aboutsummaryrefslogtreecommitdiff
path: root/test
AgeCommit message (Collapse)Author
2014-02-28Use range metadata instead of introducing selects.Dan Gohman
When GlobalOpt has determined that a GlobalVariable only ever has two values, it would convert the GlobalVariable to a boolean, and introduce SelectInsts at every load, to choose between the two possible values. These SelectInsts introduce overhead and other unpleasantness. This patch makes GlobalOpt just add range metadata to loads from such GlobalVariables instead. This enables the same main optimization (as seen in test/Transforms/GlobalOpt/integer-bool.ll), without introducing selects. The main downside is that it doesn't get the memory savings of shrinking such GlobalVariables, but this is expected to be negligible.
2014-02-28Update tests for recent changes which result in more parens being used in ↵Dan Gohman
some places.
2014-02-28Move ExpandI64 into lib/Target/JSBackend.Dan Gohman
2014-02-28Implement llvm.flt.rounds and disable the createRewriteLLVMIntrinsicsPass pass.Dan Gohman
2014-02-14Preserve alignment information when promoting integer loads and stores.Dan Gohman
2014-02-14Preserve alignment information when splitting loads and stores.Dan Gohman
2014-02-13Emscripten doesn't need to legalize pointers.Dan Gohman
2014-02-13Don't leave behind unreachable blocks with illegal instructions.Dan Gohman
2014-02-12Revamp ExpandI64.Dan Gohman
By using a reverse-postorder traversal of the basic blocks, we can perform this transform in a single pass. This eliminates the need for tricky coordination between the passes.
2014-02-12Generalize PromoteIntegers to handle arbitrary bit widths.Dan Gohman
2014-02-12Create a test directory for basic JS CodegenDan Gohman
2013-11-15PNaCl: Change exception info format to distinguish catch-all/cleanup clausesMark Seaborn
The initial version of ExceptionInfoWriter.cpp encodes a landingpad's "cleanup" clause the same as "catch i8* null" (a catch-all). But it turns out we do want to distinguish these two: If an uncaught exception occurs, we don't want the runtime to run C++ destructors (cleanups), because this involves unwinding the stack to jump to landingpads, which loses the context of where the uncaught exception was thrown from, which is unhelpful for debugging. (The C++ standard says it's optional whether destructors are run when an uncaught exception occurs.) So, change the encoding as follows: * Use 0 as the ID for "cleanup" clauses. * Add 1 to the IDs of types and "catch" clauses. (Adding 1 to both seems neater than just one of them.) * This means we can use 0 for the terminator of filter lists instead of -1, which seems a little neater. This requires a corresponding change to eh_pnacl.cc in libsupc++. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3696 TEST=NaCl's EH tests with libsupc++ change applied Review URL: https://codereview.chromium.org/69923008
2013-11-15Switch PromoteIntegers pass from clearing upper bits after converted operationsDerek Schuff
that may affect them, to clearing before operations that may be affected This is just a cleanup that doesn't have any significant performance or functionality impact, but I thought it might make fixing bug 3714 a bit simpler. R=mseaborn@chromium.org BUG= https://code.google.com/p/nativeclient/issues/detail?id=3714 Review URL: https://codereview.chromium.org/59533011
2013-10-22Make sure flatten globals doesn't insert uses of globals that previously had ↵JF Bastien
no uses. This was the deeper cause of the issues I fixed in https://codereview.chromium.org/33233002/ and is therefore a better fix. (The problem was that StripDeadPrototypes was not stripping a variable that was actually dead because the bitcast constexpr inserted by FlattenGlobals referred to it.) I reverted most of that CL's changes, though not the comment and test cleanup. R=dschuff@chromium.org TEST= pnacl-clang++ pnacl/git/libcxx/test/input.output/iostream.objects/narrow.stream.objects/clog.pass.cpp -stdlib=libc++ Review URL: https://codereview.chromium.org/34843003
2013-10-21Remove unused globals.JF Bastien
