allocated by setupterm. Without this, some folks are seeing leaked
memory whenever this routine is called more than once. Thanks to Craig
Topper for the report.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188615 91177308-0d34-0410-b5e6-96231b3b80d8
extremely subtle miscompilations (such as a load getting replaced with
the value stored *below* the load within a basic block) related to
promoting an alloca to an SSA value, there is the dim possibility that
you hit this. Please let me know if you won this unfortunate lottery.
The first half of mem2reg's core logic (as it is used both in the
standalone mem2reg pass and in SROA) builds up a mapping from
'Instruction *' to the index of that instruction within its basic block.
This allows quickly establishing which store dominate a particular load
even for large basic blocks. We cache this information throughout the
run of mem2reg over a function in order to amortize the cost of
computing it.
This is not in and of itself a strange pattern in LLVM. However, it
introduces a very important constraint: absolutely no instruction can be
deleted from the program without updating the mapping. Otherwise a newly
allocated instruction might get the same pointer address, and then end
up with a wrong index. Yes, LLVM routinely suffers from a *single
threaded* variant of the ABA problem. Most places in LLVM don't find
avoiding this an imposition because they don't both delete and create
new instructions iteratively, but mem2reg *loves* to do this... All the
time. Fortunately, the mem2reg code was really careful about updating
this cache to handle this eventuallity... except when it comes to the
debug declare intrinsic. Oops. The fix is to invalidate that pointer in
the cache when we delete it, the same as we do when deleting alloca
instructions and other instructions.
I've also caused the same bug in new code while working on a fix to
PR16867, so this seems to be a really unfortunate pattern. Hopefully in
subsequent patches the deletion of dead instructions can be consolidated
sufficiently to make it less likely that we'll see future occurences of
this bug.
Sorry for not having a test case, but I have literally no idea how to
reliably trigger this kind of thing. It may be single-threaded, but it
remains an ABA problem. It would require a really amazing number of
stars to align.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188367 91177308-0d34-0410-b5e6-96231b3b80d8
curses.h). Finding these headers is next to impossible. For example, on
Debian systems libtinfo-dev provides the terminfo reading library we
want, but *not* term.h. For the header, you have to use libncurses-dev.
And libncursesw-dev provides a *different* term.h in a different
location!
These headers aren't worth it. We want two functions the signatures of
which are clearly spec'ed in sys-v and other documentation. Just declare
them ourselves and call them. This should fix some debian builders and
provide better support for "minimal" debian systems that do want color
autodetection.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188165 91177308-0d34-0410-b5e6-96231b3b80d8
library for color support detection. This still will use a curses
library if that is all we have available on the system. This change
tries to use a smaller subset of the curses library, specifically the
subset that is on some systems split off into a separate library. For
example, if you install ncurses configured --with-tinfo, a 'libtinfo' is
install that provides just the terminfo querying functionality. That
library is now used instead of curses when it is available.
This happens to fix a build error on systems with that library because
when we tried to link ncurses into the binary, we didn't pull tinfo in
as well. =]
It should also provide an easy path for supporting the NetBSD
libterminfo library, but as I don't have access to a NetBSD system I'm
leaving adding that support to those folks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188160 91177308-0d34-0410-b5e6-96231b3b80d8
SROA-based analysis has enough information. This should work now that
both mem2reg *and* the SSAUpdater-based AllocaPromoter have been updated
to be able to promote the types of allocas that the SROA analysis
detects.
I've included tests for the AllocaPromoter that were only possible to
write once we fast-tracked promotable allocas without rewriting them.
This includes a test both for r187347 and r188145.
Original commit log for r187323:
"""
Now that mem2reg understands how to cope with a slightly wider set of uses of
an alloca, we can pre-compute promotability while analyzing an alloca for
splitting in SROA. That lets us short-circuit the common case of a bunch of
trivially promotable allocas. This cuts 20% to 30% off the run time of SROA for
typical frontend-generated IR sequneces I'm seeing. It gets the new SROA to
within 20% of ScalarRepl for such code. My current benchmark for these numbers
is PR15412, but it fits the general pattern of IR emitted by Clang so it should
be widely applicable.
"""
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188146 91177308-0d34-0410-b5e6-96231b3b80d8
the more general set of patterns that are now handled by mem2reg and that we
can detect quickly while doing SROA's initial analysis. Notably, this allows it
to promote through no-op bitcast and GEP sequences. A core part of the
SSAUpdater approach is the ability to test whether a particular instruction is
part of the set being promoted. Testing this becomes significantly more complex
in the world where the operand to every load and store isn't the alloca itself.
I ended up using the approach of walking up the def-chain until we find the
alloca. I benchmarked this against keeping a set of pointer operands and
keeping a set of the loads and stores we care about, and this one seemed faster
although the difference was very small.
