6 years ago[Support] Allow multiple paired calls to {start,stop}Timer()
Vedant Kumar [Tue, 22 Dec 2015 17:36:17 +0000 (17:36 +0000)]
[Support] Allow multiple paired calls to {start,stop}Timer()

Differential Revision: http://reviews.llvm.org/D15619

Reviewed-by: rafael
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256258 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[RS4GC] Fix crash in the case that a live variable has a constant base.
Manuel Jacob [Tue, 22 Dec 2015 16:50:44 +0000 (16:50 +0000)]
[RS4GC] Fix crash in the case that a live variable has a constant base.

Previously, RS4GC crashed in CreateGCRelocates() because it assumed
that every base is also in the array of live variables, which isn't true if a
live variable has a constant base.

This change fixes the crash by making sure CreateGCRelocates() won't try to
relocate a live variable with a constant base.  This would be unnecessary
anyway because anything with a constant base won't move.

Reviewers: reames

Subscribers: llvm-commits, sanjoy

Differential Revision: http://reviews.llvm.org/D15556

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256252 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[AArch64] Promote loads from stored
Jun Bum Lim [Tue, 22 Dec 2015 16:36:16 +0000 (16:36 +0000)]
[AArch64] Promote loads from stored

This is a recommit of r256004 which was reverted in r256160. The issue was the
incorrect promotion for half and byte loads transformed into mov instructions.
This fix will replace half and byte type loads only with bit field extracts.

Original commit message:

This change promotes load instructions which directly read from stored by
replacing them with mov instructions. If the store is wider than the load,
the load will be replaced with a bitfield extract.
For example :
  STRWui %W1, %X0, 1
  %W0 = LDRHHui %X0, 3
  STRWui %W1, %X0, 1
  %W0 = UBFMWri %W1, 16, 31

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256249 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoTypo. NFC.
Chad Rosier [Tue, 22 Dec 2015 15:06:47 +0000 (15:06 +0000)]
Typo. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256242 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[X86][AVX512] Add rcp14 and rsqrt14 intrinsics
Asaf Badouh [Tue, 22 Dec 2015 11:40:04 +0000 (11:40 +0000)]
[X86][AVX512] Add rcp14 and rsqrt14 intrinsics

Differential Revision: http://reviews.llvm.org/D15414

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256237 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[ASMPrinter] Fix missing handling of DW_OP_bit_piece
Keno Fischer [Tue, 22 Dec 2015 07:14:50 +0000 (07:14 +0000)]
[ASMPrinter] Fix missing handling of DW_OP_bit_piece

In r256077, I added printing for DIExpressions in DEBUG_VALUE comments,
but neglected to handle DW_OP_bit_piece operands. Thanks to
Mikael Holmen and Joerg Sonnenberger for spotting this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256236 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[libFuzzer] add AFL-style dictionary for C++, remove the old file with tokens
Kostya Serebryany [Tue, 22 Dec 2015 01:50:51 +0000 (01:50 +0000)]
[libFuzzer] add AFL-style dictionary for C++, remove the old file with tokens

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256229 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[MC] Don't use the architecture to govern which object file format to use
David Majnemer [Tue, 22 Dec 2015 01:39:04 +0000 (01:39 +0000)]
[MC] Don't use the architecture to govern which object file format to use

InitMCObjectFileInfo was trying to override the triple in awkward ways.
For example, a triple specifying COFF but not Windows was forced as ELF.
This makes it easy for internal invariants to get violated, such as
those which triggered PR25912.

This fixes PR25912.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256226 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoPartial fix for PR25912, see comment 13. Should fix the sanitizer bootstrap bot
Kostya Serebryany [Tue, 22 Dec 2015 01:18:49 +0000 (01:18 +0000)]
Partial fix for PR25912, see comment 13. Should fix the sanitizer bootstrap bot

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256225 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoHandle empty Subprogram list when linking metadata.
Teresa Johnson [Tue, 22 Dec 2015 01:17:19 +0000 (01:17 +0000)]
Handle empty Subprogram list when linking metadata.

Use an iterator that handles an empty subprogram list.

Fixes PR25915.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256224 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoDetermine callee's hotness and adjust threshold based on that. NFC.
Easwaran Raman [Tue, 22 Dec 2015 00:32:35 +0000 (00:32 +0000)]
Determine callee's hotness and adjust threshold based on that. NFC.

This uses the same criteria used in CFE's CodeGenPGO to identify hot and cold
callees and uses values of inlinehint-threshold and inlinecold-threshold
respectively as the thresholds for such callees.

Differential Revision: http://reviews.llvm.org/D15245

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256222 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[safestack] Add option for non-TLS unsafe stack pointer.
Evgeniy Stepanov [Tue, 22 Dec 2015 00:13:11 +0000 (00:13 +0000)]
[safestack] Add option for non-TLS unsafe stack pointer.

