David Majnemer [Mon, 13 Oct 2014 22:37:51 +0000 (22:37 +0000)]
InstCombine: Fix miscompile in X % -Y -> X % Y transform
We assumed that negation operations of the form (0 - %Z) resulted in a
negative number. This isn't true if %Z was originally negative.
Substituting the negative number into the remainder operation may result
in undefined behavior because the dividend might be INT_MIN.
Chris Bieneman [Mon, 13 Oct 2014 22:37:25 +0000 (22:37 +0000)]
Removing the static destructor from ManagedStatic.cpp by controlling the allocation and de-allocation of the mutex.
This patch adds a new llvm_call_once function which is used by the ManagedStatic implementation to safely initialize a global to avoid static construction and destruction.
Reed Kotler [Mon, 13 Oct 2014 21:46:41 +0000 (21:46 +0000)]
Make first of several changes to bring up to AArch64 fast-isel style
Summary:
Make Mips fast-isel track the form of AArch64 where practical.
This makes it easier for people to review the code, to borrow similar code, and to see how to eventually move a lot of this
target code for fast-isels into target independent code.
These are just cosmetic changes. Should be no functional difference.
Test Plan:
make check
test-suite for 4 flavors mips32 r1/r2 , -O0/-O2
Adrian Prantl [Mon, 13 Oct 2014 20:44:58 +0000 (20:44 +0000)]
Add an assertion about the integrity of the iterator.
Broken parent scope pointers in inlined DIVariables can cause
ensureAbstractVariableIsCreated to insert new abstract scopes, thus
invalidating the iterator in this loop and leading to hard-to-debug
crashes. Useful when manually reducing IR for testcases.
Bradley Smith [Mon, 13 Oct 2014 10:12:35 +0000 (10:12 +0000)]
[AArch64] Add workaround for Cortex-A53 erratum (835769)
Some early revisions of the Cortex-A53 have an erratum (835769) whereby it is
possible for a 64-bit multiply-accumulate instruction in AArch64 state to
generate an incorrect result. The details are quite complex and hard to
determine statically, since branches in the code may exist in some
circumstances, but all cases end with a memory (load, store, or prefetch)
instruction followed immediately by the multiply-accumulate operation.
The safest work-around for this issue is to make the compiler avoid emitting
multiply-accumulate instructions immediately after memory instructions and the
simplest way to do this is to insert a NOP.
This patch implements such work-around in the backend, enabled via the option
-aarch64-fix-cortex-a53-835769.
The work-around code generation is not enabled by default.
NAKAMURA Takumi [Mon, 13 Oct 2014 04:32:43 +0000 (04:32 +0000)]
Unix/Signals.inc: Let findModulesAndOffsets() built conditionally regarding to (defined(HAVE_BACKTRACE) && defined(ENABLE_BACKTRACES)). [-Wunused-function]
Richard Smith [Mon, 13 Oct 2014 00:41:03 +0000 (00:41 +0000)]
[modules] Stop excluding Support/Debug.h from the Support module. This header
has been modular since r206822, and excluding it was leading to workarounds
such as the one in r219592, which this change removes.
Simon Pilgrim [Sun, 12 Oct 2014 10:52:55 +0000 (10:52 +0000)]
[X86] Memory folding for commutative instructions.
This patch improves support for commutative instructions in the x86 memory folding implementation by attempting to fold a commuted version of the instruction if the original folding fails - if that folding fails as well the instruction is 're-commuted' back to its original order before returning.
This mainly helps the stack inliner better fold reloads of 3 (or more) operand instructions (VEX encoded SSE etc.) but by performing this in the lowest foldMemoryOperandImpl implementation it also replaces the X86InstrInfo::optimizeLoadInstr version and is now used by FastISel too.
David Majnemer [Sun, 12 Oct 2014 08:34:24 +0000 (08:34 +0000)]
InstCombine: Simplify commonIDivTransforms
A helper routine, MultiplyOverflows, was a less efficient
reimplementation of APInt's smul_ov and umul_ov. While we are here,
clean up the code so it's more uniform.
David Majnemer [Sat, 11 Oct 2014 10:20:04 +0000 (10:20 +0000)]
InstCombine: Don't fold (X <<s log(INT_MIN)) /s INT_MIN to X
Consider the case where X is 2. (2 <<s 31)/s-2147483648 is zero but we
would fold to X. Note that this is valid when we are in the unsigned
domain because we require NUW: 2 <<u 31 results in poison.
Chandler Carruth [Sat, 11 Oct 2014 05:28:30 +0000 (05:28 +0000)]
[SCEV] Fix one more caller blindly passing the latch to SCEV's
getSmallConstantTripCount even when it isn't the exiting block.
I missed this in my first audit, very sorry. This was found in LNT and
elsewhere. I don't have a test case, but it was completely obvious from
inspection that this was the problem. I'll see if I can reduce a test
case, but I'm not really hopeful, and the value seems quite low.
