Use loop deletion instead of ADCE in these tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51180 91177308-0d34-0410-b5e6-96231b3b80d8

Clean ups for loop deletion based on Chris' feedback.

Also, use SCEV to determine the trip count of the loop, which is more powerful
and accurate that Loop::getTripCount.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51179 91177308-0d34-0410-b5e6-96231b3b80d8

Use loop deletion instead of ADCE for removing loops.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51178 91177308-0d34-0410-b5e6-96231b3b80d8

Fix this test.  It was testing broken behavior in that it required ADCE to eliminate
a potentially infinite loop, which is undesirable.  Instead, test the LICM behavior
that we're really interested in.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51177 91177308-0d34-0410-b5e6-96231b3b80d8

implement PR2328.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51176 91177308-0d34-0410-b5e6-96231b3b80d8

Use common where we mean common, not weak.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51173 91177308-0d34-0410-b5e6-96231b3b80d8

Treat common as distinct from weak global on Darwin x86.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51172 91177308-0d34-0410-b5e6-96231b3b80d8

Revert the change from r51157 in
test/Verifier/2002-11-05-GetelementptrPointers.ll, which was incorrect.
Instead, fix getIndexedType to not follow pointer types, as
PointerType is a subclass of CompositeType.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51171 91177308-0d34-0410-b5e6-96231b3b80d8

Update comments based on Duncan's feedback.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51170 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a bug in LoopStrengthReduce that caused it to emit IR with
use-before-def. The problem comes up in code with multiple PHIs where
one PHI is being rewritten in terms of the other, but the other needs
to be casted first. LLVM rules requre the cast instruction to be
inserted after any PHI instructions, but when instructions were
inserted to replace the second PHI value with a function of the first,
they were ended up going before the cast instruction. Avoid this
problem by remembering the location of the cast instruction, when one
is needed, and inserting the expansion of the new value after it.

This fixes a bug that surfaced in 255.vortex on x86-64 when
instcombine was removed from the middle of the loop optimization
passes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51169 91177308-0d34-0410-b5e6-96231b3b80d8

Remove the S92 code, which really has nothing to do
with what the test is testing; makes it pass again on ppc32.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51167 91177308-0d34-0410-b5e6-96231b3b80d8

Evan has implemented this on ppc, so run the test there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51166 91177308-0d34-0410-b5e6-96231b3b80d8

Fix typos and comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51165 91177308-0d34-0410-b5e6-96231b3b80d8

Allow an extra bit for CommonLinkage.

This changes the .bc file format, but if I understand
how it works correctly, old .bc files continue to
be readable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51161 91177308-0d34-0410-b5e6-96231b3b80d8

Actually scalarize the operand to BIT_CONVERT instead of asking someone to do
something with a v1 type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51160 91177308-0d34-0410-b5e6-96231b3b80d8

Reverting r51100...I need to use VC++ 2005.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51159 91177308-0d34-0410-b5e6-96231b3b80d8

IR support for extractvalue and insertvalue instructions. Also, begin
moving toward making structs and arrays first-class types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51157 91177308-0d34-0410-b5e6-96231b3b80d8

Use static and anonymous namespaces consistently with other passes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51156 91177308-0d34-0410-b5e6-96231b3b80d8

Revert 51067 check-in for now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51155 91177308-0d34-0410-b5e6-96231b3b80d8

Remove useless check.
Patch by Matthijs Kooijman.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51154 91177308-0d34-0410-b5e6-96231b3b80d8

Disable JIT symbol table for now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51152 91177308-0d34-0410-b5e6-96231b3b80d8

Use of UINT_MAX requires climits, at least when
compiling with gcc 4.3.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51145 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a bunch of 80col violations that arose from the Create API change. Tweak makefile targets to find these better.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51143 91177308-0d34-0410-b5e6-96231b3b80d8

ignore generated files and product directories

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51142 91177308-0d34-0410-b5e6-96231b3b80d8

Make use of vector load and store operations to implement memcpy, memmove, and memset. Currently only X86 target is taking advantage of these.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51140 91177308-0d34-0410-b5e6-96231b3b80d8

