin of pow in SPLIT pass.
new eec0d80 ARM: Fix compilation of math.sinh/cosh/tanh.
new 53a285c Disable loading bytecode with an extra header (BOM or #!).
new b466641 Add links to LuaJIT mailing list.
new afc17b3 FFI: Use correct PC in FFI metamethod error message.
new 0d824b7 Fix PHI stack slot syncing.
new 886a450 FFI: Allow 'typedef _Bool int BOOL;' to make Windows users happy.
new 7d39540 FFI: Don't record test for bool result of call, if ignored.
new 6e5c872 FFI: Fix recording of test for bool result of call.
new 5cb4ad2 Fix typo.
new 649600b Update changelog.
new c8a1f27 Remove stray comment.
new 51f05d6 RELEASE LuaJIT-2.0.0-beta10
new 48567b6 ARM: Fix tonumber() argument check.
new 3fc5948 ARM: Fix modulo operator and math.floor/ceil for inf/nan.
new 74205dc ARM/PPC/MIPS: Fix tailcall from fast function (esp. tostring).
new e6bcb54 FFI: Don't call FreeLibrary() on our own EXE/DLL.
new f319704 Preserve snapshot #0 PC for root traces.
new 3a43ea5 FFI: Equality comparisons never raise an error.
new 10b9ed3 PPC/e500: Fix tailcall from fast function (esp. tostring).
new 97b7edb Fix argument checks for coroutine.create().
new fa2572d FFI: Convert io.* file handle to FILE * pointer (but as a void *).
new 5bd133d Link to LuaJIT wiki.
new 5032e6d FFI: Use CNEWI/FLOAD to box/access ffi.new("int", x) (for varargs).
new bd23a6c MIPS: Support generation of mips/mipsel bytecode object files.
new 5ed58ff Add uninstall target to top-level Makefile.
new 9ab2ae8 Split-off large-file support flags to TARGET_LFSFLAGS.
new 0bd1a66 Print version and JIT status to stdout, not stderr.
new 1543b2a Link to LuaJIT Roadmap 2012/2013.
new fad5382 Rearrange OSX vs. iOS defines. Disable callbacks on iOS.
new 7266e27 Correctly preserve snapshot #0 PC for root traces (insert NOP).
new cccf763 Move lib/* to src/jit/*.
new e75561b FFI: Fix address calculation for refs.
new 6c8aaef DynASM: Compatibility with minilua.
new 0a6c833 Add minified Lua interpreter (minilua). Used by the build process.
new 9a9509c Reorganize build process.
new 1c8ed53 Fix check for minimum GCC version.
new 9d7bd04 Avoid potential store-forwarding stalls in GC marking.
new 9f443e8 ARM/PPC: Detect more target arch variants. Detect console OS.
new 37be8a5 CONSOLE: Handle unimplemented features for console toolchains.
new b8f5727 CONSOLE: Fix PS3 build.
new d5a915c CONSOLE: Minor PS3 build fixes.
new e496a50 Pass various build, arch and OS flags to DynASM.
new 58ec704 x86/x64: Clean up interpreter.
new 5862293 Fix userdata __gc separations at state close.
new c0aaf27 PPC: Fix interpreter state setup in callbacks.
new c99439b x86/x64: More interpreter cleanups.
new 02acb39 ARM: Cleanup interpreter.
new 7da4d16 PPC: Cleanup interpreter.
new 3f1035a PPC/e500: Cleanup interpreter.
new b51d7fb MIPS: Cleanup interpreter.
new b1c2a45 OSX: Change dylib names to be consistent with other platforms.
new e089149 Improved target detection for msvcbuild.bat.
new 0528bb7 x64: Fix DynASM defines.
new 6b605bd DynASM/PPC: Add missing 64 bit rotates and mtocrf/mfocrf.
new 46dc6d3 PPC: Fix disassembly of rldcl/rldcr.
new c2159f8 ARM/PPC: Remove assembler code for lj_vm_foldarith().
new 0b3d837 PPC: Remove lj_vm_floor/ceil/trunc assembler wrappers.
new 80ff9d6 Allow parsing bcsave.lua (-b option) without FFI.
new 4f9db0f Clean up buildvm error message.
new ca6bf2d CONSOLE: Handle P64, GPR64, TOC* and PPE (PS3) in PPC interpreter.
new 946c741 FFI: Add support for parameterized C types.
new 126f990 FFI: Improve error messages for ctype arguments.
new 6cee133 FFI: Resolve metamethods for constructors, too.
new e9e4531 FFI: Do not resolve strings passed to parameterized types.
new 8b71ab1 FFI: Check for __new metamethod when calling a constructor.
new 2f844fc Fix build issues on Cygwin and MinGW with optional MSys.
new 3dc7baf PPC: Fix PPC32 build.
new 9a98ba6 Fix TDUP to HLOAD forwarding for DUALNUM builds.
new ceb76d2 Another attempt at proper target detection for msvcbuild.bat.
new 0d1056e Create string hash slots in template tables even for non-cons [...]
new a9bf455 Add HREFK forwarding. Eliminate HREFK guard for TDUP refs.
new a751c8a Fix buffer check in bytecode writer.
new a6458ce Make os.date() thread-safe.
new dd9e7a4 Fix cdecl string parsing during recording.
new eacecc9 Clean up snapshot restore.
new 89f8c92 Expand array slots in template tables with mixed const/var in [...]
new cda3630 Clean up RegSP handling for parent link instructions.
new 7ae3832 Move snapshot replay for side traces to lj_snap.c.
new 264177b Use HIOP for XSTORE in SPLIT pass.
new 5d0115e Add explicit IR_GCSTEP instruction.
new 0688de5 Add IR_PVAL instruction for non-slot parent links.
new f093a34 Handle joining IR_HIOP in SPLIT pass.
new afeb4e2 Handle initial snapshot for side traces in lj_snap.c only.
new 79e1eaa ARM: Simplify two-step constant rematerialization.
new 0af3f47 Add allocation sinking and store sinking optimization.
new 2df1904 Avoid strict aliasing issues.
new c740a30 Invoke SPLIT pass in side trace for rejoined sunk stores.
new 17d3fc4 Avoid pesky compiler warnings about C++ keywords (eh?).
new 80d0acc Fix compiler warnings.
new d5b9d84 Fix NEWREF marking in SINK pass.
new 4da7ffc Remove unneeded snapshot preps for sunk stores.
new c22d698 Add stricter check for sinking PHIs.
new bd782ce Bump some limits to accomodate SINK pass.
new 7c05648 Drop range limit for sunk stores relative to sunk allocation.
new 6a67fa8 Finally fix target detection for msvcbuild.bat.
new e3dec04 Add Lua BitOp to minilua.
new c00ffcb Change DynASM bit operations to use Lua BitOp.
new b23a783 Clean up ARM capability flags. Only set highest arch version.
new 6e4d0fd Add missing declarations for MSVC intrinsics.
new f20ff8a Avoid compiler warnings about redefined C++ keywords.
new 79ecb23 Don't allocate constants for sunk stores.
new 24fa857 Use an explicit flag to signal SINK tags for a trace.
