Commit Graph

62 Commits

Author SHA1 Message Date
Benjamin Kramer
f3a3b0fc0e Update unittest for allocator laziness.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101131 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-13 15:01:26 +00:00
Chris Lattner
796e64b546 update unit test for api change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100486 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-05 22:49:48 +00:00
Jeffrey Yasskin
39c75f2e90 Fix PR6360. It's easy for a stub's address to escape to user code, so we can't
just count references to it from JIT output to decide when to destroy it.  This
patch waits to destroy the JIT's memory of a stub until the Function it refers
to is destroyed.  External function stubs and GVIndirectSyms aren't destroyed
until the JIT itself is.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97737 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-04 19:45:09 +00:00
Jeffrey Yasskin
e57898ce55 Fix PR5291, in which a SmallPtrSet iterator was held across an insertion into
the set.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97720 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-04 06:50:01 +00:00
Jeffrey Yasskin
40966a7c68 Make it possible to create multiple JIT instances at the same time, by removing
the global TheJIT and TheJITResolver variables.  Lazy compilation is supported
by a global map from a stub address to the JITResolver that knows how to
compile it.

Patch by Olivier Meurant!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@95837 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-11 01:07:39 +00:00
Jeffrey Yasskin
f0356fe140 Kill ModuleProvider and ghost linkage by inverting the relationship between
Modules and ModuleProviders. Because the "ModuleProvider" simply materializes
GlobalValues now, and doesn't provide modules, it's renamed to
"GVMaterializer". Code that used to need a ModuleProvider to materialize
Functions can now materialize the Functions directly. Functions no longer use a
magic linkage to record that they're materializable; they simply ask the
GVMaterializer.

Because the C ABI must never change, we can't remove LLVMModuleProviderRef or
the functions that refer to it. Instead, because Module now exposes the same
functionality ModuleProvider used to, we store a Module* in any
LLVMModuleProviderRef and translate in the wrapper methods.  The bindings to
other languages still use the ModuleProvider concept.  It would probably be
worth some time to update them to follow the C++ more closely, but I don't
intend to do it.

Fixes http://llvm.org/PR5737 and http://llvm.org/PR5735.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94686 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-27 20:34:15 +00:00
Jeffrey Yasskin
9815582fad Roll r94484 (avoiding RTTI problems in tests) forward again in a way that isn't
broken by setting CXXFLAGS on the command line.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94619 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-26 23:30:46 +00:00
Bob Wilson
8dd0ab074a Revert 94484. Re-disable unittests that need RTTI.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94569 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-26 19:04:37 +00:00
Jeffrey Yasskin
f5fe3beb3a Re-enable unit tests disabled in r94164 by telling GTest about the
lack of RTTI.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94484 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-26 01:26:46 +00:00
Chris Lattner
e73a31f667 Stop building RTTI information for *most* llvm libraries. Notable
missing ones are libsupport, libsystem and libvmcore.  libvmcore is
currently blocked on bugpoint, which uses EH.  Once it stops using
EH, we can switch it off.

This #if 0's out 3 unit tests, because gtest requires RTTI information.
Suggestions welcome on how to fix this.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94164 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-22 06:49:46 +00:00
Jeffrey Yasskin
b9b88ea241 Disable JITTest.FunctionIsRecompiledAndRelinked on ARM where it's not
implemented.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91963 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-23 00:58:02 +00:00
Jeffrey Yasskin
c5818fb837 Partially revert r91626. Materializing extra functions to determine whether
they're available_externally broke VMKit, which was relying on the fact that
functions would only be materialized when they were first called.  We'll have
to wait for http://llvm.org/PR5737 to really fix this.

I also added a test for one of the F->isDeclaration() calls which wasn't
covered by anything else in the test suite.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91943 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-22 23:47:23 +00:00
Jeffrey Yasskin
92fdf4537b Fix a crash in JIT::recompileAndRelinkFunction(). It doesn't pass the MCI
argument to runJITOnFunction(), which caused a null pointer dereference at
every call.

