Commit Graph

120279 Commits

Author SHA1 Message Date
Sanjay Patel 0f51d14957 add a missing regression test for a DAGCombiner FDIV optimization
There's no test for this transform in any backend. Discovered
while debugging fast-math-flag propagation in the DAG (r244053).

llvm-svn: 244373
2015-08-07 23:19:41 +00:00
Tom Stellard fd25395c72 AMDGPU: Add pass to lower OpenCL image and sampler arguments.
The pass adds new kernel arguments for image attributes, and
resolves calls to dummy attribute and resource id getter functions.

Patch by: Zoltan Gilian

llvm-svn: 244372
2015-08-07 23:19:30 +00:00
James Y Knight 0cab80c9b3 [SPARC] Disable unsupported ExecutionEngine tests, and XFAIL a couple
of DebugInfo tests.

llvm-svn: 244371
2015-08-07 23:01:16 +00:00
Adam Nemet 155e8741f3 [LAA] Remove unused pointer partition argument from getNumberOfChecks, NFC
This is unused after filtering checks was moved to the clients.

As a result, we can just return the number of the checks in the
precomputed set.

llvm-svn: 244369
2015-08-07 22:44:21 +00:00
Adam Nemet 15840393f3 [LAA] Make the set of runtime checks part of the state of LAA, NFC
This is the full set of checks that clients can further filter. IOW,
it's client-agnostic.  This makes LAA complete in the sense that it now
provides the two main results of its analysis precomputed:

1. memory dependences via getDepChecker().getInsterestingDependences()
2. run-time checks via getRuntimePointerCheck().getChecks()

However, as a consequence we now compute this information pro-actively.
Thus if the client decides to skip the loop based on the dependences
we've computed the checks unnecessarily.  In order to see whether this
was a significant overhead I checked compile time on SPEC2k6 LTO bitcode
files.  The change was in the noise.

The checks are generated in canCheckPtrAtRT, at the same place where we
used to call groupChecks to merge checks.

llvm-svn: 244368
2015-08-07 22:44:15 +00:00
Quentin Colombet 7d8c74ff3f [AArch64][LoadStoreOptimizer] Turn a test into an assert. NFC.
At this point the given Opc must be valid, otherwise we should
not look for a matching pair to form paired load or store.

Thanks to Chad to point out this piece of code!

llvm-svn: 244366
2015-08-07 22:40:51 +00:00
Tom Stellard 8ebad11ee9 AMDGPU/SI: Use InstAlias instead of MnemonicAlias for VOPC instructions
Summary:
With InstAlias, we don't need to print the _e32 portion of the mnemonic
when we print the $dst operand.  This change makes it possible to
include vcc in the asm string when we switch VOPC over to having
implicit vcc defs.

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11813

llvm-svn: 244362
2015-08-07 22:00:56 +00:00
Sanjay Patel 8d768b5b3a redo r244360 (tighten checks...) after specifying triple
llvm-svn: 244361
2015-08-07 21:42:24 +00:00
Sanjay Patel 0e1d731631 tighten checks using update_llc_test_checks.py
llvm-svn: 244360
2015-08-07 21:38:53 +00:00
Alex Lorenz 61420f790d MIR Serialization: Serialize the base alignment for the machine memory operands.
llvm-svn: 244357
2015-08-07 20:48:30 +00:00
Alex Lorenz 83127739ff MIR Serialization: Serialize the offsets for the machine memory operands.
llvm-svn: 244356
2015-08-07 20:26:52 +00:00
Alex Lorenz dc24c1713e MIR Parser: Extract the parsing of the operand's offset into a new method. NFC.
This commit extract the code that parses the 64-bit offset from the method
'parseOperandsOffset' to a new method 'parseOffset' so that we can reuse it
when parsing the offset for the machine memory operands.

llvm-svn: 244355
2015-08-07 20:21:00 +00:00
Matt Arsenault 711b390a7c AMDGPU: Assume SMRD access for constant address space
Since r243294 these are selected to SMRD and
moved later if required.

llvm-svn: 244354
2015-08-07 20:18:34 +00:00
Rafael Espindola ee0b2b5c2b Don't look for a SHT_DYNSYM in the ELFFile's constructor.
Yet another step in not having it scan every section.

llvm-svn: 244353
2015-08-07 20:11:08 +00:00
Craig Topper a3db7d26ab Add Intel family 6 model 90 as Silvermont. Fixes PR24392.
llvm-svn: 244352
2015-08-07 20:09:42 +00:00
Rafael Espindola 4001998455 Remove the symbol iteration functions that don't take a symbol table.
Another step in making ELFFile's constructor not iterate over all sections.

llvm-svn: 244351
2015-08-07 20:07:27 +00:00
Adam Nemet 3a91e94734 [LAA] Remove unused pointer partition argument from print(), NFC
This is now handled in the client.  No need for LAA to provide this
variant.

llvm-svn: 244349
2015-08-07 19:44:48 +00:00
Chen Li eafbc9dc47 [ConstantFoldTerminator] Preserve make.implicit metadata when converting SwitchInst to BranchInst
Summary: llvm::ConstantFoldTerminator function can convert SwitchInst with single case (and default) to a conditional BranchInst. This patch adds support to preserve make.implicit metadata on this conversion.

