These instructions perform the same operation, but the semantic of which operand is destroyed is reversed. If the same register is used as both operands we can change the execution domain without worrying about this difference.
Unfortunately, this really only works in cases where the input register is killed by the instruction. If its not killed, the two address isntruction pass inserts a copy that will become a move instruction. This makes the instruction use different physical registers that contain the same data at the time the unpck/movhlps executes. I've considered using a unary pseudo instruction with tied operand to trick the two address instruction pass. We could then expand the pseudo post regalloc to get the same physical register on both inputs.
Differential Revision: https://reviews.llvm.org/D50157
llvm-svn: 338735
In this patch I add support for ARM and ARM64 break pad files. There are two flavors of ARM: Apple where FP is R7, and non Apple where FP is R11. Added minimal tests that load up ARM64 and the two flavors or ARM core files with a single thread and known register values in each register. Each register is checked for the exact value.
Differential Revision: https://reviews.llvm.org/D49750
llvm-svn: 338734
Summary:
This patch allows LLDB's Stream class to count the bytes it has written to so far.
There are two major motivations for this patch:
The first one is that this will allow us to get rid of all the handwritten byte counting code
we have in LLDB so far. Examples for this are pretty much all functions in LLDB that
take a Stream to write to and return a size_t, which usually represents the bytes written.
By moving to this centralized byte counting mechanism, we hopefully can avoid some
tricky errors that happen when some code forgets to count the written bytes while
writing something to a stream.
The second motivation is that this is needed for the migration away from LLDB's `Stream`
and towards LLVM's `raw_ostream`. My current plan is to start offering a fake raw_ostream
class that just forwards to a LLDB Stream.
However, for this raw_ostream wrapper we need to fulfill the raw_ostream interface with
LLDB's Stream, which currently lacks the ability to count the bytes written so far (which
raw_ostream exposes by it's `tell()` method). By adding this functionality it is trivial to start
rolling out our raw_ostream wrapper (and then eventually completely move to raw_ostream).
Also, once this fake raw_ostream is available, we can start replacing our own code writing
to LLDB's Stream by LLVM code writing to raw_ostream. The best example for this is the
LEB128 encoding we currently ship, which can be replaced with by LLVM's version which
accepts an raw_ostream.
From the point of view of the pure source changes this test does, we essentially just renamed
the Write implementation in Stream to `WriteImpl` while the `Write` method everyone is using
to write its raw bytes is now just forwarding and counting the written bytes.
Reviewers: labath, davide
Reviewed By: labath
Subscribers: JDevlieghere, lldb-commits
Differential Revision: https://reviews.llvm.org/D50159
llvm-svn: 338733
In according to the comment, undefined symbol should never reach there.
So, should be able to remove the check. I am assuming this is NFC.
llvm-svn: 338723
This is broken per PR36561 and PR36574, so disable it for now until
somebody interested can take a look. OMPT can still be activated manually
by passing -DLIBOMP_OMPT_SUPPORT=ON during configuration.
Differential Revision: https://reviews.llvm.org/D50086
llvm-svn: 338721
Only supported since GCC 6 and Intel 17.0. However GCC 6.3.0 is
crashing on two of the tests, so disable them as well...
Differential Revision: https://reviews.llvm.org/D50085
llvm-svn: 338720
This adds the NAN checks suggested in PR37776:
https://bugs.llvm.org/show_bug.cgi?id=37776
If both operands to maxnum are NAN, that should get constant folded, so we don't
have to handle that case. This is the same assumption as other FP ops in this
function. Returning 'false' is always conservatively correct.
Copying from the bug report:
Currently, we have this for "when is cannotBeOrderedLessThanZero
(mustBePositiveOrNaN) true for maxnum":
L
-------------------
| Pos | Neg | NaN |
------------------------
|Pos | x | x | x |
------------------------
R |Neg | x | | x |
------------------------
|NaN | x | x | x |
------------------------
The cases with (Neg & NaN) are wrong. We should have:
L
-------------------
| Pos | Neg | NaN |
------------------------
|Pos | x | x | x |
------------------------
R |Neg | x | | |
------------------------
|NaN | x | | x |
------------------------
Differential Revision: https://reviews.llvm.org/D50081
llvm-svn: 338716
It does not seem that this code is alive.
I seems was needed previously but we fixed it.
If it is still needed, it needs new tests,
but for now I do not know how to trigger it,
and so I removed it.
llvm-svn: 338713
The way address space declarations for builtins currently work
is nearly useless. The code assumes the address spaces used for
builtins is a confusingly named "target address space" from user
code using __attribute__((address_space(N))) that matches
the builtin declaration. There's no way to use this to declare
a builtin that returns a language specific address space.
The terminology used is highly cofusing since it has nothing
to do with the the address space selected by the target to use
for a language address space.
This feature is essentially unused as-is. AMDGPU and NVPTX
are the only in-tree targets attempting to use this. The AMDGPU
builtins certainly do not behave as intended (i.e. all of the
builtins returning pointers can never compile because the numbered
address space never matches the expected named address space).
