a TargetMachine to construct (and thus isn't always available), to an
analysis group that supports layered implementations much like
AliasAnalysis does. This is a pretty massive change, with a few parts
that I was unable to easily separate (sorry), so I'll walk through it.
The first step of this conversion was to make TargetTransformInfo an
analysis group, and to sink the nonce implementations in
ScalarTargetTransformInfo and VectorTargetTranformInfo into
a NoTargetTransformInfo pass. This allows other passes to add a hard
requirement on TTI, and assume they will always get at least on
implementation.
The TargetTransformInfo analysis group leverages the delegation chaining
trick that AliasAnalysis uses, where the base class for the analysis
group delegates to the previous analysis *pass*, allowing all but tho
NoFoo analysis passes to only implement the parts of the interfaces they
support. It also introduces a new trick where each pass in the group
retains a pointer to the top-most pass that has been initialized. This
allows passes to implement one API in terms of another API and benefit
when some other pass above them in the stack has more precise results
for the second API.
The second step of this conversion is to create a pass that implements
the TargetTransformInfo analysis using the target-independent
abstractions in the code generator. This replaces the
ScalarTargetTransformImpl and VectorTargetTransformImpl classes in
lib/Target with a single pass in lib/CodeGen called
BasicTargetTransformInfo. This class actually provides most of the TTI
functionality, basing it upon the TargetLowering abstraction and other
information in the target independent code generator.
The third step of the conversion adds support to all TargetMachines to
register custom analysis passes. This allows building those passes with
access to TargetLowering or other target-specific classes, and it also
allows each target to customize the set of analysis passes desired in
the pass manager. The baseline LLVMTargetMachine implements this
interface to add the BasicTTI pass to the pass manager, and all of the
tools that want to support target-aware TTI passes call this routine on
whatever target machine they end up with to add the appropriate passes.
The fourth step of the conversion created target-specific TTI analysis
passes for the X86 and ARM backends. These passes contain the custom
logic that was previously in their extensions of the
ScalarTargetTransformInfo and VectorTargetTransformInfo interfaces.
I separated them into their own file, as now all of the interface bits
are private and they just expose a function to create the pass itself.
Then I extended these target machines to set up a custom set of analysis
passes, first adding BasicTTI as a fallback, and then adding their
customized TTI implementations.
The fourth step required logic that was shared between the target
independent layer and the specific targets to move to a different
interface, as they no longer derive from each other. As a consequence,
a helper functions were added to TargetLowering representing the common
logic needed both in the target implementation and the codegen
implementation of the TTI pass. While technically this is the only
change that could have been committed separately, it would have been
a nightmare to extract.
The final step of the conversion was just to delete all the old
boilerplate. This got rid of the ScalarTargetTransformInfo and
VectorTargetTransformInfo classes, all of the support in all of the
targets for producing instances of them, and all of the support in the
tools for manually constructing a pass based around them.
Now that TTI is a relatively normal analysis group, two things become
straightforward. First, we can sink it into lib/Analysis which is a more
natural layer for it to live. Second, clients of this interface can
depend on it *always* being available which will simplify their code and
behavior. These (and other) simplifications will follow in subsequent
commits, this one is clearly big enough.
Finally, I'm very aware that much of the comments and documentation
needs to be updated. As soon as I had this working, and plausibly well
commented, I wanted to get it committed and in front of the build bots.
I'll be doing a few passes over documentation later if it sticks.
Commits to update DragonEgg and Clang will be made presently.
llvm-svn: 171681
interfaces which could be extracted from it, and must be provided on
construction, to a chained analysis group.
The end goal here is that TTI works much like AA -- there is a baseline
"no-op" and target independent pass which is in the group, and each
target can expose a target-specific pass in the group. These passes will
naturally chain allowing each target-specific pass to delegate to the
generic pass as needed.
In particular, this will allow a much simpler interface for passes that
would like to use TTI -- they can have a hard dependency on TTI and it
will just be satisfied by the stub implementation when that is all that
is available.
This patch is a WIP however. In particular, the "stub" pass is actually
the one and only pass, and everything there is implemented by delegating
to the target-provided interfaces. As a consequence the tools still have
to explicitly construct the pass. Switching targets to provide custom
passes and sinking the stub behavior into the NoTTI pass is the next
step.
llvm-svn: 171621
into their new header subdirectory: include/llvm/IR. This matches the
directory structure of lib, and begins to correct a long standing point
of file layout clutter in LLVM.
There are still more header files to move here, but I wanted to handle
them in separate commits to make tracking what files make sense at each
layer easier.
The only really questionable files here are the target intrinsic
tablegen files. But that's a battle I'd rather not fight today.
I've updated both CMake and Makefile build systems (I think, and my
tests think, but I may have missed something).