Reviewers: sanjoy, weimingz, chenli

Subscribers: mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D11841

llvm-svn: 244348
2015-08-07 19:30:12 +00:00
Frederic Riss c1c1f534ec [MCDwarf] Use more idiomatic [u]int8_t types.
llvm-svn: 244347
2015-08-07 19:19:44 +00:00
Simon Pilgrim 3815c16bf8 [InstCombine] Fix SSE2/AVX2 vector logical shift by constant
This patch fixes the sse2/avx2 vector shift by constant instcombine call to correctly deal with the fact that the shift amount is formed from the entire lower 64-bit and not just the lowest element as it currently assumes.

e.g.

%1 = tail call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)

In this case, (V)PSRLD doesn't perform a lshr by 15 but in fact attempts to shift by 64424509455 ((15 << 32) | 15) - giving a zero result.

In addition, this review also recognizes shift-by-zero from a ConstantAggregateZero type (PR23821).

Differential Revision: http://reviews.llvm.org/D11760

llvm-svn: 244341
2015-08-07 18:22:50 +00:00
Cong Hou 855ea0f71a NFC.
Use a parent class to avoid reduncant code when specializing GraphTraits for T and const T.

llvm-svn: 244340
2015-08-07 18:04:17 +00:00
Nico Weber af3f242016 Add functions to save and restore the PrettyStackTrace state.
PrettyStackTraceHead is a LLVM_THREAD_LOCAL, which means it's just a global
in LLVM_ENABLE_THREADS=NO builds.  If a CrashRecoveryContext is used with
code that uses PrettyStackEntries, and a crash happens, PrettyStackTraceHead is
currently not reset to its pre-crash value.  These functions make it possible
to add a cleanup to such code that does this.

(Not reseting the value then causes the assert in ~PrettyStackTraceEntry() to
fire if the code outside of the CrashRecoveryContext also uses
PrettyStackEntries -- for example, clang when building a module.)

Part of PR11974.

llvm-svn: 244338
2015-08-07 17:47:03 +00:00
Nico Weber 2692811d00 Add a comment.
llvm-svn: 244337
2015-08-07 17:32:06 +00:00
Rafael Espindola 7d6112b76f Delete dead code. NFC.
llvm-svn: 244335
2015-08-07 17:06:05 +00:00
Chad Rosier 9659de379d [ARM] Remove an unused reference to MachineRegisterInfo. NFC.
llvm-svn: 244334
2015-08-07 17:02:29 +00:00
Rafael Espindola d20a28ed65 Don't use a DenseMap to handle SHT_SYMTAB_SHNDX.
It is already a convenient table.

llvm-svn: 244333
2015-08-07 16:51:44 +00:00
Tom Stellard c8733e805e AMDGPU/SI: Use correct encoding of vopc for VI in the assembler
Summary: We were using the SI encoding for VI.

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11812

llvm-svn: 244332
2015-08-07 16:45:33 +00:00
Tom Stellard d37631a8ac AMDGPU/SI: Add VI checks to vop3 assembler tests
Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11811

llvm-svn: 244331
2015-08-07 16:45:30 +00:00
Eric Christopher 5a30e0328b We're actually -Wmissing-field-initializers clean thanks to the cmake
build so check and turn on -Wmissing-field-initializers. While there,
reorganize the conditional warning code based on compiler to be a bit
more obvious and inside a switch statement.

llvm-svn: 244330
2015-08-07 16:44:47 +00:00
Rafael Espindola 8ab4446703 Delete dead typedef. NFC.
llvm-svn: 244328
2015-08-07 16:36:47 +00:00
Frederic Riss 8e30197427 Use 'signed char' instead of 'char'
Hope this fixes the s390 and aarch64 bots.

llvm-svn: 244327
2015-08-07 16:31:59 +00:00
Rafael Espindola f7eb882176 add missing tests files
llvm-svn: 244323
2015-08-07 15:35:49 +00:00
Tom Stellard 85656cabfb AMDGPU/SI: v_mac_legacy_f32 does not exist on VI
Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11810

llvm-svn: 244322
2015-08-07 15:34:30 +00:00
Tom Stellard 11f19f78f0 AMDGPU/SI: Remove unused outs parameter from VOPC TableGen classes
Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11809

llvm-svn: 244321
2015-08-07 15:34:27 +00:00
Rafael Espindola e01f43bcc1 Add dynamic_table iterators back to ELF.h.
In tree they are only used by llvm-readobj, but it is also used by
https://github.com/mono/CppSharp.

