Justin Bogner [Mon, 16 Mar 2015 07:29:49 +0000 (07:29 +0000)]
InstrProf: Remove xfails for big-endian from coverage tests
This still doesn't actually work correctly for big endian input files,
but since these tests all use little endian input files they don't
actually fail. I'll be committing a real fix for big endian soon, but
I don't have proper tests for it yet.
[objc-arc] Make the ARC optimizer more conservative by forcing it to be non-safe in both direction, but mitigate the problem by noting that we just care if there was a further use.
The problem here is the infamous one direction known safe. I was
hesitant to turn it off before b/c of the potential for regressions
without an actual bug from users hitting the problem. This is that bug ;
).
The main performance impact of having known safe in both directions is
that often times it is very difficult to find two releases without a use
in-between them since we are so conservative with determining potential
uses. The one direction known safe gets around that problem by taking
advantage of many situations where we have two retains in a row,
allowing us to avoid that problem. That being said, the one direction
known safe is unsafe. Consider the following situation:
retain(x)
retain(x)
call(x)
call(x)
release(x)
Then we know the following about the reference count of x:
// rc(x) == N (for some N).
retain(x)
// rc(x) == N+1
retain(x)
// rc(x) == N+2
call A(x)
call B(x)
// rc(x) >= 1 (since we can not release a deallocated pointer).
release(x)
// rc(x) >= 0
That is all the information that we can know statically. That means that
we know that A(x), B(x) together can release (x) at most N+1 times. Lets
say that we remove the inner retain, release pair.
We knew before that A(x), B(x) could release x up to N+1 times meaning
that rc(x) may be zero at the release(x). That is not safe. On the other
hand, consider the following situation where we have a must use of
release(x) that x must be kept alive for after the release(x)**. Then we
know that:
// rc(x) == N (for some N).
retain(x)
// rc(x) == N+1
retain(x)
// rc(x) == N+2
call A(x)
call B(x)
// rc(x) >= 2 (since we know that we are going to release x and that that release can not be the last use of x).
release(x)
// rc(x) >= 1 (since we can not deallocate the pointer since we have a must use after x).
…
// rc(x) >= 1
use(x)
Thus we know that statically the calls to A(x), B(x) can together only
release rc(x) N times. Thus if we remove the inner retain, release pair:
We are still safe unless in the final … there are unbalanced retains,
releases which would have caused the program to blow up anyways even
before optimization occurred. The simplest form of must use is an
additional release that has not been paired up with any retain (if we
had paired the release with a retain and removed it we would not have
the additional use). This fits nicely into the ARC framework since
basically what you do is say that given any nested releases regardless
of what is in between, the inner release is known safe. This enables us to get
back the lost performance.
Justin Bogner [Mon, 16 Mar 2015 06:55:45 +0000 (06:55 +0000)]
InstrProf: Do a better job of reading coverage mapping data.
This code was casting regions of a memory buffer to a couple of
different structs. This is wrong in a few ways:
1. It breaks aliasing rules.
2. If the buffer isn't aligned, it hits undefined behaviour.
3. It completely ignores endianness differences.
4. The structs being defined for this aren't specifying their padding
properly, so this doesn't even represent the data properly on some
platforms.
This commit is mostly NFC, except that it fixes reading coverage for
32 bit binaries as a side effect of getting rid of the mispadded
structs. I've included a test for that.
I've also baked in that we only handle little endian more explicitly,
since that was true in practice already. I'll fix this to handle
endianness properly in a followup commit.
Frederic Riss [Mon, 16 Mar 2015 02:05:10 +0000 (02:05 +0000)]
[dsymutil] Add support to generate .debug_pubnames and .debug_pubtypes
The information gathering part of the patch stores a bit more information
than what is strictly necessary for these 2 sections. The rest will
become useful when we start emitting __apple_* type accelerator tables.
Renato Golin [Sun, 15 Mar 2015 21:15:48 +0000 (21:15 +0000)]
Adding commit msg guidelines to dev policy
After much bike shed discussions, we seem to agree to a few loose
but relevant guidelines on how to prepare a commit message. It also
points the attribution section to the new commit messages section
to deduplicate information.
Frederic Riss [Sun, 15 Mar 2015 20:45:43 +0000 (20:45 +0000)]
[dsymutil] Add support for linking line tables.
This code comes with a lot of cruft that is meant to mimic darwin's
dsymutil behavior. A much simpler approach (described in the numerous
FIXMEs that I put in there) gives the right output for the vast
majority of cases. The extra corner cases that are handled differently
need to be investigated: they seem to correctly handle debug info that
is in the input, but that info looks suspicious in the first place.
