Instead of eagerly deserializing a list of DeclIDs when we load a module file
and doing a binary search to find the redeclarations of a decl, store a list of
redeclarations of each chain before the first declaration and load it directly.
llvm-svn: 245789
Summary:
This patch adds Sphinx based documentation to libc++. The goal is to make it easier to write documentation for libc++ since writing new documentation in HTML is cumbersome. This patch rewrites the main page for libc++ along with the instructions for using, building and testing libc++.
The built documentation can be found and reviewed here: http://efcs.ca/libcxx-docs
In order to build the sphinx documentation you need to specify the cmake options `-DLLVM_ENABLE_SPHINX=ON -DLIBCXX_INCLUDE_DOCS=ON`. This will add the makefile rule `docs-libcxx-html`.
Reviewers: chandlerc, mclow.lists, danalbert, jroelofs
Subscribers: silvas, cfe-commits
Differential Revision: http://reviews.llvm.org/D12129
llvm-svn: 245788
Summary:
According to CUDA documentation, global variables declared with __device__,
__constant__ can be initialized from host code, so mark them as
externally initialized. Because __shared__ variables cannot have an
initialization as part of their declaration and since the value maybe kept
across different kernel invocation, the value of __shared__ is effectively
undefined instead of zero initialized.
Wrongly using zero initializer may cause illegitimate optimization, e.g.
removing unused __constant__ variable because it's not updated in the device
code and the value is initialized with zero.
Test Plan: test/CodeGenCUDA/address-spaces.cu
Patch by Xuetian Weng
Reviewers: jholewinski, eliben, tra, jingyue
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D12241
llvm-svn: 245786
Summary:
__shared__ variable may now emit undef value as initializer, do not
throw error on that.
Test Plan: test/CodeGen/NVPTX/global-addrspace.ll
Patch by Xuetian Weng
Reviewers: jholewinski, tra, jingyue
Subscribers: llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D12242
llvm-svn: 245785
These changes break both autoconf Mac OS X buildbot (linker errors
due to wrong Makefiles) and CMake buildbot (safestack test failures).
llvm-svn: 245784
If corresponding in-tree subdirectory exists, just ignore LLVM_EXTERNAL* stuff.
Otherwise, set LLVM_TOOL_*_BUILD ON/OFF properly according to LLVM_EXTERNAL_*.
This makes easier to walk among old revisions *without* deleteing CMakeCache.txt.
Before r242059, LLVM_EXTERNAL_* was working like;
if(EXISTS ${*_SOURCE_DIR}/CMakeLists.txt)
set(*_BUILD ON CACHE)
if(*_BUILD is ON)
add_subdirectory(*_SOURCE_DIR)
endif()
endif()
llvm-svn: 245782
all modules and reduce the number of declarations we load when loading a
redeclaration chain.
The new approach is:
* when loading the first declaration of an entity within a module file, we
first load all declarations of the entity that were imported into that
module file, and then load all the other declarations of that entity from
that module file and build a suitable decl chain from them
* when loading any other declaration of an entity, we first load the first
declaration from the same module file
As before, we complete redecl chains through name lookup where necessary.
To make this work, I also had to change the way that template specializations
are stored -- it no longer suffices to track only canonical specializations; we
now emit all "first local" declarations when emitting a list of specializations
for a template.
On one testcase with several thousand imported module files, this reduces the
total runtime by 72%.
llvm-svn: 245779
We can use the 'H' typespec modifier to use 128-bit vectors directly
in the only two users of this special-case: the vcvt f16 intrinsics.
This also lets us use more meaningful prototype modifiers.
llvm-svn: 245778
Although the basic s_load_* instructions happen to use the same
opcode, some of the special case SMRD instructions have
different opcodes.
llvm-svn: 245775
a contextually-typed expression that semantic analysis will
probably need to invasively rewrite, don't include the
RHS OVE as a separate semantic expression, and check the
operation with the original RHS expression.
There are two contextually-typed expressions that can survive
to here: overloaded function references, which are at least
safe to double-emit, and C++11 initializer list expressions,
which are not at all safe to double-emit and which often
don't update the original syntactic InitListExpr with
implicit conversions to member types, etc.
This means that the original RHS may appear, undecorated by
an OVE, in the semantic expressions. Fortunately, it will
only ever be used in a single place there, and I don't
believe there are clients that rely on being able to pick
out the original RHS from the semantic expressions.
But this could be problematic if there are clients that do
visit the entire tree and rely on not seeing the same
expression multiple times, once in the syntactic and once
in the semantic expressions. This is a very fiddly part
of the compiler.
rdar://21801088
llvm-svn: 245771
Previously we had to call __sanitizer_cov_dump() from tool-specific
callbacks - instead, let sanitizer_common library handle this in a single place.
llvm-svn: 245770
This reverts commit 7749a10ddbe22767d0e055753c674fcde7f28d39.
This commit introduces about 15-20 new test failures with windows local
targets.
llvm-svn: 245765
We had "vcvt_f16" and "VCVT_HIGH_F16": for other FP types, this naming
is used for intrinsics with integer overloads. The FP->FP conversions,
on the other hand, use the full "vcvt_f32_f64" name instead.
Use the same naming convention for the f16<->f32 conversions.
While there, reorder the definitions a little bit.
llvm-svn: 245763
Summary:
Merge functions previously relied on unsigned comparisons of pointer values to
order functions. This caused observable non-determinism in the compiler for
large bitcode programs. Basically, opt -mergefuncs program.bc | md5sum produces
different hashes when run repeatedly on the same machine. Differing output was
observed on three large bitcodes, but it was less frequent on the smallest file.
It is possible that this only manifests on the large inputs, hence remaining
undetected until now.
