This simplifies ELFObjectWriter::SymbolValue a bit more. This new version
will also be used in the COFF writer to fix pr19147.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207711 91177308-0d34-0410-b5e6-96231b3b80d8
For pattern like ((x >> C1) & Mask) << C2, DAG combiner may convert it
into (x >> (C1-C2)) & (Mask << C2), which makes pattern matching of ubfx
more difficult.
For example:
Given
%shr = lshr i64 %x, 4
%and = and i64 %shr, 15
%arrayidx = getelementptr inbounds [8 x [64 x i64]]* @arr, i64 0, %i64 2, i64 %and
%0 = load i64* %arrayidx
With current shift folding, it takes 3 instrs to compute base address:
lsr x8, x0, #1
and x8, x8, #0x78
add x8, x9, x8
If using ubfx, it only needs 2 instrs:
ubfx x8, x0, #4, #4
add x8, x9, x8, lsl #3
This fixes bug 19589
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207702 91177308-0d34-0410-b5e6-96231b3b80d8
We already do this for shstrtab, so might as well do it for strtab. This
extracts the string table building code into a separate class. The idea
is to use it for other object formats too.
I mostly wanted to do this for the general principle, but it does save a
little bit on object file size. I tried this on a clang bootstrap and
saved 0.54% on the sum of object file sizes (1.14 MB out of 212 MB for
a release build).
Differential Revision: http://reviews.llvm.org/D3533
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207670 91177308-0d34-0410-b5e6-96231b3b80d8
When we were moving from a larger vector to a smaller one but didn't
need to re-allocate, we would move-assign over uninitialized memory in
the target, then move-construct that same data again.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207663 91177308-0d34-0410-b5e6-96231b3b80d8
edge entirely within an existing SCC. Shockingly, making the connected
component more connected is ... a total snooze fest. =]
Anyways, its wired up, and I even added a test case to make sure it
pretty much sorta works. =D
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207631 91177308-0d34-0410-b5e6-96231b3b80d8
(OutBufCur + Size) might overflow if Size were large. For example on i686-linux,
OutBufCur: 0xFFFDF27D
OutBufEnd: 0xFFFDF370
Size: 0x0002BF20 (180,000)
It caused flaky error in MC/COFF/section-name-encoding.s.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207621 91177308-0d34-0410-b5e6-96231b3b80d8
bits), and discover that it's totally broken. Yay tests. Boo bug. Fix
the basic edge removal so that it works by nulling out the removed edges
rather than actually removing them. This leaves the indices valid in the
map from callee to index, and preserves some of the locality for
iterating over edges. The iterator is made bidirectional to reflect that
it now has to skip over null entries, and the skipping logic is layered
onto it.
As future work, I would like to track essentially the "load factor" of
the edge list, and when it falls below a threshold do a compaction.
An alternative I considered (and continue to consider) is storing the
callees in a doubly linked list where each element of the list is in
a set (which is essentially the classical linked-hash-table
datastructure). The problem with that approach is that either you need
to heap allocate the linked list nodes and use pointers to them, or use
a bucket hash table (with even *more* linked list pointer overhead!),
etc. It's pretty easy to get 5x overhead for values that are just
pointers. So far, I think punching holes in the vector, and periodic
compaction is likely to be much more efficient overall in the space/time
tradeoff.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207619 91177308-0d34-0410-b5e6-96231b3b80d8
wrong iterator category. These aren't comprehensive, but they have
caught the common cases for me and produce much nicer errors.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207601 91177308-0d34-0410-b5e6-96231b3b80d8
Seems MSVC wants to be able to codegen inline-definitions of virtual
functions even in TUs that don't define the key function - and it's well
within its rights to do so.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207581 91177308-0d34-0410-b5e6-96231b3b80d8
This starts in MCJIT::getSymbolAddress where the
unique_ptr<object::Binary> is release()d and (after a cast) passed to a
single caller, MCJIT::addObjectFile.
addObjectFile calls RuntimeDyld::loadObject.
RuntimeDld::loadObject calls RuntimeDyldELF::createObjectFromFile
And the pointer is never owned at this point. I say this point, because
the alternative codepath, RuntimeDyldMachO::createObjectFile certainly
does take ownership, so this seemed like a good hint that this was a/the
right place to take ownership.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207580 91177308-0d34-0410-b5e6-96231b3b80d8
Before this patch, if 'nul' was passed in input to clang, function
getStatus() (in Path.inc) always returned an instance of file_status with
field 'nFileSizeHigh' and 'nFileSizeLow' left uninitialized.
This was causing the triggering of an assertion failure in MemoryBuffer.cpp due
to an invalid FileSize for device 'nul'.
This patch fixes the assertion failure modifying the constructors of class
file_status (in llvm/Support/FileSystem.h) so that every field of the class
gets initialized to zero by default.
A clang test will be submitted on a separate patch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207575 91177308-0d34-0410-b5e6-96231b3b80d8
Change `BlockFrequency` to defer to `BranchProbability::scale()` and
`BranchProbability::scaleByInverse()`.
This removes `BlockFrequency::scale()` from its API (and drops the
ability to see the remainder), but the only user was the unit tests. If
some code in the future needs an API that exposes the remainder, we can
add something to `BranchProbability`, but I find that unlikely.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207550 91177308-0d34-0410-b5e6-96231b3b80d8
Since `BlockMass` is an implementation detail and there are no current
users of this, delete `BlockMass::operator*=(BlockMass)`. I might need
this when I try to strip out `UnsignedFloat`, but I can pull it back in
at that point.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207546 91177308-0d34-0410-b5e6-96231b3b80d8
Add API to `BranchProbability` for scaling big integers. Next job is to
rip the logic out of `BlockMass` and `BlockFrequency`.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207544 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
This calls emitOptimizationRemark from the loop unroller and vectorizer
at the point where they make a positive transformation. For the
vectorizer, it reports vectorization and interleave factors. For the
loop unroller, it reports all the different supported types of
unrolling.