Anyway, the current code needs to handle this, but I plan to revisit it
as soon as the big round of validation against the classic dsymutil is
over.
Frederic Riss [Sun, 15 Mar 2015 20:45:39 +0000 (20:45 +0000)]
[MCDwarf] Do not emit useless line table opcode.
No need to emit a DW_LNS_advance_pc with a 0 increment. Found out while
comparing dsymutil's and LLVM's line table encoding. Not a correctenss
fix, just a small encoding size optimization.
I'm not sure how to generate a sequence that triggers this, and moreover
llvm-dwardump doesn't dump the line table program, thus the effort
involved in creating a testcase for this trivial patch seemed out of
proportion.
IR: Default the Metadata::dump() argument "harder" after r232275
Use an overload instead of a default argument for `Metadata::dump()`.
The latter seems to require calling `dump(nullptr)` explicitly when
using a debugger, where as the former doesn't.
Other than utility for debugging, there's NFC here.
Frederic Riss [Sun, 15 Mar 2015 01:29:30 +0000 (01:29 +0000)]
[dsymutil] Add function size to the debug map.
The debug map embedded by ld64 in binaries conatins function sizes.
These sizes are less precise than the ones given by the debug information
(byte granularity vs linker atom granularity), but they might cover code
that is referenced in the line table but not in the DIE tree (that might
very well be a compiler bug that I need to investigate later).
Anyway, extracting that information is necessary to be able to mimic
dsymutil's behavior exactly.
Verify that debug info intrinsic arguments are valid. (These checks
will not recurse through the full debug info graph, so they don't need
to be cordoned of in `DebugInfoVerifier`.)
With those checks in place, changing the `DbgIntrinsicInst` accessors to
downcast to `MDLocalVariable` and `MDExpression` is natural (added isa
specializations in `Metadata.h` to support this).
Added tests to `test/Verifier` for the new -verify checks, and fixed the
debug info in all the in-tree tests.
If you have out-of-tree testcases that have started to fail to -verify,
hopefully the verify checks are helpful. The most likely problem is
that the expression argument is `!{}` (instead of `!MDExpression()`).
Assembler: Rewrite test for function-local metadata
This test for function-local metadata did strange things, and never
really sent in valid arguments for `llvm.dbg.declare` and
`llvm.dbg.value` intrinsics. Those that might have once been valid have
bitrotted.
Rewrite it to be a targeted test for function-local metadata --
unrelated to debug info, which is tested elsewhere -- and rename it to
better match other metadata-related tests.
(Note: the scope of function-local metadata changed drastically during
the metadata/value split, but I didn't properly clean up this testcase.
Most of the IR in this file, while invalid for debug info intrinsics,
used to provide coverage for various (now illegal) forms of
function-local metadata.)
Chris Bieneman [Sat, 14 Mar 2015 21:20:32 +0000 (21:20 +0000)]
A few minor updates based on feedback from Justin and a few things I thought were missing.
* Moved autotools configure & build example out of "Getting Started Quickly (A Summary)" and into BuildingLLVMWithAutoTools.
* Removed the annotations that CMake is the recommended process and Autotools is alternate.
* Added brief documentation about build targets under "Getting Started Quickly..."
* Added Overview text to BuildingLLVMWithAutotools
* Fixed up a broken link.
David Blaikie [Sat, 14 Mar 2015 21:11:26 +0000 (21:11 +0000)]
[opaque pointer type] gep API migration, ArgPromo
This involved threading the type-to-gep through a data structure, since
the code was relying on the pointer type to carry this information. I
imagine there will be a lot of this work across the project... slow
work chasing each use case, but the assertions will help keep me honest.
Replumb the `AsmWriter` so that `Metadata::print()` is generally useful.
(Similarly change `Metadata::printAsOperand()`.)
- `SlotTracker` now has a mode where all metadata will be correctly
numbered when initializing a `Module`. Normally, `Metadata` only
referenced from within `Function`s gets numbered when the `Function`
is incorporated.
- `Metadata::print()` and `Metadata::printAsOperand()` (and
`Metadata::dump()`) now take an optional `Module` argument. When
provided, `SlotTracker` is initialized with the new mode, and the
numbering will be complete and consistent for all calls to `print()`.