This patch adds an option, -safe-stack-no-tls, for using normal
storage instead of thread-local storage for the unsafe stack pointer.
This can be useful when SafeStack is applied to an operating system


Patch by Michael LeMay.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256221 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[PGO] Fix another comdat related issue for COFF
Xinliang David Li [Tue, 22 Dec 2015 00:11:15 +0000 (00:11 +0000)]
[PGO] Fix another comdat related issue for COFF

The linker requires that a comdat section must be associated
with a another comdat section that precedes it. This
means the comdat section's name needs to use the  profile name
var's name.

Patch tested by Johan Engelen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256220 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[Support] Timer: Use emplace_back() and range-based loops (NFC)
Vedant Kumar [Mon, 21 Dec 2015 23:41:38 +0000 (23:41 +0000)]
[Support] Timer: Use emplace_back() and range-based loops (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256217 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[Support] Timer: simplify the init() method
Vedant Kumar [Mon, 21 Dec 2015 23:27:44 +0000 (23:27 +0000)]
[Support] Timer: simplify the init() method

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256215 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[AVR] Added configuration file and machine function information class
Dylan McKay [Mon, 21 Dec 2015 23:13:15 +0000 (23:13 +0000)]
[AVR] Added configuration file and machine function information class

This commit adds the 'AVRMachineFunctionInfo' class, which simply stores
basic properties about generated machine functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256213 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix line endings after r256155. NFC.
Eric Christopher [Mon, 21 Dec 2015 23:04:27 +0000 (23:04 +0000)]
Fix line endings after r256155. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256211 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix test case comment (NFC)
Xinliang David Li [Mon, 21 Dec 2015 22:26:49 +0000 (22:26 +0000)]
Fix test case comment (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256206 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[cfi] Fix LowerBitSets on 32-bit targets.
Evgeniy Stepanov [Mon, 21 Dec 2015 22:14:04 +0000 (22:14 +0000)]
[cfi] Fix LowerBitSets on 32-bit targets.

This code attempts to truncate IntPtrTy to i32, which may be the same

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256205 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[MC, COFF] Support link /incremental conditionally
David Majnemer [Mon, 21 Dec 2015 22:09:27 +0000 (22:09 +0000)]
[MC, COFF] Support link /incremental conditionally

Today, we always take into account the possibility that object files
produced by MC may be consumed by an incremental linker.  This results
in us initialing fields which vary with time (TimeDateStamp) which harms
hermetic builds (e.g. verifying a self-host went well) and produces
sub-optimal code because we cannot assume anything about the relative
position of functions within a section (call sites can get redirected
through incremental linker thunks).

Let's provide an MCTargetOption which controls this behavior so that we
can disable this functionality if we know a-priori that the build will
not rely on /incremental.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256203 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoEnhance BranchProbabilityInfo::calcUnreachableHeuristics for InvokeInst
Jun Bum Lim [Mon, 21 Dec 2015 22:00:51 +0000 (22:00 +0000)]
Enhance BranchProbabilityInfo::calcUnreachableHeuristics for InvokeInst

This is recommit of r256028 with minor fixes in unittests:

Original commit message:

When identifying blocks post-dominated by an unreachable-terminated block
in BranchProbabilityInfo, consider only the edge to the normal destination
block if the terminator is InvokeInst and let calcInvokeHeuristics() decide
edge weights for the InvokeInst.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256202 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoResubmit r256193 with test fix: assertion failure analyzed
Xinliang David Li [Mon, 21 Dec 2015 21:52:27 +0000 (21:52 +0000)]
Resubmit r256193 with test fix: assertion failure analyzed

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256201 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRevert r256193: build bot failure triggered
Xinliang David Li [Mon, 21 Dec 2015 21:00:33 +0000 (21:00 +0000)]
Revert r256193: build bot failure triggered

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256198 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[X86][SSE] Transform truncations between vectors of integers into X86ISD::PACKUS...
Cong Hou [Mon, 21 Dec 2015 20:42:43 +0000 (20:42 +0000)]
[X86][SSE] Transform truncations between vectors of integers into X86ISD::PACKUS/PACKSS operations during DAG combine.

This patch transforms truncation between vectors of integers into
X86ISD::PACKUS/PACKSS operations during DAG combine. We don't do it in
lowering phase because after type legalization, the original truncation
will be turned into a BUILD_VECTOR with each element that is extracted
from a vector and then truncated, and from them it is difficult to do
this optimization. This greatly improves the performance of truncations
on some specific types.

Cost table is updated accordingly.

Differential revision: http://reviews.llvm.org/D14588

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256194 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[PGO] Fix profile var comdat generation problem with COFF
Xinliang David Li [Mon, 21 Dec 2015 20:41:20 +0000 (20:41 +0000)]
[PGO] Fix profile var comdat generation problem with COFF

When targeting COFF, it is required that a comdat section to
have a global obj with the same name as the comdat (except for
comdats with select kind to be associative). This fix makes
sure that the comdat is keyed on the data variable for COFF.