I've also re-sorted the includes throughout the project. I'll be
committing updates to Clang, DragonEgg, and Polly momentarily.
llvm-svn: 171366
Again, tools are trickier to pick the main module header for than
library source files. I've started to follow the pattern of using
LLVMContext.h when it is included as a stub for program source files.
llvm-svn: 169252
The TargetTransform changes are breaking LTO bootstraps of clang. I am
working with Nadav to figure out the problem, but I am reverting it for now
to get our buildbots working.
This reverts svn commits: 165665 165669 165670 165786 165787 165997
and I have also reverted clang svn 165741
llvm-svn: 166168
include/llvm/Analysis/DebugInfo.h to include/llvm/DebugInfo.h.
The reasoning is because the DebugInfo module is simply an interface to the
debug info MDNodes and has nothing to do with analysis.
llvm-svn: 159312
options, to enable easier testing of the innards of LLVM that are
enabled by such optimization strategies.
Note that this doesn't provide the (much needed) function attribute
support for -Oz (as opposed to -Os), but still seems like a positive
step to better test the logic that Clang currently relies on.
Patch by Patrik Hägglund.
llvm-svn: 156913
This is the initial checkin of the basic-block autovectorization pass along with some supporting vectorization infrastructure.
Special thanks to everyone who helped review this code over the last several months (especially Tobias Grosser).
llvm-svn: 149468
the X86 asmparser to produce ranges in the one case that was annoying me, for example:
test.s:10:15: error: invalid operand for instruction
movl 0(%rax), 0(%edx)
^~~~~~~
It should be straight-forward to enhance filecheck, tblgen, and/or the .ll parser to use
ranges where appropriate if someone is interested.
llvm-svn: 142106
developers can see if their driver changed any cl::Option's. The
current implementation isn't perfect but handles most kinds of
options. This is nice to have when decomposing the stages of
compilation and moving between different drivers. It's also a good
sanity check when comparing results produced by different command line
invocations that are expected to produce the comparable results.
Note: This is not an attempt to prolong the life of cl::Option. On the
contrary, it's a placeholder for a feature that must exist when
cl::Option is replaced by a more appropriate framework. A new
framework needs: a central option registry, dynamic name lookup,
non-global containers of option values (e.g. per-module,
per-function), *and* the ability to print options values and their defaults at
any point during compilation.
llvm-svn: 128910
The check to not allow -analyze and -disable-output at the same time was done
before parsing the command line flags. Therefore it never triggered, and in case
both options where used opt segfaulted. Fix this by moving this check a after
command line parsing.
llvm-svn: 120732
A RegionPass is executed like a LoopPass but on the regions detected by the
RegionInfo pass instead of the loops detected by the LoopInfo pass.
llvm-svn: 116905
must be called in the pass's constructor. This function uses static dependency declarations to recursively initialize
the pass's dependencies.
Clients that only create passes through the createFooPass() APIs will require no changes. Clients that want to use the
CommandLine options for passes will need to manually call the appropriate initialization functions in PassInitialization.h
before parsing commandline arguments.
I have tested this with all standard configurations of clang and llvm-gcc on Darwin. It is possible that there are problems
with the static dependencies that will only be visible with non-standard options. If you encounter any crash in pass
registration/creation, please send the testcase to me directly.
llvm-svn: 116820
Follow the same logic in the LoopPass, ModulePass and CallGraphSCCPass printers,
as it was already used in the BasicBlockPass and FunctionPass printers. This is
more consistent.
The other option would have been to completely disable dumping the analysis
information. However, as this information is the only information printed if the
-analysis flag is set, calling opt would not do anything at all.
llvm-svn: 113360
of a base class.
This makes it possible to unregister the file from FilesToRemove when
the file is done. Also, this eliminates the need for
formatted_tool_output_file.
llvm-svn: 112706
of problems with output files being left behind or output streams
being left unclosed. Fix llvm-mc to respect the -o option in all
modes, rather than hardcoding outs() in some cases.
llvm-svn: 111603
callgraph SCC's. This makes it match what the node itself would print. Also,
"indirect callgraph node" doesn't make sense - it has nothing particularly to
do with indirect calls.
llvm-svn: 105730
to CallGraphSCCPass's instead of passing around a
std::vector<CallGraphNode*>. No functionality change,
but now we have a much tidier interface.
llvm-svn: 101558
Modules and ModuleProviders. Because the "ModuleProvider" simply materializes
GlobalValues now, and doesn't provide modules, it's renamed to
"GVMaterializer". Code that used to need a ModuleProvider to materialize
Functions can now materialize the Functions directly. Functions no longer use a
magic linkage to record that they're materializable; they simply ask the
GVMaterializer.
Because the C ABI must never change, we can't remove LLVMModuleProviderRef or
the functions that refer to it. Instead, because Module now exposes the same
functionality ModuleProvider used to, we store a Module* in any
LLVMModuleProviderRef and translate in the wrapper methods. The bindings to
other languages still use the ModuleProvider concept. It would probably be
worth some time to update them to follow the C++ more closely, but I don't
intend to do it.
