r231483 taught ConstantRange::multiply to be clever about signed vs unsigned ranges. For example, an unsigned range could be full-set while the signed range is more specific than that.
In looking at the allocations trace for LTO'ing verify-uselistorder (see r236629 for details), millions of allocations are from APInt, many of which come from ConstantRange's.
This change tries to avoid some (3.2 million) allocations by returning the unsigned range if its suitable. The checks here are that it should not be a wrapping range, and should be positive. That should be enough to check for ranges such as [1, 10) which the signed range will be equal to, if we were to calculate it.
Differential Revision: http://reviews.llvm.org/D20723
Reviewed by James Molloy
llvm-svn: 271020
When we traced through a phi node looking for floating-point reductions, we
forgot whether we'd ever seen an instruction without fast-math flags (that
would block vectorization). This propagates it through to the end.
llvm-svn: 271015
Summary: Sample profile pass need to have instcombine pass. A related change is http://reviews.llvm.org/D17742. But we should not explicitly add dependency between to non-analysis passes. So we add the dependency here.
Reviewers: davidxl, dnovillo
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D20502
llvm-svn: 271010
unwind_phase1 and unwind_phase2 allocate their own copies of unw_cursor_t buffers
on the stack. This can blow-up stack usage of the unwinder depending on how these
two functions get inlined into _Unwind_RaiseException. Clang seems to inline
unwind_phase1 into _Unwind_RaiseException but not unwind_phase2, thus creating
two unw_cursor_t buffers on the stack.
One way to work-around this problem is to mark both unwind_phase1 and
unwind_phase2 as noinline. This patch takes the less compiler-dependent approach
and explicitly allocate a unw_cursor_t buffer and pass that into unwind_phase1
and unwind_phase2 functions.
A follow-up patch will replicate this behavior for the non-EHABI and non-SJLJ
implementations.
Reviewers: jroelofs, bcraig.
Differential revision: http://reviews.llvm.org/D20320
llvm-svn: 271004
Summary:
Turn off lifetime-start-on-first-use enhancement for the moment
pending a fix for bug 27903.
Bug: 27903
Reviewers: tejohnson, wmi, qcolombet, gbiv
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D20731
llvm-svn: 271003
Summary:
Target regions require globals to be captured. This patch fixes a bug exposed when that happens in a template function.
Reviewers: hfinkel, arpith-jacob, kkwli0, carlo.bertolli, ABataev
Subscribers: guansong, ABataev, cfe-commits, caomhin, fraggamuffin
Differential Revision: http://reviews.llvm.org/D18110
llvm-svn: 271001
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
Pete Cooper