This patch adds the commandline option -mcompact-branches={never,optimal,always),
which controls how LLVM generates compact branches for MIPSR6 targets. By default,
the compact branch policy is 'optimal' where LLVM will generate the most
appropriate branch for any situation. The 'never' and 'always' policy will disable
or always generate compact branches wherever possible respectfully.
Reviewers: dsanders, vkalintiris, atanasyan
Differential Revision: http://reviews.llvm.org/D20729
llvm-svn: 271000
MergedInputSection::getOffset is the busiest function in LLD if string
merging is enabled and input files have lots of mergeable sections.
It is usually the case when creating executable with debug info,
so it is pretty common.
The reason why it is slow is because it has to do faily complex
computations. For non-mergeable sections, section contents are
contiguous in output, so in order to compute an output offset,
we only have to add the output section's base address to an input
offset. But for mergeable strings, section contents are split for
merging, so they are not contigous. We've got to do some lookups.
We used to do binary search on the list of section pieces.
It is slow because I think it's hostile to branch prediction.
This patch replaces it with hash table lookup. Seems it's working
pretty well. Below is "perf stat -r10" output when linking clang
with debug info. In this case this patch speeds up about 4%.
Before:
6584.153205 task-clock (msec) # 1.001 CPUs utilized ( +- 0.09% )
238 context-switches # 0.036 K/sec ( +- 6.59% )
0 cpu-migrations # 0.000 K/sec ( +- 50.92% )
1,067,675 page-faults # 0.162 M/sec ( +- 0.15% )
18,369,931,470 cycles # 2.790 GHz ( +- 0.09% )
9,640,680,143 stalled-cycles-frontend # 52.48% frontend cycles idle ( +- 0.18% )
<not supported> stalled-cycles-backend
21,206,747,787 instructions # 1.15 insns per cycle
# 0.45 stalled cycles per insn ( +- 0.04% )
3,817,398,032 branches # 579.786 M/sec ( +- 0.04% )
132,787,249 branch-misses # 3.48% of all branches ( +- 0.02% )
6.579106511 seconds time elapsed ( +- 0.09% )
After:
6312.317533 task-clock (msec) # 1.001 CPUs utilized ( +- 0.19% )
221 context-switches # 0.035 K/sec ( +- 4.11% )
1 cpu-migrations # 0.000 K/sec ( +- 45.21% )
1,280,775 page-faults # 0.203 M/sec ( +- 0.37% )
17,611,539,150 cycles # 2.790 GHz ( +- 0.19% )
10,285,148,569 stalled-cycles-frontend # 58.40% frontend cycles idle ( +- 0.30% )
<not supported> stalled-cycles-backend
18,794,779,900 instructions # 1.07 insns per cycle
# 0.55 stalled cycles per insn ( +- 0.03% )
3,287,450,865 branches # 520.799 M/sec ( +- 0.03% )
72,259,605 branch-misses # 2.20% of all branches ( +- 0.01% )
6.307411828 seconds time elapsed ( +- 0.19% )
Differential Revision: http://reviews.llvm.org/D20645
llvm-svn: 270999
Summary:
There are no llvm backend tests* for EABI and no EABI buildbots. There were only
three clang tests, all of which checked that -mabi=eabi was passed to the
assembler.
*There is a single backend test that specifies EABI but it actually tests MIPS16.
Reviewers: atanasyan
Subscribers: emaste, sdardis, atanasyan, cfe-commits
Differential Revision: http://reviews.llvm.org/D20679
llvm-svn: 270998
The isMemWithSimmOffset predicate rejects relocations which is incorrect
behaviour. Linkers and other tools should handle|warn|error when the
field overflows.
Reviewers: dsanders, vkalintiris
Differential Revision: http://reviews.llvm.org/D20727
llvm-svn: 270995
Currently we consider that each constant has itself as a base value. I.e "base(const) = const".
This introduces couple of problems when we are trying to avoid reporting constants in statepoint live sets:
1. When querying "base( phi(const1, const2) )" we will get "phi(const1, const2)" as a base pointer. Since
it's not a constant we will record it in a stack map. However on practice we don't want this to happen
(constant are never relocated).
2. base( phi(const, gc ptr) ) = phi( const, base(gc ptr) ). This particular case imposes challenge on our
runtime - we don't expect to see constant base pointers other than null. This problems can be avoided
by treating all constant as if they were derived from null pointer base. I.e in a first case we will
not include constant pointer in a stack map at all. In a second case we will get "phi(null, base(gc ptr))"
as a base pointer which is a lot more convenient.