No test case yet because currently the rewriting always "fixes" the inputs to
not require this. The next patch which re-enables early promotion of easy cases
in SROA will include a test case that specifically exercises this aspect of the
alloca promoter.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188145 91177308-0d34-0410-b5e6-96231b3b80d8
our visiting datastructures in the AllocaPromoter/SSAUpdater path of
SROA. Also shift the order if clears around to be more consistent.
No functionality changed here, this is just a cleanup.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188144 91177308-0d34-0410-b5e6-96231b3b80d8
LLVMConfig.cmake file that is (I think) used in the stand-alone Clang
build, and causing link errors there w.r.t. curses.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187950 91177308-0d34-0410-b5e6-96231b3b80d8
r187874 seems to have been missed by the build bot infrastructure, and
the subsequent commits to compiler-rt don't seem to be queuing up new
build requsets. Hopefully this will.
As it happens, having the space here is the more common formatting. =]
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187879 91177308-0d34-0410-b5e6-96231b3b80d8
using it to detect whether or not a terminal supports colors. This
replaces a particularly egregious hack that merely compared the TERM
environment variable to "dumb". That doesn't really translate to
a reasonable experience for users that have actually ensured their
terminal's capabilities are accurately reflected.
This makes testing a terminal for color support somewhat more expensive,
but it is called very rarely anyways. The important fast path when the
output is being piped somewhere is already in place.
The global lock may seem excessive, but the spec for calling into curses
is *terrible*. The whole library is terrible, and I spent quite a bit of
time looking for a better way of doing this before convincing myself
that this was the fundamentally correct way to behave. The damage of the
curses library is very narrowly confined, and we continue to use raw
escape codes for actually manipulating the colors which is a much sane
system than directly using curses here (IMO).
If this causes trouble for folks, please let me know. I've tested it on
Linux and will watch the bots carefully. I've also worked to account for
the variances of curses interfaces that I could finde documentation for,
but that may not have been sufficient.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187874 91177308-0d34-0410-b5e6-96231b3b80d8
infrastructure to do promotion without a domtree the same smarts about
looking through GEPs, bitcasts, etc., that I just taught mem2reg about.
This way, if SROA chooses to promote an alloca which still has some
noisy instructions this code can cope with them.
I've not used as principled of an approach here for two reasons:
1) This code doesn't really need it as we were already set up to zip
through the instructions used by the alloca.
2) I view the code here as more of a hack, and hopefully a temporary one.
The SSAUpdater path in SROA is a real sore point for me. It doesn't make
a lot of architectural sense for many reasons:
- We're likely to end up needing the domtree anyways in a subsequent
pass, so why not compute it earlier and use it.
- In the future we'll likely end up needing the domtree for parts of the
inliner itself.
- If we need to we could teach the inliner to preserve the domtree. Part
of the re-work of the pass manager will allow this to be very powerful
even in large SCCs with many functions.
- Ultimately, computing a domtree has gotten significantly faster since
the original SSAUpdater-using code went into ScalarRepl. We no longer
use domfrontiers, and much of domtree is lazily done based on queries
rather than eagerly.
- At this point keeping the SSAUpdater-based promotion saves a total of
0.7% on a build of the 'opt' tool for me. That's not a lot of
performance given the complexity!
So I'm leaving this a bit ugly in the hope that eventually we just
remove all of this nonsense.
I can't even readily test this because this code isn't reachable except
through SROA. When I re-instate the patch that fast-tracks allocas
already suitable for promotion, I'll add a testcase there that failed
before this change. Before that, SROA will fix any test case I give it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187347 91177308-0d34-0410-b5e6-96231b3b80d8
standards for LLVM. Remove duplicated comments on the interface from the
implementation file (implementation comments are left there of course).
Also clean up, re-word, and fix a few typos and errors in the commenst
spotted along the way.
This is in preparation for changes to these files and to keep the
uninteresting tidying in a separate commit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187335 91177308-0d34-0410-b5e6-96231b3b80d8
uses of an alloca, we can pre-compute promotability while analyzing an
alloca for splitting in SROA. That lets us short-circuit the common case
of a bunch of trivially promotable allocas. This cuts 20% to 30% off the
run time of SROA for typical frontend-generated IR sequneces I'm seeing.
It gets the new SROA to within 20% of ScalarRepl for such code. My
current benchmark for these numbers is PR15412, but it fits the general
pattern of IR emitted by Clang so it should be widely applicable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187323 91177308-0d34-0410-b5e6-96231b3b80d8
useful in a subsequent patch, but causes an unfortunate amount of noise,
so I pulled it out into a separate patch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187322 91177308-0d34-0410-b5e6-96231b3b80d8
do in the SDag when lowering references to the GOT: use
ARMConstantPoolSymbol rather than creating a dummy global variable. The
computation of the alignment still feels weird (it uses IR types and
datalayout) but it preserves the exact previous behavior. This change
fixes the memory leak of the global variable detected on the valgrind
leak checking bot.