Also improved test coverage for this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256193 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[ValueTracking] Properly handle non-sized types in isAligned function.
Michael Zolotukhin [Mon, 21 Dec 2015 20:38:18 +0000 (20:38 +0000)]
[ValueTracking] Properly handle non-sized types in isAligned function.

Reviewers: apilipenko, reames, sanjoy, hfinkel

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D15597

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256192 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix PR24563 (LiveDebugVariables unconditionally propagates all DBG_VALUEs)
Adrian Prantl [Mon, 21 Dec 2015 20:03:00 +0000 (20:03 +0000)]
Fix PR24563 (LiveDebugVariables unconditionally propagates all DBG_VALUEs)

LiveDebugVariables unconditionally propagates all DBG_VALUE down the
dominator tree, which happens to work fine if there already is another
DBG_VALUE or the DBG_VALUE happends to describe a single-assignment vreg
but is otherwise wrong if the DBG_VALUE is coming from only one of the

In r255759 we introduced a proper data flow analysis scheduled after
LiveDebugVariables that correctly propagates DBG_VALUEs across basic block
boundaries. With the new pass in place, the incorrect propagation in
LiveDebugVariables can be retired witout loosing any of the benefits
where LiveDebugVariables happened to do the right thing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256188 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoConvert the CodeGen/ARM/sched-it-debug-nodes.ll testcase from IR -> MIR.
Adrian Prantl [Mon, 21 Dec 2015 19:44:42 +0000 (19:44 +0000)]
Convert the CodeGen/ARM/sched-it-debug-nodes.ll testcase from IR -> MIR.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256187 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoTeach ARMLoadStoreOptimizer to ignore DBG_VALUE instructions when merging
Adrian Prantl [Mon, 21 Dec 2015 19:25:03 +0000 (19:25 +0000)]
Teach ARMLoadStoreOptimizer to ignore DBG_VALUE instructions when merging

As noted in PR24563.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256183 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agofix leak in a test, make the sanitizer bot green
Kostya Serebryany [Mon, 21 Dec 2015 19:09:01 +0000 (19:09 +0000)]
fix leak in a test, make the sanitizer bot green

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256179 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAMDGPU/SI: Fix encoding for FLAT_SCRATCH registers on VI
Tom Stellard [Mon, 21 Dec 2015 18:44:27 +0000 (18:44 +0000)]
AMDGPU/SI: Fix encoding for FLAT_SCRATCH registers on VI

These register has different encodings on CI and VI, so we add pseudo
FLAT_SCRACTH registers to be used before MC, and subtarget specific
registers to be used by the MC layer.

Reviewers: arsenm

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15661

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256178 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAMDGPU/SI: Change assembly name for flat scratch registers to flat_scratch
Tom Stellard [Mon, 21 Dec 2015 18:44:21 +0000 (18:44 +0000)]
AMDGPU/SI: Change assembly name for flat scratch registers to flat_scratch

This matches what the assembler accepts.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256177 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[AArch64] Add additional extract-extend patterns for smov
Matthew Simpson [Mon, 21 Dec 2015 18:31:25 +0000 (18:31 +0000)]
[AArch64] Add additional extract-extend patterns for smov

This patch adds to the target description two additional patterns for matching
extract-extend operations to SMOV. The patterns catch the v16i8-to-i64 and
v8i16-to-i64 cases. The existing patterns miss these cases because the
extracted elements must first be legalized to i32, resulting in any_extend

This was originally implemented as a DAG combine (r255895), but was reverted
due to failing out-of-tree tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256176 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAdd testcase for r256161 (PR25907)
Teresa Johnson [Mon, 21 Dec 2015 18:24:35 +0000 (18:24 +0000)]
Add testcase for r256161 (PR25907)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256174 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRemove extra whitespace. NFC.
Chad Rosier [Mon, 21 Dec 2015 18:08:05 +0000 (18:08 +0000)]
Remove extra whitespace. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256173 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[ThinLTO] Rename variable to reflect bulk importing change (NFC)
Teresa Johnson [Mon, 21 Dec 2015 17:33:24 +0000 (17:33 +0000)]
[ThinLTO] Rename variable to reflect bulk importing change (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256171 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[WebAssembly] Convert a regular for loop to a range-based for loop.
Dan Gohman [Mon, 21 Dec 2015 17:22:02 +0000 (17:22 +0000)]
[WebAssembly] Convert a regular for loop to a range-based for loop.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256169 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[WebAssembly] Clean up comments and fix a missing #include dependency.
Dan Gohman [Mon, 21 Dec 2015 17:19:31 +0000 (17:19 +0000)]
[WebAssembly] Clean up comments and fix a missing #include dependency.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256168 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[WebAssembly] Remove an unneeded empty destructor.
Dan Gohman [Mon, 21 Dec 2015 17:12:40 +0000 (17:12 +0000)]
[WebAssembly] Remove an unneeded empty destructor.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256167 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[WebAssembly] Enclose the operand variables for load and store instructions in braces.
Dan Gohman [Mon, 21 Dec 2015 16:58:49 +0000 (16:58 +0000)]
[WebAssembly] Enclose the operand variables for load and store instructions in braces.

