Split mergeInGraph into two methods.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20035 91177308-0d34-0410-b5e6-96231b3b80d8

Add some new members

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20034 91177308-0d34-0410-b5e6-96231b3b80d8

add new member

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20033 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the Regression/Transforms/DSAnalysis/recursion.ll regression.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20031 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a case where were incorrectly compiled cast from short to int on 64-bit
targets.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20030 91177308-0d34-0410-b5e6-96231b3b80d8

alignment

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20028 91177308-0d34-0410-b5e6-96231b3b80d8

get alignment printing correctly and get rid of __main hack

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20027 91177308-0d34-0410-b5e6-96231b3b80d8

fix constant pointer outputing on 64 bit machines

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20026 91177308-0d34-0410-b5e6-96231b3b80d8

* Fix broken link to directions to get llvm, llvm-gcc, and llvm-test from CVS
* Remove duplicated CVS directions and refer reader to the CVS section
* Clean up directions to be brief

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20023 91177308-0d34-0410-b5e6-96231b3b80d8

FP fixes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20019 91177308-0d34-0410-b5e6-96231b3b80d8

Refactor getFunctionArgumentsForCall out of mergeInGraph.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20018 91177308-0d34-0410-b5e6-96231b3b80d8

Add a new method.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20017 91177308-0d34-0410-b5e6-96231b3b80d8

* Clearly mark LLVM_LIB_SEARCH_PATH as being optional
* llvmgcc and llvmg++ aliases are no longer needed (binaries have llvm- prefix)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20016 91177308-0d34-0410-b5e6-96231b3b80d8

This is no longer needed.  Global variables with undef initializers can be
initialized to anything, including garbage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20010 91177308-0d34-0410-b5e6-96231b3b80d8

* Instead of fixing the version numbers before every release, mark them as x.y
* Eliminate the redundant "This is the..." in released file listing
* Fix grammar

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20005 91177308-0d34-0410-b5e6-96231b3b80d8

Store fix

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20004 91177308-0d34-0410-b5e6-96231b3b80d8

oops

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20003 91177308-0d34-0410-b5e6-96231b3b80d8

prevent register allocator from using the stack pointer :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@20002 91177308-0d34-0410-b5e6-96231b3b80d8

fix loading of floats

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19997 91177308-0d34-0410-b5e6-96231b3b80d8

Put libraries in a common directory

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19995 91177308-0d34-0410-b5e6-96231b3b80d8

marked mem* as not supported

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19992 91177308-0d34-0410-b5e6-96231b3b80d8

Add additional source of Win32 GNU utilities.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19991 91177308-0d34-0410-b5e6-96231b3b80d8

Fix crash on MallocInsts of unsized types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19988 91177308-0d34-0410-b5e6-96231b3b80d8

fix Load bug

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19987 91177308-0d34-0410-b5e6-96231b3b80d8

Fix yet another memset issue.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19986 91177308-0d34-0410-b5e6-96231b3b80d8

try to make a bug bugpointable, add yet more constant pool stuff, fixup constant loads for FP

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19985 91177308-0d34-0410-b5e6-96231b3b80d8

better constant handling, should fix many remaining cases

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19984 91177308-0d34-0410-b5e6-96231b3b80d8

Add variable for bugpoint.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19981 91177308-0d34-0410-b5e6-96231b3b80d8

Eliminate some duplicated debug code

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19980 91177308-0d34-0410-b5e6-96231b3b80d8

Eliminate self-recursion as a special case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19979 91177308-0d34-0410-b5e6-96231b3b80d8

Eliminate use of DSCallSiteIterator in key loop.  This is a half step to
a tasty speedup.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19978 91177308-0d34-0410-b5e6-96231b3b80d8

remove dead method

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19977 91177308-0d34-0410-b5e6-96231b3b80d8

fix FP arg passing bug, Add unsigned to/from int, fix SELECT, fix Constant pool

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19976 91177308-0d34-0410-b5e6-96231b3b80d8

Print the Constant pool

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19975 91177308-0d34-0410-b5e6-96231b3b80d8

Make cmov work right and loads for fp from constant pool

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19974 91177308-0d34-0410-b5e6-96231b3b80d8

Correct stack stuff for FP

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19973 91177308-0d34-0410-b5e6-96231b3b80d8

try to match alpha pattern

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19972 91177308-0d34-0410-b5e6-96231b3b80d8

fix register names

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19971 91177308-0d34-0410-b5e6-96231b3b80d8

Signficantly speed up printing by not emitting the same file twice with
different names.  Large SCC's tend to be big, so this saves a lot of time.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19970 91177308-0d34-0410-b5e6-96231b3b80d8

Fix some bugs andrew noticed legalizing memset for alpha

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19969 91177308-0d34-0410-b5e6-96231b3b80d8

Do not revisit nodes in the SCC traversal.  This speeds up the BU pass a bit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19968 91177308-0d34-0410-b5e6-96231b3b80d8

Put finishing touches on GettingStartedVS.html and link it to the index.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19967 91177308-0d34-0410-b5e6-96231b3b80d8

Fix test/Regression/Assembler/2005-01-31-CallingAggregateFunction.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19966 91177308-0d34-0410-b5e6-96231b3b80d8

Apparently := confuses makellvm

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19965 91177308-0d34-0410-b5e6-96231b3b80d8

This bug crashes the assembler, distilled from a testcase produced by Andrew.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19964 91177308-0d34-0410-b5e6-96231b3b80d8

pecimise loads, put indirect call addr in right register.  still doesn't fix methcall

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19963 91177308-0d34-0410-b5e6-96231b3b80d8

Update speedup and shrinkage info

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19962 91177308-0d34-0410-b5e6-96231b3b80d8

Updates for new use list changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19961 91177308-0d34-0410-b5e6-96231b3b80d8

Update for API change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19960 91177308-0d34-0410-b5e6-96231b3b80d8

API change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19959 91177308-0d34-0410-b5e6-96231b3b80d8

Adjust to changes in APIs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19958 91177308-0d34-0410-b5e6-96231b3b80d8

Hacks to make this ugly ugly code work with the new use lists.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19957 91177308-0d34-0410-b5e6-96231b3b80d8

Switch from using an ilist for uses to using a custom doubly linked list.
This list does not provide the ability to go backwards in the list (its
more of an unordered collection, stored in the shape of a list).