- `Value::print()` uses the new `SlotTracker` mode when printing
intrinsics with `MDNode` operands, `MetadataAsValue` operands, or the
bodies of functions. Thus, metadata numbering will be consistent
between calls to `Metadata::print()` and `Value::print()`.
- `Metadata::print()` (and `Metadata::dump()`) now print the full
definition of `MDNode`s:
!5 = !{!6, !"abc", !7}
This matches behaviour for `Value::print()`, which includes the name
of instructions.
- Updated call sites in `Verifier` to call `print()` instead of
`printAsOperand()`.
All this, so that `Verifier` can print out useful failure messages that
involve `Metadata` for PR22777.
Note that `Metadata::printAsOperand()` previously took an optional
`bool` and `Module` operand. The former was cargo-culted from
`Value::printAsOperand()` and wasn't doing anything useful. The latter
didn't give consistent results (without the new `SlotTracker` mode).
David Blaikie [Sat, 14 Mar 2015 19:24:04 +0000 (19:24 +0000)]
[opaque pointer type] more gep API migrations
Adding nullptr to all the IRBuilder stuff because it's the first thing
that fails to build when testing without the back-compat functions, so
I'll keep having to re-add these locally for each chunk of migration I
do. Might as well check them in to save me the churn. Eventually I'll
have to migrate these too, but I'm going breadth-first.
Recover the ability to 'b CheckFailed' after r231577
Given that the stated purpose of `CheckFailed()` is to provide a nice
spot for a breakpoint, it'd be nice not to have to use a regex to break
on it. Recover the ability to simply use `b CheckFailed` by
specializing the message-only version, and by changing the variadic
version to call into the message-only version.
Frederic Riss [Sat, 14 Mar 2015 15:49:07 +0000 (15:49 +0000)]
[dsymutil] Add support for debug_loc section.
There is no need to look into the location expressions to transfer them,
the only modification to apply is to patch their base address to reflect
the linked function address.
Daniel Jasper [Sat, 14 Mar 2015 10:58:38 +0000 (10:58 +0000)]
[MachineLICM] First steps of sinking GEPs near calls.
Specifically, if there are copy-like instructions in the loop header
they are moved into the loop close to their uses. This reduces the live
intervals of the values and can avoid register spills.
This is working towards a fix for http://llvm.org/PR22230.
Review: http://reviews.llvm.org/D7259
Next steps:
- Find a better cost model (which non-copy instructions should be sunk?)
- Make this dependent on register pressure
Frederic Riss [Sat, 14 Mar 2015 03:46:51 +0000 (03:46 +0000)]
[dsymutil] Generate debug_aranges section.
This actually shares most of its implementation with the generation
of the debug_ranges (the absence of 'a' is not a typo) contribution
for the unit's DW_AT_ranges attribute.
Frederic Riss [Sat, 14 Mar 2015 03:46:40 +0000 (03:46 +0000)]
[dsymutil] Identify each CompileUnit with a unique ID.
The ID can eg. de used in MCSymbol names to differentiate the ones
that need to be created for every unit.
The ID is a constructor parameter and not a static class member so
there is no issue with counter updates if we decide to thread that
code.
David Blaikie [Sat, 14 Mar 2015 01:53:18 +0000 (01:53 +0000)]
[opaque pointer type] Start migrating GEP creation to explicitly specify the pointee type
I'm just going to migrate these in a pretty ad-hoc & incremental way -
providing the backwards compatible API for now, then locally removing
it, fixing a few callers, adding it back in and commiting those callers.
Rinse, repeat.
The assertions should ensure that if I get this wrong we'll find out
about it and not just have one giant patch to revert, recommit, revert,
recommit, etc.
LowerBitSets: Do not export symbols for bit set referenced globals on Darwin.
The linker on that platform may re-order symbols or strip dead symbols, which
will break bit set checks. Avoid this by hiding the symbols from the linker.
Frederic Riss [Fri, 13 Mar 2015 23:55:29 +0000 (23:55 +0000)]
[dsymutil] Fix typo in comment.
Next time, when I fix a typo, I'll take the time to reread the whole
comment instead of waiting for the commit email to realize that there
is another one two words later...
Rafael Espindola [Fri, 13 Mar 2015 22:18:18 +0000 (22:18 +0000)]
Use add32ri8 and friends on fast isel.
This fixes pr22854.
The core issue on the bug is that there are multiple instructions that
print the same in assembly. In fact, there doesn't seem to be any
syntax for specifying that a constant that fits in 8 bits should use a 32 bit
immediate.