Differential Revision: http://reviews.llvm.org/D20584
llvm-svn: 270993
It was: "Recommit 270977 - [llvm-mc] - Teach llvm-mc to generate zlib styled compression sections."
Fix:
since test requires no zlib available and r270987 changed the
compression flag for llvm-mc to mandatory specify the compression style,
then just add 2 available styles to this test.
llvm-svn: 270992
Register numbers may be specified as assembly-time expressions.
This feature can be useful in macros and alike. However, expressions
are supported within sqare braces only.
Sqare braces were initially intended to support specifying of multiple
(pairs/quads...) registers. Syntax like v[8:8] which specifies single register
is also supported. That allows expressions but looks a bit unnatural.
This change supports syntax REG[EXPR].
Tests added.
Differential Revision: http://reviews.llvm.org/D20588
llvm-svn: 270990
Fix: updated clang code which was not updated by mistake.
Original commit message:
[llvm-mc] - Teach llvm-mc to generate zlib styled compression sections.
This patch is strongly based on previously reverted D20331.
(because of gnuutils < 2.26 does not support compressed debug sections in non zlib-gnu style)
Difference that this patch supports both zlib and zlib-gnu styles.
-compress-debug-sections option now supports next values:
-compress-debug-sections=zlib-gnu
-compress-debug-sections=zlib
-compress-debug-sections=none
Previously specifying -compress-debug-sections enabled zlib-gnu compression,
so anyone can put "-compress-debug-sections=zlib-gnu" to restore the behavior
that was before this patch for case when compression was enabled.
Differential revision: http://reviews.llvm.org/D20676
llvm-svn: 270987
Summary:
This unifies mips/mipsel and mips64/mips64el into a single class so that we can
later support O32 on mips64/mips64el and N32/N64 on mips/mipsel (when an
appropriate CPU selected).
Reviewers: atanasyan
Subscribers: atanasyan, jfb, cfe-commits, dschuff
Differential Revision: http://reviews.llvm.org/D20678
llvm-svn: 270984
It broke buildbot:
http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-ubuntu-fast/builds/13585/steps/build/logs/stdio
Initial commit message:
[llvm-mc] - Teach llvm-mc to generate zlib styled compression sections.
This patch is strongly based on previously reverted D20331.
(because of gnuutils < 2.26 does not support compressed debug sections in non zlib-gnu style)
Difference that this patch supports both zlib and zlib-gnu styles.
-compress-debug-sections option now supports next values:
-compress-debug-sections=zlib-gnu
-compress-debug-sections=zlib
-compress-debug-sections=none
Previously specifying -compress-debug-sections enabled zlib-gnu compression,
so anyone can put "-compress-debug-sections=zlib-gnu" to restore the behavior
that was before this patch for case when compression was enabled.
Differential revision: http://reviews.llvm.org/D20676
llvm-svn: 270978
This patch is strongly based on previously reverted D20331.
(because of gnuutils < 2.26 does not support compressed debug sections in non zlib-gnu style)
Difference that this patch supports both zlib and zlib-gnu styles.
-compress-debug-sections option now supports next values:
-compress-debug-sections=zlib-gnu
-compress-debug-sections=zlib
-compress-debug-sections=none
Previously specifying -compress-debug-sections enabled zlib-gnu compression,
so anyone can put "-compress-debug-sections=zlib-gnu" to restore the behavior
that was before this patch for case when compression was enabled.
Differential revision: http://reviews.llvm.org/D20676
llvm-svn: 270977
It seems that suffix '@4HA' was omitted for unknown reason. It is
non-cont non-volatile 'int' type of normal variable TSS.
Differential revision: http://reviews.llvm.org/D20683
llvm-svn: 270974
This patch removes the llvm intrinsics VPMOVSX and (V)PMOVZX sign/zero extension intrinsics and auto-upgrades to SEXT/ZEXT calls instead. We already did this for SSE41 PMOVSX sometime ago so much of that implementation can be reused.
A companion patch (D20684) removes/auto-upgrade the clang intrinsics.
Differential Revision: http://reviews.llvm.org/D20686
llvm-svn: 270973
Cross unwinding requires larger sizes for unw_context_t and unw_cursor_t
buffers (large enough to hold the VRS of any architecture). This is not
desirable for the most common situation where libunwind is used for native
unwinding only.