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D3456
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207528 91177308-0d34-0410-b5e6-96231b3b80d8
This patch centralizes the handling of the thumb bit around
MCStreamer::isThumbFunc and makes isThumbFunc handle aliases.
This fixes a corner case, but the main advantage is having just one
way to check if a MCSymbol is thumb or not. This should still be
refactored to be ARM only, but at least now it is just one predicate
that has to be refactored instead of 3 (isThumbFunc,
ELF_Other_ThumbFunc, and SF_ThumbFunc).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207522 91177308-0d34-0410-b5e6-96231b3b80d8
requiring full control over the various parameters to the std::iterator
concept / trait thing. This is a precursor for adjusting these things to
where you can write a bidirectional iterator wrapping a random access
iterator with custom increment and decrement logic.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207487 91177308-0d34-0410-b5e6-96231b3b80d8
When evaluating an assembly expression for a relocation, we want to
stop at MCSymbols that are in the symbol table, even if they are variables.
This is needed since the semantics may require that the relocation use them.
That is not the case when computing the value of a symbol in the symbol table.
There are no relocations in this case and we have to keep going until we hit
a section or find out that the expression doesn't have an assembly time
value.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207445 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r207287, reapplying r207286.
I'm hoping that declaring an explicit struct and instantiating
`addBlockEdges()` directly works around the GCC crash from r207286.
This is a lot more boilerplate, though.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207438 91177308-0d34-0410-b5e6-96231b3b80d8
This commit provides the necessary C/C++ APIs and infastructure to enable fine-
grain progress report and safe suspension points after each pass in the pass
manager.
Clients can provide a callback function to the pass manager to call after each
pass. This can be used in a variety of ways (progress report, dumping of IR
between passes, safe suspension of threads, etc).
The run listener list is maintained in the LLVMContext, which allows a multi-
threaded client to be only informed for it's own thread. This of course assumes
that the client created a LLVMContext for each thread.
This fixes <rdar://problem/16728690>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207430 91177308-0d34-0410-b5e6-96231b3b80d8
domtree. When finding a nearest common dominator, if neither A dominates
B nor B dominates A, we immediately resorted to a tree walk. The tree
walk here is *particularly* expensive because we have to build
a (potentially very large) set for one side's dominators and compare it
with the other side's.
If at any point we have DFS info, we don't need to do any of this. We
can just walk up one side's immediate dominators and return the first
one which dominates the other side. Because of the DFS info, the
dominates queries are trivially constant time.
This reduces the optimizers time in the test case on PR19499 by 70%. It
now optimizes in about 30 seconds for me. And there is still more to be
done for this case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207406 91177308-0d34-0410-b5e6-96231b3b80d8
This introduces a target specific streamer, X86WinCOFFStreamer, which handles
the target specific behaviour (e.g. WinEH). This is mostly to ensure that
differences between ARM and X86 remain disjoint and do not accidentally cross
boundaries. This is the final staging change for enabling object emission for
Windows on ARM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207344 91177308-0d34-0410-b5e6-96231b3b80d8
This is in preparation for promoting WinCOFFStreamer to a base class which will
be shared by the X86 and ARM specific target COFF streamers. Also add a new
getOrCreateSymbolData interface (like MCELFStreamer) for the ARM COFF Streamer.
This makes the COFFStreamer more similar to the ELFStreamer.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207343 91177308-0d34-0410-b5e6-96231b3b80d8
API requirements much more obvious.
The key here is that there are two totally different use cases for
mutating the graph. Prior to doing any SCC formation, it is very easy to
mutate the graph. There may be users that want to do small tweaks here,
and then use the already-built graph for their SCC-based operations.
This method remains on the graph itself and is documented carefully as
being cheap but unavailable once SCCs are formed.
Once SCCs are formed, and there is some in-flight DFS building them, we
have to be much more careful in how we mutate the graph. These mutation
operations are sunk onto the SCCs themselves, which both simplifies
things (the code was already there!) and helps make it obvious that
these interfaces are only applicable within that context. The other
primary constraint is that the edge being mutated is actually related to
the SCC on which we call the method. This helps make it obvious that you
cannot arbitrarily mutate some other SCC.
I've tried to write much more complete documentation for the interesting
mutation API -- intra-SCC edge removal. Currently one aspect of this
documentation is a lie (the result list of SCCs) but we also don't even
have tests for that API. =[ I'm going to add tests and fix it to match
the documentation next.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207339 91177308-0d34-0410-b5e6-96231b3b80d8
Otherwise the legalizer would just scalarize everything. Support for
mulhi in the targets isn't that great yet so on most targets we get
exactly the same scalarized output. Add a test for x86 vector udiv.
I had to disable the mulhi nodes on ARM because there aren't any patterns
for it. As far as I know ARM has instructions for getting the high part of
a multiply so this should be fixed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207315 91177308-0d34-0410-b5e6-96231b3b80d8
them, just skip over any DFS-numbered nodes when finding the next root
of a DFS. This allows the entry set to just be a vector as we populate
it from a uniqued source. It also removes the possibility for a linear
scan of the entry set to actually do the removal which can make things
go quadratic if we get unlucky.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207312 91177308-0d34-0410-b5e6-96231b3b80d8
makes working through the worklist much cleaner, and makes it possible
to avoid the 'bool-to-continue-the-outer-loop' hack. Not a huge
difference, but I think this is approaching as polished as I can make
it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207310 91177308-0d34-0410-b5e6-96231b3b80d8
processed in the DFS out of the stack completely. Keep it exclusively in
a variable. Re-shuffle some code structure to make this easier. This can
have a very dramatic effect in some cases because call graphs tend to
look like a high fan-out spanning tree. As a consequence, there are
a large number of leaf nodes in the graph, and this technique causes
leaf nodes to never even go into the stack. While this only reduces the
max depth by 1, it may cause the total number of round trips through the
stack to drop by a lot.