new 53e28d7 Limit number of parent links handled by backend.
new 0e2f5be FFI: Show address pointed to for tostring(ref), too.
new bf2d4ac Correct dispatch table modifications for return hooks.
new 63bb052 From Lua 5.2: Add -E command line option (ignore env vars).
new e44cb53 ARM: Invoke SPLIT pass for leftover IR_TOBIT.
new 08d8ec0 x86/x64: Fix fusion of unsigned byte comparisons with swapped ops.
new c172e7a Workaround for MSVC conversion bug (double -> uint32_t -> int32_t).
new 2139c67 Fix FOLD rule (i-j)-i => 0-j.
new 4d9c29a FFI: Box all accessed or returned enums.
new f1b4fa8 Fix restore of sunk CNEWI with implicit P32 to P64 conversion.
new 834ff6d Fix slot flags in snapshot replay.
new 3636a72 Turn loads from immutable upvalues into constants.
new e8af6e9 Don't treat all constified cdata content as constant.
new 2d58872 DynASM/ARM: Add VFP instructions.
new b98c158 ARM: Add VFP instructions to ARM disassembler.
new e84c2a9 Handle cross-compiles with FPU/no-FPU or hard-fp/soft-fp ABI [...]
new 85f29ff Need to handle TBAR in SINK pass.
new 23abbd9 Only replay sunk stores up to the snapshot reference.
new a373fdd ARM: Add VFP and hard-float ABI variants to interpreter.
new 4b0af61 ARM: Add hard-float ABI support to the FFI (interpreter).
new 77183e7 FFI: Fix alignment of C call structure.
new 0a750fe FFI: Initialize all fields of standard types.
new 806822b FFI: Improve error message for bad indexing attempts.
new 5e22786 Add missing change for last commit.
new e2373c1 FFI: Compile assignments from enums.
new 2717623 FFI: Compile ffi.typeof(cdata).
new e6fac91 Never compile DWARF unwinder on Windows.
new 4dc9e22 ARM: Add VFP and hard-float ABI variants to JIT compiler.
new 85a02f9 ARM: Enable VFP and hard-float ABI support.
new 1878d17 Update cross-compilation docs.
new ece23a4 Do not use DWARF unwinder on Windows.
new 1a3e5e1 FFI: Don't keep CType * across call. Callback may reallocate table.
new 5e18c91 x86: Simplify error message for CPUs lacking CMOV.
new ff0a1f3 FOLD x / 2^k ==> x * 2^-k.
new 0dd175d Fix last commit.
new d61e9e5 Fix shrinking of direct mapped block in builtin allocator.
new f58a84d ARM: Fix calls to hard-float floor/ceil/trunc.
new 2ac083c MIPS: Fix calls to floor/ceil/trunc.
new 1481263 ARM, MIPS, PPC: Rematerialize FPR invariants first.
new ceaa60c ARM: Don't use argument GPRs to rematerialize FPR arguments.
new 1fef2df MIPS: Don't use argument GPRs to rematerialize FPR arguments.
new af2cb02 ARM: Fix hard-float lj_vm_trunc() (used by compiled math.modf).
new f2479a9 Minor fixes to FOLD rules.
new 099635c ARM: Fix commit ceaa60c0.
new 52ca2cc ARM: Fix base register coalescing.
new 653facd Add support for '%a' and '%A' to string.format.
new 4c882fe Replace strtod() with builtin string to number conversion.
new dad6645 Remove old Makefile rules.
new cf3a263 Fix compilation with FFI disabled.
new c0efa6f Replace divisions with simpler code.
new 774280d Replace some trivial uses of fprintf() with fputs.
new 1e477e6 Fix underflow handling in builtin string to number conversion.
new 30f458f ARM, PPC, MIPS: Improve XLOAD operand fusion and register hinting.
new 76b18b2 Add table of IR type sizes.
new c7826af FFI: Detect type punning through unions.
new 751cd9d Don't constify upvalues that may retain large amounts of memory.
new ff00a78 Limit recursion depth in string.match() et al.
new d369999 Add check for unsupported MIPS soft-float targets.
new a6d636c Preserve snapshot #0 PC for all traces (potential gcstep exit).
new 90ec1f9 Fix recording of ctype() constructors for pointers.
new 0648fd4 FFI: Handle __pairs/__ipairs metamethods for cdata objects.
new 21cea85 FFI: Always resolve metamethods for pointers to structs.
new 76ab370 FFI: Correctly propagate alignment when interning nested types.
new 039bf85 Add more assertions for stack consistency during recording.
new 23932a6 Use 0/1 macro for Lua 5.2 compatibility.
new aed2009 Fix despecialization of ITERN when already running.
new ead325b From Lua 5.2: Add goto and ::label:: statements.
new 998c656 From Lua 5.2: debug.getupvalue() returns "" for C function upvalues.
new 7d49b19 From Lua 5.2: debug.getlocal()/setlocal() treats slot < 0 as vararg.
new c687d01 From Lua 5.2: debug.getlocal() accepts function arg, too.
new 1afcea1 From Lua 5.2: 'break' allowed anywhere.
new 16d8c83 From Lua 5.2: debug.setmetatable() returns object.
new 2d3c196 Disable LUA_COMPAT_GFIND and LUA_COMPAT_MOD in Lua 5.2 mode.
new 965694b Fix 'f' and 'L' options for debug.getinfo() and lua_getinfo().
new a9baead From Lua 5.2: debug.getinfo(..., "u") returns nparams and isvararg.
new 8352335 From Lua 5.2: Add debug.upvalueid() and debug.upvaluejoin().
new 40afe71 Fix last commit.
new 98f0580 Fix package.searchpath().
new 3dceaa9 Move load/dump functions to lj_load.c. Add load modes.
new 2ba1686 From Lua 5.2: Add mode and env arguments to load*().
new 2ff4afe Allow conditional inclusion of builtin library functions.
new d9c5eba Remove some library functions for no-JIT/no-FFI builds.
new 4c7c2c8 Rearrange library functions to get a fixed FF_next.
new ca0bb48 From Lua 5.2: Add rawlen(). Needs -DLUAJIT_ENABLE_LUA52COMPAT.
new 125cc87 From Lua 5.2: Add debug.getuservalue() and debug.setuservalue().
new cfca926 From Lua 5.2: Extended results from os.execute() and pipe:close().
new 0ecdff4 x86: Fix register allocation for calls returning register pair.
new d4d9807 Fold KPTR + offset in SPLIT pass.
new e013206 FFI: ctypeid of ctype object is immutable.
new 2684129 From Lua 5.2: Add table.pack(). Needs -DLUAJIT_ENABLE_LUA52COMPAT.
new 7a6bb9a From Lua 5.2: Return file object for io.write() and file:write().
new 550ac44 From Lua 5.2: Remove error for ambiguous function call syntax.
new d08e7bd From Lua 5.2: Allow mixed metamethods for ordered comparisons.
new 3ad6168 Update dependencies.
new fcddd5a From Lua 5.2: Add luaL_traceback().
new b66ab96 Move a GC macro.
new 0d7094f From Lua 5.2: Return nil for bad position in string.find().