Patch by Gianluca Guida!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91939 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-22 23:18:18 +00:00
Jeffrey Yasskin
aad0d52c5b Don't codegen available_externally functions. Fixes http://llvm.org/PR5735.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91626 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-17 21:35:29 +00:00
Jeffrey Yasskin
898e9df8db Reinstate r91208 to fix available_externally linkage for globals, with
nlewycky's fix to add -rdynamic so the JIT can look symbols up in Linux builds
of the JITTests binary.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91250 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-13 20:30:32 +00:00
Jeffrey Yasskin
9b8ff28870 Revert r91208. Something on Linux prevents the JIT from looking up a symbol
defined in the test, and I don't have time tonight to figure it out.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91209 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-12 06:18:46 +00:00
Jeffrey Yasskin
5cc966a6c1 Fix available_externally linkage for globals. It's probably still not
supported by emitGlobals, but I don't have a test case for that.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@91208 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-12 05:58:14 +00:00
Jeffrey Yasskin
630382adc8 Oops. Re-disable JITTest.NoStubs on ARM and PPC since they still use stubs to
make far calls work.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@89733 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-24 02:11:14 +00:00
Jeffrey Yasskin
108c838093 * Move stub allocation inside the JITEmitter, instead of exposing a
way for each TargetJITInfo subclass to allocate its own stubs. This
means stubs aren't as exactly-sized anymore, but it lets us get rid of
TargetJITInfo::emitFunctionStubAtAddr(), which lets ARM and PPC
support the eager JIT, fixing http://llvm.org/PR4816.

* Rename the JITEmitter's stub creation functions to describe the kind
of stub they create. So far, all of them create lazy-compilation
stubs, but they sometimes get used when far-call stubs are needed.
Fixing http://llvm.org/PR5201 will involve fixing this.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@89715 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-23 23:35:19 +00:00
Jeffrey Yasskin
feada9462b Revert the test from r88984. It relies on being able to mmap 16GB of
address space (though it only uses a small fraction of that), and the
buildbots disallow that.

Also add a comment to the Makefile's ulimit line warning future
developers that changing it won't work.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@88994 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-16 23:32:30 +00:00
Jeffrey Yasskin
d1ba06bf13 Make X86-64 in the Large model always emit 64-bit calls.
The large code model is documented at
http://www.x86-64.org/documentation/abi.pdf and says that calls should
assume their target doesn't live within the 32-bit pc-relative offset
that fits in the call instruction.

To do this, we turn off the global-address->target-global-address
conversion in X86TargetLowering::LowerCall(). The first attempt at
this broke the lazy JIT because it can separate the movabs(imm->reg)
from the actual call instruction. The lazy JIT receives the address of
the movabs as a relocation and needs to record the return address from
the call; and then when that call happens, it needs to patch the
movabs with the newly-compiled target. We could thread the call
instruction into the relocation and record the movabs<->call mapping
explicitly, but that seems to require at least as much new
complication in the code generator as this change.

To fix this, we make lazy functions _always_ go through a call
stub. You'd think we'd only have to force lazy calls through a stub on
difficult platforms, but that turns out to break indirect calls
through a function pointer. The right fix for that is to distinguish
between calls and address-of operations on uncompiled functions, but
that's complex enough to leave for someone else to do.

Another attempt at this defined a new CALL64i pseudo-instruction,
which expanded to a 2-instruction sequence in the assembly output and
was special-cased in the X86CodeEmitter's emitInstruction()
function. That broke indirect calls in the same way as above.

This patch also removes a hack forcing Darwin to the small code model.
Without far-call-stubs, the small code model requires things of the
JITMemoryManager that the DefaultJITMemoryManager can't provide.

Thanks to echristo for lots of testing!



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@88984 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-16 22:41:33 +00:00
Benjamin Kramer
74c10e7d8d This test doesn't work on arm either.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@88794 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-14 15:15:39 +00:00
Bill Wendling
0c2749f3e2 Disable the JITTest.NoStubs test for Darwin PPC. It apparently doesn't implement
emitFunctionStubAtAddr.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@88708 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-13 21:58:54 +00:00
Eric Christopher
116664a697 Use stubs when we have them, otherwise use code we already have,
otherwise create a stub.

Add a test to make sure we don't create extraneous stubs.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86941 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-12 03:12:18 +00:00
Jeffrey Yasskin
b069c91191 Fix JITTest.ModuleDeletion in -Asserts mode (which turns off JITEmitDebugInfo
by default).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86807 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-11 05:30:02 +00:00
Jeffrey Yasskin
6f348e4586 Remove dlsym stubs, with Nate Begeman's permission.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86606 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-09 22:34:19 +00:00
Jeffrey Yasskin
b235224ee7 Fix the ModuleDeletion test on PPC and ARM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85352 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 00:28:31 +00:00
Jeffrey Yasskin
18fec73e29 Revert the API changes from r85295 to make it easier for people to build
against both 2.6 and HEAD.  The default is still changed to eager jitting.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85330 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-27 22:39:42 +00:00
Jeffrey Yasskin
dc85724f70 Change the JIT to compile eagerly by default as agreed in
http://llvm.org/PR5184, and beef up the comments to describe what both options
do and the risks of lazy compilation in the presence of threads.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85295 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-27 20:30:28 +00:00
Chris Lattner
f3523592b2 Type.h doesn't need to #include LLVMContext.h
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85254 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-27 17:08:31 +00:00
Jeffrey Yasskin
7a9034c4db Automatically do the equivalent of freeMachineCodeForFunction(F) when F is
being destroyed. This allows users to run global optimizations like globaldce
even after some functions have been jitted.