This allows the AsmMatcherEmitter to properly tokenize the AsmStrings for
load and store instructions. This is a step towards asm parsing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256166 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[WebAssembly] Mark the ARGUMENT pseudo-instructions as CodeGenOnly.
Dan Gohman [Mon, 21 Dec 2015 16:53:29 +0000 (16:53 +0000)]
[WebAssembly] Mark the ARGUMENT pseudo-instructions as CodeGenOnly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256165 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[WebAssembly] Add some comments and make some minor source cleanups.
Dan Gohman [Mon, 21 Dec 2015 16:50:41 +0000 (16:50 +0000)]
[WebAssembly] Add some comments and make some minor source cleanups.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256164 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoTeach MCOperand::print how to print FPImm operands.
Dan Gohman [Mon, 21 Dec 2015 16:47:10 +0000 (16:47 +0000)]
Teach MCOperand::print how to print FPImm operands.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256163 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRemove unused functions from ModuleLinker (NFC)
Teresa Johnson [Mon, 21 Dec 2015 15:49:59 +0000 (15:49 +0000)]
Remove unused functions from ModuleLinker (NFC)

Remove a couple ModuleLinker methods and a related static function that
are no longer used after the linker split.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256162 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRemove overly strict new assert in BitcodeReader.
Teresa Johnson [Mon, 21 Dec 2015 15:38:13 +0000 (15:38 +0000)]
Remove overly strict new assert in BitcodeReader.

This fixes a bug introduced by the ThinLTO metadata linking patch
r255909. The assert is overly-strict and while useful in development of
the patch, doesn't seem interesting to keep.

Fixes PR25907.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256161 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRevert "[AArch64] Promote loads from stores"
Jun Bum Lim [Mon, 21 Dec 2015 15:36:49 +0000 (15:36 +0000)]
Revert "[AArch64] Promote loads from stores"

This reverts commit r256004 due to a failure in cortex-a53.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256160 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[LIR] Refactor code to enable future patch. NFC.
Chad Rosier [Mon, 21 Dec 2015 14:49:32 +0000 (14:49 +0000)]
[LIR] Refactor code to enable future patch. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256159 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[AArch64] Enable PostRAScheduler for AArch64 generic build.
Chad Rosier [Mon, 21 Dec 2015 14:43:45 +0000 (14:43 +0000)]
[AArch64] Enable PostRAScheduler for AArch64 generic build.

Disable post-ra scheduler for perturbed tests to appease the bots and to
preserve the history of the tests.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256158 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAVX512BW: Enable AND/OR/XOR vector byte/word paked operation by promoting to qword...
Igor Breger [Mon, 21 Dec 2015 14:40:36 +0000 (14:40 +0000)]
AVX512BW: Enable AND/OR/XOR vector byte/word paked operation by promoting to qword that natively suppored.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256157 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[llvm-objdump] Use appropriate helper. NFC.
Davide Italiano [Mon, 21 Dec 2015 14:10:54 +0000 (14:10 +0000)]
[llvm-objdump] Use appropriate helper. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256156 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoImplemented Support of IA interrupt and exception handlers:
Amjad Aboud [Mon, 21 Dec 2015 14:07:14 +0000 (14:07 +0000)]
Implemented Support of IA interrupt and exception handlers:

Differential Revision: http://reviews.llvm.org/D15567

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256155 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[mips][microMIPS] Implement DERET and DI instructions and check size operand for...
Zlatko Buljan [Mon, 21 Dec 2015 13:08:58 +0000 (13:08 +0000)]
[mips][microMIPS] Implement DERET and DI instructions and check size operand for EXT and DEXT* instructions
Differential Revision: http://reviews.llvm.org/D15570

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256152 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agocheck-llvm: Tweak the feature "timestamps" for autoconf.
NAKAMURA Takumi [Mon, 21 Dec 2015 08:46:12 +0000 (08:46 +0000)]
check-llvm: Tweak the feature "timestamps" for autoconf.

Note, ENABLE_TIMESTAMPS is either 1 or 0 in Makefile.config.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256138 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[MC, COFF] Unbreak support for COFF timestamps
David Majnemer [Mon, 21 Dec 2015 08:03:07 +0000 (08:03 +0000)]
[MC, COFF] Unbreak support for COFF timestamps

Support for COFF timestamps was unintentionally broken in r246905 when
it was conditionally available depending on whether or not LLVM was
configured with LLVM_ENABLE_TIMESTAMPS.  However, Config/config.h was
never included which essentially broke the feature.  Due to lax testing,
the breakage was never identified until we observed strange failures
during incremental links of Chromium.