new faf05c3 From Lua 5.2: '%s' option to string.format() behaves like tos [...]
new 8d8dbf2 Fix x64 build.
new 2e9f4ef Fix scope for resolving break labels.
new 0d62e2e Clarify comments in Makefile on -DLUAJIT_ENABLE_LUA52COMPAT.
new 0561a56 From Lua 5.2: Add math.log(x, base).
new 2f5ed5d From Lua 5.2: Add string.rep(s, n, sep).
new c681e00 PPC: Fix string.sub() range check.
new da682b0 FFI: Add support for copy constructors.
new eabfdfe Don't use stack unwinding for lua_yield().
new 0452e97 Add FOLD rules for CONV.int.i8/u8/i16/u16 of KINT.
new 5ebe499 Add LJ_TARGET_UNALIGNED.
new a7d1dba FFI: Optimize ffi.copy() and ffi.fill().
new 4a97fae FFI: Compile array/struct copies.
new 1c62611 FFI: Compile ffi.sizeof(), ffi.alignof() and ffi.offsetof().
new 0b55e05 FFI: Compile ffi.gc().
new 425f67c FFI: Update docs on FFI semantics.
new fd292c9 Fix assertion.
new fff1297 From Lua 5.2: Add format options to io.lines().
new b50375a From Lua 5.2: Add package.loadlib(libname, "*").
new a0ee70d Documentation layout changes.
new d70d775 Documentation cleanup and reorganization.
new c25d4da Add note to recalcitrant distro maintainers.
new 6d36234 Update Android install docs.
new 93e4040 Workaround for broken Android sprintf("%g", -0.0).
new 1447ee6 ARM, MIPS: Fix workaround for argument GPRs vs. FPR remat.
new 894d2d6 DynASM/ARM: Fix conditional VFP instruction encoding.
new 2621617 ARM: Drop hard-fp variants of floor/ceil/trunc.
new a58b86d PPC: Compile math.sqrt() to fsqrt instruction.
new bbe35ad MIPS: Compile math.sqrt() to sqrt.d instruction.
new a0f83f6 Update changelog.
new 4baa01b RELEASE LuaJIT-2.0.0-beta11
new e32cc2a Fix Android/x86 build.
new f1fe96e Extended Android build instructions.
new 614e78d Fix recording of equality comparisons with __eq metamethods.
new f6536c6 Fix detection of immutable upvalues.
new 1d5c2ce Replace error with PANIC for callbacks from JIT-compiled code.
new 1fef908 Release planning update.
new f874452 Remove strict.lua.
new c799006 Fix builtin string to number conversion for INT_MIN.
new 4e7fe81 Document tonumber() enhancements.
new b0dde5b FFI: Fix code generation for replay of sunk float fields.
new b3afb0a Install bin/luajit symlink for non-beta releases.
new 1f372e1 Update changelog.
new 4b9441c Update docs for release candidate.
new eb6f890 RELEASE LuaJIT-2.0.0-rc1
new 69906d1 Don't create unneeded array part for template tables.
new ae1987d Fix CONV.num.int sinking.
new c022692 Don't propagate implicitly widened number to index metamethods.
new 8a1a609 FFI: Fix signedness of bool.
new 4e5ef81 FFI: Fix recording of bool call result check on x86/x64.
new e941caa RELEASE LuaJIT-2.0.0-rc2
new 8a11869 FFI: Fix stack-adjustment for __thiscall callbacks.
new 0613358 ARM: Fix ordered comparisons for number <=> non-number.
new 07406a5 FFI: Mention __thiscall attribute in docs.
new 2156925 Improve iOS install docs.
new 74feb25 FFI: Clarify callback docs.
new 87d74a8 RELEASE LuaJIT-2.0.0-rc3
new 1cd13f6 x64: Don't fuse implicitly 32-to-64 extended operands.
new caed566 Mention -Osink in docs.
new 32fa68d Fix assertion.
new c77a9e4 Don't clear frame for out-of-memory error.
new 2689f32 Prevent erroneous x64 builds with non-native exception handling.
new 16f2345 FFI: Change priority of table initializer variants for structs.
new 2ad9834 Add missing GC steps for template table creation.
new 405020c FFI: Add debug define to check for unanchored CTypes.
new ee748dd ARM: Fix armhf call argument handling.
new 41738e3 ARM: Fix code generation for integer math.min/math.max.
new e422ae2 Leave hook when resume catches error thrown from hook.
new 9827650 ARM: Handle armv7s arch defines.
new 8b97dc7 FFI: Fix code generation for bool call result check on x86/x64.
new fb10145 Allow using MSVC intrinsics with older versions.
new f22d5d9 Fix discharge order of comparisons in Lua parser.
new e434ab5 Fix PS3 build.
new f0a0a23 Remove fictitious Solaris define.
new 8a3be5d Remove redundant code in jit/*.lua.
new ad36fda PPC/e500: Remove assembler code for lj_vm_foldarith().
new fec750c Another fix for PS3 build.
new 2e8e587 x86/x64: Always mark ref for shift count as non-weak.
new 2c293a9 FFI: Load FFI library on-demand for bytecode with cdata literals.
new 89e4650 Add XBox 360 port.
new 1651684 Add missing FOLD rules for U32 conversions.
new a3db8f3 Avoid leaking memory on kernels with recalcitrant mmap() behavior.
new f772bed Fix XBox 360 port.
new 4eb4b8a Keep line number for KPRI to KNIL optimization.
new fe9934f FFI: Fix handling of qualified transparent structs/unions.
new 8b46013 FFI: Stricter parsing of declaration specifiers.
new 4a44c4f Bump copyright date to 2013.
new 250b24f Improve buffer handling for io.read().
new 339b77e OSX: Add support for Mach-O object files to -b option.
new 4e35c02 PPC/e500: Fix lj_vm_floor() for inf/nan.
new c7f0b6b Fix capitalization of 'Xbox'.
new e822bdb Update changelog.
new 92699e9 RELEASE LuaJIT-2.0.1
new e7633db RELEASE LuaJIT-2.0.1 (fixed)
new d4bc6ab Fix memory access check for fast string interning.
new 61fb587 BRANCH LuaJIT-2.1
new 57768cd x86: Remove x87 support from interpreter.
new c3219b7 x86/x64: Drop xmm register/memory move tuning for K8.
new e20157c Add support for embedding LuaJIT bytecode for builtins.
new 19d5651 Replace math.deg/math.rad with builtin Lua function.
new fdc0ce8 Don't modify generated libbc header if unchanged.
new b359ce8 Remove obsolete non-truncating number to integer conversions.
new 73ef845 Add special bytecodes for builtins.
new 60e380f Replace table.getn/foreach/foreachi with bytecode builtins.
new 9b7d2e9 Properly fix MSVC intrinsics for older versions.
new 41f3ae2 Merge branch 'master' into v2.1
new b8abb4b Fix genlibbc.
new 5e60189 Replace table.remove with bytecode builtin.
new cea899b Allow changing *.dll/*.lib name in msvcbuild.bat.
new d44337a Merge branch 'master' into v2.1
new 28cfcf7 String buffer refactoring, part 1.
new 116cdd7 String buffer refactoring, part 2.