libc++'s iostream values are extern, and never actually used in the headers (unlike libstdc++'s) which means that including iostream and doing something like (void)std::clog used to leave a global external ostream object declaration without a definition, which cause PNaCl's module ABI verifier to fail ('has no initializer' and 'is not a valid external symbol'). R=dschuff@chromium.org BUG= http://code.google.com/p/nativeclient/issues/detail?id=3623 TEST= globalcleanup.ll Review URL: https://codereview.chromium.org/33233002
2013-10-16Add PNaClSjLjEH pass to implement C++ exception handling using ↵Mark Seaborn
setjmp()+longjmp() There are two parts to this: * PNaClSjLjEH.cpp expands out the "invoke", "landingpad" and "resume" instructions, modifying the control flow to use setjmp(). * ExceptionInfoWriter.cpp lowers landingpads' clause lists to data that PNaCl's C++ runtime library will interpret. This part will be reused when we drop the SjLj part and create a stable ABI for zero-cost EH. This pass isn't enabled in PNaClABISimplify yet: I'll do that in a separate change. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3696 TEST=*.ll tests (also tested end-to-end: plumbing for this will follow later) Review URL: https://codereview.chromium.org/24777002
2013-10-11Cherry-pick LLVM 187787 to prevent tail calls on x86-32 when not appropriate.Jan Voung
See: http://llvm.org/viewvc/llvm-project?view=revision&revision=187787 The newer version of newlib tickles this x86-32 bug when building the exception handling tests, which don't strip the "tail" attribute. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3702 Waiting on trybots, but it seems to have fixed the minimal reproducer I have: http://chromegw.corp.google.com/i/tryserver.nacl/builders/nacl-toolchain-linux-pnacl-x86_64/builds/922 http://chromegw.corp.google.com/i/tryserver.nacl/builders/nacl-toolchain-linux-pnacl-x86_32/builds/870 http://chromegw.corp.google.com/i/tryserver.nacl/builders/nacl-toolchain-mac-pnacl-x86_32/builds/875 R=jfb@chromium.org Review URL: https://codereview.chromium.org/26538008
2013-10-11Fix bug in rewriting of library calls to intrinsics + new regression test.Eli Bendersky
The pass gets confused in some cases when library functions get passed to other functions as arguments, because use_iterator returns the call instruction. The existing test (rewrite-longjmp-noncall-uses.ll) did not catch this problem because there a bitcast constexpr was applied to the library function pointer, and it came up as the use instead of the containing call. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3706 R=mseaborn@chromium.org Review URL: https://codereview.chromium.org/26952003
2013-10-11Apply upstream: [mips] Fix a bug in MipsLongBranch::replaceBranchPetar Jovanovic
Cherry-pick r191978 from upstream. Original commit message: Author: Akira Hatanaka <ahatanaka@mips.com> Date: Fri Oct 4 20:51:40 2013 +0000 [mips] Fix a bug in MipsLongBranch::replaceBranch, which was erasing instructions in delay slots along with the original branch instructions This has to be cherrypicked, as it is a bug in backend. It was exposed in a long function inside of llc, which caused llc.nexe to work incorrectly. TBR= mseaborn@chromium.org, dschuff@chromium.org BUG= bug in MIPS backend Review URL: https://codereview.chromium.org/26933005
2013-10-11Apply upstream: [mips] Implement llvm.trap intrinsic.Petar Jovanovic
Cherry-pick r187244 from upstream. Original commit message: Author: Akira Hatanaka <ahatanaka@mips.com> Date: Fri Jul 26 20:58:55 2013 +0000 [mips] Implement llvm.trap intrinsic. Patch by Sasa Stankovic. This has to be cherrypicked, as two tests fail due to missing llvm.trap intrinsic. The tests are: - run_sysbrk_test - run_abi_types_test TBR= mseaborn@chromium.org, dschuff@chromium.org BUG= sysbrk and abi_types tests fail for MIPS Review URL: https://codereview.chromium.org/26953003
2013-10-11Apply upstream: Add missing ATOMIC_CMP_SWAP case.Petar Jovanovic