This issue is resolved by simply including Config/config.h in
WinCOFFObjectWriter and teaching lit that the MC/COFF/timestamp.s test
is conditionally supported depending on LLVM_ENABLE_TIMESTAMPS.  With
this in place, we can strengthen the test to ensure that it will not
accidentally get broken in the future.

This fixes PR25891.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256137 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[Cygwin] Enable TLS as emutls.
NAKAMURA Takumi [Mon, 21 Dec 2015 02:37:23 +0000 (02:37 +0000)]
[Cygwin] Enable TLS as emutls.

It resolves clang selfhosting with std::once() for Cygwin.

FIXME: It may be EmulatedTLS-generic also for X86-Android.
FIXME: Pass EmulatedTLS to LLVM CodeGen from Clang with -femulated-tls.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256134 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[RS4GC] Add an assert which fails if there is a (yet unsupported) addrspacecast.
Manuel Jacob [Mon, 21 Dec 2015 01:26:46 +0000 (01:26 +0000)]
[RS4GC] Add an assert which fails if there is a (yet unsupported) addrspacecast.

The slightly strange indentation comes from clang-format.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256132 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[InstCombine] Fix indentation. NFC.
Craig Topper [Mon, 21 Dec 2015 01:02:28 +0000 (01:02 +0000)]
[InstCombine] Fix indentation. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256131 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[AVR] Added AVRCallingConv.td
Dylan McKay [Sun, 20 Dec 2015 23:17:44 +0000 (23:17 +0000)]
[AVR] Added AVRCallingConv.td

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256130 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[X86] Use range-based for loop. NFC
Craig Topper [Sun, 20 Dec 2015 18:41:57 +0000 (18:41 +0000)]
[X86] Use range-based for loop. NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256127 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[X86] Prevent constant hoisting for a couple compare immediates that the selection...
Craig Topper [Sun, 20 Dec 2015 18:41:54 +0000 (18:41 +0000)]
[X86] Prevent constant hoisting for a couple compare immediates that the selection DAG knows how to optimize into a shift.

This allows "icmp ugt %a, 4294967295" and "icmp uge %a, 4294967296" to be optimized into right shifts by 32 which can fold the immediate into the shift instruction. These patterns show up with some regularity in real code.

Unfortunately, since getImmCost can't see the icmp predicate we can't be tell if we're only catching these specific cases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256126 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAdd AVR.td and AVRRegisterInfo.td
Dylan McKay [Sun, 20 Dec 2015 12:16:20 +0000 (12:16 +0000)]
Add AVR.td and AVRRegisterInfo.td

This adds the core AVR TableGen file, along with the register descriptions.

Lines in AVR.td which require other TableGen files which haven't been committed
yet are commented out.

This is a fairly trivial patch, and should only require a quick review.

I kept the line width smaller than 80 columns, but there are a few exceptions
because I'm not sure how to split a string over several lines.

Reviewers: stoklund

Subscribers: dylanmckay, agnat

Differential Revision: http://reviews.llvm.org/D14684

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256120 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[llvm-objdump] Move COFF function to where it belongs.
Davide Italiano [Sun, 20 Dec 2015 09:54:34 +0000 (09:54 +0000)]
[llvm-objdump] Move COFF function to where it belongs.

Ideally much more stuff should be moved out of llvm-objdump.cpp, but that
will happen later.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256118 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix a bug in test case -- duplicate entries
Xinliang David Li [Sun, 20 Dec 2015 08:49:31 +0000 (08:49 +0000)]
Fix a bug in test case -- duplicate entries

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256117 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix a latent UAF bug in profwriter
Xinliang David Li [Sun, 20 Dec 2015 08:46:18 +0000 (08:46 +0000)]
Fix a latent UAF bug in profwriter

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256116 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix mapping of @llvm.arm.ssat/usat intrinsics to ssat/usat instructions for Thumb2
Weiming Zhao [Sun, 20 Dec 2015 06:41:44 +0000 (06:41 +0000)]
Fix mapping of @llvm.arm.ssat/usat intrinsics to ssat/usat instructions for Thumb2

r250697 fixed the mapping for ARM mode. We have to do the same for Thumb2 otherwise the same llvm.arm.ssat() will generate different saturating amount for ARM and Thumb.

r250697: http://reviews.llvm.org/rL250697

Reviewers: rmaprath

Subscribers: aemerson, llvm-commits, rengolin

Differential Revision: http://reviews.llvm.org/D15653

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256115 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[PGO] Improve Indexed Profile Reader efficiency
Xinliang David Li [Sun, 20 Dec 2015 06:22:13 +0000 (06:22 +0000)]
[PGO] Improve Indexed Profile Reader efficiency

With the support of value profiling added, the Indexed prof
reader gets less efficient. The prof reader initialization
used to be just reading the file header, but with VP support
added, initialization needs to walk through all profile keys
of ondisk hash table resulting in very poor locality and large
memory increase (keys are stored together with the profile data
in the mapped profile buffer). Even worse, when the reader is
used by the compiler (not llvm-profdata too), the penalty becomes
very high as compilation of each single module requires touching
profile data buffer for the whole program.

