In general, it's difficult to poke the ConstantExpr code in CFLAA, since
LLVM is so great at eagerly reducing ConstantExprs. :)
Sadly, this only shows a functional difference from before the patch
because CFLAA has some special logic around taking loads of non-pointers
into account. Namely, with the broken select behavior, CFLAA will
completely fail to take note of @g3. Since CFLAA doesn't have any record
about @g3 when we do an alias query for @g3 and %a, it conservatively
answers MayAlias. When we properly take @g3 into account with the new
select logic, we get NoAlias for this query.
I suspect that the aforementioned "special logic" isn't completely
correct, but this test-case should prevent future wonky aliasing results
from appearing for these flavors of ConstantExprs, so I think it's still
worth having.
llvm-svn: 332017
Accessing the members of a large data structures needs a lot of GEPs which
usually have large offsets due to the size of the underlying data structure. If
the offsets are too large to fit into the r+i addressing mode, these GEPs cannot
be sunk to their users' blocks and many extra registers are needed then to carry
the values of these GEPs.
This patch tries to split a large data struct starting from %base like the
following.
Before:
BB0:
%base =
BB1:
%gep0 = gep %base, off0
%gep1 = gep %base, off1
%gep2 = gep %base, off2
BB2:
%load1 = load %gep0
%load2 = load %gep1
%load3 = load %gep2
After:
BB0:
%base =
%new_base = gep %base, off0
BB1:
%new_gep0 = %new_base
%new_gep1 = gep %new_base, off1 - off0
%new_gep2 = gep %new_base, off2 - off0
BB2:
%load1 = load i32, i32* %new_gep0
%load2 = load i32, i32* %new_gep1
%load3 = load i32, i32* %new_gep2
In the above example, the struct is split into two parts. The first part still
starts from %base and the second part starts from %new_base. After the
splitting, %new_gep1 and %new_gep2 have smaller offsets and then can be sunk to
BB2 and folded into their users.
The algorithm to split data structure is simple and very similar to the work of
merging SExts. First, it collects GEPs that have large offsets when iterating
the blocks. Second, it splits the underlying data structures and updates the
collected GEPs to use smaller offsets.
Differential Revision: https://reviews.llvm.org/D42759
llvm-svn: 332015
Summary: The final -wasm component has been the default for some time now.
Subscribers: jfb, dschuff, jgravelle-google, eraman, aheejin, JDevlieghere, sunfish, llvm-commits
Differential Revision: https://reviews.llvm.org/D46342
llvm-svn: 332007
This is a follow up to the rL330983. The patch teaches ld, sd, and lld
commands accept 32-bit memory offsets by replacing `mem_simm16` operand
to `mem_simmptr`. In fact, these commands should accept 64-bit offsets,
but so large offsets require another command expanding and will be
supported by a separate patch.
Differential Revision: https://reviews.llvm.org/D46629
llvm-svn: 331997
This is a follow up to the rL330983. The patch teaches lh and lhu
commands accepts 32-bit memory offsets by replacing `mem_simm16` operand
to `mem_simmptr`.
Differential Revision: https://reviews.llvm.org/D46513
llvm-svn: 331996
This diff slightly reorganizes the tests and improves
the test coverage of help messages / error reports.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D46589
llvm-svn: 331993
With nnan, there's no need for the masked merge / blend
sequence (that probably costs much more than the min/max
instruction).
Somewhere between clang 5.0 and 6.0, we started producing
these intrinsics for fmax()/fmin() in C source instead of
libcalls or fcmp/select. The backend wasn't prepared for
that, so we regressed perf in those cases.
Note: it's possible that other targets have similar problems
as seen here.
Noticed while investigating PR37403 and related bugs:
https://bugs.llvm.org/show_bug.cgi?id=37403
The IR FMF propagation cases still don't work. There's
a proposal that might fix those cases in D46563.
llvm-svn: 331992
Summary:
This change teaches DSE that the atomic memory intrinsics can be overwriten
partially in the same way as the non-atomic forms. Specifically, that the
atomic memcpy & memset can be shortened at the end and that the atomic memset
can be shortened at the beginning, if they partially overwritten
by later stores.