new 87c51e7 Cleanup lexer source code.
new 3c0157f String buffer refactoring, part 3.
new 9ec869b String buffer refactoring, part 4.
new 3e286c1 ARM: Fix cache flush/sync for exit stubs of JIT-compiled code.
new 78c97bc MIPS: Fix cache flush/sync for JIT-compiled code jump area.
new b65196b Merge branch 'master' into v2.1
new dac2825 Add missing GC steps for io.* functions.
new b2edd5c Merge branch 'master' into v2.1
new 7950894 ARM: Rearrange global_State to satisfy imm12 encoding constraints.
new eea48c8 Fix rotate definitions for variable shifts.
new 3e8f5ac Merge branch 'master' into v2.1
new a98aede FFI: Add 64 bit bitwise operations.
new d147eed DynASM: Improve validation of local backwards relocations.
new 18d7c97 Merge branch 'master' into v2.1
new d1645c8 Clean up TValue to buffer conversions.
new deb61e0 Use string buffer for table.concat().
new d38d10a Use string buffer for os.date().
new fecde1b Use string buffer for string.dump().
new f1dbd65 Use string buffer for string.format().
new 172bd95 FFI: Fix snapshot substitution in SPLIT pass.
new 3b0b3af Merge branch 'master' into v2.1
new 59329cf Optimize and refactor printing of numbers to buffers.
new df94b81 Tune string concatenation.
new 6ba65ce FFI/x86: Fix register allocation for 64 bit comparisons.
new 389822d Fix spurious red zone overflows in machine code generation.
new ac14d88 Merge branch 'master' into v2.1
new b5bbacd Fix jump-range constrained mcode allocation.
new 48912a2 Merge branch 'master' into v2.1
new f17b038 Inhibit DSE for implicit loads via calls.
new b1327bc Document support for Lua 5.2 extension of loadfile().
new e92e29d Merge branch 'master' into v2.1
new 9ead735 Add generic load/store with offset to assembler backends.
new 7b629b7 Move buffer shrinking to atomic GC phase.
new 5f1781a Compile string concatenations (BC_CAT).
new 988e183 Reorganize generic operations common to all assembler backends.
new 2cd4ce6 Use same HREF+EQ/NE optimization in all assembler backends.
new 611ec41 Fix compiler warning.
new 2ab5e7c Merge branch 'master' into v2.1
new a2c7881 Combine IR instruction dispatch for all assembler backends.
new 1ae3e52 Fix PHI barrier in BUFPUT FOLD rule.
new 2cdf90f Always emit TOSTR for non-string CAT operands. Fuse in backend.
new 557391c Fuse string creation into concats. Optimize single-char concats.
new 255326a Use explicit conversion type for IR_TOSTR. Add char conversion.
new 39e53e8 Compile string.char().
new 7c28448 Fix os.date() broken by d38d10a3.
new eeb204c Fix FOLD rules for appending to string buffers.
new 61cb25b Refactor string.reverse(), string.lower(), string.upper().
new b6adfff Compile string.reverse(), string.lower(), string.upper().
new 895a8bc Tighten requirements for non-default optimization flags.
new bdaf14b Merge branch 'master' into v2.1
new 5fd659f Fix MSVC build.
new 47fa9a8 Change semantics of buffer ops to simplify CSE and DCE.
new f9421f2 Replace string.len with bytecode builtin.
new a98e6a7 Abstract out some common buffer operations.
new 9b8db40 Refactor string.rep().
new 068783b Compile string.rep().
new 723574d Optimize BUFHDR code generation.
new 64d2883 Refactor table.concat().
new c776808 Compile table.concat().
new 8b3a320 Fix BC_ISTYPE for WIN64 calling conventions.
new 4ac25a9 Compile __concat metamethod.
new 87993a6 Inhibit CSE for table.concat().
new 09aeca6 Fix builtin string to number conversion for overflow digits.
new 633d91c Merge branch 'master' into v2.1
new 5d6db41 FFI: Fix tailcall in lowest frame to C function with bool result.
new 2be1c26 Merge branch 'master' into v2.1
new 43de451 Partially refactor string.find().
new 87b560b Compile string.find() for fixed string searches (no patterns).
new c5d7666 Fix optional argument handling while recording builtins.
new b030788 Fix optional argument handling in table.concat().
new 71ce1e8 Merge branch 'master' into v2.1
new bb2cc1d Merge branch 'master' into v2.1
new 5bb1f0e Refactor string.format().
new 7d5acc2 Refactor bit.tohex().
new 19a9206 Refactor internal string formatting.
new 625ffca Refactor raw object to pointer or string conversions.
new 8f90a12 Big renaming of string buffer/formatting/conversion functions.
new 23aa02f FFI: Ignore CTF_LONG in ffi.istype().
new cc1e342 Merge branch 'master' into v2.1
new 101780c Fix compiler warnings.
new 2369f5d Abstract out default BUFHDR emitter.
new 2ef2c34 Adapt -jdump to unified string.format("%x", -1) behavior.
new 23e47f2 Fix bit.tohex() broken by 7d5acc29.
new acda75a Refactor CCallInfo representation for split arguments.
new 09ef8d1 FFI: Load FFI library on-demand for -jdump.
new 4c39e9f Compile string.format().
new a860177 Another fix for bit.tohex().
new fd616ca Compile bit.tohex().
new 1a5fd52 Add partial support for building with MingW64 GCC 4.8-SEH.
new 5dabdb2 FFI: Fix calling conventions for 32 bit OSX and iOS simulator.
new 2b94827 Merge branch 'master' into v2.1
new 1c7650f FFI: Fix calling conventions for ARM hard-float EABI.
new a946f67 Merge branch 'master' into v2.1
new 0f79d47 Handle calls with max. args in backends even after SPLIT.
new 647cc46 Merge branch 'master' into v2.1
new fe87736 Get rid of module() in all internal modules.
new e2db3f6 Remove use of printf() in luajit.c.
new 1593e77 FFI: Improve error messages for arithmetic and comparison operators.
new 82390d1 Merge branch 'master' into v2.1
new 6f451c6 Don't try to reinvent memcpy().
new 38cfb6a Refactor lj_cdata_newv().
new 890754e FFI: Insert no-op type conversion for pointer to integer cast.
new 55772e3 Merge branch 'master' into v2.1
new 7ef2b55 FFI: Fix unroll limit for ffi.fill().
new 2c3e80d FFI: Must sink XBAR together with XSTOREs.
new b3369f3 Merge branch 'master' into v2.1
new f1f7e40 FFI: Compile VLA/VLS and large cdata allocs with default init [...]
new d686217 Add missing PHI barrier to string.sub(str, a, b) == kstr FOLD rule.
new 4c1f9dd Merge branch 'master' into v2.1
new 992f7d4 FFI: Refactor ffi.gc()/__gc. Compile ffi.gc(cdata, nil), too.
new 5a261dd Fix compatibility issues with Illumos.
new 26e4287 Merge branch 'master' into v2.1
new 6850d79 Minor change to lj_lib_pushcc().
new b6882a5 Fix -jdump broken by 2ef2c34d.