In this patch, the icall target values (MD5hash) are no longer eargerly
converted back to name strings when the data is read into memory. New
interface is added to to profile reader so that InstrProfSymtab can be
lazily created for Indexed profile reader on-demand. Creating of the
symtab is intended to be used by llvm-profdata tool for symbolic dumping
of  VP data. It can be used with compiler (for legacy out of tree uses)
too but not recommended due to compile time and memory reasons
mentioned above.

Some other cleanups are also included: Function Addr to md5 map is now
consolated into InstrProfSymtab. InstrProfStringtab is no longer used and

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256114 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoMinor clean up -- move large single use method out of header(NFC)
Xinliang David Li [Sun, 20 Dec 2015 05:15:45 +0000 (05:15 +0000)]
Minor clean up -- move large single use method out of header(NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256113 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRevert r219171, "llvm/test/lit.cfg: Suppress dwarf stuff for targeting x86_64-mingw32...
NAKAMURA Takumi [Sun, 20 Dec 2015 03:48:23 +0000 (03:48 +0000)]
Revert r219171, "llvm/test/lit.cfg: Suppress dwarf stuff for targeting x86_64-mingw32 while investigating since r219108."

It has been fixed since r219280 by David Majnemer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256112 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoThreadPool unittests: do not hold mutex when calling condition_variable:notify()
Mehdi Amini [Sat, 19 Dec 2015 22:56:24 +0000 (22:56 +0000)]
ThreadPool unittests: do not hold mutex when calling condition_variable:notify()

From: Mehdi Amini <mehdi.amini@apple.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256111 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoNonnull elements in OperandBundleCallSites are not all Instructions
Sanjoy Das [Sat, 19 Dec 2015 22:40:28 +0000 (22:40 +0000)]
Nonnull elements in OperandBundleCallSites are not all Instructions

`CloneAndPruneIntoFromInst` sometimes RAUW's dead instructions with
`undef` before erasing them (to avoid deleting instructions that still
have uses).  This changes the `WeakVH` in `OperandBundleCallSites` to
hold an `undef`, and we need to guard for this situation in eventuality
in `llvm::InlineFunction`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256110 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[Deopt bundles] Fix a test case
Sanjoy Das [Sat, 19 Dec 2015 22:40:22 +0000 (22:40 +0000)]
[Deopt bundles] Fix a test case

The `CHECK-NOT` line was incorrect, and would not have caught a

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256109 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[llvm-objdump] Fail early if we can't parse the object header.
Davide Italiano [Sat, 19 Dec 2015 22:09:40 +0000 (22:09 +0000)]
[llvm-objdump] Fail early if we can't parse the object header.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256108 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoDelete APIs that have been deprecated since 2010.
Rafael Espindola [Sat, 19 Dec 2015 21:42:07 +0000 (21:42 +0000)]
Delete APIs that have been deprecated since 2010.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256107 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoThe PS4 baton passes.
Paul Robinson [Sat, 19 Dec 2015 20:04:03 +0000 (20:04 +0000)]
The PS4 baton passes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256106 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAssert that we have all use/users in the getters.
Rafael Espindola [Sat, 19 Dec 2015 20:03:23 +0000 (20:03 +0000)]
Assert that we have all use/users in the getters.

An error that is pretty easy to make is to use the lazy bitcode reader
and then do something like

if (V.use_empty())

The problem is that uses in unmaterialized functions are not accounted

This patch adds asserts that all uses are known.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256105 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoWebAssembly: add vtable test
JF Bastien [Sat, 19 Dec 2015 18:55:18 +0000 (18:55 +0000)]
WebAssembly: add vtable test

The test will mainly be useful to check that the .s file assembles and relocates properly because vtables reference functions in their data section.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256102 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRemove unnecessary casts. NFC.
Manuel Jacob [Sat, 19 Dec 2015 18:38:42 +0000 (18:38 +0000)]
Remove unnecessary casts.  NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256101 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRemove double blanks. NFC.
Manuel Jacob [Sat, 19 Dec 2015 18:26:53 +0000 (18:26 +0000)]
Remove double blanks.  NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256100 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoImprove InstrProfSymtab test coverage
Xinliang David Li [Sat, 19 Dec 2015 18:20:09 +0000 (18:20 +0000)]
Improve InstrProfSymtab test coverage

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256099 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoSelectionDAG: Cleanup integer bin op promotion functions.
Matt Arsenault [Sat, 19 Dec 2015 17:18:43 +0000 (17:18 +0000)]
SelectionDAG: Cleanup integer bin op promotion functions.