Thanks to Benjamin Kramer for pointing me at ARMConstantPoolSymbol to
handle this use case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187303 91177308-0d34-0410-b5e6-96231b3b80d8
me) should start watching this bot more as its catching lots of bugs.
The fix here is to not construct the global if we aren't going to need
it. That's cheaper anyways, and globals have highly predictable types in
practice. I've added an assert to catch skew between our manual testing
of the type and the actual type just for paranoia's sake.
Note that this pattern is actually fine in most globals because when you
build a global with a module it automatically is moved to be owned by
that module. But here, we're in isel and don't really want to do that.
The solution of not creating a global is simpler anyways.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187302 91177308-0d34-0410-b5e6-96231b3b80d8
There doesn't appear to be any reason to put this variable on the heap.
I'm suspicious of the LexicalScope above that we stuff in a map and then
delete afterward, but I'm just trying to get the valgrind bot clean.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187301 91177308-0d34-0410-b5e6-96231b3b80d8
than once, and the second time through we leaked memory. Found thanks to
the vg-leak bot, but I can't locally reproduce it with valgrind. The
debugger confirms that it is in fact leaking here.
This whole code is totally gross. Why is initialize being called on each
runOnFunction??? Why aren't these OwningPtr<>s, and why aren't their
lifetimes better defined? Anyways, this is just a surgical change to
help out the leak checking bots.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187299 91177308-0d34-0410-b5e6-96231b3b80d8
their being optimized out in debug mode. Realistically, this just isn't
going to be the slow part anyways. This also fixes unused variable
warnings that are breaking LLD build bots. =/ I didn't see these at
first, and kept losing track of the fact that they were broken.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187297 91177308-0d34-0410-b5e6-96231b3b80d8
analysis of the alloca. We don't need to visit all the users twice for
this. We build up a kill list during the analysis and then just process
it afterward. This recovers the tiny bit of performance lost by moving
to the visitor based analysis system as it removes one entire use-list
walk from mem2reg. In some cases, this is now faster than mem2reg was
previously.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187296 91177308-0d34-0410-b5e6-96231b3b80d8
robust. It now uses an InstVisitor and worklist to actually walk the
uses of the Alloca transitively and detect the pattern which we can
directly promote: loads & stores of the whole alloca and instructions we
can completely ignore.
Also, with this new implementation teach both the predicate for testing
whether we can promote and the promotion engine itself to use the same
code so we no longer have strange divergence between the two code paths.
I've added some silly test cases to demonstrate that we can handle
slightly more degenerate code patterns now. See the below for why this
is even interesting.
Performance impact: roughly 1% regression in the performance of SROA or
ScalarRepl on a large C++-ish test case where most of the allocas are
basically ready for promotion. The reason is because of silly redundant
work that I've left FIXMEs for and which I'll address in the next
commit. I wanted to separate this commit as it changes the behavior.
Once the redundant work in removing the dead uses of the alloca is
fixed, this code appears to be faster than the old version. =]
So why is this useful? Because the previous requirement for promotion
required a *specific* visit pattern of the uses of the alloca to verify:
we *had* to look for no more than 1 intervening use. The end goal is to
have SROA automatically detect when an alloca is already promotable and
directly hand it to the mem2reg machinery rather than trying to
partition and rewrite it. This is a 25% or more performance improvement
for SROA, and a significant chunk of the delta between it and
ScalarRepl. To get there, we need to make mem2reg actually capable of
promoting allocas which *look* promotable to SROA without have SROA do
tons of work to massage the code into just the right form.
This is actually the tip of the iceberg. There are tremendous potential
savings we can realize here by de-duplicating work between mem2reg and
SROA.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187191 91177308-0d34-0410-b5e6-96231b3b80d8
schedule an alloca for another iteration in SROA. This only showed up
with a mixture of promotable and unpromotable selects and phis. Added
a test case for this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187031 91177308-0d34-0410-b5e6-96231b3b80d8
pending speculation for a phi node. The problem here is that we were
using growth of the specluation set as an indicator of whether
speculation would occur, and if the phi node is already in the set we
don't see it grow. This is a symptom of the fact that this signal is
a total hack.
Unfortunately, I couldn't really come up with a non-hacky way of
signaling that promotion remains valid *after* speculation occurs, such
that we only speculate when all else looks good for promotion. In the
end, I went with at least a much more explicit approach of doing the
work of queuing inside the phi and select processing and setting
a preposterously named flag to convey that we're in the special state of
requiring speculating before promotion.