Cherry-pick r185186 from upstream. Original commit message: Author: Lang Hames <lhames@gmail.com> Date: Fri Jun 28 18:36:42 2013 +0000 Add missing case to switch statement - DAGTypeLegalizer::ExpandIntegerResult should expand ATOMIC_CMP_SWAP nodes the same way that it does for ATOMIC_SWAP. Since ATOMIC_LOADs on some targets (e.g. older ARM variants) get legalized to ATOMIC_CMP_SWAPs, the missing case had been causing i64 atomic loads to crash during isel. This has to be cherry-picked, as we have experienced the same bug described in the original message. Missing case caused MIPS 64 atomics to crash. TBR= mseaborn@chromium.org, dschuff@chromium.org BUG= crash for MIPS atomics Review URL: https://codereview.chromium.org/26958002
2013-10-11Apply upstream: [mips] Trap on integer division by zero.Petar Jovanovic
Cherry-pick r182306 from upstream. Original commit message: Author: Akira Hatanaka <ahatanaka@mips.com> Date: Mon May 20 18:07:43 2013 +0000 [mips] Trap on integer division by zero. By default, a teq instruction is inserted after integer divide. No divide-by-zero checks are performed if option "-mnocheck-zero-division" is used. TBR= mseaborn@chromium.org, dschuff@chromium.org BUG= missing trap for MIPS Review URL: https://codereview.chromium.org/26846007
2013-10-03PNaCl bitcode: Simplify how the writer elides castsMark Seaborn
The PNaCl bitcode writer had some complex logic for deciding when to omit casts in the output. This was left over from when the writer was trying to leave in the casts in some but not all cases (as part of incrementally removing casts). This is no longer needed now that the writer just omits all inttoptrs, all ptrtoints, and all pointer bitcasts. This cleanup also fixes the writer so that it elides an inttoptr of a ptrtoint. This sequence is allowed by the PNaCl ABI verifier, but never occurred in practice because ReplacePtrsWithInts' SimplifyCasts() function converts this sequence to an equivalent bitcast. Before this change, the writer would give this error for an inttoptr-of-ptrtoint: LLVM ERROR: Illegal (PNaCl ABI) pointer cast : %1 = ptrtoint [4 x i8]* @bytes to i32 BUG=https://code.google.com/p/nativeclient/issues/detail?id=3590 TEST=toolchain trybots Review URL: https://codereview.chromium.org/25817002
2013-09-20Support mul binary operator in integer promotion passDerek Schuff
Its handling is the same as add (may overflow, may set upper bits) R=jvoung@chromium.org, mseaborn@chromium.org BUG= https://code.google.com/p/nativeclient/issues/detail?id=3599 Review URL: https://codereview.chromium.org/24244008
2013-09-18Only allow PNaCl version 2 bitcode files.Karl Schimpf
BUG= https://code.google.com/p/nativeclient/issues/detail?id=3590 R=jvoung@chromium.org Review URL: https://codereview.chromium.org/23503071
2013-09-16Work around a gcc 4.6.3 / 4.7 bug.Jim Stichnoth
GCC bug http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58416 causes the bf64-1.c test to fail in the GCC torture test suite. This provides an upstreamable workaround. Inspection of the LLVM code base showed no other instances of the pattern that triggers the gcc bug. This can also be upstreamed as soon as I can get a working x86-32 upstream build working to verify/test against. In the meantime, we can make one pnacl-fyi bot go green again. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3685 R=jfb@chromium.org, jfb@google.com Review URL: https://codereview.chromium.org/23437037
2013-09-09PNaCl bitcode: Fix reader to handle pointer type in is.lock.free intrinsicMark Seaborn
BUG=https://code.google.com/p/nativeclient/issues/detail?id=3671 TEST=test/NaCl/Bitcode/*.ll + NaCl's toolchain_tests with bitcode v2 enabled Review URL: https://codereview.chromium.org/23523041
2013-09-09PNaCl bitcode: Remove TYPE_CODE_POINTER entries from type tableMark Seaborn
There are now no uses of pointer type IDs in PNaCl bitcode, so we can stop outputting pointer types into the type table. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3671 TEST=test/NaCl/Bitcode/*.ll Review URL: https://codereview.chromium.org/23600013
2013-09-09PNaCl bitcode: Change FORWARDTYPEREF to never use pointer typesMark Seaborn