This patch also removes the Function* parameter to
JITEventListener::NotifyFreeingMachineCode() since it can cause that to be
called when the Function is partially destroyed. This change will be even more
helpful later when I think we'll want to allow machine code to actually outlive
its Function.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85182 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-27 00:03:05 +00:00
Jeffrey Yasskin
23e5fcfec4 Fix http://llvm.org/PR4822: allow module deletion after a function has been
compiled.

When functions are compiled, they accumulate references in the JITResolver's
stub maps. This patch removes those references when the functions are
destroyed.  It's illegal to destroy a Function when any thread may still try to
call its machine code.

This patch also updates r83987 to use ValueMap instead of explicit CallbackVHs
and fixes a couple "do stuff inside assert()" bugs from r84522.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84975 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-23 22:37:43 +00:00
Jeffrey Yasskin
1e86132122 Move the Function*->allocated blocks map from the JITMemoryManager to the
JITEmitter.

I'm gradually making Functions auto-remove themselves from the JIT when they're
destroyed. In this case, the Function needs to be removed from the JITEmitter,
but the map recording which Functions need to be removed lived behind the
JITMemoryManager interface, which made things difficult.

This patch replaces the deallocateMemForFunction(Function*) method with a pair
of methods deallocateFunctionBody(void *) and deallocateExceptionTable(void *)
corresponding to the two startFoo/endFoo pairs.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84651 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-20 18:13:21 +00:00
Daniel Dunbar
b576beaaa8 PowerPC ifdef'ing considered more complicated than one might like.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84603 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-20 05:33:23 +00:00
Nick Lewycky
13555eae17 Correct test for PowerPC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84595 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-20 04:09:50 +00:00
Daniel Dunbar
e9869d8eeb Also check for __POWERPC__ when skipping these tests.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84482 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-19 09:19:32 +00:00
Benjamin Kramer
06b386a68d Disable another unittest that doesn't work on arm and ppc.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84186 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-15 16:49:16 +00:00
Nick Lewycky
6c79e2c9aa The ARM and PowerPC jits are broken in this regard.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84128 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-14 20:04:41 +00:00
Jeffrey Yasskin
e5f879825f Keep track of stubs that are created. This fixes PR5162 and probably PR4822 and
4406. Patch by Nick Lewycky!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84032 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-13 21:32:57 +00:00
Jeffrey Yasskin
4c5b23b24f Make the ExecutionEngine automatically remove global mappings on when their
GlobalValue is destroyed.  Function destruction still leaks machine code and
can crash on leaked stubs, but this is some progress.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83987 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-13 17:42:08 +00:00
Jeffrey Yasskin
c89d27a440 ExecutionEngine::clearGlobalMappingsFromModule failed to remove reverse
mappings, which could cause errors and assert-failures.  This patch fixes that,
adds a test, and refactors the global-mapping-removal code into a single place.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83678 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-09 22:10:27 +00:00
Jeffrey Yasskin
0f2ba783d9 Fix illegal cross-type aliasing. Found by baldrick on a newer gcc.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83401 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-06 19:06:16 +00:00
Jeffrey Yasskin
ea5ed00ea3 Fix http://llvm.org/PR5116 by rolling back r60822. This passes `make unittests
check-lit` on both x86-64 Linux and x86-32 Darwin.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83353 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-06 00:35:55 +00:00
Owen Anderson
1d0be15f89 Push LLVMContexts through the IntegerType APIs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78948 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-13 21:58:54 +00:00
Jeffrey Yasskin
0a962314fb Make ExecutionEngine::updateGlobalMapping(GV, NULL) properly remove GV's old
address from the reverse mapping, and add a test that this works now.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78127 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-04 23:53:16 +00:00
Benjamin Kramer
feba7562ec Update unittest for LLVM API change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77730 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-31 20:56:31 +00:00
Owen Anderson
debcb01b0f Move types back to the 2.5 API.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77516 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-29 22:17:13 +00:00
Owen Anderson
eed707b1e6 Revert the ConstantInt constructors back to their 2.5 forms where possible, thanks to contexts-on-types. More to come.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77011 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-24 23:12:02 +00:00
Reid Kleckner
535c9c3c9d Fixing unittests on 32-bit Darwin, using 0x...ULL instead of 0x...U .
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@76904 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-23 22:27:18 +00:00
Reid Kleckner
10b4fc552f Re-committing r76828 with the JIT memory manager changes now that the build
bots like the BumpPtrAllocator changes.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@76902 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-23 21:46:56 +00:00