Thanks to Richard Trieu and Nick Lewycky for the excellent work reducing
a testcase for this from a pretty giant, nasty assert in a big
application. =] The testcase was excellent.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187029 91177308-0d34-0410-b5e6-96231b3b80d8
helper function. This leaves both trivial cases handled entirely in
helper functions and merely manages the list of allocas to process in
the run method.
The next step will be to handle all of the trivial promotion work prior
to even creating the core class and the subsequent simplifications that
enables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186784 91177308-0d34-0410-b5e6-96231b3b80d8
a single block into the helper routine. This takes advantage of the fact
that we can directly replace uses prior to any store with undef to
simplify matters and unconditionally promote allocas only used within
one block.
I've removed the special handling for the case of no stores existing.
This has no semantic effect but might slow things down. I'll fix that in
a later patch when I refactor this entire thing to be easier to manage
the different cases.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186783 91177308-0d34-0410-b5e6-96231b3b80d8
handles the general cases.
The hope is to refactor this so that we don't end up building the entire
class for the trivial cases. I also want to lift a lot of the early
pre-processing in the initial segment of run() into a separate routine,
and really none of it needs to happen inside the primary promotion
class.
These routines in particular used none of the actual state in the
promotion class, so they don't really make sense as members.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186781 91177308-0d34-0410-b5e6-96231b3b80d8
This struct is nicely independent of everything else, and we already
needed a foward declaration here. It's simpler to just define it
immediately.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186780 91177308-0d34-0410-b5e6-96231b3b80d8
implementation of the SROA algorithm. We were using the term 'partition'
in many places that no longer ever represented an actual partition, but
rather just an arbitrary slice of an alloca.
No functionality change intended here. Mostly just renaming of types,
functions, variables, and rewording of comments. Several comments were
rewritten to make a lot more sense in the new structure of things.
The stats are still weird and not reflective of how this really works.
I'll fix those up in a separate patch as it is a touch more semantic of
a change...
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186659 91177308-0d34-0410-b5e6-96231b3b80d8
SROA.
The crux of the issue is that now we track uses of a partition of the
alloca in two places: the iterators over the partitioning uses and the
previously collected split uses vector. We weren't accounting for the
fact that the split uses might invalidate integer widening in ways other
than due to their width (in this case due to being volatile).
Further reduced testcase added to the tests.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186655 91177308-0d34-0410-b5e6-96231b3b80d8
end of a vector. This was found with ASan. I've had one other report of
a crasher, but thus far been unable to reproduce the crash. It may well
be fixed with this version, and if not I'd like to get more information
from the build bots about what is happening.
See r186316 for the full commit log for the new implementation of the
SROA algorithm.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186565 91177308-0d34-0410-b5e6-96231b3b80d8
a bot.
This reverts the commit which introduced a new implementation of the
fancy SROA pass designed to reduce its overhead. I'll skip the huge
commit log here, refer to r186316 if you're looking for how this all
works and why it works that way.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186332 91177308-0d34-0410-b5e6-96231b3b80d8
different core implementation strategy.
Previously, SROA would build a relatively elaborate partitioning of an
alloca, associate uses with each partition, and then rewrite the uses of
each partition in an attempt to break apart the alloca into chunks that
could be promoted. This was very wasteful in terms of memory and compile
time because regardless of how complex the alloca or how much we're able
to do in breaking it up, all of the datastructure work to analyze the
partitioning was done up front.
The new implementation attempts to form partitions of the alloca lazily
and on the fly, rewriting the uses that make up that partition as it
goes. This has a few significant effects:
1) Much simpler data structures are used throughout.
2) No more double walk of the recursive use graph of the alloca, only
walk it once.
3) No more complex algorithms for associating a particular use with
a particular partition.
4) PHI and Select speculation is simplified and happens lazily.
5) More precise information is available about a specific use of the
alloca, removing the need for some side datastructures.
Ultimately, I think this is a much better implementation. It removes
about 300 lines of code, but arguably removes more like 500 considering
that some code grew in the process of being factored apart and cleaned
up for this all to work.
I've re-used as much of the old implementation as possible, which
includes the lion's share of code in the form of the rewriting logic.
The interesting new logic centers around how the uses of a partition are
sorted, and split into actual partitions.
Each instruction using a pointer derived from the alloca gets
a 'Partition' entry. This name is totally wrong, but I'll do a rename in
a follow-up commit as there is already enough churn here. The entry
describes the offset range accessed and the nature of the access. Once
we have all of these entries we sort them in a very specific way:
increasing order of begin offset, followed by whether they are
splittable uses (memcpy, etc), followed by the end offset or whatever.
Sorting by splittability is important as it simplifies the collection of
uses into a partition.
Once we have these uses sorted, we walk from the beginning to the end
building up a range of uses that form a partition of the alloca.