Remove tail duplication pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51139 91177308-0d34-0410-b5e6-96231b3b80d8

Move the operator new and operator delete out of line.  This fixes an issue with
operator new() referring to the static initTags function, which has to be in the
same linkage unit as any file including User.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51136 91177308-0d34-0410-b5e6-96231b3b80d8

Don't assume underlying APInt type is limited
to 64 bits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51135 91177308-0d34-0410-b5e6-96231b3b80d8

Situations can arise when you have a function called that returns a 'void', but
is bitcast to return a floating point value. The result of the instruction may
not be used by the program afterwards, and LLVM will happily remove all
instructions except the call. But, on some platforms, if a value is returned as
a floating point, it may need to be removed from the stack (like x87). Thus, we
can't get rid of the bitcast even if there isn't a use of the value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51134 91177308-0d34-0410-b5e6-96231b3b80d8

Remove undefined behavior in hex string->APFloat
conversion.  Try 0x1.0000a4p+0f.  Neil, please review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51132 91177308-0d34-0410-b5e6-96231b3b80d8

Use a better idiom to silence compiler warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51131 91177308-0d34-0410-b5e6-96231b3b80d8

rename SimplifyCFG.cpp -> SimplifyCFGPass.cpp

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51130 91177308-0d34-0410-b5e6-96231b3b80d8

Silence warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51129 91177308-0d34-0410-b5e6-96231b3b80d8

Add support to IR builder for new vicmp, vfcmp routines

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51127 91177308-0d34-0410-b5e6-96231b3b80d8

Really silence compiler warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51126 91177308-0d34-0410-b5e6-96231b3b80d8

Don't generate unused variables in a no-assert build
Add some checks to the new vicmp,vfcmp instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51125 91177308-0d34-0410-b5e6-96231b3b80d8

Really silence compiler warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51123 91177308-0d34-0410-b5e6-96231b3b80d8

CommonLinkage (missed a file)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51120 91177308-0d34-0410-b5e6-96231b3b80d8

Generated files for CommonLinkage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51119 91177308-0d34-0410-b5e6-96231b3b80d8

Add CommonLinkage; currently tentative definitions
are represented as "weak", but there are subtle differences
in some cases on Darwin, so we need both.  The intent
is that "common" will behave identically to "weak" unless
somebody changes their target to do something else.
No functional change as yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51118 91177308-0d34-0410-b5e6-96231b3b80d8

Silence some compiler warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51115 91177308-0d34-0410-b5e6-96231b3b80d8

Simplify internalize pass. Add test case.
Patch by Matthijs Kooijman!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51114 91177308-0d34-0410-b5e6-96231b3b80d8

When bit-twiddling CondCode values for integer comparisons produces
SETOEQ, is it does with (SETEQ & SETULE), map it to SETEQ.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51112 91177308-0d34-0410-b5e6-96231b3b80d8

Recover nestedloop regression reported by nightly tester.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51110 91177308-0d34-0410-b5e6-96231b3b80d8

Check if llvm-gcc is available before running tests. Patch by Matthijs Kooijman!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51108 91177308-0d34-0410-b5e6-96231b3b80d8

Detabification. Fixed indentation and spacing.
Changed cout to DOUT, and TODOs to FIXMEs.
Other changes as per coding conventions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51105 91177308-0d34-0410-b5e6-96231b3b80d8

Do not generate by TableGen the hard-coded standard, target-independent part of
DAG instruction selectors. Introudce a dedicated header file for this part:
include/llvm/CodeGen/DAGISelHeader.h

TableGen now only generates the include preprocessor directive to include this
new header.

This is a preparation for supporting multiple implementations of instruction
selectors in the future.