Now, most of this isn't really relevant for the incremental version. =]
But I wanted to prototype it first here as this variant is in ways more
complex. As long as I can get the code factored well here, I'll next
make the primary walk look the same. There are several refactorings this
exposes I think.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207306 91177308-0d34-0410-b5e6-96231b3b80d8
a helper function. Also factor the other two places where we did the
same thing into the helper function. =] Much cleaner this way. NFC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207300 91177308-0d34-0410-b5e6-96231b3b80d8
Previously, irreducible backedges were ignored. With this commit,
irreducible SCCs are discovered on the fly, and modelled as loops with
multiple headers.
This approximation specifies the headers of irreducible sub-SCCs as its
entry blocks and all nodes that are targets of a backedge within it
(excluding backedges within true sub-loops). Block frequency
calculations act as if we insert a new block that intercepts all the
edges to the headers. All backedges and entries to the irreducible SCC
point to this imaginary block. This imaginary block has an edge (with
even probability) to each header block.
The result is now reasonable enough that I've added a number of
testcases for irreducible control flow. I've outlined in
`BlockFrequencyInfoImpl.h` ways to improve the approximation.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207286 91177308-0d34-0410-b5e6-96231b3b80d8
This adds support for an -mattr option to the gold plugin and to llvm-lto. This
allows the caller to specify details of the subtarget architecture, like +aes,
or +ssse3 on x86. Note that this requires a change to the include/llvm-c/lto.h
interface: it adds a function lto_codegen_set_attr and it increments the
version of the interface.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207279 91177308-0d34-0410-b5e6-96231b3b80d8
Actually use the `reference` typedef, and remove the private
redefinition of `pointer` since it has no users.
Using `reference` exposes a problem with r207257, which specified the
wrong `value_type` to `iterator_facade_base` (fixed that too).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207270 91177308-0d34-0410-b5e6-96231b3b80d8
buildbot - do not insert debug intrinsics before phi nodes.
Debug info for optimized code: Support variables that are on the stack and
described by DBG_VALUEs during their lifetime.
Previously, when a variable was at a FrameIndex for any part of its
lifetime, this would shadow all other DBG_VALUEs and only a single
fbreg location would be emitted, which in fact is only valid for a small
range and not the entire lexical scope of the variable. The included
dbg-value-const-byref testcase demonstrates this.
This patch fixes this by
Local
- emitting dbg.value intrinsics for allocas that are passed by reference
- dropping all dbg.declares (they are now fully lowered to dbg.values)
SelectionDAG
- renamed constructors for SDDbgValue for better readability.
- fix UserValue::match() to handle indirect values correctly
- not inserting an MMI table entries for dbg.values that describe allocas.
- lowering dbg.values that describe allocas into *indirect* DBG_VALUEs.
CodeGenPrepare
- leaving dbg.values for an alloca were they are (see comment)
Other
- regenerated/updated instcombine.ll testcase and included source
rdar://problem/16679879
http://reviews.llvm.org/D3374
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207269 91177308-0d34-0410-b5e6-96231b3b80d8
Move a lot of the loop-related logic that was sprinkled around the code
into `LoopData`.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207258 91177308-0d34-0410-b5e6-96231b3b80d8
Use the fancy new `iterator_facade_base` to add
`scc_iterator::operator->()`. Remove other definitions where
`iterator_facade_base` does the right thing.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207257 91177308-0d34-0410-b5e6-96231b3b80d8
These are long functions that really shouldn't be inlined. Otherwise,
no functionality change.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207249 91177308-0d34-0410-b5e6-96231b3b80d8
Functions declared in line in a class are inlined by default. There's
no reason for the `inline` keyword.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207248 91177308-0d34-0410-b5e6-96231b3b80d8
This intrinsic is no longer needed with the new @llvm.arm.hint(i32) intrinsic
which provides a generic, extensible manner for adding hint instructions. This
functionality can now be represented as @llvm.arm.hint(i32 5).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207246 91177308-0d34-0410-b5e6-96231b3b80d8
Introduce the llvm.arm.hint(i32) intrinsic that can be used to inject hints into
the instruction stream. This is particularly useful for generating IR from a
compiler where the user may inject an intrinsic (e.g. __yield). These are then
pattern substituted into the correct instruction which already existed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207242 91177308-0d34-0410-b5e6-96231b3b80d8
AllocaInst that was missing in one location.
Debug info for optimized code: Support variables that are on the stack and
described by DBG_VALUEs during their lifetime.
Previously, when a variable was at a FrameIndex for any part of its
lifetime, this would shadow all other DBG_VALUEs and only a single
fbreg location would be emitted, which in fact is only valid for a small
range and not the entire lexical scope of the variable. The included
dbg-value-const-byref testcase demonstrates this.
This patch fixes this by
Local
- emitting dbg.value intrinsics for allocas that are passed by reference
- dropping all dbg.declares (they are now fully lowered to dbg.values)
SelectionDAG
- renamed constructors for SDDbgValue for better readability.
- fix UserValue::match() to handle indirect values correctly
- not inserting an MMI table entries for dbg.values that describe allocas.
- lowering dbg.values that describe allocas into *indirect* DBG_VALUEs.