Before, FORWARDTYPEREFs used i8* type rather than i32 if they referenced an "alloca" instruction. Clean this up so that FORWARDTYPEREFs use i32 instead in this case. Note that this means that a forward-referenced "alloca" can be used via an ptrtoint+inttoptr, rather than a bitcast, but that's no problem. This is a step towards removing TYPE_CODE_POINTER from the types table, to simplify the PNaCl bitcode format. Rename NormalizeParamType() to NormalizeScalarType() to reflect that it's used in more cases; make it public. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3671 TEST=test/NaCl/Bitcode/*.ll Review URL: https://codereview.chromium.org/23719016
2013-09-06PNaCl bitcode: Strip pointer types from intrinsic declarations' parametersMark Seaborn
Change the writer to strip pointer types from intrinsics' argument and return types, replacing them with i32. This simplifies the PNaCl bitcode format so that pointer types don't need to be represented here. Change the reader to restore the pointer types so that the intrinsic declarations pass the LLVM and PNaCl verifiers. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3671 TEST=intrinsic tests in call-elide.ll + intrinsic-pointer-args.ll + run small_tests with bitcode v2 enabled Review URL: https://codereview.chromium.org/23793005
2013-09-06Clean up vestigial code for constants in PNaCl bitcode files.Karl Schimpf
* Removes code that is no longer used because global variables are handled separately. * Removes CST_CODE_NULL. * Adds special abbreviations for constant integer 0, replacing most of what was lost when CST_CODE_NULL is removed. * Adds abbreviation for floating point constants. * Removes CST_CODE_AGGREGATE and CST_CODE_DATA, except for reading old PNaCl version 1 bitcode files. * Removes call to ParseConstants() in ParseModule() since the module block no longer contains a constants block (globals are handled separately). * Removes getConstantFwdRef() method, since it is no longer needed. * Remove ConstantPlaceHolder class, since it is no longer needed. * Remove ResolvedConstants and ResolveConstantForwardRefs() from the bitcode reader, since it is no longer needed. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3668 R=mseaborn@chromium.org Review URL: https://codereview.chromium.org/23522024
2013-09-06PNaCl bitcode: Fix some tests to not use unsupported pointer-typed argsMark Seaborn
Remove an unused "i32* %ptr" argument from a couple of test cases. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3671 TEST=call-elide.ll Review URL: https://codereview.chromium.org/23945004
2013-09-05PNaCl bitcode: Change test to use an intrinsic to make it more realisticMark Seaborn
Pointer-typed arguments are only allowed for intrinsics, so change the test of pointer-args to call an intrinsic. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3671 TEST=call-elide.ll Review URL: https://codereview.chromium.org/23735008
2013-09-05PNaCl bitcode: Indirect calls: Store return type instead of function typeMark Seaborn
For indirect call instructions (INST_CALL_INDIRECT), it's not necessary to store the full function type. The argument types are already known from the arguments in the instruction. We only need to store the return type to be able to reconstruct the full function type. Storing only the return type ID will make the bitcode a little more compact. Return type IDs will be frequently-used scalar types, which can be given smaller type IDs than function types, which are less frequently used. This potentially makes the writer simpler: In principle, the writer no longer needs to make a pass across all functions' bodies to determine which function types are used in order to build the type table. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3544 TEST=*.ll tests Review URL: https://codereview.chromium.org/23521005
2013-09-05PNaCl bitcode: Don't output the LABEL type in the type tableMark Seaborn
The type ID for the "label" type is never referenced. The "label" type was only being added to the type table because EnumerateType() treats BasicBlock and Value operands the same. However, the rest of the reader and writer treat BasicBlock operands specially and not like other Values. Change some tests to use wildcards for some type IDs. This is so that I don't have to update all the type ID numbers now that the generated type tables have changed. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3590 TEST=run small_tests with v2 bitcode format enabled Review URL: https://codereview.chromium.org/23530031