While at it, add some missing error checking.

llvm-svn: 244320
2015-08-07 15:25:20 +00:00
Frederic Riss a5e1453ac3 [dsymutil] Use the new MCDwarfLineTableParams customization to emit linetables
llvm-dsymutil has to be able to process debug info produced by other compilers
which use different line table settings. The testcase wasn't generated by
another compiler, but by a modified clang.

llvm-svn: 244319
2015-08-07 15:14:13 +00:00
Frederic Riss a5ab8443c1 [MC/Dwarf] Allow to specify custom parameters for linetable emission.
NFC patch for current users, but llvm-dsymutil will use the new
functionality to adapt to the input linetable.

Based on a patch by Adrian Prantl.

llvm-svn: 244318
2015-08-07 15:14:08 +00:00
Silviu Baranga a07090f7fa Fix unused variable warning introduced in r244314
llvm-svn: 244315
2015-08-07 12:05:46 +00:00
Silviu Baranga 3e8e51c1a9 [ARM] Update ReconstructShuffle to handle mismatched types
Summary:
Port the ReconstructShuffle function from AArch64 to ARM
to handle mismatched incoming types in the BUILD_VECTOR
node.

This fixes an outstanding FIXME in the ReconstructShuffle
code.

Reviewers: t.p.northover, rengolin

Subscribers: aemerson, llvm-commits, rengolin

Differential Revision: http://reviews.llvm.org/D11720

llvm-svn: 244314
2015-08-07 11:40:46 +00:00
John Brawn 64e5a66794 Revert "Make global aliases have symbol size equal to their type"
This reverts r242520, as it caused pr24379. Also removes part of the test added
by r243874 that checks the size of alias symbols.

llvm-svn: 244313
2015-08-07 10:56:21 +00:00
NAKAMURA Takumi b918a404c3 Tweak llvm/test/tools/dsymutil/arch-option.test to avoid globbing on mingw-w64.
llvm-svn: 244311
2015-08-07 08:38:22 +00:00
NAKAMURA Takumi 8dbe161502 ShrinkWrap.cpp: Tweak r244235 for a non-functional member, PredicateFtor. [-Wdocumentation]
llvm-svn: 244309
2015-08-07 07:40:23 +00:00
JF Bastien 315cc06840 WebAssembly: textual emission uses expected opcode names
Summary: WebAssembly's tablegen instructions have the names WebAssembly expects, but by LLVM convention they're uppercase and suffixed with their type after an underscore. Leave the C++ code that way, but print outt he names WebAssembly expects (lowercase, no type). We could teach tablegen to do this later, maybe by using `!cast<string>(node)` in the .td files.

Reviewers: sunfish

Subscribers: jfb, llvm-commits

Differential Revision: http://reviews.llvm.org/D11776

llvm-svn: 244305
2015-08-07 01:57:03 +00:00
Tom Stellard f594fcad73 ELF: Add AMDGPU specific defintions
Reviewers: rafael

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11458

llvm-svn: 244303
2015-08-07 01:35:24 +00:00
Duncan P. N. Exon Smith 8c9dcace0d ValueMapper: Resolve uniquing cycles more aggressively
As a follow-up to r244181, resolve uniquing cycles underneath distinct
nodes on the fly.  This prevents uniquing cycles in early operands from
affecting later operands.  It also removes an iteration through distinct
nodes' operands.

No real functional change here, just more prompt resolution of temporary
nodes.

llvm-svn: 244302
2015-08-07 00:44:55 +00:00
Duncan P. N. Exon Smith c9fdbdb78d ValueMapper: Pull out helper to resolve cycles, NFC
Pull out a helper for resolving uniquing cycles of `Metadata` to remove
the boiler-plate of downcasting to `MDNode`.

llvm-svn: 244301
2015-08-07 00:39:26 +00:00
Davide Italiano db8803b05a [llvm-readobj] Convert to range-loops.
llvm-svn: 244300
2015-08-07 00:35:28 +00:00
Alex Lorenz cba8c5fe31 MIR Serialization: Fix serialization of unnamed IR block references.
The block address machine operands can reference IR blocks in other functions.
This commit fixes a bug where the references to unnamed IR blocks in other
functions weren't serialized correctly.

llvm-svn: 244299
2015-08-06 23:57:04 +00:00
Alex Lorenz 3fb77686c1 MIR Parser: Simplify the token's string value handling.
This commit removes the 'StringOffset' and 'HasStringValue' fields from the
MIToken struct and simplifies the 'stringValue' method which now returns
the new 'StringValue' field.

This commit also adopts a different way of initializing the lexed tokens -
instead of constructing a new MIToken instance, the lexer resets the old token
using the new 'reset' method and sets its attributes using the new
'setStringValue', 'setOwnedStringValue', and 'setIntegerValue' methods.

Reviewers: Sean Silva

Differential Revision: http://reviews.llvm.org/D11792

llvm-svn: 244295
2015-08-06 23:17:42 +00:00
Sean Silva a11903215f [LangRef] Cleanup some stray whitespace inconsistencies.
Patch by Debora Setton Sanches! <setton.debora@gmail.com>

llvm-svn: 244294
2015-08-06 22:56:48 +00:00