CodeGenPrepare
- leaving dbg.values for an alloca were they are (see comment)
Other
- regenerated/updated instcombine.ll testcase and included source
rdar://problem/16679879
http://reviews.llvm.org/D3374
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207235 91177308-0d34-0410-b5e6-96231b3b80d8
Remove the concepts of "forward" and "general" mass distributions, which
was wrong. The split might have made sense in an early version of the
algorithm, but it's definitely wrong now.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207195 91177308-0d34-0410-b5e6-96231b3b80d8
Make `getPackagedNode()` a member function of
`BlockFrequencyInfoImplBase` so that it's available for templated code.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207183 91177308-0d34-0410-b5e6-96231b3b80d8
Continue refactoring to make `LoopData` first-class. Here I'm making
the `LoopData` hierarchy explicit, instead of bouncing back and forth
with `WorkingData`. This simplifies the logic and better matches the
`LoopInfo` design. (Eventually, `LoopInfo` should be restructured so
that it supports this pass, and `LoopData` can be removed.)
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207180 91177308-0d34-0410-b5e6-96231b3b80d8
As pointed out by David Blaikie in code review, a `std::list<T>` is
simpler than a `std::vector<std::unique_ptr<T>>`. Another option is a
`std::deque<T>` (which allocates in chunks), but I'd like to leave open
the option of inserting in the middle of the sequence for handling
irreducible control flow on the fly.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207177 91177308-0d34-0410-b5e6-96231b3b80d8
AllocaInst that was missing in one location.
Debug info for optimized code: Support variables that are on the stack and
described by DBG_VALUEs during their lifetime.
Previously, when a variable was at a FrameIndex for any part of its
lifetime, this would shadow all other DBG_VALUEs and only a single
fbreg location would be emitted, which in fact is only valid for a small
range and not the entire lexical scope of the variable. The included
dbg-value-const-byref testcase demonstrates this.
This patch fixes this by
Local
- emitting dbg.value intrinsics for allocas that are passed by reference
- dropping all dbg.declares (they are now fully lowered to dbg.values)
SelectionDAG
- renamed constructors for SDDbgValue for better readability.
- fix UserValue::match() to handle indirect values correctly
- not inserting an MMI table entries for dbg.values that describe allocas.
- lowering dbg.values that describe allocas into *indirect* DBG_VALUEs.
CodeGenPrepare
- leaving dbg.values for an alloca were they are (see comment)
Other
- regenerated/updated instcombine.ll testcase and included source
rdar://problem/16679879
http://reviews.llvm.org/D3374
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207165 91177308-0d34-0410-b5e6-96231b3b80d8
This is similar to the 'tail' marker, except that it guarantees that
tail call optimization will occur. It also comes with convervative IR
verification rules that ensure that tail call optimization is possible.
Reviewers: nicholas
Differential Revision: http://llvm-reviews.chandlerc.com/D3240
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207143 91177308-0d34-0410-b5e6-96231b3b80d8
rather than by adding an overload and hoping that it's declared before the code
that calls it. (In a modules build, it isn't.)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207133 91177308-0d34-0410-b5e6-96231b3b80d8
described by DBG_VALUEs during their lifetime.
Previously, when a variable was at a FrameIndex for any part of its
lifetime, this would shadow all other DBG_VALUEs and only a single
fbreg location would be emitted, which in fact is only valid for a small
range and not the entire lexical scope of the variable. The included
dbg-value-const-byref testcase demonstrates this.
This patch fixes this by
Local
- emitting dbg.value intrinsics for allocas that are passed by reference
- dropping all dbg.declares (they are now fully lowered to dbg.values)
SelectionDAG
- renamed constructors for SDDbgValue for better readability.
- fix UserValue::match() to handle indirect values correctly
- not inserting an MMI table entries for dbg.values that describe allocas.
- lowering dbg.values that describe allocas into *indirect* DBG_VALUEs.
CodeGenPrepare
- leaving dbg.values for an alloca were they are (see comment)
Other
- regenerated/updated instcombine-intrinsics testcase and included source
rdar://problem/16679879
http://reviews.llvm.org/D3374
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207130 91177308-0d34-0410-b5e6-96231b3b80d8
This patch:
- Adds two new X86 builtin intrinsics ('int_x86_rdtsc' and
'int_x86_rdtscp') as GCCBuiltin intrinsics;
- Teaches the backend how to lower the two new builtins;
- Introduces a common function to lower READCYCLECOUNTER dag nodes
and the two new rdtsc/rdtscp intrinsics;
- Improves (and extends) the existing x86 test 'rdtsc.ll'; now test 'rdtsc.ll'
correctly verifies that both READCYCLECOUNTER and the two new intrinsics
work fine for both 64bit and 32bit Subtargets.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207127 91177308-0d34-0410-b5e6-96231b3b80d8
I discovered this const-hole while attempting to coalesnce the Symbol
and SymbolMap data structures. There's some pending issues with that,
but I figured this change was easy to flush early.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207124 91177308-0d34-0410-b5e6-96231b3b80d8
algorithm here: http://dl.acm.org/citation.cfm?id=177301.
The idea of isolating the roots has even more relevance when using the
stack not just to implement the DFS but also to implement the recursive
step. Because we use it for the recursive step, to isolate the roots we
need to maintain two stacks: one for our recursive DFS walk, and another
of the nodes that have been walked. The nice thing is that the latter
will be half the size. It also fixes a complete hack where we scanned
backwards over the stack to find the next potential-root to continue
processing. Now that is always the top of the DFS stack.
While this is a really nice improvement already (IMO) it further opens
the door for two important simplifications:
1) De-duplicating some of the code across the two different walks. I've
actually made the duplication a bit worse in some senses with this
patch because the two are starting to converge.
2) Dramatically simplifying the loop structures of both walks.