Fixes http://llvm.org/PR5737 and http://llvm.org/PR5735.
llvm-svn: 94686
missing ones are libsupport, libsystem and libvmcore. libvmcore is
currently blocked on bugpoint, which uses EH. Once it stops using
EH, we can switch it off.
This #if 0's out 3 unit tests, because gtest requires RTTI information.
Suggestions welcome on how to fix this.
llvm-svn: 94164
to a console, when --analyze is used.
Similarly, avoid creating an empty output file when --disable-output is used.
Print a warning when the -o option appears with either --analyze or
--disable-output, to indicate that the option is being ignored.
llvm-svn: 93685
working. To support this, add an is_displayed() function to raw_ostream,
and generalize Process::StandardOutIsDisplayed and friends in order to
support it.
Also, call RemoveFileOnSignal before creating a file instead of after, so
that the file isn't left behind if the program is interrupted between when
the file is created and RemoveFileOnSignal is called.
While here, add a -S to llvm-extract and port it to IRReader so that it
supports assembly input.
llvm-svn: 81568
argpromotion and structretpromote. Basically, when replacing
a function, they used the 'changeFunction' api which changes
the entry in the function map (and steals/reuses the callgraph
node).
This has some interesting effects: first, the problem is that it doesn't
update the "callee" edges in any callees of the function in the call graph.
Second, this covers for a major problem in all the CGSCC pass stuff, which
is that it is completely broken when functions are deleted if they *don't*
reuse a CGN. (there is a cute little fixme about this though :).
This patch changes the protocol that CGSCC passes must obey: now the CGSCC
pass manager copies the SCC and preincrements its iterator to avoid passes
invalidating it. This allows CGSCC passes to mutate the current SCC. However
multiple passes may be run on that SCC, so if passes do this, they are now
required to *update* the SCC to be current when they return.
Other less interesting parts of this patch are that it makes passes update
the CG more directly, eliminates changeFunction, and requires clients of
replaceCallSite to specify the new callee CGN if they are changing it.
llvm-svn: 80527
This is conventional command-line tool behavior. -f now just means
"enable binary output on terminals".
Add a -f option to llvm-extract and llvm-link, for consistency.
Remove F_Force from raw_fd_ostream and enable overwriting and
truncating by default. Introduce an F_Excl flag to permit users to
enable a failure when the file already exists. This flag is
currently unused.
Update Makefiles and documentation accordingly.
llvm-svn: 79990
instead of as two bools. Use this to add a F_Append flag
which has the obvious behavior.
Other unrelated changes conflated into this patch:
1. REmove EH stuff from llvm-dis and llvm-as, the try blocks
are dead.
2. Simplify the filename inference code in llvm-as/llvm-dis,
because raw_fd_ostream does the right thing with '-'.
3. Switch machine verifier to use raw_ostream instead of ostream
(Which is the thing that needed append in the first place).
llvm-svn: 79807
an existing file is considered an error. Convert several tools
to use raw_fd_ostream instead of std::ostream, and to use this
new option instead of doing a manual check.
llvm-svn: 75801
to work out (in a very simplistic way) which function
arguments (pointer arguments only) are only dereferenced
and so do not escape. Mark such arguments 'nocapture'.
llvm-svn: 61525
I placed it just before GVN because that it is the
pass most likely to benefit from it. Some quick
and dirty testing confirms that this is a decent
place for it.
llvm-svn: 56144
the end. It is now run at the same moment as in
llvm-gcc. Also, run StripDeadPrototypes at the
end, just before running DeadTypeElimination.
This may be useful when doing LTO. Note that
llvm-gcc runs StripDeadPrototypes but not
DeadTypeElimination.
llvm-svn: 49848
Note: FastDSE now equals or exceeds the results of old DSE on all of SPEC2000 and SPEC2006. Unless major problems
show up in the testers, it will likely completely replace old DSE in the near future.
llvm-svn: 39986
api's look like this:
ModuleProvider *getBytecodeModuleProvider(
const std::string &Filename, ///< Name of file to be read
BCDecompressor_t *BCDC = Compressor::decompressToNewBuffer,
std::string* ErrMsg = 0, ///< Optional error message holder
BytecodeHandler* H = 0 ///< Optional handler for reader events
);
This is ugly, but allows a client to say:
getBytecodeModuleProvider("foo", 0);
If they do this, there is no dependency on the compression libraries, saving
codesize.
llvm-svn: 34012
This patch replaces the SymbolTable class with ValueSymbolTable which does
not support types planes. This means that all symbol names in LLVM must now
be unique. The patch addresses the necessary changes to deal with this and
removes code no longer needed as a result. This completes the bulk of the
changes for this PR. Some cleanup patches will follow.