The attached patch changes fast isel to consider i16immSExt8,
i32immSExt8, and i64immSExt8. They were disabled because fastisel didn’t know
to call the predicate back in the day.
David Blaikie [Fri, 13 Mar 2015 21:03:36 +0000 (21:03 +0000)]
[opaque pointer type] Bitcode support for explicit type parameter on the gep operator
This happened to be fairly easy to support backwards compatibility based
on the number of operands (old format had an even number, new format has
one more operand so an odd number).
test/Bitcode/old-aliases.ll already appears to test old gep operators
(if I remove the backwards compatibility in the BitcodeReader, this and
another test fail) so I'm not adding extra test coverage here.
David Blaikie [Fri, 13 Mar 2015 21:03:34 +0000 (21:03 +0000)]
Turn assertion into bitcode reading error
I don't think we test invalid bitcode records in any detail, so no test
here - just a change for consistency with existing error checks in
surrounding code.
Robert Lougher [Fri, 13 Mar 2015 20:53:01 +0000 (20:53 +0000)]
Reapply "[Reassociate] Add initial support for vector instructions."
This reapplies the patch previously committed at revision 232190. This was
reverted at revision 232196 as it caused test failures in tests that did not
expect operands to be commuted. I have made the tests more resilient to
reassociation in revision 232206.
As a follow-up to r232200, add an `-instcombine` to canonicalize scalar
allocations to `i32 1`. Since r232200, `iX 1` (for X != 32) are only
created by RAUWs, so this shouldn't fire too often. Nevertheless, it's
a cheap check and a nice cleanup.
instcombine: alloca: Limit array size type promotion
Move type promotion of the size of the array allocation to the end of
`simplifyAllocaArraySize()`. This avoids promoting the type of the
array size if it's a `ConstantInt`, since the next -instcombine
iteration will drop it to a scalar allocation anyway. Similarly, this
avoids promoting the type if it's an `UndefValue`, in which case the
alloca gets RAUW'ed.
This is NFC when considered over the lifetime of -instcombine, since
it's just reducing the number of iterations needed to reach fixed point.
Write the `alloca` array size explicitly when it's non-canonical.
Previously, if the array size was `iX 1` (where X is not 32), the type
would mutate to `i32` when round-tripping through assembly.
The testcase I added fails in `verify-uselistorder` (as well as
`FileCheck`), since the use-lists for `i32 1` and `i64 1` change.
(Manman Ren came across this when running `verify-uselistorder` on some
non-trivial, optimized code as part of PR5680.)
The type mutation started with r104911, which allowed array sizes to be
something other than an `i32`. Starting with r204945, we
"canonicalized" to `i64` on 64-bit platforms -- and then on every
round-trip through assembly, mutated back to `i32`.
I bundled a fixup for `-instcombine` to avoid r204945 on scalar
allocations. (There wasn't a clean way to sequence this into two
commits, since the assembly change on its own caused testcase churn, and
the `-instcombine` change can't be tested without the assembly changes.)
An obvious alternative fix -- change `AllocaInst::AllocaInst()`,
`AsmWriter` and `LLParser` to treat `intptr_t` as the canonical type for
scalar allocations -- was rejected out of hand, since this required
teaching them each about the data layout.
A follow-up commit will add an `-instcombine` to canonicalize the scalar
allocation array size to `i32 1` rather than leaving `iX 1` alone.
Manman Ren [Fri, 13 Mar 2015 19:24:30 +0000 (19:24 +0000)]
Add a parameter for getLazyBitcodeModule to lazily load Metadata.
We only defer loading metadata inside ParseModule when ShouldLazyLoadMetadata
is true and we have not loaded any Metadata block yet.
This commit implements all-or-nothing loading of Metadata. If there is a
request to load any metadata block, we will load all deferred metadata blocks.
We make sure the deferred metadata blocks are loaded before we materialize any
function or a module.
The default value of the added parameter ShouldLazyLoadMetadata for
getLazyBitcodeModule is false, so the default behavior stays the same.
We only set the parameter to true when creating LTOModule in local contexts.
These can only really be used for parsing symbols, so it's unnecessary to ever
load the metadata blocks.
If we are going to enable lazy-loading of Metadata for other usages of
getLazyBitcodeModule, where deferred metadata blocks need to be loaded, we can
expose BitcodeReader::materializeMetadata to Module, similar to
Module::materialize.
instcombine: alloca: Split out simplifyAllocaArraySize(), NFC
Follow-up commits will change some of the logic here. Splitting into a
separate function simplifies the logic by allowing early returns instead
of deeper nesting.