Overlapping unsplittable uses are merged into a single partition while
splittable uses are broken apart and carried from one partition to the
next. A partition is also introduced to bridge splittable uses between
the unsplittable regions when necessary.
I've looked at the performance PRs fairly closely. PR15471 no longer
will even load (the module is invalid). Not sure what is up there.
PR15412 improves by between 5% and 10%, however it is nearly impossible
to know what is holding it up as SROA (the entire pass) takes less time
than reading the IR for that test case. The analysis takes the same time
as running mem2reg on the final allocas. I suspect (without much
evidence) that the new implementation will scale much better however,
and it is just the small nature of the test cases that makes the changes
small and noisy. Either way, it is still simpler and cleaner I think.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186316 91177308-0d34-0410-b5e6-96231b3b80d8
is executed within the same second as the inputs for the test are
checked out from the source tree, it will fail to update due to being
below the resolution of the 'mtime' test used.
Now, this may seem improbably to you... ok, maybe *really* improbable,
but consider a system which does distributed execution of tests by
shipping their inputs to another machine and runs them. That might cause
the mtime to be quite recent during the test run. ;]
Instead, create two files directly in the test (allowing all platforms
to see the problem) and add either a use of the 'touch' command that
forces one mtime to some time quite a bit in the past, or it sleeps for
just over a second to be outside of the precision window.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186282 91177308-0d34-0410-b5e6-96231b3b80d8
against a constant."
This reverts commit r186107. It didn't handle wrapping arithmetic in the
loop correctly and thus caused the following C program to count from
0 to UINT64_MAX instead of from 0 to 255 as intended:
#include <stdio.h>
int main() {
unsigned char first = 0, last = 255;
do { printf("%d\n", first); } while (first++ != last);
}
Full test case and instructions to reproduce with just the -indvars pass
sent to the original review thread rather than to r186107's commit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186152 91177308-0d34-0410-b5e6-96231b3b80d8
CGSCC pass manager. This should insulate the inlining decisions from the
vectorization decisions, however it may have both compile time and code
size problems so it is just an experimental option right now.
Adding this based on a discussion with Arnold and it seems at least
worth having this flag for us to both run some experiments to see if
this strategy is workable. It may solve some of the regressions seen
with the loop vectorizer.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184698 91177308-0d34-0410-b5e6-96231b3b80d8
There is some hope of eventually supporting a unified build with it, but
until then this lets me (and others) check it out in this location
without things breaking.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184697 91177308-0d34-0410-b5e6-96231b3b80d8
I couldn't touch this file and not clean it up some. These reformattings
brought to you by clang-format, with some minor adjustments by me. More
spring cleaning to follow here.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179004 91177308-0d34-0410-b5e6-96231b3b80d8
internal linkage and so wasn't a patent bug, it doesn't make any sense
here. We can avoid even calling operator<< by just embedding the newline
in the string literals that were already being streamed out. It also
gives the impression of some line-ending agnosticisms which is not
present, and that flushing happens when it doesn't.
If we want to use std::endl, we could do that, but honestly it doesn't
seem remotely worth it. Using '\n' directly is much more clear when
working with raw_ostream.
It also happens to fix builds with old crufty GCC STL implementations
that include std::endl into the global namespace (or headers written to
be compatible with such atrocities).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179003 91177308-0d34-0410-b5e6-96231b3b80d8
The fix for PR14972 in r177055 introduced a real think-o in the *store*
side, likely because I was much more focused on the load side. While we
can arbitrarily widen (or narrow) a loaded value, we can't arbitrarily
widen a value to be stored, as that changes the width of memory access!
Lock down the code path in the store rewriting which would do this to
only handle the intended circumstance.
All of the existing tests continue to pass, and I've added a test from
the PR.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178974 91177308-0d34-0410-b5e6-96231b3b80d8
that work on the LLVMBuild based dependency specification didn't
actually work, we just now maintain dependencies in *3* places instead
of 2. Yay.
There may still be some missing dependencies, I'm still sifting through
the bots and my builds, but this is a step in the right direction.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177988 91177308-0d34-0410-b5e6-96231b3b80d8
its own library. These functions are bridging between the bitcode reader
and the ll parser which are in different libraries. Previously we didn't
have any good library to do this, and instead played fast and loose with
a "header only" set of interfaces in the Support library. This really
doesn't work well as evidenced by the recent attempt to add timing logic
to the these routines.
As part of this, make them normal functions rather than weird inline
functions, and sink the implementation into the library. Also clean up
the header to be nice and minimal.
This requires updating lots of build system dependencies to specify that
the IRReader library is needed, and several source files to not
implicitly rely upon the header file to transitively include all manner
of other headers.