This patch fixes this by removing (almost, see below) all places where
comparisons between pointers are used to order functions. Most of these changes
are local, but the comparison of global values requires assigning an identifier
to each local in the order it is visited. This is very similar to the way the
comparison function identifies Value*'s defined within a function. Because the
order of visiting the functions and their subparts is deterministic, the
identifiers assigned to the globals will be as well, and the order of functions
will be deterministic.
With these changes, there is no more observed non-determinism. There is also
only minor slowdowns (negligible to 4%) compared to the baseline, which is
likely a result of the fact that global comparisons involve hash lookups and not
just pointer comparisons.
The one caveat so far is that programs containing BlockAddress constants can
still be non-deterministic. It is not clear what the right solution is here. In
particular, even if the global numbers are used to order by function, we still
need a way to order the BasicBlock*'s. Unfortunately, we cannot just bail out
and fail to order the functions or consider them equal, because we require a
total order over functions. Note that programs with BlockAddress constants are
relatively rare, so the impact of leaving this in is minor as long as this pass
is opt-in.
Author: jrkoenig
Reviewers: nlewycky, jfb, dschuff
Subscribers: jevinskie, llvm-commits, chapuni
Differential revision: http://reviews.llvm.org/D12168
llvm-svn: 245762
SCEV expansion can invalidate previously expanded values. For example
in SCEVExpander::ReuseOrCreateCast, if we already have the requested
cast value but it's not at the desired location, a new cast is inserted
and the old cast will be invalidated.
Therefore, when expanding the bounds for the pointers, a later entry can
invalidate the IR value for an earlier one. The fix is to store a value
handle rather than the value itself.
The newly added test has a more detailed description of how the bug
triggers.
This bug can have a negative but potentially highly variable performance
impact in Loop Distribution. Because one of the bound values was
invalidated and is an undef expression now, InstCombine is free to
transform the array overlap check:
Start0 <= End1 && Start1 <= End0
into:
Start0 <= End1
So depending on the runtime location of the arrays, we would detect a
conflict and fall back on the original loop of the versioned loop.
Also tested compile time with SPEC2006 LTO bc files.
llvm-svn: 245760
This allows us to remove a bunch of code in LTOCodeGenerator and llvm-lto
and has the side effect of improving error handling in the libLTO C API.
llvm-svn: 245756
We can wait on either VM, EXP or LGKM.
The waits are independent.
Without this patch, a wait inserted because of one of them
would also wait for all the previous others.
This patch makes s_wait only wait for the ones we need for the next
instruction.
Here's an example of subtle perf reduction this patch solves:
This is without the patch:
buffer_load_format_xyzw v[8:11], v0, s[44:47], 0 idxen
buffer_load_format_xyzw v[12:15], v0, s[48:51], 0 idxen
s_load_dwordx4 s[44:47], s[8:9], 0xc
s_waitcnt lgkmcnt(0)
buffer_load_format_xyzw v[16:19], v0, s[52:55], 0 idxen
s_load_dwordx4 s[48:51], s[8:9], 0x10
s_waitcnt vmcnt(1)
buffer_load_format_xyzw v[20:23], v0, s[44:47], 0 idxen
The s_waitcnt vmcnt(1) is useless.
The reason it is added is because the last
buffer_load_format_xyzw needs s[44:47], which was issued
by the first s_load_dwordx4. It waits for all VM
before that call to have finished.
Internally after every instruction, 3 counters (for VM, EXP and LGTM)
are updated after every instruction. For example buffer_load_format_xyzw
will
increase the VM counter, and s_load_dwordx4 the LGKM one.
Without the patch, for every defined register,
the current 3 counters are stored, and are used to know
how long to wait when an instruction needs the register.
Because of that, the s[44:47] counter includes that to use the register
you need to wait for the previous buffer_load_format_xyzw.
Instead this patch stores only the counters that matter for the
register,
and puts zero for the other ones, since we don't need any wait for them.
Patch by: Axel Davy
Differential Revision: http://reviews.llvm.org/D11883
llvm-svn: 245755
The original checkin was buggy, this change has a fix.
Original commit message:
[InstCombine] Transform A & (L - 1) u< L --> L != 0
Summary:
This transform is never a pessimization at the IR level (since it
replaces an `icmp` with another), and has potentiall payoffs:
1. It may make the `icmp` fold away or become loop invariant.
2. It may make the `A & (L - 1)` computation dead.
This shows up in Java, in range checks generated by array accesses of
the form `a[i & (a.length - 1)]`.
Reviewers: reames, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D12210
llvm-svn: 245753
clang-cl does not yet support C++ exceptions, so these tests will
not even compile.
Re-enabling these tests is tracked by llvm.org/pr24538
llvm-svn: 245747
When PPCVSXFMAMutate would look at the input addend register, it would get its
input value number. This would fail, however, if the register was undef,
causing a segfault. Don't segfault (just skip such FMA instructions).
Fixes the test case from PR24542 (although that may have been over-reduced).
llvm-svn: 245741
This commit extends the 'SlotMapping' structure and includes mappings for named
and numbered types in it. The LLParser is extended accordingly to fill out
those mappings at the end of module parsing.
This information is useful when we want to parse standalone constant values
at a later stage using the 'parseConstantValue' method. The constant values
can be constant expressions, which can contain references to types. In order
to parse such constant values, we have to restore the internal named and
numbered mappings for the types in LLParser, otherwise the parser will report
a parsing error. Therefore, this commit also introduces a new method called
'restoreParsingState' to LLParser, which uses the slot mappings to restore
some of its internal parsing state.
This commit is required to serialize constant value pointers in the machine
memory operands for the MIR format.
Reviewers: Duncan P. N. Exon Smith
llvm-svn: 245740
Richard Smith