I wanted to do those separately as they'll be essentially *just* CFG
restructuring. This patch on the other hand actually uses different
datastructures to implement the algorithm itself.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207098 91177308-0d34-0410-b5e6-96231b3b80d8
a SmallPtrSet. Currently, there is no need for stable iteration in this
dimension, and I now thing there won't need to be going forward.
If this is ever re-introduced in any form, it needs to not be
a SetVector based solution because removal cannot be linear. There will
be many SCCs with large numbers of parents. When encountering these, the
incremental SCC update for intra-SCC edge removal was quadratic due to
linear removal (kind of).
I'm really hoping we can avoid having an ordering property here at all
though...
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207091 91177308-0d34-0410-b5e6-96231b3b80d8
own CRTP base class for more general purpose use. Add some clarifying
comments for the exact way in which the adaptor uses it. Hopefully this
will help us write increasingly full featured iterators. This is
becoming important as they start to be used heavily inside of ranges.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207072 91177308-0d34-0410-b5e6-96231b3b80d8
Boost's iterator_adaptor, and a specific adaptor which iterates over
pointees when wrapped around an iterator over pointers.
This is the result of a long discussion on IRC with Duncan Smith, Dave
Blaikie, Richard Smith, and myself. Essentially, I could use some subset
of the iterator facade facilities often used from Boost, and everyone
seemed interested in having the functionality in a reasonably generic
form. I've tried to strike a balance between the pragmatism and the
established Boost design. The primary differences are:
1) Delegating to the standard iterator interface names rather than
special names that then make up a second iterator-like API.
2) Using the name 'pointee_iterator' which seems more clear than
'indirect_iterator'. The whole business of calling the '*p' operation
'pointer indirection' in the standard is ... quite confusing. And
'dereference' is no better of a term for moving from a pointer to
a reference.
Hoping Duncan, and others continue to provide comments on this until
we've got a nice, minimal abstraction.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207069 91177308-0d34-0410-b5e6-96231b3b80d8
than functions. So far, this access pattern is *much* more common. It
seems likely that any user of this interface is going to have nodes at
the point that they are querying the SCCs.
No functionality changed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207045 91177308-0d34-0410-b5e6-96231b3b80d8
GCOV provides an option to prepend output file names with the source
file name, to disambiguate between covered data that's included from
multiple sources. Add a flag to llvm-cov that does the same.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207035 91177308-0d34-0410-b5e6-96231b3b80d8
For now it contains a single flag, SanitizeAddress, which enables
AddressSanitizer instrumentation of inline assembly.
Patch by Yuri Gorshenin.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206971 91177308-0d34-0410-b5e6-96231b3b80d8
This implements the core functionality necessary to remove an edge from
the call graph and correctly update both the basic graph and the SCC
structure. As part of that it has to run a tiny (in number of nodes)
Tarjan-style DFS walk of an SCC being mutated to compute newly formed
SCCs, etc.
This is *very rough* and a WIP. I have a bunch of FIXMEs for code
cleanup that will reduce the boilerplate in this change substantially.
I also have a bunch of simplifications to various parts of both
algorithms that I want to make, but first I'd like to have a more
holistic picture. Ideally, I'd also like more testing. I'll probably add
quite a few more unit tests as I go here to cover the various different
aspects and corner cases of removing edges from the graph.
Still, this is, so far, successfully updating the SCC graph in-place
without disrupting the identity established for the existing SCCs even
when we do challenging things like delete the critical edge that made an
SCC cycle at all and have to reform things as a tree of smaller SCCs.
Getting this to work is really critical for the new pass manager as it
is going to associate significant state with the SCC instance and needs
it to be stable. That is also the motivation behind the return of the
newly formed SCCs. Eventually, I'll wire this all the way up to the
public API so that the pass manager can use it to correctly re-enqueue
newly formed SCCs into a fresh postorder traversal.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206968 91177308-0d34-0410-b5e6-96231b3b80d8
up the stack finishing the exploration of each entries children before
we're finished in addition to accounting for their low-links. Added
a unittest that really hammers home the need for this with interlocking
cycles that would each appear distinct otherwise and crash or compute
the wrong result. As part of this, nuke a stale fixme and bring the rest
of the implementation still more closely in line with the original
algorithm.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206966 91177308-0d34-0410-b5e6-96231b3b80d8
parents of an SCC, and add a lookup method for finding the SCC for
a given function. These aren't used yet, but will be used shortly in
some unit tests I'm adding and are really part of the broader intended
interface for the analysis.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206959 91177308-0d34-0410-b5e6-96231b3b80d8
resisted this for too long. Just with the basic testing here I was able
to exercise the analysis in more detail and sift out both type signature
bugs in the API and a bug in the DFS numbering. All of these are fixed
here as well.
The unittests will be much more important for the mutation support where
it is necessary to craft minimal mutations and then inspect the state of
the graph. There is just no way to do that with a standard FileCheck
test. However, unittesting these kinds of analyses is really quite easy,
especially as they're designed with the new pass manager where there is
essentially no infrastructure required to rig up the core logic and
exercise it at an API level.
As a minor aside about the DFS numbering bug, the DFS numbering used in
LCG is a bit unusual. Rather than numbering from 0, we number from 1,
and use 0 as the sentinel "unvisited" state. Other implementations often
use '-1' for this, but I find it easier to deal with 0 and it shouldn't
make any real difference provided someone doesn't write silly bugs like
forgetting to actually initialize the DFS numbering. Oops. ;]
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206954 91177308-0d34-0410-b5e6-96231b3b80d8
into a helper function. I plan to re-use it for doing incremental
DFS-based updates to the SCCs when we mutate the call graph.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206948 91177308-0d34-0410-b5e6-96231b3b80d8
the Callee list. This is going to be quite important to prevent removal
from going quadratic. No functionality changed at this point, this is
one of the refactoring patches I've broken out of my initial work toward
mutation updates of the call graph.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206938 91177308-0d34-0410-b5e6-96231b3b80d8
from places like MCCodeEmitter() in the MC backend when the
MCContext is const.