SDIV and UDIV had special handling, but this is the same handling
that min/max need.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256098 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[unittests] ThreadPool: Remove redundant loop, NFC
Vedant Kumar [Sat, 19 Dec 2015 09:54:27 +0000 (09:54 +0000)]
[unittests] ThreadPool: Remove redundant loop, NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256097 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[unittests] ThreadPool: Guard updates to MainThreadReady
Vedant Kumar [Sat, 19 Dec 2015 09:49:09 +0000 (09:49 +0000)]
[unittests] ThreadPool: Guard updates to MainThreadReady

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256096 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRe-reapply "[IR] Move optional data in llvm::Function into a hungoff uselist"
Vedant Kumar [Sat, 19 Dec 2015 08:52:49 +0000 (08:52 +0000)]
Re-reapply "[IR] Move optional data in llvm::Function into a hungoff uselist"

Make personality functions, prefix data, and prologue data hungoff
operands of Function.

This is based on the email thread "[RFC] Clean up the way we store
optional Function data" on llvm-dev.

Thanks to sanjoyd, majnemer, rnk, loladiro, and dexonsmith for feedback!

Includes a fix to scrub value subclass data in dropAllReferences. Does not
use binary literals.

Differential Revision: http://reviews.llvm.org/D13829

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256095 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRevert "Reapply "[IR] Move optional data in llvm::Function into a hungoff uselist""
Vedant Kumar [Sat, 19 Dec 2015 08:48:43 +0000 (08:48 +0000)]
Revert "Reapply "[IR] Move optional data in llvm::Function into a hungoff uselist""

This reverts commit r256093.

This broke lld-x86_64-win7 because of -Werror,-Wc++1y-extensions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256094 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoReapply "[IR] Move optional data in llvm::Function into a hungoff uselist"
Vedant Kumar [Sat, 19 Dec 2015 08:29:51 +0000 (08:29 +0000)]
Reapply "[IR] Move optional data in llvm::Function into a hungoff uselist"

Make personality functions, prefix data, and prologue data hungoff
operands of Function.

This is based on the email thread "[RFC] Clean up the way we store
optional Function data" on llvm-dev.

Thanks to sanjoyd, majnemer, rnk, loladiro, and dexonsmith for feedback!

Includes a fix to scrub value subclass data in dropAllReferences.

Differential Revision: http://reviews.llvm.org/D13829

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256093 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[PGO] Add hash to name mapping in InstrProfSymtab
Xinliang David Li [Sat, 19 Dec 2015 07:44:57 +0000 (07:44 +0000)]
[PGO] Add hash to name mapping in InstrProfSymtab

Creator and lookup interfaces are added to this symtab class.
The new interfaces will be used by InstrProf Readers and writer.

A unit test is also added for the new APIs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256092 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoRevert "[IR] Move optional data in llvm::Function into a hungoff uselist"
Vedant Kumar [Sat, 19 Dec 2015 07:30:44 +0000 (07:30 +0000)]
Revert "[IR] Move optional data in llvm::Function into a hungoff uselist"

This reverts commit r256090.

This broke llvm-clang-lld-x86_64-debian-fast.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256091 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[IR] Move optional data in llvm::Function into a hungoff uselist
Vedant Kumar [Sat, 19 Dec 2015 07:08:56 +0000 (07:08 +0000)]
[IR] Move optional data in llvm::Function into a hungoff uselist

Make personality functions, prefix data, and prologue data hungoff
operands of Function.

This is based on the email thread "[RFC] Clean up the way we store
optional Function data" on llvm-dev.

Thanks to sanjoyd, majnemer, rnk, loladiro, and dexonsmith for feedback!

Differential Revision: http://reviews.llvm.org/D13829

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256090 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoThreadPool unittest: reimplement concurrency test, deterministically this time.
Mehdi Amini [Sat, 19 Dec 2015 05:12:07 +0000 (05:12 +0000)]
ThreadPool unittest: reimplement concurrency test, deterministically this time.

Follow-up to r256056.

From: Mehdi Amini <mehdi.amini@apple.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256087 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[libFuzzer] deprecate -save_minimized_corpus, -merge can be used instead
Kostya Serebryany [Sat, 19 Dec 2015 03:42:16 +0000 (03:42 +0000)]
[libFuzzer] deprecate -save_minimized_corpus, -merge can be used instead

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256086 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[libFuzzer] split the tests to run them in parallel, remove one redundant test
Kostya Serebryany [Sat, 19 Dec 2015 03:35:30 +0000 (03:35 +0000)]
[libFuzzer] split the tests to run them in parallel, remove one redundant test

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256085 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoHopefully fix debug-info-blocks.ll test on win32 bot
Keno Fischer [Sat, 19 Dec 2015 03:32:23 +0000 (03:32 +0000)]
Hopefully fix debug-info-blocks.ll test on win32 bot

llc_dwarf adds an mtriple, which forces this to use COFF, causing
the test to fail. Hopefully using regular llc without the triple
will work fine everywhere

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256084 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAMDGPU/SI: Fix implemenation of isSourceOfDivergence() for graphics shaders
Tom Stellard [Sat, 19 Dec 2015 02:54:15 +0000 (02:54 +0000)]
AMDGPU/SI: Fix implemenation of isSourceOfDivergence() for graphics shaders

The analysis of shader inputs was completely wrong.  We were passing the
wrong index to AttributeSet::hasAttribute() and the logic for which
inputs where in SGPRs was wrong too.