If you are using IRReader.h, this commit will break you (the header
moved) and you'll need to also update your library usage to include
'irreader'. I will commit the corresponding change to Clang momentarily.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177971 91177308-0d34-0410-b5e6-96231b3b80d8
the ARM build bots, and it adds a weird case to the test suite where
a test uses as inputs files in the parent directory.
Talked about this with Dave on IRC and he's fine with this approach even
though it isn't optimal.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177700 91177308-0d34-0410-b5e6-96231b3b80d8
-time-ir-parsing flag
This breaks the layering of the Support library. We can't add an
implementation side to IRReader because it refers directly to entities
only accessible as part of the IR, AsmParser, and BitcodeReader
libraries. It can only be used in a context where all of those libraries
will be available.
We'll need to find some other way to get this functionality, and
hopefully solve the long-standing layering problem of IRReader.h...
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177695 91177308-0d34-0410-b5e6-96231b3b80d8
header.
This method is called in the hot path for *many* passes, SROA is what
caught my interest. A common pattern is that which branch of the switch
should be taken is known in the callsite and so it is a very good
candidate for inlining and simplification. Moving it into the header
allows the optimizer to fold a lot of boring, repeatitive code in
callers of this routine.
I'm seeing pretty significant speedups in parts of SROA and I suspect
other passes will see similar speedups if they end up working with type
sizes frequently. I've not seen any significant growth of the binaries
as a consequence, but let me know if you see anything suspicious here.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177632 91177308-0d34-0410-b5e6-96231b3b80d8
The key part of this is ensuring that name prefixes remain in a Twine
form until we get to a point where we can nuke them under NDEBUG. This
is tricky using the old APIs as they played fast and loose with Twine,
which is prone to serious error. The inserter is much cleaner as it is
actually in the call stack leading to the setName call, and so has
a good opportunity to prepend the prefix.
This matters more than you might imagine because most runs over an
alloca find a single partition, and rewrite 3 or 4 instructions
referring to it. As a consequence doing this lazily and exclusively with
Twine allows the optimizer to delete more of it and shaves another 2% to
3% off of the release build's SROA run time for PR15412. I also think
the APIs are cleaner, and the use of Twine is more reliable, so
I consider it a win-win despite the churn required to reach this state.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177631 91177308-0d34-0410-b5e6-96231b3b80d8
This is espcially important because the new SROA pass goes to great
lengths to provide helpful names for debugging, and as a consequence
they can become very slow to render.
Good for between 5% and 15% of the SROA runtime on some slow test cases
such as the one in PR15412.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177495 91177308-0d34-0410-b5e6-96231b3b80d8
The fundamental problem is that SROA didn't allow for overly wide loads
where the bits past the end of the alloca were masked away and the load
was sufficiently aligned to ensure there is no risk of page fault, or
other trapping behavior. With such widened loads, SROA would delete the
load entirely rather than clamping it to the size of the alloca in order
to allow mem2reg to fire. This was exposed by a test case that neatly
arranged for GVN to run first, widening certain loads, followed by an
inline step, and then SROA which miscompiles the code. However, I see no
reason why this hasn't been plaguing us in other contexts. It seems
deeply broken.
Diagnosing all of the above took all of 10 minutes of debugging. The
really annoying aspect is that fixing this completely breaks the pass.
;] There was an implicit reliance on the fact that no loads or stores
extended past the alloca once we decided to rewrite them in the final
stage of SROA. This was used to encode information about whether the
loads and stores had been split across multiple partitions of the
original alloca. That required threading explicit tracking of whether
a *use* of a partition is split across multiple partitions.
Once that was done, another problem arose: we allowed splitting of
integer loads and stores iff they were loads and stores to the entire
alloca. This is a really arbitrary limitation, and splitting at least
some integer loads and stores is crucial to maximize promotion
opportunities. My first attempt was to start removing the restriction
entirely, but currently that does Very Bad Things by causing *many*
common alloca patterns to be fully decomposed into i8 operations and
lots of or-ing together to produce larger integers on demand. The code
bloat is terrifying. That is still the right end-goal, but substantial
work must be done to either merge partitions or ensure that small i8
values are eagerly merged in some other pass. Sadly, figuring all this
out took essentially all the time and effort here.
So the end result is that we allow splitting only when the load or store
at least covers the alloca. That ensures widened loads and stores don't
hurt SROA, and that we don't rampantly decompose operations more than we
have previously.
All of this was already fairly well tested, and so I've just updated the
tests to cover the wide load behavior. I can add a test that crafts the
pass ordering magic which caused the original PR, but that seems really
brittle and to provide little benefit. The fundamental problem is that
widened loads should Just Work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177055 91177308-0d34-0410-b5e6-96231b3b80d8
memory intrinsics in the SDAG builder.