I was going to use this in my change for r206669 but Jim convinced
me to use an assert there. But this still is a good tweak.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206923 91177308-0d34-0410-b5e6-96231b3b80d8
r206916 was not logically the same as the previous code because the
goto statements did not create loop. This should be the same as the
previous code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206918 91177308-0d34-0410-b5e6-96231b3b80d8
Goto statements jumping into previous inner blocks are pretty confusing
to read even though in this case they are valid. No reason to not use
while loops there.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206916 91177308-0d34-0410-b5e6-96231b3b80d8
diagnostic that includes location information.
Currently if one has this assembly:
.quad (0x1234 + (4 * SOME_VALUE))
where SOME_VALUE is undefined ones gets the less than
useful error message with no location information:
% clang -c x.s
clang -cc1as: fatal error: error in backend: expected relocatable expression
With this fix one now gets a more useful error message
with location information:
% clang -c x.s
x.s:5:8: error: expected relocatable expression
.quad (0x1234 + (4 * SOME_VALUE))
^
To do this I plumbed the SMLoc through the MCObjectStreamer
EmitValue() and EmitValueImpl() interfaces so it could be used
when creating the MCFixup.
rdar://12391022
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206906 91177308-0d34-0410-b5e6-96231b3b80d8
Store pointers directly to loops inside the nodes. This could have been
done without changing the type stored in `std::vector<>`. However,
rather than computing the number of loops before constructing them
(which `LoopInfo` doesn't provide directly), I've switched to a
`vector<unique_ptr<LoopData>>`.
This adds some heap overhead, but the number of loops is typically
small.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206857 91177308-0d34-0410-b5e6-96231b3b80d8
define below all header includes in the lib/CodeGen/... tree. While the
current modules implementation doesn't check for this kind of ODR
violation yet, it is likely to grow support for it in the future. It
also removes one layer of macro pollution across all the included
headers.
Other sub-trees will follow.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206837 91177308-0d34-0410-b5e6-96231b3b80d8
ELFEntityIterator does not implement RandomAccessIterator. It does
not even implement BidirectionalIterator.
This patch fixes LLD build issue when compiled with MSVC2013 with
debug: MSVC's find_if checks if the start iterator is before the end
iterator in the sense of operator< if it declares implementing
RandomAccessIterator. If a class does not have operator<, it fails
to compile.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206825 91177308-0d34-0410-b5e6-96231b3b80d8
behavior based on other files defining DEBUG_TYPE, which means it cannot
define DEBUG_TYPE at all. This is actually better IMO as it forces folks
to define relevant DEBUG_TYPEs for their files. However, it requires all
files that currently use DEBUG(...) to define a DEBUG_TYPE if they don't
already. I've updated all such files in LLVM and will do the same for
other upstream projects.
This still leaves one important change in how LLVM uses the DEBUG_TYPE
macro going forward: we need to only define the macro *after* header
files have been #include-ed. Previously, this wasn't possible because
Debug.h required the macro to be pre-defined. This commit removes that.
By defining DEBUG_TYPE after the includes two things are fixed:
- Header files that need to provide a DEBUG_TYPE for some inline code
can do so by defining the macro before their inline code and undef-ing
it afterward so the macro does not escape.
- We no longer have rampant ODR violations due to including headers with
different DEBUG_TYPE definitions. This may be mostly an academic
violation today, but with modules these types of violations are easy
to check for and potentially very relevant.
Where necessary to suppor headers with DEBUG_TYPE, I have moved the
definitions below the includes in this commit. I plan to move the rest
of the DEBUG_TYPE macros in LLVM in subsequent commits; this one is big
enough.
The comments in Debug.h, which were hilariously out of date already,
have been updated to reflect the recommended practice going forward.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206822 91177308-0d34-0410-b5e6-96231b3b80d8
Change `PositiveFloat` to `UnsignedFloat`, and fix some of the comments
to indicate that it's disappearing eventually.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206771 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r206707, reapplying r206704. The preceding commit
to CalcSpillWeights should have sorted out the failing buildbots.
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206766 91177308-0d34-0410-b5e6-96231b3b80d8
We normally don't drop functions from the C API's, but in this case I think we
can:
* The old implementation of getFileOffset was fairly broken
* The introduction of LLVMGetSymbolFileOffset was itself a C api breaking
change as it removed LLVMGetSymbolOffset.
* It is an incredibly specialized use case. The only reason MCJIT needs it is
because of its odd position of being a dynamic linker of .o files.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206750 91177308-0d34-0410-b5e6-96231b3b80d8
LazyCallGraph analysis framework. Wire it up all the way through the opt
driver and add some very basic testing that we can build pass pipelines
including these components. Still a lot more to do in terms of testing
that all of this works, but the basic pieces are here.
There is a *lot* of boiler plate here. It's something I'm going to
actively look at reducing, but I don't have any immediate ideas that
don't end up making the code terribly complex in order to fold away the
boilerplate. Until I figure out something to minimize the boilerplate,
almost all of this is based on the code for the existing pass managers,
copied and heavily adjusted to suit the needs of the CGSCC pass
management layer.
The actual CG management still has a bunch of FIXMEs in it. Notably, we
don't do *any* updating of the CG as it is potentially invalidated.
I wanted to get this in place to motivate the new analysis, and add
update APIs to the analysis and the pass management layers in concert to
make sure that the *right* APIs are present.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206745 91177308-0d34-0410-b5e6-96231b3b80d8
It could even be made non-virtual if it weren't for bad compiler
warnings.