new c5ebfa0 Refactor string buffer growth helpers.
new 3eeb7d3 FFI: Preserve intermediate string for const char * conversion.
new 45f3ee8 Merge branch 'master' into v2.1
new 338e986 PPC: Add plt suffix for external calls from assembler code.
new e92d64a Merge branch 'master' into v2.1
new 742ba5c Update changelog.
new 21af151 RELEASE LuaJIT-2.0.2
new 1fa97be Merge branch 'master' into v2.1
new 9211a66 Improve ULOAD forwarding for open upvalues.
new edc7f35 Merge branch 'master' into v2.1
new 053041a FFI: Add missing GC steps for callback argument conversions.
new 88188d5 Merge branch 'master' into v2.1
new b2a13dd DynASM/x86: Add lock prefix.
new 519ef0c Add support for multilib distro builds.
new 146807b Merge branch 'master' into v2.1
new 9aa5d04 Fix compilation on NetBSD/x64.
new dd44018 Merge branch 'master' into v2.1
new 1ddf568 Fix compiler warning.
new ec96d8b DynASM/x64: Add VREG support to mov64.
new c86f47a Merge branch 'master' into v2.1
new 6dc0245 FFI: Properly unload loaded DLLs.
new 96d5478 Merge branch 'master' into v2.1
new 6a878b0 ARM: Fix register allocation when rematerializing FPRs.
new 92a3cf4 Merge branch 'master' into v2.1
new 99d588b Compatibility fix for old GCC versions.
new f5fba48 Merge branch 'master' into v2.1
new 362260b Fix jit.flush(func|true).
new 9b14426 Merge branch 'master' into v2.1
new fd1a8dd Fix argument checks for math.deg()/math.rad().
new 5451089 Merge branch 'master' into v2.1
new a6936be Fix compiler warning.
new b33e96f Merge branch 'master' into v2.1
new 5120240 Use g->jit_base for on/off-trace detection.
new 517500b Save currently executing lua_State in g->cur_L.
new 19c2dd1 Add common helper for pointer to string formatting.
new 4dce22c Add low-overhead profiler. Part 1: interpreter, low-level C API.
new 8a9519a Add internal API to add sub-modules to package.preload.
new b186fb8 Add low-overhead profiler. Part 2: low-level Lua API.
new eca0914 Add low-overhead profiler. Part 3: command line interface.
new a4fa4dc Install new jit.* modules, too.
new fffe75a Fix frame traversal for backtraces.
new 7873c8d Recent GCC versions need more memory.
new b6ec7c7 Fix quoting in xedkbuild.bat.
new f8cecff Merge branch 'master' into v2.1
new f8e3653 Fix string.format("%#X", 1).
new d3d30d3 Add notice when no samples were collected by -jp.
new d1194a8 Low-overhead profiler, part 4: JIT compiler support.
new a13b920 Load jit.util module on demand (saves 1 KB).
new 93c2b93 FFI: Fix argument checks for ffi.string().
new 0d41601 Merge branch 'master' into v2.1
new 782787f FFI/x64: Fix passing of vector arguments to calls.
new cb33699 Merge branch 'master' into v2.1
new 1fd2048 Switch from -jp=n<top_n> to -jp=m<min_pct> as a threshold.
new 2befb8b Low-overhead profiler, part 5: add annotation mode for -jp.
new 2907851 Add -jp=fl mode and minor fixes.
new 31e7cd1 Low-overhead profiler, part 6: documentation.
new 7dc6750 Fix PS3 build.
new 803d4dd Fix non-FFI build.
new 483f823 Low-overhead profiler, part 7: console ports.
new 1f2eff7 Fix -jp=a mode.
new 0953b9a Fix profiler stack dump for string chunks.
new 1f7b326 DynASM: Emit #line <n> <file> instead of #<n> <file>.
new 5f24b1d Merge branch 'master' into v2.1
new f1e7a44 Fix snapshot restore for exit to function header.
new da14565 Merge branch 'master' into v2.1
new 474bb21 Fix ABC elimination.
new 8201fda Merge branch 'master' into v2.1
new b30a789 Protect g->hookmask with lock when using profiler thread.
new a51b8bf Fix build with profiler but without JIT compiler.
new eec06a0 Fix profiler/hook interaction.
new 54af4e7 Fix loading of embedded bytecode.
new 47df3ae Properly fix loading of embedded bytecode.
new c8cfca0 Add table.new().
new 88c7785 FFI: Compile lightuserdata to void * conversion.
new e6e760c Compile getfenv(0).
new 1837d07 Compile BC_TSETM, e.g. {1,2,3,f()}.
new 3cdeb54 Check for MAP_32BIT presence instead of checking for Linux.
new 4ccb2dd Merge branch 'master' into v2.1
new 5d25645 FFI: Rehash finalizer table after GC cycle, if needed.
new d0b48ec Merge branch 'master' into v2.1
new 4fba08a Fix line number for relocated bytecode after closure fixup.
new 5706153 Merge branch 'master' into v2.1
new a014d5e FFI: Compile conversions from functions to function pointers.
new 7e538b5 Check for failure to mark memory as executable (restricted kernels).
new d8f0535 Merge branch 'master' into v2.1
new 8944e27 Fix handling of redundant PHIs.
new 0315b87 Merge branch 'master' into v2.1
new 7b4d530 Fix GC steps threshold handling when called by JIT-compiled code.
new 5c55b76 Merge branch 'master' into v2.1
new 8941b19 Fix for last commit
new 8a2b2be Merge branch 'master' into v2.1
new 5a25a73 Fix type punning alias analysis for constified pointers.
new ffabcc4 Merge branch 'master' into v2.1
new db64b2b FFI: Clarify ffi.cdef behavior.
new 9f77a31 Merge branch 'master' into v2.1
new 751eace Improve msvcbuild.bat static linking.
new 94bfa7c Merge branch 'master' into v2.1
new 9d90988 Follow-up fix for GC step threshold fix.
new b00a026 Merge branch 'master' into v2.1
new f371ed6 Respect jit.off(func) when returning to a function, too.
new a8c3862 Merge branch 'master' into v2.1
new c378f7d Abstract out post-registration handling of pre-registered modules.
new 4593fb5 Add table.clear().
new 916f09d x64: Fix store to upvalue for lightuserdata values.
new 5c7db3c Merge branch 'master' into v2.1
new 0ded8e8 Fix call unroll checks in the presence of metamethod frames.
new 585dd44 Merge branch 'master' into v2.1
new 279673d Fix MinGW build.
new b03cebe Merge branch 'master' into v2.1
new 89e8d70 Fix compilation of string.byte(s, nil, n).
new e52119b Merge branch 'master' into v2.1
new ca67c89 MIPS: Cosmetic fix for interpreter.
new 6e02c21 Merge branch 'master' into v2.1
new b5d741f Add trace stitching.
new c41e654 Show trailing NYI item for stitched trace with -jv.
new c1328e9 x86: Fix stack slot counting for IR_CALLA (affects table.new).
new dce3639 Fix final snapshot for trace stitching.
new b779f31 Fix MinGW build.
new 87b33b9 Merge branch 'master' into v2.1
new 6964a77 Fix ABC elimination.
