Matthias Braun [Wed, 25 Nov 2015 00:50:47 +0000 (00:50 +0000)]
Doxygen: Use mathjax to create formulas.
The main motivation is to not require a latex installation when building
the documentation. I would also expect a better image quality and the
ability to copy&paste from formulas with a javascript based solution for
displaying the math.
Teresa Johnson [Tue, 24 Nov 2015 22:55:46 +0000 (22:55 +0000)]
[ThinLTO] Add option to limit importing based on instruction count
Add a simple initial heuristic to control importing based on the number
of instructions recorded in the function's summary. Add option to
control the limit, and test using option.
Rong Xu [Tue, 24 Nov 2015 22:50:34 +0000 (22:50 +0000)]
[PGO] Relax test cases in PGO instrumentation
Fix buildbot failure for clang-x86_64-linux-selfhost-modules.
http://lab.llvm.org:8011/builders/clang-x86_64-linux-selfhost-modules/builds/8866
The failing test cases are newly added from r254021. It seems the IR has a
different order in this platform. In this patch, I temporarily relax the test
case to make the build green. I'll have a complete fix (more robust way to test)
soon.
Diego Novillo [Tue, 24 Nov 2015 22:38:37 +0000 (22:38 +0000)]
SamplePGO - Add test for hot/cold inlined functions.
When the original binary is executed and sampled, the resulting profile
contains information on the original inline stack. We currently follow
the original inline plan if we notice that the inlined callsite has more
than 0 samples to it.
A better way is to determine whether the callsite is actually worth
inlining. If the callsite accumulates a small fraction of the samples
spent in the parent function, then we don't want to bother inlining it
(as it means that the callsite is actually cold).
This patch introduces a threshold expressed in percentage of samples
in relation to the parent function. If the callsite uses less than N%
of the total samples used by its parent, the original inline decision is
not re-applied.
I've set the threshold to the very arbitrary value of 5%. I'm yet to do
any actual experiments to see what's a good value. I wanted to separate
the basic mechanism from the tuning.
Evgeniy Stepanov [Tue, 24 Nov 2015 21:44:16 +0000 (21:44 +0000)]
[msan] Relax origin-alignment test.
Change origin-alignment test to test only the alignment of the origin
store, and not the exact instruction sequence used to compute the
address. This makes the test less fragile and, in particular, lets it
pass both with the old and new MSan ABIs.
Rong Xu [Tue, 24 Nov 2015 21:31:25 +0000 (21:31 +0000)]
[PGO] MST based PGO instrumentation infrastructure
This patch implements a minimum spanning tree (MST) based instrumentation for
PGO. The use of MST guarantees minimum number of CFG edges getting
instrumented. An addition optimization is to instrument the less executed
edges to further reduce the instrumentation overhead. The patch contains both the
instrumentation and the use of the profile to set the branch weights.
Sanjoy Das [Tue, 24 Nov 2015 20:37:01 +0000 (20:37 +0000)]
[RuntimeDyld] Fix a class of arithmetic errors introduced in r253918
r253918 had refactored expressions like "A - B.Address + C" to "A -
B.getAddressWithOffset(C)". This is incorrect, since the latter really
computes "A - B.Address - C".
None of the tests I can run locally on x86 broke due to this bug, but it
is the current suspect for breakage on the AArch64 buildbots.
Simon Pilgrim [Tue, 24 Nov 2015 20:31:46 +0000 (20:31 +0000)]
[X86][FMA] Optimize FNEG(FMA) Patterns
X86 needs to use its own FMA opcodes, preventing the standard FNEG(FMA) pattern table recognition method used by other platforms. This patch adds support for lowering FNEG(FMA(X,Y,Z)) into a single suitably negated FMA instruction.
Teresa Johnson [Tue, 24 Nov 2015 19:55:04 +0000 (19:55 +0000)]
[ThinLTO] Enable iterative importing in FunctionImport pass
Analyze imported function bodies and add any new external calls to
the worklist for importing. Currently no controls on the importing
so this will end up importing everything possible in the call tree
below the importing module. Basic profitability checks coming next.
Update test to check for iteratively inlined functions.
Cong Hou [Tue, 24 Nov 2015 19:51:26 +0000 (19:51 +0000)]
[X86] Fix several issues related to X86's psadbw instruction.