2013-09-04Remove ARRAY/VECTOR types from PNaCl bitcode files.Karl Schimpf
The value selector list for switch statements are represented using ARRAY/VECTOR constants, but this information is not put into the bitcode file. This CL changes the value enumerator to not emit these constants. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3649 R=mseaborn@chromium.org Review URL: https://codereview.chromium.org/23653013
2013-09-04PNaCl bitcode reader: Disallow pointer-typed arguments in indirect callsMark Seaborn
Make the bitcode reader stricter, so that it disallows pointer arguments in indirect function calls, which are disallowed by the PNaCl ABI checker. Pointer arguments in function calls are only allowed in intrinsic calls, and calls to intrinsics must always be direct calls, not indirect calls. This involves removing two tests that specifically test for pointer args. This is in preparation for tweaking how indirect calls are represented, so that they store the call's return type rather than the function type. BUG=https://code.google.com/p/nativeclient/issues/detail?id=3544 TEST=*.ll tests Review URL: https://codereview.chromium.org/23660005
2013-09-03Remove generating STRUCT_ANON records in PNaCl bitcode files.Karl Schimpf
Don't generate types for elided cast instructions, since they are never put into the bitcode file. In addition, do not generate a type id for the types of global variables, because they are never needed. Don't allow STRUCT_ANON in bitcode files PNaClVersion==2. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3648 R=jvoung@chromium.org Review URL: https://codereview.chromium.org/23455023
2013-09-03Remove all remaining pointer casts from PNaCl bitcode files.Karl Schimpf
For PNaClVersion==2, removes remaining pointer casts from bitcode files. This includes: * Return value can be casted to a scalar value. * Intrinsic calls may return an inherent pointer, which may need casting to a scalar value. Also modifies tests bitcast-elide.ll and inttoptr-elide.ll by removing tests that assumed there were remaining pointer bitcasts that do not get removed. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3544 R=jvoung@chromium.org Review URL: https://codereview.chromium.org/23524003
2013-08-30Revert some ARM byval localmods since byval+varargs are not in stable pexes.Jan Voung
Localmods came from: https://codereview.chromium.org/10825082/, and earlier. (1) The original change was so that byval parameters always go on the stack. That part was added because the original ARM code was buggy, and did not actually make a copy of the value, modifying the caller's struct (ouch!). (2) Then came a localmod to make all arguments following a byval go on the stack and to make the var-args code aware of that. This is so that arguments stay in the correct order for var-args to pick up. For (1) there has been some work upstream to make it work better. In any case, clang with --target=armv7a-...-gnueabi only used byval in some limited cases -- when the size of the struct is > 64 bytes where the backend will know that part of it could be in regs, and the rest can be memcpy'ed to the stack. For le32, clang will still generate byval without satisfying the same ARM condition (only for structs bigger than 64 bytes), so it could be *very bad* if we didn't have the ABI simpification passes rewrite the byval and try to let the ARM backend do things with byval... TEST=the GCC torture tests: va-arg-4.c, and 20030914-2.c and the example in issue 2746 still pass. BUG=none, cleanup R=dschuff@chromium.org Review URL: https://codereview.chromium.org/23691009
2013-08-30Revert "Remove generating STRUCT_ANON records in PNaCl bitcode files."Karl Schimpf
This reverts commit 2302e5d39e2302962d1a0e45d60e00ed47b9b061. BUG= R=eliben@chromium.org Review URL: https://codereview.chromium.org/23827002