new ba9dde2 Merge branch 'master' into v2.1
new ce9e8ed Fix KBASE for Lua functions below stitched fast functions.
new ef59e54 Bump copyright date to 2014.
new a9d4543 Merge branch 'master' into v2.1
new 45c067d FFI: Fix cts->L for cdata unsinking in snapshot restore.
new 926686a Fix OSX build issue.
new 4e78b9b Merge branch 'master' into v2.1
new 3db3cc1 Fix compiler warnings.
new dd910f0 Fix initial maxslot for down-recursive traces.
new 9247541 Fix OSX build issue.
new bcd2703 Merge branch 'master' into v2.1
new 5a5a340 Don't start stitched trace with CALLMT, too.
new a2abf2e Don't start a trace on side exit to non-Lua function.
new b20ff7a Use own buffer for LOOP substitution table.
new 2be62dc Fix exception unwind struct declaration.
new ec5546d Merge branch 'master' into v2.1
new e3d5b21 Fix quoting in xedkbuild.bat (again).
new 2bc63bb Prevent BASE register coalescing if parent uses IR_RETF.
new 1994a24 FFI: Disallow nested VLS.
new 998651f Don't purge modified function from stack slots in BC_RET.
new 46ce034 Fix Xbox 360 build.
new 2863b10 Merge branch 'master' into v2.1
new 6a47e12 Add PS4 port.
new 9695c5a Merge branch 'master' into v2.1
new 900d0ce Fix for -jdump docs.
new b342b7a Merge branch 'master' into v2.1
new ef3cc47 Fix recording of BC_VARG.
new e941508 Fix frame depth display for bytecode dump in -jdump.
new fd3354a Merge branch 'master' into v2.1
new 19d53ed Specialize to ffid for non-monomorphic builtins.
new 416abff Don't access dangling reference to reallocated IR.
new 0e9bf2f Merge branch 'master' into v2.1
new 80efb07 DynASM: Emit version check after .arch directive.
new 03bf26e Merge branch 'master' into v2.1
new 30c5757 Update changelog.
new 880ca30 RELEASE LuaJIT-2.0.3
new 14e0f5a Merge branch 'master' into v2.1
new 1ca794e Fix compiler warnings.
new ce477ff Merge branch 'master' into v2.1
new c03be12 x86/x64: Avoid use of SAHF instruction.
new 3500c35 Merge branch 'master' into v2.1
new 095344c Fix check for reuse of SCEV results in FORL.
new 4401af8 Merge branch 'master' into v2.1
new c735293 FFI: Fix pseudo type conversions for type punning.
new b65515f Merge branch 'master' into v2.1
new 19b69f2 FFI: Allow non-scalar cdata to be compared for equality by address.
new caefd06 Merge branch 'master' into v2.1
new 423f10f Fix alias analysis for table.len vs. table.clear.
new b293063 Correct fix for reuse of SCEV results in FORL.
new a972d37 Merge branch 'master' into v2.1
new 111b2e8 Fix top slot calculation for snapshots with continuations.
new d3e3700 Merge branch 'master' into v2.1
new 4ed3ee1 FFI: Fix cdata equality comparison against strings.
new 347b49e Merge branch 'master' into v2.1
new cc5075e FFI: Fix cdata equality comparison against other Lua types.
new 27c453e Merge branch 'master' into v2.1
new 43dc06b Fix FOLD of string concatenations.
new 65626ea PPC: Don't use mcrxr on PPE.
new 39acdb8 Merge branch 'master' into v2.1
new f3374b5 x64: Allow building with LUAJIT_USE_SYSMALLOC and LUAJIT_USE_ [...]
new 18309b0 Prevent adding side traces for stack checks.
new 2715fe3 Prevent GC estimate miscalculation due to buffer growth.
new c8d1aff Merge branch 'master' into v2.1
new 7a39be0 Add PS Vita port.
new e26dc8a Merge branch 'master' into v2.1
new ef82bb0 FFI: Another fix for cdata equality comparisons.
new f7e209c Merge branch 'master' into v2.1
new 2917288 x64: Workaround for MSVC build issue.
new 392b6c9 Merge branch 'master' into v2.1
new 624260f Fix frame traversal while searching for error function.
new 159cfd6 Merge branch 'master' into v2.1
new 8a38af2 Fix FOLD rule for STRREF of SNEW.
new fe288e7 FFI: Fix recording of indexing a struct pointer ctype object itself.
new 52b0815 Merge branch 'master' into v2.1
new e157651 Fix FOLD rule for string.sub(s, ...) == k.
new 945af49 Merge branch 'master' into v2.1
new fc76d54 x86: Fix code generation for unused result of math.random().
new 49d3157 PPC: Fix red zone overflow in machine code generation.
new 7400e2c Merge branch 'master' into v2.1
new 7fb503d Fix compatibility issues with Illumos.
new 9597759 Merge branch 'master' into v2.1
new e7c6a1a Fix debug info for main chunk of stripped bytecode.
new afcc3c0 Merge branch 'master' into v2.1
new 7dc9e5a x86: lj_math_random_step() clobbers XMM regs on OSX Clang.
new 9d67be8 Merge branch 'master' into v2.1
new 5e7b9bb FFI: Fix frame traversal for backtraces with FFI callbacks.
new 89f61de Merge branch 'master' into v2.1
new bb5dfbb Typo.
new ccd456a Merge branch 'master' into v2.1
new 1cb5929 Compile debug.getmetatable().
new a6c34b8 Fix ABC elimination.
new 0ac4c43 Merge branch 'master' into v2.1
new f948fe0 FFI: Fix compilation of reference field access.
new 5f370d9 Merge branch 'master' into v2.1
new 59d2906 Invalidate backpropagation cache after DCE.
new 92c32b7 Merge branch 'master' into v2.1
new 66515a0 FFI: Fix __index/__newindex metamethod resolution for ctypes.
new e107525 Merge branch 'master' into v2.1
new 247ead6 FFI: Fix for cdata vs. non-cdata arithmetic and comparisons.
new 41156fe FFI: Fix initialization of unions of subtypes.
new 72c661e Merge branch 'master' into v2.1
new b7d29b6 Eliminate hmask guard for forwarded HREFK.
new ae7d4d5 DynASM/x86: Add shld/shrd opcodes.
new 357defa Merge branch 'master' into v2.1
new cb886b5 DynASM/x86: Add fldenv, f[n]stenv, fxsave, fxrstor opcodes.
new 0de431e Merge branch 'master' into v2.1
new 4846a71 Fix DragonFly build (unsupported).
new be64823 Merge branch 'master' into v2.1
new 6d0654d Fix fused constant loads under high register pressure.
new d2809be Merge branch 'master' into v2.1
new d6ff3af FFI: Add ffi.typeinfo().
new a31c317 FFI: No meta fallback when indexing pointer to incomplete struct.
new 42ed4e5 Merge branch 'master' into v2.1
new c75c62b Fix snapshot #0 handling for traces with a stack check on entry.
new 8cc8933 Merge branch 'master' into v2.1
new 3f2e4ec DynASM/ARM: Fix description shown for multi-element templates.