Chandler Carruth [Sat, 11 Oct 2014 01:04:40 +0000 (01:04 +0000)]
Guard the definition of the stack tracing function with the same macros
that guard its usage. Without this, we can get unused function warnings
when backtraces are disabled.
Chandler Carruth [Sat, 11 Oct 2014 00:12:11 +0000 (00:12 +0000)]
[SCEV] Add some asserts to the recently improved trip count computation
routines and fix all of the bugs they expose.
I hit a test case that crashed even without these asserts due to passing
a non-exiting latch to the ExitingBlock parameter of the trip count
computation machinery. However, when I add the nice asserts, it turns
out we have plenty of coverage of these bugs, they just didn't manifest
in crashers.
The core problem seems to stem from an assumption that the latch *is*
the exiting block. While this is often true, and somewhat the "normal"
way to think about loops, it isn't necessarily true. The correct way to
call the trip count routines in a *generic* fashion (that is, without
a particular exit in mind) is to just use the loop's single exiting
block if it has one. The trip count can't be computed generically unless
it does. This works great for the loop vectorizer. The loop unroller
actually *wants* to select the latch when it has to chose between
multiple exits because for unrolling it is the latch trips that matter.
But if this is the desire, it needs to explicitly guard for non-exiting
latches and check for the generic trip count in that case.
I've added the asserts, and added convenience APIs for querying the trip
count generically that check for a single exit block. I've kept the APIs
consistent between computing trip count and trip multiples.
Thansk to Mark for the help debugging and tracking down the *right* fix
here!
Sanjay Patel [Fri, 10 Oct 2014 23:00:21 +0000 (23:00 +0000)]
Return undef on FP <-> Int conversions that overflow (PR21330).
The LLVM Lang Ref states for signed/unsigned int to float conversions:
"If the value cannot fit in the floating point value, the results are undefined."
And for FP to signed/unsigned int:
"If the value cannot fit in ty2, the results are undefined."
This matches the C definitions.
The existing behavior pins to infinity or a max int value, but that may just
lead to more confusion as seen in:
http://llvm.org/bugs/show_bug.cgi?id=21130
Returning undef will hopefully lead to a less silent failure.
Alexey Samsonov [Fri, 10 Oct 2014 22:06:59 +0000 (22:06 +0000)]
Re-land r219354: Use llvm-symbolizer to symbolize LLVM/Clang crash dumps.
In fact, symbolization is now expected to work only on Linux and
FreeBSD/NetBSD, where we have dl_iterate_phdr and can learn the
main executable name without argv0 (it will be possible on BSD systems
after http://reviews.llvm.org/D5693 lands). #ifdef-out the code for
all the rest Unix systems.
Sanjoy Das [Fri, 10 Oct 2014 21:22:34 +0000 (21:22 +0000)]
This patch teaches ScalarEvolution to pick and use !range metadata.
It also makes it more aggressive in querying range information by
adding a call to isKnownPredicateWithRanges to
isLoopBackedgeGuardedByCond and isLoopEntryGuardedByCond.
Rafael Espindola [Fri, 10 Oct 2014 18:33:51 +0000 (18:33 +0000)]
llvm-ar: Start adding support for mri scripts.
I was quiet surprised to find this feature being used. Fortunately the uses
I found look fairly simple. In fact, they are just a very verbose version
of the regular ar commands.
Start implementing it then by parsing the script and setting the command
variables as if we had a regular command line.
This patch adds just enough support to create an empty archive and do a bit
of error checking. In followup patches I will implement at least addmod
and addlib.
From the description in the manual, even the more general case should not
be too hard to implement if needed. The features that don't map 1:1 to
the simple command line are
* Reading from multiple archives.
* Creating multiple archives.
Mark Heffernan [Fri, 10 Oct 2014 17:39:11 +0000 (17:39 +0000)]
This patch de-pessimizes the calculation of loop trip counts in
ScalarEvolution in the presence of multiple exits. Previously all
loops exits had to have identical counts for a loop trip count to be
considered computable. This pessimization was implemented by calling
getBackedgeTakenCount(L) rather than getExitCount(L, ExitingBlock)
inside of ScalarEvolution::getSmallConstantTripCount() (see the FIXME
in the comments of that function). The pessimization was added to fix
a corner case involving undefined behavior (pr/16130). This patch more
precisely handles the undefined behavior case allowing the pessimization
to be removed.
ControlsExit replaces IsSubExpr to more precisely track the case where
undefined behavior is expected to occur. Because undefined behavior is
tracked more precisely we can remove MustExit from ExitLimit. MustExit
was used to track the case where the limit was computed potentially
assuming undefined behavior even if undefined behavior didn't necessarily
occur.