Reviewed and approved by Evan and Dan.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51102 91177308-0d34-0410-b5e6-96231b3b80d8

Upgrading clang VC++ solution to VC++ express 2008.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51100 91177308-0d34-0410-b5e6-96231b3b80d8

Make this test pass on x86-32 linux.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51099 91177308-0d34-0410-b5e6-96231b3b80d8

Adding missing files to Transforms and VMCore VC++ projects.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51098 91177308-0d34-0410-b5e6-96231b3b80d8

Add documentation for tail call optimization to CodeGenerator.html. Add a link
referring to it to LangRef.html.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51097 91177308-0d34-0410-b5e6-96231b3b80d8

Added configure switches for PIC16 in configure.ac.
Regenerated configure.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51096 91177308-0d34-0410-b5e6-96231b3b80d8

Fix typo in ParameterAttribute fields usage. Add an include
to make the Cpp backend output compilable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51095 91177308-0d34-0410-b5e6-96231b3b80d8

Fixed the file description header at the top to remove the developer name.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51094 91177308-0d34-0410-b5e6-96231b3b80d8

Commit the header I accidentally left out of 51083.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51093 91177308-0d34-0410-b5e6-96231b3b80d8

Doh. Alignment is in bytes, not in bits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51092 91177308-0d34-0410-b5e6-96231b3b80d8

Change target-specific classes to use more precise static types.
This eliminates the need for several awkward casts, including
the last dynamic_cast under lib/Target.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51091 91177308-0d34-0410-b5e6-96231b3b80d8

Move RemoveFromVector out of the global namespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51090 91177308-0d34-0410-b5e6-96231b3b80d8

Whitespace cleanups.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51089 91177308-0d34-0410-b5e6-96231b3b80d8

Make PreVerifyID, IntSigsEnd, and KillSigsEnd const.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51088 91177308-0d34-0410-b5e6-96231b3b80d8

Make getNumContainedManagers and getNumContainedPasses const.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51087 91177308-0d34-0410-b5e6-96231b3b80d8

Make PassInfo noncopyable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51085 91177308-0d34-0410-b5e6-96231b3b80d8

Do not run instruction combiner in middle of loop optimization passes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51084 91177308-0d34-0410-b5e6-96231b3b80d8

Split the loop unroll mechanism logic out into a utility function.
Patch by Matthijs Kooijman!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51083 91177308-0d34-0410-b5e6-96231b3b80d8

Fix Analysis/BasicAA/pure-const-dce.ll.  This turned out to be a correctness
bug as well as a missed optimization.  We weren't properly checking for local
dependencies before moving on to non-local ones when doing non-local read-only
call CSE.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51082 91177308-0d34-0410-b5e6-96231b3b80d8

Merge of r51073-51074 from use-diet branch.

Do not rely on std::swap<Use>, provide a (faster) member function instead.
This change is primarily necessitated by MSVC++'s incompatibility with
declaring std::swap<Use> to be a friend of Use.

Also contains some minor tweaks to Use inline functions,
to undo pointless changes that sneaked in with the last merge.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51078 91177308-0d34-0410-b5e6-96231b3b80d8

Make this function public.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51077 91177308-0d34-0410-b5e6-96231b3b80d8

Dominance Frontier is cfg only pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51075 91177308-0d34-0410-b5e6-96231b3b80d8

Fix memdep's handling of invokes when finding the dependency of another call
instruction.  This fixes some Ada miscompiles reported in PR2324.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51069 91177308-0d34-0410-b5e6-96231b3b80d8

Update the Win32 project files, patch by Razvan Aciu!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51067 91177308-0d34-0410-b5e6-96231b3b80d8

Fix for PR 2323, infinite loop in tail dup.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51063 91177308-0d34-0410-b5e6-96231b3b80d8

add a note

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51062 91177308-0d34-0410-b5e6-96231b3b80d8

- Fix the pasto in the fix for a previous pasto.
- Incorporate Chris' comment suggestion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51061 91177308-0d34-0410-b5e6-96231b3b80d8

add a note

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51060 91177308-0d34-0410-b5e6-96231b3b80d8

s/indicies/indices/ and clarify the extractvalue and insertvalue are
working with struct field or array element values. Thanks Duncan!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51059 91177308-0d34-0410-b5e6-96231b3b80d8

Fix one more encoding bug.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51057 91177308-0d34-0410-b5e6-96231b3b80d8

Added configure switches for PIC16 backend.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51056 91177308-0d34-0410-b5e6-96231b3b80d8

- Don't treat anyext 16-bit load as a 32-bit load if it's volatile.
- Correct a pasto.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51054 91177308-0d34-0410-b5e6-96231b3b80d8

Add thin layer over StringMap to form StringSet. By Mikhail Glushenkov.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51048 91177308-0d34-0410-b5e6-96231b3b80d8

Make the non-local CSE safety checks slightly more thorough.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51035 91177308-0d34-0410-b5e6-96231b3b80d8

Adding files for Microchip's PIC16 target.