This patch makes native unwinding the default build configuration. This
will start testing the tigher (compile-time) bounds of unw_context_t
and unw_cursor_t buffers for each architecture.
Change-Id: Ic61c476a75603ca4812959c233cfe56f83dc0b00
llvm-svn: 270972
This will be needed in order to consistently return an Error
to clients of the API being developed in D20268.
Differential Revision: http://reviews.llvm.org/D20550
llvm-svn: 270967
We would previously accept `--threads=4`, but this option just turns on
threading and does not specify a number of threads.
I ran into this by accident because I was passing `--threads=<n>` but
the number didn't seem to affect anything.
llvm-svn: 270963
r259714 introduces the transport method into the
URL passed to the gdb-remote stub. On debugserver,
this is not supported and prevented debugserver from
being launched by lldb-server in platform mode.
This change skips the transport method addition from
r259714 when on Apple hosts.
llvm-svn: 270961
Since we want to move toward zero-copy access to stream data, we
want to remove all instances of copying operations. So get rid
of some of those here.
Differential Revision: http://reviews.llvm.org/D20720
Reviewed By: ruiu
llvm-svn: 270960
APInt::operator+(uint64_t) just forwarded to operator+(const APInt&).
Constructing the APInt for the RHS takes an allocation which isn't
required. Also, for APInt's in the slow path, operator+ would
call add() internally which iterates over both arrays of values. Instead
we can use add_1 and sub_1 which only iterate while there is something to do.
Using the memory for 'opt -O2 verify-uselistorder.lto.opt.bc -o opt.bc'
(see r236629 for details), this reduces the number of allocations from
23.9M to 22.7M.
llvm-svn: 270959
Add a new CMake function (``add_libfuzzer_test()``) to simplify
declaration of executables for testing LibFuzzer and use it to
reorganise how tests are declared.
Note that configuration of the lit configuration files has been moved
as late as possible because we are going to need to disable some tests
for some platforms and we will need to propagate this information into
the lit configuration.
Note the code for custom mains was removed because no tests are
currently written for this and Kostya seems happy to remove this.
Differential Revision: http://reviews.llvm.org/D20706
llvm-svn: 270958
type_traits header in libstdc++ 4.8 does not define is_trivially_contructible
so the code doesn't compile with it.
In this file we are using the trait for assertion to provide a better
error message. Removing it doesn't change the meaning of the code.
Differential Revision: http://reviews.llvm.org/D20719
llvm-svn: 270957
This comment was included in Peter Collingbourne's original version of
StringError (see http://reviews.llvm.org/D20550), where it made sense. It was
accidentally copied over with the rest of the class, but no longer applies.
llvm-svn: 270956
objc_storeStrong can be formed from a sequence such as
%0 = tail call i8* @objc_retain(i8* %p) nounwind
%tmp = load i8*, i8** @x, align 8
store i8* %0, i8** @x, align 8
tail call void @objc_release(i8* %tmp) nounwind
The code was already looking through bitcasts for most of the values
involved, but had missed one case where the pointer operand for the
store was a bitcast. Ultimately the pointer for the load and store
have to be the same value, after stripping casts.
llvm-svn: 270955
_InterlockedIncrement and _InterlockedDecrement have 'long' in their
prototypes. We assumed 'long' was the same size as an i32 which is
incorrect for other targets.
This fixes PR27892.
llvm-svn: 270953
PDBs can be extremely large. We're already mapping the entire
PDB into the process's address space, but to make matters worse
the blocks of the PDB are not arranged contiguously. So, when
we have something like an array or a string embedded into the
stream, we have to make a copy. Since it's convenient to use
traditional data structures to iterate and manipulate these
records, we need the memory to be contiguous.
As a result of this, we were using roughly twice as much memory
as the file size of the PDB, because every stream was copied
out and re-stitched together contiguously.
This patch addresses this by improving the MappedBlockStream
to allocate from a BumpPtrAllocator only when a read requires
a discontiguous read. Furthermore, it introduces some data
structures backed by a stream which can iterate over both
fixed and variable length records of a PDB. Since everything
is backed by a stream and not a buffer, we can read almost
everything from the PDB with zero copies.
Differential Revision: http://reviews.llvm.org/D20654
Reviewed By: ruiu
llvm-svn: 270951
Simon Dardis