Reviewers: arsenm

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15608

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256082 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[libFuzzer] make CrossOver just one of the other mutations
Kostya Serebryany [Sat, 19 Dec 2015 02:49:09 +0000 (02:49 +0000)]
[libFuzzer] make CrossOver just one of the other mutations

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256081 91177308-0d34-0410-b5e6-96231b3b80d8

6 years ago[RS4GC] Remove an overly strong assertion
Philip Reames [Sat, 19 Dec 2015 02:38:22 +0000 (02:38 +0000)]
[RS4GC] Remove an overly strong assertion

As shown by the included test case, it's reasonable to end up with constant references during base pointer calculation.  The code actually handled this case just fine, we only had the assert to help isolate problems under the belief that constant references shouldn't be present in IR generated by managed frontends. This turned out to be wrong on two fronts: 1) Manual Jacobs is working on a language with constant references, and b) we found a case where the optimizer does create them in practice.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256079 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoClean up the processing of dbg.value in various places
Keno Fischer [Sat, 19 Dec 2015 02:02:44 +0000 (02:02 +0000)]
Clean up the processing of dbg.value in various places

First up is instcombine, where in the dbg.declare -> dbg.value conversion,
the llvm.dbg.value needs to be called on the actual loaded value, rather
than the address (since the whole point of this transformation is to be
able to get rid of the alloca). Further, now that that's cleaned up, we
can remove a hack in the backend, that would add an implicit OP_deref if
the argument to dbg.value was an alloca. This stems from before the
existence of DIExpression and is no longer necessary since the deref can
be expressed explicitly.

Now, in order to make sure that the tests pass with this change, we need to
correct the printing of DEBUG_VALUE comments to take into account the
expression, which wasn't taken into account before.

Unfortunately, for both these changes, there were a number of incorrect
test cases (mostly the wrong number of DW_OP_derefs, but also a couple
where the test itself was broken more badly). aprantl and I have gone
through and adjusted these test case in order to make them pass with
these fixes and in some cases to make sure they're actually testing
what they are meant to test.

Reviewers: aprantl

Subscribers: dsanders

Differential Revision: http://reviews.llvm.org/D14186

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256077 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAMDGPU: Switch barrier intrinsics to using convergent
Matt Arsenault [Sat, 19 Dec 2015 01:46:41 +0000 (01:46 +0000)]
AMDGPU: Switch barrier intrinsics to using convergent

noduplicate prevents unrolling of small loops that happen to have
barriers in them. If a loop has a barrier in it, it is OK to duplicate
it for the unroll.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256075 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoFix broken type legalization of min/max
Matt Arsenault [Sat, 19 Dec 2015 01:39:48 +0000 (01:39 +0000)]
Fix broken type legalization of min/max

This was using an anyext when promoting the type
when zext/sext is required.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256074 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAMDGPU/SI: use S_MOV_B64 for larger copies in copyPhysReg
Nicolai Haehnle [Sat, 19 Dec 2015 01:36:26 +0000 (01:36 +0000)]
AMDGPU/SI: use S_MOV_B64 for larger copies in copyPhysReg

Reviewers: arsenm, tstellarAMD

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15629

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256073 91177308-0d34-0410-b5e6-96231b3b80d8

6 years agoAMDGPU: fix overlapping copies in copyPhysReg
Nicolai Haehnle [Sat, 19 Dec 2015 01:16:06 +0000 (01:16 +0000)]
AMDGPU: fix overlapping copies in copyPhysReg

When copying aggregate registers within the same register class, there may
be an overlap between source and destination that forces us to do the copy

Do the simplest possible thing that guarantees the correct order of moves
when there are overlaps, and does whatever when there is no overlap. (The
last part forces some trivial adjustments to test cases.)

Together with r255906, this fixes a VM fault in Unreal Elemental Demo.

While at it, change the generation of kill and def flags to something that
looks more reasonable. This method is used very late during compilation, so
it probably doesn't matter in practice, and to be honest, I don't know if
this change is actually correct because the semantics in connection with
aggregate registers vs. sub-registers are not clear to me.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93264

Reviewers: arsenm, tstellarAMD

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15622

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256072 91177308-0d34-0410-b5e6-96231b3b80d8