This patch fixes the following issues:
1. Fix the return type of X86psadbw: it should not be the same type of inputs.
For vNi8 inputs the output should be vMi64, where M = N/8.
2. Fix the return type of int_x86_avx512_psad_bw_512 accordingly.
3. Fix the definiton of PSADBW, VPSADBW, and VPSADBWY accordingly.
4. Adjust the return type when building a DAG node of X86ISD::PSADBW type.
5. Update related tests.
Teresa Johnson [Tue, 24 Nov 2015 19:46:58 +0000 (19:46 +0000)]
[ThinLTO] Handle previously imported and promoted locals in module linker
The new function import pass exposed an issue when we import references
to local values on multiple importing passes. They are renamed on each
import pass, and we need to ensure that the already promoted and renamed
references existing in the dest module are correctly identified and
updated so that they aren't spuriously renamed again (due to a perceived
conflict with the newly linked reference).
The closure is designed to abstact away two types of value profile
data:
- InstrProfRecord which is the primary data structure used to
represent profile data in host tools (reader, writer, and profile-use)
- value profile runtime data structure suitable to be used by C
runtime library.
Both sources of data need to serialize to disk/memory-buffer in common
format: ValueProfData.
The abstraction allows compiler-rt's raw profiler writer to share
the same code with indexed profile writer.
As shown in the test case and PR25554:
https://llvm.org/bugs/show_bug.cgi?id=25554
This causes us to miss reusing an operand because later passes don't know these 'movq' are the same instruction.
This patch deletes one pair of these defs.
Sadly, this won't fix the original test case in the bug report. Something else is still broken.
Matt Arsenault [Tue, 24 Nov 2015 12:05:03 +0000 (12:05 +0000)]
AMDGPU: Split x8 and x16 vector loads instead of scalarize
The one regression in the builtin tests is in the read2 test which now
(again) has many extra copies, but this should be solved once the pass
is replaced with a DAG combine.
Pavel Labath [Tue, 24 Nov 2015 09:46:01 +0000 (09:46 +0000)]
Fix non-PIC build after 253959
CMAKE_EXE_LINKER_FLAGS is a string. Appending a flag using list(APPEND) introduces an extra
semicolon which breaks stuff. Change this to append the value in the same way that everyone else
seems to be doing.
Cong Hou [Tue, 24 Nov 2015 08:51:23 +0000 (08:51 +0000)]
Let SelectionDAG start to use probability-based interface to add successors.
The patch in http://reviews.llvm.org/D13745 is broken into four parts:
1. New interfaces without functional changes.
2. Use new interfaces in SelectionDAG, while in other passes treat probabilities
as weights.
3. Use new interfaces in all other passes.
4. Remove old interfaces.
This the second patch above. In this patch SelectionDAG starts to use
probability-based interfaces in MBB to add successors but other MC passes are
still using weight-based interfaces. Therefore, we need to maintain correct
weight list in MBB even when probability-based interfaces are used. This is
done by updating weight list in probability-based interfaces by treating the
numerator of probabilities as weights. This change affects many test cases
that check successor weight values. I will update those test cases once this
patch looks good to you.
Chris Bieneman [Tue, 24 Nov 2015 08:04:59 +0000 (08:04 +0000)]
[CMake] When disabling PIC, also pass -fno-pie when linking if it is supported.
Building clang with -fno-pie generates slightly faster code. In my not-very-rigorous testing I saw about a 4% speed up using the clang test-suite sources.
Cong Hou [Tue, 24 Nov 2015 05:44:19 +0000 (05:44 +0000)]
[X86][SSE] Detect AVG pattern during instruction combine for SSE2/AVX2/AVX512BW.
This patch detects the AVG pattern in vectorized code, which is simply
c = (a + b + 1) / 2, where a, b, and c have the same type which are vectors of
either unsigned i8 or unsigned i16. In the IR, i8/i16 will be promoted to
i32 before any arithmetic operations. The following IR shows such an example:
%1 = zext <N x i8> %a to <N x i32>
%2 = zext <N x i8> %b to <N x i32>
%3 = add nuw nsw <N x i32> %1, <i32 1 x N>
%4 = add nuw nsw <N x i32> %3, %2
%5 = lshr <N x i32> %N, <i32 1 x N>
%6 = trunc <N x i32> %5 to <N x i8>
and with this patch it will be converted to a X86ISD::AVG instruction.