Reviewers: mkazantsev, skatkov, apilipenko, efriedma, rsmith, spatel, filcab, sanjoy
Reviewed By: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D45584
llvm-svn: 331991
Fixes bug https://bugs.llvm.org/show_bug.cgi?id=37339.
InlineAsm is only uniqued if the FunctionTypes are exactly the
same, while cmpTypes() for example considers all pointer types
in the default address space to be the same. For this reason
the end of cmpInlineAsm() can be reached.
This patch replaces the unreachable assertion with a check that
the function types are not identical.
Differential Revision: https://reviews.llvm.org/D46495
Reviewers: jfb
llvm-svn: 331990
Clang 6.0 was updated to create these intrinsics rather than
libcalls or fcmp/select, but the backend wasn't prepared to
handle that optimally.
This bug is not the primary reason for PR37403:
https://bugs.llvm.org/show_bug.cgi?id=37403
...but it's probably more important for x86 perf.
llvm-svn: 331988
Summary:
The combine in rebuildSetCC may be combined to another
node leaving our references stale. Keep a handle on
it to avoid stale references.
Fixes PR36602.
Reviewers: dbabokin, RKSimon, eli.friedman, davide
Subscribers: hiraditya, uabelho, JesperAntonsson, qcolombet, llvm-commits
Differential Revision: https://reviews.llvm.org/D46404
llvm-svn: 331985
Reviewed by: dblaikie, JDevlieghere, espindola
Differential Revision: https://reviews.llvm.org/D44560
Summary:
The .debug_line parser previously reported errors by printing to stderr and
return false. This is not particularly helpful for clients of the library code,
as it prevents them from handling the errors in a manner based on the calling
context. This change switches to using llvm::Error and callbacks to indicate
what problems were detected during parsing, and has updated clients to handle
the errors in a location-specific manner. In general, this means that they
continue to do the same thing to external users. Below, I have outlined what
the known behaviour changes are, relating to this change.
There are two levels of "errors" in the new error mechanism, to broadly
distinguish between different fail states of the parser, since not every
failure will prevent parsing of the unit, or of subsequent unit. Malformed
table errors that prevent reading the remainder of the table (reported by
returning them) and other minor issues representing problems with parsing that
do not prevent attempting to continue reading the table (reported by calling a
specified callback funciton). The only example of this currently is when the
last sequence of a unit is unterminated. However, I think it would be good to
change the handling of unrecognised opcodes to report as minor issues as well,
rather than just printing to the stream if --verbose is used (this would be a
subsequent change however).
I have substantially extended the DwarfGenerator to be able to handle
custom-crafted .debug_line sections, allowing for comprehensive unit-testing
of the parser code. For now, I am just adding unit tests to cover the basic
error reporting, and positive cases, and do not currently intend to test every
part of the parser, although the framework should be sufficient to do so at a
later point.
Known behaviour changes:
- The dump function in DWARFContext now does not attempt to read subsequent
tables when searching for a specific offset, if the unit length field of a
table before the specified offset is a reserved value.
- getOrParseLineTable now returns a useful Error if an invalid offset is
encountered, rather than simply a nullptr.
- The parse functions no longer use `WithColor::warning` directly to report
errors, allowing LLD to call its own warning function.
- The existing parse error messages have been updated to not specifically
include "warning" in their message, allowing consumers to determine what
severity the problem is.
- If the line table version field appears to have a value less than 2, an
informative error is returned, instead of just false.
- If the line table unit length field uses a reserved value, an informative
error is returned, instead of just false.
- Dumping of .debug_line.dwo sections is now implemented the same as regular
.debug_line sections.
- Verbose dumping of .debug_line[.dwo] sections now prints the prologue, if
there is a prologue error, just like non-verbose dumping.
As a helper for the generator code, I have re-added emitInt64 to the
AsmPrinter code. This previously existed, but was removed way back in r100296,
presumably because it was dead at the time.