This demonstrates that ArgList objects aren't destroyed polymorphically
and possibly that they aren't even used polymorphically. If that's the
case, it might be possible to refactor the two ArgList types more
separately and simplify the Arg ownership model. *continues
experimenting*
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206727 91177308-0d34-0410-b5e6-96231b3b80d8
This might be able to be simplified further by using Arg as a value type
in a linked list (to maintain pointer validity), but here's something
simple to start with.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206724 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r206677, reapplying my BlockFrequencyInfo rewrite.
I've done a careful audit, added some asserts, and fixed a couple of
bugs (unfortunately, they were in unlikely code paths). There's a small
chance that this will appease the failing bots [1][2]. (If so, great!)
If not, I have a follow-up commit ready that will temporarily add
-debug-only=block-freq to the two failing tests, allowing me to compare
the code path between what the failing bots and what my machines (and
the rest of the bots) are doing. Once I've triggered those builds, I'll
revert both commits so the bots go green again.
[1]: http://bb.pgr.jp/builders/ninja-x64-msvc-RA-centos6/builds/1816
[2]: http://llvm-amd64.freebsd.your.org/b/builders/clang-i386-freebsd/builds/18445
<rdar://problem/14292693>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206704 91177308-0d34-0410-b5e6-96231b3b80d8
Win64 stack unwinder gets confused when execution flow "falls through" after
a call to 'noreturn' function. This fixes the "missing epilogue" problem by
emitting a trap instruction for IR 'unreachable' on x86_x64-pc-windows.
A secondary use for it would be for anyone wanting to make double-sure that
'noreturn' functions, indeed, do not return.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206684 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r206666, as planned.
Still stumped on why the bots are failing. Sanitizer bots haven't
turned anything up. If anyone can help me debug either of the failures
(referenced in r206666) I'll owe them a beer. (In the meantime, I'll be
auditing my patch for undefined behaviour.)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206677 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r206628, reapplying r206622 (and r206626).
Two tests are failing only on buildbots [1][2]: i.e., I can't reproduce
on Darwin, and Chandler can't reproduce on Linux. Asan and valgrind
don't tell us anything, but we're hoping the msan bot will catch it.
So, I'm applying this again to get more feedback from the bots. I'll
leave it in long enough to trigger builds in at least the sanitizer
buildbots (it was failing for reasons unrelated to my commit last time
it was in), and hopefully a few others.... and then I expect to revert a
third time.
[1]: http://bb.pgr.jp/builders/ninja-x64-msvc-RA-centos6/builds/1816
[2]: http://llvm-amd64.freebsd.your.org/b/builders/clang-i386-freebsd/builds/18445
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206666 91177308-0d34-0410-b5e6-96231b3b80d8
This adds support for an indexed instrumentation based profiling
format, which is just a small header and an on disk hash table. This
format will be used by clang's -fprofile-instr-use= for PGO.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206656 91177308-0d34-0410-b5e6-96231b3b80d8
Immutable DILineInfo doesn't bring any benefits and complicates
code. Also, use std::string instead of SmallString<16> for file
and function names - their length can vary significantly.
No functionality change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206654 91177308-0d34-0410-b5e6-96231b3b80d8
This changes the on-disk hash to get the type to use for offsets from
the Info type, so that clients can be more flexible with the size of
table they support.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206643 91177308-0d34-0410-b5e6-96231b3b80d8
This changes the on-disk hash to get the size of a hash value from the
Info type, so that clients can be more flexible with the types of hash
they use.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206642 91177308-0d34-0410-b5e6-96231b3b80d8
This warning is disabled for the LLVM build,
but external users of the header can still
run into this.
Patch by Ke Bai
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206629 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r206622 and the MSVC fixup in r206626.
Apparently the remotely failing tests are still failing, despite my
attempt to fix the nondeterminism in r206621.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206628 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r206556, effectively reapplying commit r206548 and
its fixups in r206549 and r206550.
In an intervening commit I've added target triples to the tests that
were failing remotely [1] (but passing locally). I'm hoping the mystery
is solved? I'll revert this again if the tests are still failing
remotely.
[1]: http://bb.pgr.jp/builders/ninja-x64-msvc-RA-centos6/builds/1816
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206622 91177308-0d34-0410-b5e6-96231b3b80d8
Doesn't make sense to restrict this to BumpPtrAllocator. While there
replace an explicit loop with std::equal. Some standard libraries know
how to compile this down to a ::memcmp call if possible.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206615 91177308-0d34-0410-b5e6-96231b3b80d8
Reality is that we're never going to copy one of these. Supporting this
was becoming a nightmare because nothing even causes it to compile most
of the time. Lots of subtle errors built up that wouldn't have been
caught by any "normal" testing.
Also, make the move assignment actually work rather than the bogus swap
implementation that would just infloop if used. As part of that, factor
out the graph pointer updates into a helper to share between move
construction and move assignment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206583 91177308-0d34-0410-b5e6-96231b3b80d8
implementation of the SpecificBumpPtrAllocator -- we have to actually
move the subobject. =] Noticed when using this code more directly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206582 91177308-0d34-0410-b5e6-96231b3b80d8
LazyCallGraph. This is the start of the whole point of this different
abstraction, but it is just the initial bits. Here is a run-down of
what's going on here. I'm planning to incorporate some (or all) of this
into comments going forward, hopefully with better editing and wording.
=]
The crux of the problem with the traditional way of building SCCs is
that they are ephemeral. The new pass manager however really needs the
ability to associate analysis passes and results of analysis passes with
SCCs in order to expose these analysis passes to the SCC passes. Making
this work is kind-of the whole point of the new pass manager. =]
So, when we're building SCCs for the call graph, we actually want to
build persistent nodes that stick around and can be reasoned about
later. We'd also like the ability to walk the SCC graph in more complex
ways than just the traditional postorder traversal of the current CGSCC
walk. That means that in addition to being persistent, the SCCs need to
be connected into a useful graph structure.