When alignment is zero, the lang ref says that *no* alignment
assumptions can be made. This is the exact opposite of the internal API
contracts of the DAG where alignment 0 indicates that the alignment can
be made to be anything desired.
There is another, more explicit alignment that is better suited for the
role of "no alignment at all": an alignment of 1. Map the intrinsic
alignment to this early so that we don't end up generating aligned DAGs.
It is really terrifying that we've never seen this before, but we
suddenly started generating a large number of alignment 0 memcpys due to
the new code to do memcpy-based copying of POD class members. That patch
contains a bug that rounds bitfield alignments down when they are the
first field. This can in turn produce zero alignments.
This fixes weird crashes I've seen in library users of LLVM on 32-bit
hosts, etc.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@176022 91177308-0d34-0410-b5e6-96231b3b80d8
fields were only ever set in the constructor. The create method retains
its consistent interface so that these bits can be re-threaded through
the emitter if they're ever needed.
This was found by the -Wunused-private-field Clang warning.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175482 91177308-0d34-0410-b5e6-96231b3b80d8
infrastructure on MCStreamer to test for whether there is an
MCELFStreamer object available.
This is just a cleanup on the AsmPrinter side of things, moving ad-hoc
tests of random APIs to a direct type query. But the AsmParser
completely broken. There were no tests, it just blindly cast its
streamer to an MCELFStreamer and started manipulating it.
I don't have a test case -- this actually failed on LLVM's own
regression test suite. Unfortunately the failure only appears when the
stars, compilers, and runtime align to misbehave when we read a pointer
to a formatted_raw_ostream as-if it were an MCAssembler. =/
UBSan would catch this immediately.
Many thanks to Matt for doing about 80% of the debugging work here in
GDB, Jim for helping to explain how exactly to fix this, and others for
putting up with the hair pulling that ensued during debugging it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@174118 91177308-0d34-0410-b5e6-96231b3b80d8
caught this, but I want that in a separate commit in case there is
a need to revert the actual functional bit as part of reverting other
patches. This way, the commits relating to just getting the RTTI bits in
place are separate from the functional changes that start using them.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@174117 91177308-0d34-0410-b5e6-96231b3b80d8
isa<> and dyn_cast<>. In several places, code is already hacking around
the absence of this, and there seem to be several interfaces that might
be lifted and/or devirtualized using this.
This change was based on a discussion with Jim Grosbach about how best
to handle testing for specific MCStreamer subclasses. He said that this
was the correct end state, and everything else was too hacky so
I decided to just make it so.
No functionality should be changed here, this is just threading the kind
through all the constructors and setting up the classof overloads.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@174113 91177308-0d34-0410-b5e6-96231b3b80d8
loops over instructions in the basic block or the use-def list of the
value, neither of which are really efficient when repeatedly querying
about values in the same basic block.
What's more, we already know that the CondBB is small, and so we can do
a much more efficient test by counting the uses in CondBB, and seeing if
those account for all of the uses.
Finally, we shouldn't blanket fail on any such instruction, instead we
should conservatively assume that those instructions are part of the
cost.
Note that this actually fixes a bug in the pass because
isUsedInBasicBlock has a really terrible bug in it. I'll fix that in my
next commit, but the fix for it would make this code suddenly take the
compile time hit I thought it already was taking, so I wanted to go
ahead and migrate this code to a faster & better pattern.
The bug in isUsedInBasicBlock was also causing other tests to test the
wrong thing entirely: for example we weren't actually disabling
speculation for floating point operations as intended (and tested), but
the test passed because we failed to speculate them due to the
isUsedInBasicBlock failure.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173417 91177308-0d34-0410-b5e6-96231b3b80d8
that can be specialized by targets.
The goal here is not to be more aggressive, but to just be more accurate
with very obvious cases. There are instructions which are known to be
truly free and which were not being modeled as such in this code -- see
the regression test which is distilled from an inner loop of zlib.
Everywhere the TTI cost model is insufficiently conservative I've added
explicit checks with FIXME comments to go add proper modelling of these
cost factors.
If this causes regressions, the likely solution is to make TTI even more
conservative in its cost estimates, but test cases will help here.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173342 91177308-0d34-0410-b5e6-96231b3b80d8
a cost fuction that seems both a bit ad-hoc and also poorly suited to
evaluating constant expressions.
Notably, it is missing any support for trivial expressions such as
'inttoptr'. I could fix this routine, but it isn't clear to me all of
the constraints its other users are operating under.
The core protection that seems relevant here is avoiding the formation
of a select instruction wich a further chain of select operations in
a constant expression operand. Just explicitly encode that constraint.
Also, update the comments and organization here to make it clear where
this needs to go -- this should be driven off of real cost measurements
which take into account the number of constants expressions and the
depth of the constant expression tree.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173340 91177308-0d34-0410-b5e6-96231b3b80d8
terms of cost rather than hoisting a single instruction.