A brief description about PIC16:
===============================
PIC16 is an 8-bit microcontroller with only one 8-bit register which is the
accumulator. All arithmetic/load/store operations are 8-bit only.
The architecture has two address spaces: program and data. The program memory
is divided into 2K pages and the data memory is divided into banks of 128 byte, with only 80 usable bytes, resulting in an non-contiguous data memory.

It supports direct data memory access (by specifying the address as part of the instruction) and indirect data and program memory access (in an unorthodox fashion which utilize a 16 bit pointer register).

Two classes of registers exist: (8-bit class which is only one
accumulator) (16-bit class, which contains one or more 16 bit
pointer(s))

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51027 91177308-0d34-0410-b5e6-96231b3b80d8

Instead of a vector load, shuffle and then extract an element. Load the element from address with an offset.
        pshufd $1, (%rdi), %xmm0
        movd %xmm0, %eax
=>
        movl 4(%rdi), %eax

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51026 91177308-0d34-0410-b5e6-96231b3b80d8

Add a testcase for non-local CSE of read-only calls.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51025 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for non-local CSE of read-only calls.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51024 91177308-0d34-0410-b5e6-96231b3b80d8

Derive GetResultInst from UnaryInstruction, this simplifies code and removes a FIXME.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51023 91177308-0d34-0410-b5e6-96231b3b80d8

Change class' public PassInfo variables to by initialized with the
address of the PassInfo directly instead of calling getPassInfo.
This eliminates a bunch of dynamic initializations of static data.

Also, fold RegisterPassBase into PassInfo, make a bunch of its
data members const, and rearrange some code to initialize data
members in constructors instead of using setter member functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51022 91177308-0d34-0410-b5e6-96231b3b80d8

80 col / tabs fixes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51021 91177308-0d34-0410-b5e6-96231b3b80d8

Fix and encoding error in the psrad xmm, imm8 instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51020 91177308-0d34-0410-b5e6-96231b3b80d8

On x86, it's safe to treat i32 load anyext as a normal i32 load. Ditto for i8 anyext load to i16.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51019 91177308-0d34-0410-b5e6-96231b3b80d8

Clean up the use of static and anonymous namespaces. This turned up
several things that were neither in an anonymous namespace nor static
but not intended to be global.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51017 91177308-0d34-0410-b5e6-96231b3b80d8

Initial documentation for first-class aggregates changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51013 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a malformed %gt;

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51011 91177308-0d34-0410-b5e6-96231b3b80d8

Teach Legalize how to scalarize VSETCC
Teach X86 a few more vsetcc patterns.  Custom lowering for unsupported ones is next.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51009 91177308-0d34-0410-b5e6-96231b3b80d8

Xform bitconvert(build_pair(load a, load b)) to a single load if the load locations are at the right offset from each other.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51008 91177308-0d34-0410-b5e6-96231b3b80d8

New test for tail merging

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51007 91177308-0d34-0410-b5e6-96231b3b80d8

Be more aggressive about tail-merging small blocks
if those blocks consist entirely of common instructions;
merging will not add an extra branch in this case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51006 91177308-0d34-0410-b5e6-96231b3b80d8

Constify isSourceDefinedByImplicitDef function. Otherwise, just formatting
changes that don't change functionality.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51004 91177308-0d34-0410-b5e6-96231b3b80d8

Constify the machine instruction passed into the
"is{Trivially,Really}ReMaterializable" methods.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51001 91177308-0d34-0410-b5e6-96231b3b80d8

Initial X86 codegen support for VSETCC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51000 91177308-0d34-0410-b5e6-96231b3b80d8

Further rework of tail merge algorithm.  Not quite
semantically identical, but little difference in
either results or execution speed; but it's much
easier to read, at least IMO.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@50999 91177308-0d34-0410-b5e6-96231b3b80d8