This change also requires a change to LLD, which will be committed separately.
llvm-svn: 331971
Const/local/shared address spaces are all < 4GB and we can always use
32-bit pointers to access them. This has substantial performance impact
on kernels that uses shared memory for intermediary results.
The feature is disabled by default.
Differential Revision: https://reviews.llvm.org/D46147
llvm-svn: 331941
This is a follow-up to D45986. As suggested there, we should match the "all-bits-set"
pattern in addition to "any-bits-set".
This was a little more complicated than I thought it would be initially because the
"and 1" instruction can be anywhere in the chain. Hopefully, the code comments make
that logic understandable, but if you see a way to simplify or improve that, it's
most appreciated.
This transforms patterns that emerge from bitfield tests as seen in PR37098:
https://bugs.llvm.org/show_bug.cgi?id=37098
I think it would also help reduce the large test from:
D46336
D46595
but we need something to reassociate that case to the forms we're expecting here first.
Differential Revision: https://reviews.llvm.org/D46649
llvm-svn: 331937
The previous handling for guard widening in InstCombine was extremely restrictive. In particular, it didn't handle the common case where we had two guards separated by a single icmp. Handle this by scanning through a small fixed window of instructions to find the next guard if needed.
Differential Revision: https://reviews.llvm.org/D46203
llvm-svn: 331935
This is safe as long as the udiv is not exact. The pattern is not common in
C++ code, but comes up all the time in code generated by XLA's GPU backend.
Differential Revision: https://reviews.llvm.org/D46647
llvm-svn: 331933
The second source operand of G_SHL, G_ASHR, and G_LSHR must preserve its
value as a (small) unsigned integer, therefore its incorrect to widen it
in any way but by zero extending it.
G_SHL was using G_ANYEXT and G_ASHR - G_SEXT (which is correct for their
destination and first source operands, but not the "number of bits to
shift" operand).
Generally, shifts aren't as similar to regular binary operations as it
might seem, for instance, they aren't commutative nor associative and
the second source operand usually requires a special treatment.
Reviewers: bogner, javed.absar, aivchenk, rovka
Reviewed By: bogner
Subscribers: igorb, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D46413
llvm-svn: 331926
If a multiply is truncated, SimplifyDemandedBits
sometimes turns a zero_extend of the inputs into an
any_extend, which makes the known bits computation unhelpful.
Ignore these and compute known bits for the underlying value,
since we insert the correct extend type after.
llvm-svn: 331919
This is an extension of an existing combine to reduce wider
shls if the result fits in the final result type. This
introduces the same combine, but reduces the shift to a middle
sized type to avoid the slow 64-bit shift.
llvm-svn: 331916
If the truncate is only accessing the first element of the vector,
we can use the original source value.
This helps with some combine ordering issues after operations are
lowered to integer operations between bitcasts of build_vector.
In particular it stops unnecessarily materializing the unused
top half of a vector in some cases.
llvm-svn: 331909
When preprocessing resource scripts (which can easily be done outside
of llvm-rc), included headers can leave behind C declarations (despite
preprocessing with -DRC_INVOKED), that can't be parsed by a resource
compiler.
This is handled in all of rc.exe, by parsing the preprocessor output
line markers and ignoring content from files named *.h and *.c,
documented at [1].
In addition to this filtering, strip out any other preprocessor directive
that is left behind (like pragmas) which also can't be handled by the
tokenizer.
The added test uses both standard #line markers (supported by rc.exe) and
GNU style extended line markers, thus this test doesn't pass with rc.exe,
but passes with GNU windres. (Windres on the other hand doesn't filter
out files named *.c, only *.h.)
Differential Revision: https://reviews.llvm.org/D46579
[1] https://msdn.microsoft.com/en-us/library/windows/desktop/aa381033(v=vs.85).aspx
llvm-svn: 331903
This is the same as any other user defined resource, but with
a specific allocated resource type number.