new 1fc9cd0 Merge branch 'master' into v2.1
new f49c61a DynASM/ARM64: Initial commit of ARM64 module.
new e03df1e x86/x64: Call external symbols directly from interpreter code.
new ad03eba x86/x64: Drop internal x87 math functions. Use libm functions.
new 6fddb9b Avoid error messages when building with Clang.
new cca81a3 Merge branch 'master' into v2.1
new c6e6924 Gracefully handle lua_error() for a suspended coroutine.
new b93b624 Merge branch 'master' into v2.1
new 881f48f Cleanup of frame handling. No functional changes.
new c6130e6 Avoid dependency on frame layout for LJ_POST_FIXCOMP.
new 9679a94 Disable debug info for generated *.S to avoid confusing gdb.
new 82e6e5f x86: Fix stack slot reservation for FP math functions.
new 6e9145a Cleanup of memory vs. GC sizes. No functional changes.
new 10caad0 x86: Fix argument checks for ipairs() iterator.
new 5cb6e2e Cleanup of TValue setters. No functional changes.
new ecaa4ea Merge branch 'master' into v2.1
new 1b774d9 Fix corner case in string to number conversion.
new 5482656 ARM: Fix excess stack growth in interpreter.
new 16f910b ARM: Fix write barrier check in BC_USETS.
new 6319e03 PPC: Fix excess stack growth in interpreter.
new f45d11e MIPS: Fix excess stack growth in interpreter.
new db7cb5a DynASM/ARM: Fix rollback for variant templates.
new 7a76d43 Merge branch 'master' into v2.1
new a13dfd6 DynASM/ARM64: Various fixes.
new 054e6ab Add LJ_FR2 mode: Two-slot frame info.
new cb481dd Add LJ_GC64 mode: 64 bit GC object references.
new f307d0a ARM64: Add build infrastructure and initial port of interpreter.
new 9caaf65 Fix docs.
new 968725c Merge branch 'master' into v2.1
new dfe84fd PPC/e500: Drop support for this architecture.
new e1261a0 ARM64: Compatibility fixes for Clang.
new 7bdadeb ARM64: Update docs.
new 31e6683 x86: Minor interpreter optimization.
new 4d9e8e2 ARM: Minor interpreter optimization.
new fcae87e Merge branch 'master' into v2.1
new 71ecc05 Fix MSVC build.
new 999f57d Fix BC_UCLO/BC_JMP join optimization in Lua parser.
new c71cdf4 Merge branch 'master' into v2.1
new 86913b9 Bump copyright date to 2015.
new 0a5045c Merge branch 'master' into v2.1
new 1b1bd08 DynASM/ARM64: Fix checks for scaled immediates.
new 9647aab LJ_FR2: Fix lua_settable() and lua_setfield().
new ce1a5ee iOS/ARM64: The frame pointer is required.
new 33f0c24 ARM64: Add FFI support.
new 962d2aa LJ_GC64: Return true for ffi.abi("gc64").
new 3f19b34 iOS/ARM64: Fix target OS detection.
new 3e17038 DynASM/PPC: Add missing PPC64 instructions and various extensions.
new b876d6d OpenBSD/x86: Better executable memory allocation for W^X mode.
new 7f01300 Don't compile IR_RETF after CALLT to ff with-side effects.
new 43cb818 Merge branch 'master' into v2.1
new dfa173b Fix lexer error for chunks without tokens.
new ca5bbdf Merge branch 'master' into v2.1
new 04dc64b LJ_FR2: Fix bytecode generation for method lookups.
new 704280f FFI: Prevent DSE across ffi.string().
new 063f3d5 Merge branch 'master' into v2.1
new 4fa51af x86/x64: Fix code generation for fused test/arith ops.
new 576ddf1 Merge branch 'master' into v2.1
new e6b0875 Fix string to number conversion.
new 906008d Merge branch 'master' into v2.1
new 7ff4768 Properly fail unsupported cross-compile to MIPS64.
new 5c0e33c ARM: Handle more arch defines.
new e1a8f3b Merge branch 'master' into v2.1
new 1f578bd FFI: Fix FOLD rule for TOBIT + CONV num.u32.
new 2c0748f Merge branch 'master' into v2.1
new 8b6af89 Fix Lua/C API typecheck error for special indexes.
new c821e0a Merge branch 'master' into v2.1
new 361827c PPC64: Add build infrastructure.
new a9fd686 Doc cleanup.
new 4e52d1f Merge branch 'master' into v2.1
new 8002871 PPC: Fix write barrier in BC_TSETR.
new dec4acc PPC: Fix cross-endian builds.
new b2a5cc8 DynASM/PPC: Add support for parameterized shifts/masks.
new 1c968d5 DynASM/PPC: Add sub/shift/rotate/clear instruction aliases.
new 1e67a4c ARM64: Fix math.floor/math.ceil for string args.
new 11106aa Fix stack check in narrowing optimization.
new 74caac9 Merge branch 'master' into v2.1
new c2924c3 Disable trace stitching (for now) due to a design mistake.
new 5caf53d Fix broken Windows build.
new a92e730 x64: Add LJ_GC64 mode interpreter.
new 9622d6b x86/x64: Fix argument check for bit shifts.
new 4cec6a5 Merge branch 'master' into v2.1
new a836b7e Cleanup.
new cd4c59f Update changelog.
new 69e5342 RELEASE LuaJIT-2.0.4
new d8cfc37 Merge branch 'master' into v2.1
new b82fc3d Bump table allocations retroactively if they grow later on.
new bca5da8 Fix for last commit.
new 4a146ca LJ_GC64: Fix frame unwinding.
new 60fb3fe Fix table allocation bump optimization.
new 5de95ed x64/LJ_GC64: Fix setmetatable().
new 4da1bb6 x64/LJ_GC64: Fix upvalue write barrier.
new 7234789 x64: Allow building without external unwinder.
new fb19df9 Merge branch 'master' into v2.1
new c5d262f Fix table allocation bump optimization (again).
new 08e100e x64/LJ_GC64: Fix ipairs().
new 7f454ae Fix narrowing of TOBIT.
new 5081e38 Merge branch 'master' into v2.1
new 458a40b Fix table allocation bump optimization (sigh).
new 26b95a9 Add Xbox One port.
new fe56522 Disable table allocation bump optimization (for now).
new 28d8728 Fix Xbox One port.
new 575bfe5 Limit number of arguments given to io.lines() and fp:lines().
new 5874c21 Merge branch 'master' into v2.1
new ba6344d ARM64: Fix __call metamethod handling for tail calls.
new a5b1c4d FFI: Don't propagate qualifiers into subtypes of complex.
new 0dcd2d1 Merge branch 'master' into v2.1
new eb050f9 Parse binary number literals (0bxxx).
new 424940f Fix NYICF error message.
new b86fc2b Merge branch 'master' into v2.1
new 8715ae6 Properly handle OOM in trace_save().
new ca78889 Merge branch 'master' into v2.1
new 9d18567 Remove LuaJIT 1.x change history.
new 718af62 Add note about bytecode compatibility wrt. LJ_GC64/LJ_FR2.
new ba617df Update changelog.