The NVPTX builtins are missing tests for some, and the others
seem to rely on an implicit addrspacecast.
Change the used address space for builtins based on a target
hook to allow using a language address space for a builtin.
This allows the same builtin declaration to be used for multiple
languages with similarly purposed address spaces (e.g. the same
AMDGPU builtin can be used in OpenCL and CUDA even though the
constant address spaces are arbitarily different).
This breaks the possibility of using arbitrary numbered
address spaces alongside the named address spaces for builtins.
If this is an issue we probably need to introduce another builtin
declaration character to distinguish language address spaces from
so-called "target address spaces".
llvm-svn: 338707
Summary:
Previously, clang-format would crash if it tried to wrap an overlong
single line comment, because two parts of the code inserted a break in
the same location.
/** heregoesalongcommentwithnospace */
This wasn't previously noticed as it could only trigger for an overlong
single line comment that did have no breaking opportunities except for a
whitespace at the very beginning.
This also introduces a check for JavaScript to not ever wrap a comment
before an opening curly brace:
/** @mods {donotbreakbeforethecurly} */
This is because some machinery parsing these tags sometimes supports
breaks before a possible `{`, but in some other cases does not.
Previously clang-format was careful never to wrap a line with certain
tags on it. The better solution is to specifically disable wrapping
before the problematic token: this allows wrapping and aligning comments
but still avoids the problem.
Reviewers: krasimir
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D50177
llvm-svn: 338706
Some lit tests that call llvm-ar use the 'r' flag. If the target archive
already exists and is in a corrupt state, this can cause the test to fail. We
have added 'rm -f' calls before the llvm-ar calls to increase the
robustness of the tests.
Differential revision: https://reviews.llvm.org/D49184
llvm-svn: 338705
Corrected and simplified the help text.
It was clearly too difficult to maintain before (see e.g. @227296) making it
simpler and more consistent it should help people keep it up to date.
Differential Revision: https://reviews.llvm.org/D48577
llvm-svn: 338703
We don't need to use a map to store ResourceState objects. The number of
processor resources is known statically from the scheduling model. We can
therefore use a vector, and reserve a slot for each processor resource that we
want to simulate.
Every time the ResourceManager queries the ResourceState vector, the index to
the vector of ResourceState objects can be easily computed from the processor
resource mask.
This drastically reduces the time complexity of method ResourceManager::use() and
method ResourceManager::release(). This patch gives an average speedup of 12%.
llvm-svn: 338702
Patch by Konstantin Schwarz!
If more than a single output section is added to a PT_LOAD header,
only the first section should set the LMAOffset of the segment.
Otherwise, we get a load-address overlap error
Differential revision: https://reviews.llvm.org/D50133
llvm-svn: 338697
Summary: In order to exploit the potential of LLVM's new ItaniumPartialDemangler for indexing in LLDB, we expect conceptual changes in the implementation of the InitNameIndexes function. Here is a unit test that aims at covering all relevant code paths in that function.
Reviewers: labath, jingham, JDevlieghere
Reviewed By: JDevlieghere
Subscribers: friss, teemperor, davide, clayborg, erik.pilkington, lldb-commits, mgorny
Differential Revision: https://reviews.llvm.org/D49909
llvm-svn: 338695
Patch by Konstantin Schwarz!
If both the MemRegion and LMARegion are set for an output section in
a linker script, we should only increase the LMARegion if it is
different from the MemRegion. Otherwise, we reserve the memory twice.
Differential revision: https://reviews.llvm.org/D50065
llvm-svn: 338684
Patch by Konstantin Schwarz!
The condition to create a new phdr must also check the usage of "AT>"
linker script command, and create a new PT_LOAD header if a new LMARegion is used.
This fixes PR38307
Differential revision: https://reviews.llvm.org/D50052
llvm-svn: 338679
Adds some cleaned up debug messages from back when I was writing this.
Hopefully useful to others (and myself) as to why unroll and jam is not
transforming as expected.
Differential Revision: https://reviews.llvm.org/D50062
llvm-svn: 338676
As a part of adding the tiny codemodel, we need to support ldr's with :got:
relocations on them. This seems to be mostly already done, just needs the
relocation type support.
Differential Revision: https://reviews.llvm.org/D50137
llvm-svn: 338673
This method has three callers, each of which wanted distinct handling:
1) Sinking into a loop is moving an instruction known to execute before a loop into the loop. We don't need to worry about introducing a fault at all in this case.
2) Hoisting from a loop into a preheader already duplicated the check in the caller.
3) Sinking from the loop into an exit block was the only true user of the code within the routine. For the moment, this has just been lifted into the caller, but up next is examining the logic more carefully. Whitelisting of loads and calls - while consistent with the previous code - is rather suspicious. Either way, a behavior change is worthy of it's own patch.
llvm-svn: 338671
Craig Topper