Differential Revision: https://reviews.llvm.org/D46636
llvm-svn: 331902
-1 is commonly used as ID for controls that one don't want to
refer to later. For DIALOG resources, the IDs are 16 bit numbers,
and -1 gets interpreted as UINT32_MAX earlier, which then later is
too large to write into a uint16_t.
Differential Revision: https://reviews.llvm.org/D46506
llvm-svn: 331901
Previously if !LegalOperations we would blindly call getBitcast and hope that getNode would constant fold it. But if the conversion is between a vector and a scalar, getNode has no simplification.
This means we would just get back the original N. We would then return that N which would make the caller of visitBITCAST think that we used CombineTo and did our own worklist management. This prevents target specific optimizations from being called for vector/scalar bitcasts until after legal operations.
llvm-svn: 331896
With custom lowering for vector MULLH{S,U}, it is now profitable to
vectorize a divide by constant loop for the custom types (v16i8, v8i16,
and v4i32). The cost if based on TargetLowering::Build{S,U}DIV which
uses a multiply by constant plus adjustment to express a divide by
constant.
Both {u,s}mull{2} are expressed as Instruction::Mul and shifts by
Instruction::AShr.
llvm-svn: 331873
Its only two uses were removed in r311730.
Effectively reverts r304851 (but that code has removed around a bit since then).
https://reviews.llvm.org/D46619
clang side done in r331871.
llvm-svn: 331872
The new verifier check has found an error in the
debug-names-name-collisions.ll test on the PS4 bot:
error: Name Index @ 0x0: Entry @ 0xdc: mismatched Name of DIE @ 0x23: index - _ZN3foo3fooE; debug_info - foo.
Reverting while I investigate whether this is a bug in the verifier or
the generator.
This reverts commit r331868.
llvm-svn: 331869
Summary:
This patch implements a check which makes sure all entries required by
the DWARF v5 specification are present in the Name Index. The algorithm
tries to follow the wording of Section 6.1.1.1 of the spec as closely as
possible.
The main deviation from it is that instead of a whitelist-based approach
in the spec "The name index must contain an entry for each debugging
information entry that defines a named subprogram, label, variable,
type, or namespace" I chose a blacklist-based one, where I consider
everything to be "in" and then remove the entries that don't make sense.
I did this because it has more potential for catching interesting cases
and the above is a bit vague (it uses plain words like "variable" and
"subprogram", but the rest of the section speaks about specific TAGs).
This approach has raised some interesting questions, the main one being
whether enumerator values should be indexed. The consensus seems to be
that they should, although it does not follow from section 6.1.1.1.
For the time being I made the verifier ignore these, as LLVM does not do
this yet, and I wanted to get a clean run when verifying generated debug
info.
Another interesting case was the DW_TAG_imported_declaration. It was not
immediately clear to me whether this should go in or not, but currently
it is not indexed, and (unlike the enumerators) in does not seem to cause
problems for LLDB, so I've also ignored it.
Reviewers: JDevlieghere, aprantl, dblaikie
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D46583
llvm-svn: 331868
MOVNTPD/MOVNTPS should be WriteFStore
Standardized BDW/HSW/SKL/SKX WriteFStore/WriteVecStore - fixes some missed instregex patterns. (V)MASKMOVDQU was already using the default, its costs gets increased but is still nowhere near the real cost of that nasty instruction....
llvm-svn: 331864
The operator == used for exporting a function with a different
name in the DLL compared to the name in the import library
(which is useful for adding linker level aliases for function
in the import library) is a feature distinct and different from
the operator = used for exporting a function with a different
name (both in import library and DLL) than in the implementation
producing the DLL.
When creating an import library using dlltool, from a def file that
contains forwards (Func = OtherDll.Func), this shouldn't affect the
produced import library, which should still behave just as if it
was a normal exported function.
This clears a lot of confusion and subtle misunderstandings, and
avoids a parameter that was used to avoid creating weak aliases
when invoked from lld. (This parameter was added previously due to
the existing conflation of the two features.)
Differential Revision: https://reviews.llvm.org/D46245
llvm-svn: 331859
George Burgess IV