Hal Finkel [Fri, 10 Oct 2014 17:06:20 +0000 (17:06 +0000)]
[MiSched] Fix a logic error in tryPressure()
Fixes a logic error in the MachineScheduler found by Steve Montgomery (and
confirmed by Andy). This has gone unfixed for months because the fix has been
found to introduce some small performance regressions. However, Andy has
recommended that, at this point, we fix this to avoid further dependence on the
incorrect behavior (and then follow-up separately on any regressions), and I
agree.
This change depends on the ApplePropertyString helper that I sent spearately.
Not sure how you want this tested: as a tool test by adding a binary to dump, or as an llvm test starting from an IR file?
Frederic Riss [Fri, 10 Oct 2014 15:51:02 +0000 (15:51 +0000)]
[dwarfdump] Resolve also variable specifications/abstract_origins.
DW_AT_specification and DW_AT_abstract_origin resolving was only performed
on subroutine DIEs because it used the getSubroutineName method. Introduce
a more generic getName() and use it to dump the reference attributes.
Testcases have been updated to check the printed names instead of the offsets
except when the name could be ambiguous.
Bill Schmidt [Fri, 10 Oct 2014 15:09:28 +0000 (15:09 +0000)]
[PowerPC] Add feature for Power8 vector extensions
The current VSX feature for PowerPC specifies availability of the VSX
instructions added with the 2.06 architecture version. With 2.07, the
architecture adds new instructions to both the Category:Vector and
Category:VSX instruction sets. Additionally, unaligned vector storage
operations have improved performance.
This patch adds a feature to provide access to the new instructions
and performance capabilities of Power8. For compatibility with GCC,
the feature is controlled via a new -mpower8-vector switch, and the
feature causes the __POWER8_VECTOR__ builtin define to be generated by
the preprocessor.
There is a companion patch for cfe being committed at the same time.
Chandler Carruth [Fri, 10 Oct 2014 08:27:19 +0000 (08:27 +0000)]
Don't use an unqualified 'abs' function call with a builtin type.
This is dangerous for numerous reasons. The primary risk here is with
floating point or double types where if the wrong header files are
included in a strange order this can implicitly convert to integers and
then call the C abs function on the integers. There is a secondary risk
that even impacts integers where if the namespace the code is written in
ever defines an abs overload for types within that namespace the global
abs will be hidden. The correct form is to call std::abs or write 'using
std::abs' for builtin types (and only the latter is correct in any
generic context).
I've also added the requisite header to be a bit more explicit here.
David Majnemer [Fri, 10 Oct 2014 06:58:11 +0000 (06:58 +0000)]
Object, COFF: Relax aux symbols for section definitions
We, I suppose naïvely, believed the COFF specification with regard to
auxiliary symbol records which defined sections: they specified that the
symbol value should be zero. However, dumpbin and MinGW's objdump do
not consider the symbol value as a restriction. Relaxing this allows us
to properly dump MinGW linked executables.
Chandler Carruth [Fri, 10 Oct 2014 05:14:12 +0000 (05:14 +0000)]
[ADT] Replace the logb implementation with the simpler and much closer
to what we actually want ilogb implementation. This makes everything
*much* easier to deal with and is actually what we want when using it
anyways.
because the simplifycfg transformation into selects would happen to happen
before the simplifycfg transformation that removes unreachable control flow
(We have 'unreachable control flow' due to the store to null which is undefined
behavior).
The existing transformation that removes unreachable control flow in simplifycfg
is:
/// If BB has an incoming value that will always trigger undefined behavior
/// (eg. null pointer dereference), remove the branch leading here.
static bool removeUndefIntroducingPredecessor(BasicBlock *BB)
Now we generate:
define void @test6(i1 %cond, i8* %ptr) {
store i8 2, i8* %ptr.2, align 8
ret void
}
I did not see any impact on the test-suite + externals.
Sanjay Patel [Thu, 9 Oct 2014 21:26:35 +0000 (21:26 +0000)]
Improve sqrt estimate algorithm (fast-math)
This patch changes the fast-math implementation for calculating sqrt(x) from:
y = 1 / (1 / sqrt(x))
to:
y = x * (1 / sqrt(x))
This has 2 benefits: less code / faster code and one less estimate instruction
that may lose precision.
The only target that will be affected (until http://reviews.llvm.org/D5658 is approved)
is PPC. The difference in codegen for PPC is 2 less flops for a single-precision sqrtf
or vector sqrtf and 4 less flops for a double-precision sqrt.
We also eliminate a constant load and extra register usage.
Samuel Antao [Thu, 9 Oct 2014 20:42:56 +0000 (20:42 +0000)]
Fix bug in GPR to FPR moves in PPC64LE.
The current implementation of GPR->FPR register moves uses a stack slot. This mechanism writes a double word and reads a word. In big-endian the load address must be displaced by 4-bytes in order to get the right value. In little endian this is no longer required. This patch fixes the issue and adds LE regression tests to fast-isel-conversion which currently expose this problem.
projects / llvm / log