However, we still want the SCCs to be formed lazily where possible.
These constraints are quite hard to satisfy with the SCC iterator. Also,
using that would bypass our ability to actually add data to the nodes of
the call graph to facilite implementing the Tarjan walk. So I've
re-implemented things in a more direct and embedded way. This
immediately makes it easy to get the persistence and connectivity
correct, and it also allows leveraging the existing nodes to simplify
the algorithm. I've worked somewhat to make this implementation more
closely follow the traditional paper's nomenclature and strategy,
although it is still a bit obtuse because it isn't recursive, using
an explicit stack and a tail call instead, and it is interruptable,
resuming each time we need another SCC.
The other tricky bit here, and what actually took almost all the time
and trials and errors I spent building this, is exactly *what* graph
structure to build for the SCCs. The naive thing to build is the call
graph in its newly acyclic form. I wrote about 4 versions of this which
did precisely this. Inevitably, when I experimented with them across
various use cases, they became incredibly awkward. It was all
implementable, but it felt like a complete wrong fit. Square peg, round
hole. There were two overriding aspects that pushed me in a different
direction:
1) We want to discover the SCC graph in a postorder fashion. That means
the root node will be the *last* node we find. Using the call-SCC DAG
as the graph structure of the SCCs results in an orphaned graph until
we discover a root.
2) We will eventually want to walk the SCC graph in parallel, exploring
distinct sub-graphs independently, and synchronizing at merge points.
This again is not helped by the call-SCC DAG structure.
The structure which, quite surprisingly, ended up being completely
natural to use is the *inverse* of the call-SCC DAG. We add the leaf
SCCs to the graph as "roots", and have edges to the caller SCCs. Once
I switched to building this structure, everything just fell into place
elegantly.
Aside from general cleanups (there are FIXMEs and too few comments
overall) that are still needed, the other missing piece of this is
support for iterating across levels of the SCC graph. These will become
useful for implementing #2, but they aren't an immediate priority.
Once SCCs are in good shape, I'll be working on adding mutation support
for incremental updates and adding the pass manager that this analysis
enables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206581 91177308-0d34-0410-b5e6-96231b3b80d8
Previously module verification was always enabled, with no way to turn it off.
As of this commit, module verification is on by default in Debug builds, and off
by default in release builds. The default behaviour can be overridden by calling
setVerifyModules(bool) on the JIT instance (this works for both the old JIT, and
MCJIT).
<rdar://problem/16150008>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206561 91177308-0d34-0410-b5e6-96231b3b80d8
Rewrite the shared implementation of BlockFrequencyInfo and
MachineBlockFrequencyInfo entirely.
The old implementation had a fundamental flaw: precision losses from
nested loops (or very wide branches) compounded past loop exits (and
convergence points).
The @nested_loops testcase at the end of
test/Analysis/BlockFrequencyAnalysis/basic.ll is motivating. This
function has three nested loops, with branch weights in the loop headers
of 1:4000 (exit:continue). The old analysis gives non-sensical results:
Printing analysis 'Block Frequency Analysis' for function 'nested_loops':
---- Block Freqs ----
entry = 1.0
for.cond1.preheader = 1.00103
for.cond4.preheader = 5.5222
for.body6 = 18095.19995
for.inc8 = 4.52264
for.inc11 = 0.00109
for.end13 = 0.0
The new analysis gives correct results:
Printing analysis 'Block Frequency Analysis' for function 'nested_loops':
block-frequency-info: nested_loops
- entry: float = 1.0, int = 8
- for.cond1.preheader: float = 4001.0, int = 32007
- for.cond4.preheader: float = 16008001.0, int = 128064007
- for.body6: float = 64048012001.0, int = 512384096007
- for.inc8: float = 16008001.0, int = 128064007
- for.inc11: float = 4001.0, int = 32007
- for.end13: float = 1.0, int = 8
Most importantly, the frequency leaving each loop matches the frequency
entering it.
The new algorithm leverages BlockMass and PositiveFloat to maintain
precision, separates "probability mass distribution" from "loop
scaling", and uses dithering to eliminate probability mass loss. I have
unit tests for these types out of tree, but it was decided in the review
to make the classes private to BlockFrequencyInfoImpl, and try to shrink
them (or remove them entirely) in follow-up commits.
The new algorithm should generally have a complexity advantage over the
old. The previous algorithm was quadratic in the worst case. The new
algorithm is still worst-case quadratic in the presence of irreducible
control flow, but it's linear without it.
The key difference between the old algorithm and the new is that control
flow within a loop is evaluated separately from control flow outside,
limiting propagation of precision problems and allowing loop scale to be
calculated independently of mass distribution. Loops are visited
bottom-up, their loop scales are calculated, and they are replaced by
pseudo-nodes. Mass is then distributed through the function, which is
now a DAG. Finally, loops are revisited top-down to multiply through
the loop scales and the masses distributed to pseudo nodes.
There are some remaining flaws.
- Irreducible control flow isn't modelled correctly. LoopInfo and
MachineLoopInfo ignore irreducible edges, so this algorithm will
fail to scale accordingly. There's a note in the class
documentation about how to get closer. See also the comments in
test/Analysis/BlockFrequencyInfo/irreducible.ll.
- Loop scale is limited to 4096 per loop (2^12) to avoid exhausting
the 64-bit integer precision used downstream.
- The "bias" calculation proposed on llvmdev is *not* incorporated
here. This will be added in a follow-up commit, once comments from
this review have been handled.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206548 91177308-0d34-0410-b5e6-96231b3b80d8