This does *not* change the cost model! We still set the cost threshold
at 1 here, it's just that we track it by accumulating cost rather than
by storing an instruction.
The primary advantage is that we no longer leave no-op intrinsics in the
basic block. For example, this will now move both debug info intrinsics
and a single instruction, instead of only moving the instruction and
leaving a basic block with nothing bug debug info intrinsics in it, and
those intrinsics now no longer ordered correctly with the hoisted value.
Instead, we now splice the entire conditional basic block's instruction
sequence.
This also places the code for checking the safety of hoisting next to
the code computing the cost.
Currently, the only observable side-effect of this change is that debug
info intrinsics are no longer abandoned. I'm not sure how to craft
a test case for this, and my real goal was the refactoring, but I'll
talk to Dave or Eric about how to add a test case for this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173339 91177308-0d34-0410-b5e6-96231b3b80d8
Previously, the code would scan the PHI nodes and build up a small
setvector of candidate value pairs in phi nodes to go and rewrite. Once
certain the rewrite could be performed, the code walks the set, and for
each one re-scans the entire PHI node list looking for nodes to rewrite
operands.
Instead, scan the PHI nodes once to check for hazards, and then scan it
a second time to rewrite the operands to selects. No set vector, and
a max of two scans.
The only downside is that we might form identical selects, but
instcombine or anything else should fold those easily, and it seems
unlikely to happen often.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173337 91177308-0d34-0410-b5e6-96231b3b80d8
pretty in doxygen, adding some of the details actually present in
a classic example where this matters (a loop from gzip and many other
compression algorithms), and a cautionary note about the risks inherent
in the transform. This has come up on the mailing lists recently, and
I suspect folks reading this code could benefit from going and looking
at the MI pass that can really deal with these issues.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173329 91177308-0d34-0410-b5e6-96231b3b80d8
generic function calls and intrinsics. This is somewhat overlapping with
an existing intrinsic cost method, but that one seems targetted at
vector intrinsics. I'll merge them or separate their names and use cases
in a separate commit.
This sinks the test of 'callIsSmall' down into TTI where targets can
control it. The whole thing feels very hack-ish to me though. I've left
a FIXME comment about the fundamental design problem this presents. It
isn't yet clear to me what the users of this function *really* care
about. I'll have to do more analysis to figure that out. Putting this
here at least provides it access to proper analysis pass tools and other
such. It also allows us to more cleanly implement the baseline cost
interfaces in TTI.
With this commit, it is now theoretically possible to simplify much of
the inline cost analysis's handling of calls by calling through to this
interface. That conversion will have to happen in subsequent commits as
it requires more extensive restructuring of the inline cost analysis.
The CodeMetrics class is now really only in the business of running over
a block of code and aggregating the metrics on that block of code, with
the actual cost evaluation done entirely in terms of TTI.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173148 91177308-0d34-0410-b5e6-96231b3b80d8
is free. The whole CodeMetrics API should probably be reworked more, but
this is enough to allow deleting the duplicate code there for computing
whether an instruction is free.
All of the passes using this have been updated to pull in TTI and hand
it to the CodeMetrics stuff. Further, a dead CodeMetrics API
(analyzeFunction) is nuked for lack of users.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173036 91177308-0d34-0410-b5e6-96231b3b80d8
analysis. How cute that it wasn't previously. ;]
Part of this confusion stems from the flattened header file tree. Thanks
to Benjamin for pointing out the goof on IRC, and we're considering
un-flattening the headers, so speak now if that would bug you.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173033 91177308-0d34-0410-b5e6-96231b3b80d8
old CodeMetrics system. TTI has the specific advantage of being
extensible and customizable by targets to reflect target-specific cost
metrics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173032 91177308-0d34-0410-b5e6-96231b3b80d8
depend on and use other analyses (as long as they're either immutable
passes or CGSCC passes of course -- nothing in the pass manager has been
fixed here). Leverage this to thread TargetTransformInfo down through
the inline cost analysis.
No functionality changed here, this just threads things through.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173031 91177308-0d34-0410-b5e6-96231b3b80d8
a dynamic analysis done on each call to the routine. However, now it can
use the standard pass infrastructure to reference other analyses,
instead of a silly setter method. This will become more interesting as
I teach it about more analysis passes.
This updates the two inliner passes to use the inline cost analysis.
Doing so highlights how utterly redundant these two passes are. Either
we should find a cheaper way to do always inlining, or we should merge
the two and just fiddle with the thresholds to get the desired behavior.
I'm leaning increasingly toward the latter as it would also remove the
Inliner sub-class split.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173030 91177308-0d34-0410-b5e6-96231b3b80d8
Chandler Carruth