llvm-svn: 33918
Step 1: Copy gccas functionality to opt. This endows opt with a new
-std-compile-opts option to get the set of optimization passes that
gccas used. It also adds -disable-inlining and -disable-opt which
both apply only if -std-compile-opts is given. The -strip-debug option
was also removed. It just makes sure that "-strip" gets done early and
is mostly there for compatibility with gccas. Finally, a new
-verify-each option will cause the verify pass to be run after each pass.
llvm-svn: 33786
adding a temporary wrapper around the ostream to make it friendly to
functions expecting an LLVM stream. This should be fixed in the future.
llvm-svn: 31990
Turn on -Wunused and -Wno-unused-parameter. Clean up most of the resulting
fall out by removing unused variables. Remaining warnings have to do with
unused functions (I didn't want to delete code without review) and unused
variables in generated code. Maintainers should clean up the remaining
issues when they see them. All changes pass DejaGnu tests and Olden.
llvm-svn: 31380
Close out this long standing bug by removing the remaining overloaded
virtual functions in LLVM. The -Woverloaded-virtual option is now turned on.
llvm-svn: 29934
Make sys::Program::ExecuteAndWait not throw exceptions and update any
affected code. It now return -9999 to signal that the program couldn't be
executed. Only one case (in bugpoint) actually examines the result code.
llvm-svn: 29785
Rid the Assembly Parser of exceptions. This is a really gross hack but it
will do until the Assembly Parser is re-written as a recursive descent.
The basic premise is that wherever the old "ThrowException" function was
called (new name: GenerateError) we set a flag (TriggerError). Every
production checks that flag and calls YYERROR if it is set. Additionally,
each call to ThrowException in the grammar is replaced with GEN_ERROR
which calls GenerateError and then YYERROR immediately. This prevents
the remaining production from continuing after an error condition.
llvm-svn: 29763
Shrinkify LLVM's footprint by removing the analyze tool and moving its
functionality into the opt tool. THis eliminates one of the largest tools
from LLVM and doesn't make opt much bigger because it already included
most of the analysis passes. To get the old analyze functionality pass
the -analyze option to opt. Note that the integeration here is dead
simple. The "main" of analyze was just copied to opt and invoked if the
-analyze option was given. There may be opportunities for further
integration such as removing the distinction between transform passes
and analysis passes.
To use the analysis functionality, if you previously did this:
analyze $FNAME -domset -disable-verify
you would now do this:
opt -analyze $FNAME -domset -disable-verify
Pretty simple.
llvm-svn: 29762
1. Add #includes to LinkAllVMCore.h to get Mangler.o and InlineAsm.o
2. Make Mangler.h and InlineAsm.h use the macros to ensure linkage
3. Make each of the tools with --load options include LinkAllVMCore.h
This should be the last set of changes for this bug and 800.
llvm-svn: 28719
into the LLVMAnalysis library.
This allows LLVMTranform and LLVMTransformUtils to be archives and linked
with LLVMAnalysis.a, which provides any missing definitions.
llvm-svn: 24036
SparcV9 JIT.
2. Make LLVMTransformUtils a relinked object file and always link it before
LLVMAnalysis.a. These two libraries have circular dependencies on each
other which creates problem when building the SparcV9 JIT. This change
fixes the dependency on all platforms problems with a minimum of fuss.
llvm-svn: 24023
1. Get rid of TOOLLINKOPTS as it is a hold over from llvm-test and only
used to communicate additional libraries to the linker. The *standard*
way to do that is with the LIBS variable which this change supports.
2. Allow the TARGETS_TO_BUILD variable to be set from the configuration
substitution. This is the result of the --enable-target= parameter to
the configure script.
llvm-svn: 21449
* Place a try/catch block around the entire tool to Make sure std::string
exceptions are caught and printed before exiting the tool.
* Make sure we catch unhandled exceptions at the top level so that we don't
abort with a useless message but indicate than an unhandled exception was
generated.
llvm-svn: 19192
Move include/Config and include/Support into include/llvm/Config,
include/llvm/ADT and include/llvm/Support. From here on out, all LLVM
public header files must be under include/llvm/.
llvm-svn: 16137
- Rename runOnMethod to runOnFunction
* Transform getAnalysisUsageInfo into getAnalysisUsage
- Method is now const
- It now takes one AnalysisUsage object to fill in instead of 3 vectors
to fill in
- Pass's now specify which other passes they _preserve_ not which ones
they modify (be conservative!)
- A pass can specify that it preserves all analyses (because it never
modifies the underlying program)
* s/Method/Function/g in other random places as well
llvm-svn: 2333
written output file. This is important because crashing testcases often write
part of a file out, and the testing harness decides the file is up-to-date next
time the test is run.
llvm-svn: 2303
Jakub Staszak