2013-08-30Remove generating STRUCT_ANON records in PNaCl bitcode files.Karl Schimpf
Don't generate types for elided cast instructions, since they are never put into the bitcode file. In addition, do not generate a type id for the types of global variables, because they are never needed. Don't allow and STRUCT records in bitcode files when PNaClVersion==2. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3648 R=mseaborn@chromium.org Review URL: https://codereview.chromium.org/23431008
2013-08-29PNaCl bitcode: Remove handling of named struct typesMark Seaborn
Named struct types should not appear in LLVM IR that passes the PNaCl ABI verifier. Remove the test struct-types.ll because it no longer passes. Handling of TYPE_CODE_STRUCT_ANON must remain for now until this issue is fixed: https://code.google.com/p/nativeclient/issues/detail?id=3648 BUG=https://code.google.com/p/nativeclient/issues/detail?id=3590 TEST=PNaCl toolchain trybots Review URL: https://codereview.chromium.org/23490018
2013-08-28Handle pointer conversions for call instructions.Karl Schimpf
This also should complete the changes associated with removing pointer cast instructions from the PNaCl bitcode file. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3544 R=dschuff@chromium.org, jvoung@chromium.org Review URL: https://codereview.chromium.org/23482002
2013-08-26Elide pointer to int casts on phi nodes.Karl Schimpf
Handles the eliding of pointer to integer casts operands of the phi node. Also caches unelided casts generated in the reader (removing duplicates within the same block). This reduces the size of thawed pnacl-llc.pexe by about 2%. BUG= https://code.google.com/p/nativeclient/issues/detailid=3544 R=mseaborn@chromium.org Review URL: https://codereview.chromium.org/22909016
2013-08-14Remove ptrtoint instructions from the PNaCl bitcode file.Karl Schimpf
Removes ptrtoint instructions when applicable (currently only in stores), and add them back just before their use. Note: This code does not handle ptrtoint casts for calls and phi nodes, binary operators, etc. because handling of casts for these instructions has not been added yet. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3544 R=jvoung@chromium.org Review URL: https://codereview.chromium.org/22633002
2013-08-14Allow record-level printing by pnacl-bcanalyzer.Karl Schimpf
BUG= https://code.google.com/p/nativeclient/issues/detail?id=3627 R=jvoung@chromium.org Review URL: https://codereview.chromium.org/23060004
2013-08-09Properly support 16-bit atomics on x86-32.JF Bastien
16-bit atomics aren't handled properly by the current validator, this patch changes the translator so that on x86-32 it emulates 16-bit atomics with 32-bit atomics in a compare-exchange loop. TEST= ./scons run_synchronization_sync_test bitcode=1 platform=x86-32 BUG= https://code.google.com/p/nativeclient/issues/detail?id=3579 BUG= https://code.google.com/p/nativeclient/issues/detail?id=2981 BUG= https://code.google.com/p/nativeclient/issues/detail?id=3475 R=dschuff@chromium.org Review URL: https://codereview.chromium.org/22760002
2013-08-07Add the new @llvm.nacl.atomic.fence.all intrinsicJF Bastien
This is a follow-up to: https://codereview.chromium.org/22240002/ And requires the Clang changes from: https://codereview.chromium.org/22294002/ This new intrinsic represents ``asm("":::"~{memory}")`` as well as ``__sync_synchronize()``, and in IR it corresponds to a sequentially-consistent fence surrounded by ``call void asm sideeffect "", "~{memory}"()``. R=jvoung@chromium.org TEST= ninja check-all BUG= https://code.google.com/p/nativeclient/issues/detail?id=3475 Review URL: https://codereview.chromium.org/22474008
2013-08-05Fix handling of the volatile bit of loads/stores in PNaCl bitcode files.Karl Schimpf
Fixes so that the volatile bit is no longer put into the bitcode file, since the volatile bit is not in the PNaCl ABI. BUG= https://code.google.com/p/nativeclient/issues/detail?id=3610 R=jvoung@chromium.org Review URL: https://codereview.chromium.org/21949006
generated by cgit v1.2.3-18-g5258 (git 2.32.0) at 2024-11-21 21:05:17 +0000