Summary: Hot callsites should have higher threshold than inline hints. This patch uses separate threshold parameter for hot callsites.
Reviewers: davidxl, eraman
Subscribers: Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D22368
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277860 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Chrome on Linux uses WholeProgramDevirt for speed ups, and it's
important to detect regressions on both sides: the toolchain,
if fewer methods get devirtualized after an update, and Chrome,
if an innocently looking change caused many hot methods become
virtual again.
The need to track devirtualized methods is not Chrome-specific,
but it's probably the only user of the pass at this time.
Reviewers: kcc
Differential Revision: https://reviews.llvm.org/D23219
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277856 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
This is another refactoring to break up the one function into three logical components functions.
Another non-functional change before we start added in features.
Reviewers: nadav, mehdi_amini, majnemer
Subscribers: twoh, freik, llvm-commits
Differential Revision: https://reviews.llvm.org/D23102
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277855 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
If a profile has no samples for a function, then the function "entry count" is set to the value 0. Several places in the code test that if the Function::getEntryCount is defined at all. Here we change to treat a 0 entry count the same as undefined.
In particular, this fixes a problem in getLayoutSuccessorProbThreshold in MachineBlockPlacement.cpp where we use a different and inferior heuristic for laying out basic blocks.
Reviewers: danielcdh, dnovillo
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23082
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277849 91177308-0d34-0410-b5e6-96231b3b80d8
This is where an LC_SEGMENT load command has a fileoff field that
extends past the end of the file.
Also fix llvm-nm and llvm-size to remove the errorToErrorCode() call so error messages are printed.
And needed to update a few test cases now that they do print the error messages just a
bit differently.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277845 91177308-0d34-0410-b5e6-96231b3b80d8
Test is just reordering the existing functions (it would trigger for any
function after one with a phi).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277841 91177308-0d34-0410-b5e6-96231b3b80d8
There were two locations where fast-isel would generate a LFD instruction
with a target register class VSFRC instead of F8RC when VSX was enabled.
This can ccause invalid registers to be used in certain cases, like:
lfd 36, ...
instead of using a VSX load instruction. The wrong register number gets
silently truncated, causing invalid code to be generated.
The first place is PPCFastISel::PPCEmitLoad, which had multiple problems:
1.) The IsVSSRC and IsVSFRC flags are not initialized correctly, since they
are computed from resultReg, which is still zero at this point in many cases.
Fixed by changing the helper routines to operate on a register class instead
of a register and passing in UseRC.
2.) Even with this fixed, Is64VSXLoad is still wrong due to a typo:
bool Is32VSXLoad = IsVSSRC && Opc == PPC::LFS;
bool Is64VSXLoad = IsVSSRC && Opc == PPC::LFD;
The second line needs to use isVSFRC (like PPCEmitStore does).
3.) Once both the above are fixed, we're now generating a VSX instruction --
but an incorrect one, since generation of an indexed instruction with null
index is wrong. Fixed by copying the code handling the same issue in
PPCEmitStore.
The second place is PPCFastISel::PPCMaterializeFP, where we would emit an
LFD to load a constant from the literal pool, and use the wrong result
register class. Fixed by hardcoding a F8RC class even on systems
supporting VSX.
Fixes: https://llvm.org/bugs/show_bug.cgi?id=28630
Differential Revision: https://reviews.llvm.org/D22632
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277823 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Add instruction formats E, RSI, SSd, SSE, and SSF.
Added BRXH, BRXLE, PR, MVCK, STRAG, and ECTG instructions to test out
those formats.
Reviewers: uweigand
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23179
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277822 91177308-0d34-0410-b5e6-96231b3b80d8
This generated IR based on the order of evaluation, which is different
between GCC and Clang. With that in mind you get bootstrap miscompares
if you compare a Clang built with GCC-built Clang vs. Clang built with
Clang-built Clang. Diagnosing that made my head hurt.
This also reverts commit r277337, which "fixed" the test case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277820 91177308-0d34-0410-b5e6-96231b3b80d8
Preparation for adding 2 input support so we want to avoid unnecessary references to the input value type.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277817 91177308-0d34-0410-b5e6-96231b3b80d8
Preparation for adding 2 input support so we want to avoid unnecessary references to the input value type.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277814 91177308-0d34-0410-b5e6-96231b3b80d8
We should not use double (or float) in the LLVM, unless it is really needed. x87 FP register doesn't preserve SNaN to move the value.
FIXME: APFloat() may have the constructor by raw bit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277813 91177308-0d34-0410-b5e6-96231b3b80d8
This differs from the previous version by being more careful about template
instantiation/specialization in order to prevent errors when building with
clang -Werror. Specifically:
* begin is not defined in the template and is instead instantiated when Head
is. I think the warning when we don't do that is wrong (PR28815) but for now
at least do it this way to avoid the warning.
* Instead of performing template specializations in LLVM_INSTANTIATE_REGISTRY
instead provide a template definition then do explicit instantiation. No
compiler I've tried has problems with doing it the other way, but strictly
speaking it's not permitted by the C++ standard so better safe than sorry.
Original commit message:
Currently the Registry class contains the vestiges of a previous attempt to
allow plugins to be used on Windows without using BUILD_SHARED_LIBS, where a
plugin would have its own copy of a registry and export it to be imported by
the tool that's loading the plugin. This only works if the plugin is entirely
self-contained with the only interface between the plugin and tool being the
registry, and in particular this conflicts with how IR pass plugins work.