new 0a3cd94 ARM64: Add support for saving bytecode as object files.
new fb77f7d RELEASE LuaJIT-2.1.0-beta1
new ad29314 ARM64: Fix ELF bytecode saving.
new a3a6866 Re-enable trace stitching.
new 55c3b29 Parse Unicode string escape \u{XX...}.
new 076d625 Fix Cygwin support. Note: this is not a supported target.
new 49427df Merge branch 'master' into v2.1
new 52ea1a3 FFI: Add ssize_t declaration.
new ef087aa Fix unsinking check.
new 0b09c97 Merge branch 'master' into v2.1
new 86c21bd Add collectgarbage("isrunning").
new 2e85af8 Flush symbol tables in jit.dump on trace flush.
new 22a9ed8 Merge branch 'master' into v2.1
new 4808e62 Drop marks from replayed instructions when sinking.
new 776c693 Merge branch 'master' into v2.1
new e5f4cb1 FFI: Fix ipairs() recording.
new b20642c Merge branch 'master' into v2.1
new efaf24f Fix pairs() recording.
new 9dc785e DynASM/x86: Fix some SSE instruction templates.
new 7579b16 DynASM/x86: Restrict shld/shrd to operands with same width.
new e54ca42 Merge branch 'master' into v2.1
new 7e22082 DynASM/x86: Add AES-NI opcodes.
new 8a13c9c DynASM/x86: Add AVX and AVX2 opcodes.
new d62dc01 DynASM: Bump version to 1.4.
new 4c08158 Whitespace.
new 6a54411 x86/x64: Disassemble AVX/AVX2 instructions.
new 3639ff4 iOS: Disable os.execute() when building for iOS >= 8.0.
new ffb124e Merge branch 'master' into v2.1
new f416cf6 OSX: Switch to Clang as the default compiler.
new 1393b2f Update build instructions.
new 3ad7734 Merge branch 'master' into v2.1
new 42b9c98 iOS/ARM64: Add build instructions.
new 0029110 Update changelog.
new 7991a66 Extend Valgrind suppressions for non-inlined C code.
new 367cba2 DynASM/x86: Add rdpmc instruction.
new 0aa337a Merge branch 'master' into v2.1
new 4f87367 Document minimum compiler requirements for 2.1 branch.
new 3a0b2a9 x64/LJ_GC64: Fix stack growth in vararg function setup.
new 3d4c9f9 FFI: Fix SPLIT pass for CONV i64.u64.
new 126e55d Merge branch 'master' into v2.1
new 3f5c724 MIPS soft-float, part 1: Add soft-float support to interpreter.
new 0345f36 MIPS soft-float, part 2: Add soft-float FFI support.
new 20f4141 FFI: Properly unsink non-standard cdata allocations.
new a687a60 DynASM/x64: Add full VREG support.
new f61148c ARM: Add external frame unwinding.
new cfae384 DynASM/x86: Add AVX AES instructions.
new 5ce6399 x86: Improve disassembly of AVX shift instructions.
new 52ebe02 x86: Disassemble AES instructions.
new 22e7b00 DynASM/x64: Fix for full VREG support.
new 04d2806 Rollback due to HREFK + load fwd must restore guardemit state.
new 39eddd3 Always merge snapshots without instructions inbetween.
new 89982b7 FFI: Parse #line NN and #NN.
new 2f6b296 Add proper DynASM dependency.
new 60de2f3 MIPS: Switch to dual-number mode. Fix soft-float interpreter.
new bfbcd72 PS4: Switch default build to amalgamated and LJ_GC64 mode.
new 825dcdc Generalize LJ_SOFTFP dependencies in lj_asm.c.
new f547a14 MIPS: Add soft-float support to JIT compiler backend.
new a443889 Don't allocate unused 2nd result register in JIT compiler backend.
new 6cb38f7 Merge branch 'master' into v2.1
new 339a1fd Windows/x64/LJ_GC64: Fix math.frexp() and math.modf().
new 18f6aa9 Use internal implementation for converting FP numbers to strings.
new 1c45c6a Fix build with JIT disabled.
new 6f3eed9 Fix Android/x86 build.
new db1b399 Bump copyright date to 2016.
new f423194 Merge branch 'master' into v2.1
new 8290076 Update changelog.
new 3e4a196 RELEASE LuaJIT-2.1.0-beta2
new 75d046d DynASM/x86: Add ADX instructions.
new 64c6da6 MIPS soft-float: Fix code generation for HREF.
new be25592 ARM: Fix build problem with external frame unwinding.
new ddadbe8 Fix compiler warnings.
new cf3e01e Merge branch 'master' into v2.1
new 7e05355 Fix install for cross-builds.
new 713e340 Merge branch 'master' into v2.1
new c730540 Fix formatting of some small denormals at low precision.
new e23fc10 Fix display of NULL (light)userdata in -jdump.
new df7bb5b Merge branch 'master' into v2.1
new e03e597 Fix compiler warnings.
new 9531eb2 Windows: Remove intermediate files at end of build.
new d714561 Merge branch 'master' into v2.1
new c24c8e5 x64/LJ_GC64: Fix JIT glue code in interpreter.
new 6801e71 x86: Detect BMI2 instruction support.
new 892887e x86: Generate BMI2 shifts and rotates, if available.
new 62af101 MIPS: Fix use of ffgccheck delay slots in interpreter.
new 6e623b9 Merge branch 'master' into v2.1
new 296f0ca Windows/x64/LJ_GC64: Fix math.frexp() and math.modf() (again).
new 4ab6367 Cygwin: Allow cross-builds to non-Cygwin targets.
new 25b3779 Merge branch 'master' into v2.1
new 1c6fd13 Fix recording of select(n, ...) with off-trace varargs
new d150fbf Merge branch 'master' into v2.1
new ac9193c x86: Improve disassembly of BMI2 instructions.
new 096a7cf x64/LJ_GC64: Fix BC_UCLO check for fast-path.
new e5b5e07 MIPS: Fix BC_ISNEXT fallback path.
new 101115d Merge branch 'master' into v2.1
new 0c6fdc1 Rewrite memory block allocator.
new 73680a5 x86/x64: Search for exit jumps with instruction length decoder.
new d13d420 Merge branch 'master' into v2.1
new ea9a962 stub out the stuff to enable the JIT for ARM64
new c525b86 WIP
new 677781f WIP
The 1983 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
This is an automated email from the git hooks/post-receive script.
unknown user pushed a change to branch master
in repository clang.
from 231ca7b [Clang][AVX512][Builtin] Adding intrinsics for vcvttpd2udq in [...]
new e9c6482 [Clang][AVX512][Builtin] Adding intrinsics for vcvt{ph|ps}2{p [...]
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "adds" were already present in the repository and have only
been added to this reference.
Summary of changes:
include/clang/Basic/BuiltinsX86.def | 4 +++
lib/Headers/avx512vlintrin.h | 59 +++++++++++++++++++++++++++++++++++++
test/CodeGen/avx512vl-builtins.c | 48 ++++++++++++++++++++++++++++++
3 files changed, 111 insertions(+)
--
To stop receiving notification emails like this one, please contact
the administrator of this repository.