The pattern recognition is done when combining instructions just before type
legalization during instruction selection. We do it here because after type
legalization, it is much more difficult to do pattern recognition based
on many instructions that are doing type conversions. Therefore, for
target-specific instructions (like X86ISD::AVG), we need to take care of type
legalization by ourselves. However, as X86ISD::AVG behaves similarly to
ISD::ADD, I am wondering if there is a way to legalize operands and result
types of X86ISD::AVG together with ISD::ADD. It seems that the current design
doesn't support this idea.
Tests are added for SSE2, AVX2, and AVX512BW and both i8 and i16 types of
variant vector sizes.
Andy Ayers [Mon, 23 Nov 2015 22:17:44 +0000 (22:17 +0000)]
findDeadCallerSavedReg needs to pay attention to calling convention
Caller saved regs differ between SysV and Win64. Use the tail call available set to scavenge from.
Refactor register info to create new helper to get at tail call GPRs. Added a new test case for windows. Fixed up a number of X64 tests since now RCX is preferred over RDX on SysV.
With the '=' suffix now indicating which operands are output operands, it's
no longer as important to distinguish between a call's inputs and its outputs
using operand ordering, so we can go back to printing them in the normal order.
Dan Gohman [Mon, 23 Nov 2015 21:55:57 +0000 (21:55 +0000)]
[WebAssembly] Suffix output operands with '='.
This distinguishes input operands from output operands. This is something of
a syntactic experiment to see whether the mild amount of clutter this adds is
outweighed by the extra information it conveys to the reader.
Sanjoy Das [Mon, 23 Nov 2015 21:47:41 +0000 (21:47 +0000)]
[RuntimeDyld] Add accessors to `SectionEntry`; NFC
Summary:
Remove naked access to the data members in `SectionEntry` and route
accesses through accessor functions. This makes it obvious how the
instances of the class are used, and will also facilitate adding bounds
checking to `advanceStubOffset` in a later change.
Diego Novillo [Mon, 23 Nov 2015 20:12:21 +0000 (20:12 +0000)]
SamplePGO - Add coverage tracking for samples.
The existing coverage tracker counts the number of records that were used
from the input profile. An alternative view of coverage is to check how
many available samples were applied.
This way, if the profile contains several records with few samples, it
doesn't really matter much that they were not applied. The more
interesting records to apply are the ones that contribute many samples.
Dan Gohman [Mon, 23 Nov 2015 19:30:43 +0000 (19:30 +0000)]
[WebAssembly] Don't use set_local instructions explicitly.
The current approach to using get_local and set_local is to use them
implicitly, as register uses and defs. Introduce new copy instructions
which are themselves no-ops except for the get_local and set_local
that they imply, so that we use get_local and set_local consistently.
Dan Gohman [Mon, 23 Nov 2015 19:12:37 +0000 (19:12 +0000)]
[WebAssembly] Always print loop end labels
WebAssembly is currently using labels to end scopes, so for example a
loop scope looks like this:
BB0_0:
loop BB0_1
...
BB0_1:
with BB0_0 being the label of the first block not in the loop. This
requires that the label be printed even when it's only reachable via
fallthrough. To arrange this, insert a no-op LOOP_END instruction in
such cases at the end of the loop.
Dan Gohman [Mon, 23 Nov 2015 16:19:56 +0000 (16:19 +0000)]
[WebAssembly] Use dominator information to improve BLOCK placement
Always starting blocks at the top of their containing loops works, but creates
unnecessarily deep nesting because it makes all blocks in a loop overlap.
Refine the BLOCK placement algorithm to start blocks at nearest common
dominating points instead, which significantly shrinks them and reduces
overlapping.
[Support] Fix SaturatingMultiply<T>() to be correct (and fast), Re-enable Unit Tests
Summary:
This change fixes the SaturatingMultiply<T>() function template to not cause undefined behavior with T=uint16_t.
Thanks to Richard Smith's contribution, it also no longer requires an integer division.
projects / llvm / log