This patch changes things so that instead the add_node function of the registry
is exported by the tool and then imported by the plugin, which solves this
problem and also means that instead of every plugin having to export every
registry they use instead LLVM only has to export the add_node functions. This
allows plugins that use a registry to work on Windows if
LLVM_EXPORT_SYMBOLS_FOR_PLUGINS is used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277806 91177308-0d34-0410-b5e6-96231b3b80d8
This patch fixes passing long double type arguments to function in
soft float mode. If there is less than 4 argument registers free
(long double type is mapped in 4 gpr registers in soft float mode)
long double type argument must be passed through stack.
Differential Revision: https://reviews.llvm.org/D20114.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277804 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Turn (select C, (sext A), B) into (sext (select C, A, B')) when A is i1 and
B is a compatible constant, also for zext instead of sext. This will then be
further folded into logical operations.
The transformation would be valid for non-i1 types as well, but other parts of
InstCombine prefer to have sext from non-i1 as an operand of select.
Motivated by the shader compiler frontend in Mesa for AMDGPU, which emits i32
for boolean operations. With this change, the boolean logic is fully
recovered.
Reviewers: majnemer, spatel, tstellarAMD
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D22747
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277801 91177308-0d34-0410-b5e6-96231b3b80d8
The patch splits a complex && if condition into easier to read and understand
logic. That wrong early exit condition was letting some instructions with not
all operands available pass through when HoistingGeps was true.
Differential Revision: https://reviews.llvm.org/D23174
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277785 91177308-0d34-0410-b5e6-96231b3b80d8
Add a generalized IRBuilderCallbackInserter, which is just given a
callback to execute after insertion. This can be used to get rid of
the custom inserter in InstCombine, which will in turn allow me to add
target specific InstCombineCalls API for intrinsics without horrible
layering violations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277784 91177308-0d34-0410-b5e6-96231b3b80d8
Shifts with a uniform but non-constant count were considered very expensive to
vectorize, because the splat of the uniform count and the shift would tend to
appear in different blocks. That made the splat invisible to ISel, and we'd
scalarize the shift at codegen time.
Since r201655, CodeGenPrepare sinks those splats to be next to their use, and we
are able to select the appropriate vector shifts. This updates the cost model to
to take this into account by making shifts by a uniform cheap again.
Differential Revision: https://reviews.llvm.org/D23049
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277782 91177308-0d34-0410-b5e6-96231b3b80d8
when we are pointed at real data.
David Blaikie pointed out some odd logic in the case the Err value was a nullptr and
Lang Hames suggested it could be cleaned it up with an assert to know that Err is
not a nullptr when we are pointed at real data. As only in the case of constructing
the sentinel value by pointing it at null data is Err is permitted to be a nullptr,
since no error could occur in that case.
With this change the testing for “if (Err)” is removed from the constructor’s logic
and *Err is used directly without any check after the assert().
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277776 91177308-0d34-0410-b5e6-96231b3b80d8
These are the operations that are trivially identical. Division is omitted for
now because you need to use the correct sign/zero extension.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277775 91177308-0d34-0410-b5e6-96231b3b80d8
This is the forth patch in the coroutine series. CoroEaly pass now lowers coro.resume
and coro.destroy intrinsics by replacing them with an indirect call to an address
returned by coro.subfn.addr intrinsic. This is done so that CGPassManager recognizes
devirtualization when CoroElide replaces a call to coro.subfn.addr with an appropriate
function address.
Patch by Gor Nishanov!
Differential Revision: https://reviews.llvm.org/D22998
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277765 91177308-0d34-0410-b5e6-96231b3b80d8
Previously, FastISel for WebAssembly wasn't checking the return value of
`getRegForValue` in certain cases, which would generate instructions
referencing NoReg. This patch fixes this behavior.
Patch by Dominic Chen
Differential Revision: https://reviews.llvm.org/D23100
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277742 91177308-0d34-0410-b5e6-96231b3b80d8
I'm removing a misplaced pair of more specific folds from InstCombine in this patch as well,
so we know where those folds are happening in InstSimplify.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277738 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
TargetBaseAlign is no longer required since LSV checks if target allows misaligned accesses.
A constant defining a base alignment is still needed for stack accesses where alignment can be adjusted.
Previous patch (D22936) was reverted because tests were failing. This patch also fixes the cause of those failures:
- x86 failing tests either did not have the right target, or the right alignment.
- NVPTX failing tests did not have the right alignment.
- AMDGPU failing test (merge-stores) should allow vectorization with the given alignment but the target info
considers <3xi32> a non-standard type and gives up early. This patch removes the condition and only checks
for a maximum size allowed and relies on the next condition checking for %4 for correctness.
This should be revisited to include 3xi32 as a MVT type (on arsenm's non-immediate todo list).
Note that checking the sizeInBits for a MVT is undefined (leads to an assertion failure),
so we need to create an EVT, hence the interface change in allowsMisaligned to include the Context.
Reviewers: arsenm, jlebar, tstellarAMD
Subscribers: jholewinski, arsenm, mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D23068
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@277735 91177308-0d34-0410-b5e6-96231b3b80d8