This change means that use iterators are now only forward iterators, not
bidirectional.

This improves the memory usage of use lists from '5 + 4*#use' per value to
'1 + 4*#use'.  While it would be better to reduce the multiplied factor,
I'm not smart enough to do so.  This list also has slightly more efficient
operators for manipulating list nodes (a few less loads/stores), due to not
needing to be able to iterate backwards through the list.

This change reduces the memory footprint required to hold 176.gcc from
66.025M -> 57.687M, a 14% reduction.  It also speeds up the compiler,
7.73% in the case of bytecode loading alone (release build loading 176.gcc).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19956 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a problem where we could infinitely recurse on phi nodes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19955 91177308-0d34-0410-b5e6-96231b3b80d8

Fix hyphenation in output comment

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19954 91177308-0d34-0410-b5e6-96231b3b80d8

Implement InstCombine/cast.ll:test25, a case that occurs many times
in spec

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19953 91177308-0d34-0410-b5e6-96231b3b80d8

New testcase

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19952 91177308-0d34-0410-b5e6-96231b3b80d8

Add LLVM with Visual Studio overview

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19951 91177308-0d34-0410-b5e6-96231b3b80d8

Implement the trivial cases in InstCombine/store.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19950 91177308-0d34-0410-b5e6-96231b3b80d8

New testcase

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19949 91177308-0d34-0410-b5e6-96231b3b80d8

Rename these to start with %c, as they are the only ones that are checked.
No change to the tests themselves.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19948 91177308-0d34-0410-b5e6-96231b3b80d8

Implement Transforms/InstCombine/cast-load-gep.ll, which allows us to devirtualize
11 indirect calls in perlbmk.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19947 91177308-0d34-0410-b5e6-96231b3b80d8

New testcase

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19946 91177308-0d34-0410-b5e6-96231b3b80d8

indirect call fix

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19945 91177308-0d34-0410-b5e6-96231b3b80d8

fp to int and back conversion sequences

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19944 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the regressions my User changes introduced.  Apparently some parts of
LLVM make the very reasonable assumption that constant expressions will
have at least one operand! :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19943 91177308-0d34-0410-b5e6-96231b3b80d8

Rename variables to work with VC++'s hokey scoping rules.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19942 91177308-0d34-0410-b5e6-96231b3b80d8

Fix some scary bugs that VC++ detected.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19941 91177308-0d34-0410-b5e6-96231b3b80d8

Make things more const-correct, adjust to changes in DSA interfaces.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19940 91177308-0d34-0410-b5e6-96231b3b80d8

* Make some methods more const correct.
* Change the FunctionCalls and AuxFunctionCalls vectors into std::lists.
  This makes many operations on these lists much more natural, and avoids
  *exteremely* expensive copying of DSCallSites (e.g. moving nodes around
  between lists, erasing a node from not the end of the vector, etc).

With a profile build of analyze, this speeds up BU DS from 25.14s to
12.59s on 176.gcc.  I expect that it would help TD even more, but I don't
have data for it.

This effectively eliminates removeIdenticalCalls and children from the
profile, going from 6.53 to 0.27s.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19939 91177308-0d34-0410-b5e6-96231b3b80d8

added fp extend and removed a forgotten assert in more than 6 arg support (should break somewhere else now :) ) and fix an incorrect asm sequence for indirect calls

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19938 91177308-0d34-0410-b5e6-96231b3b80d8

Silence VC++ warnings about using 'this' in base member initializations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19937 91177308-0d34-0410-b5e6-96231b3b80d8

This code is really unreachable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19934 91177308-0d34-0410-b5e6-96231b3b80d8

Fix warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19933 91177308-0d34-0410-b5e6-96231b3b80d8

support for larger calls

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19932 91177308-0d34-0410-b5e6-96231b3b80d8

Improve spelling

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19931 91177308-0d34-0410-b5e6-96231b3b80d8

Improve conformance with the Misha spelling benchmark suite

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19930 91177308-0d34-0410-b5e6-96231b3b80d8

Make this work on systems where size_t == unsigned and where they are not
the same.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19929 91177308-0d34-0410-b5e6-96231b3b80d8

Make this work on systems where size_t is not the same as unsigned.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19928 91177308-0d34-0410-b5e6-96231b3b80d8

Unbreak the build  :(

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19926 91177308-0d34-0410-b5e6-96231b3b80d8

Adjust to ilist changes.

Based on the ilist changes avoid allocating an entire Use object for the
end of the Use chain.  This saves 8 bytes of memory for each Value allocated
in the program.  For 176.gcc, this reduces us from 69.5M -> 66.0M, a 5.3%
memory savings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19925 91177308-0d34-0410-b5e6-96231b3b80d8

adjust to ilist changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19924 91177308-0d34-0410-b5e6-96231b3b80d8

Adjust to ilist changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19923 91177308-0d34-0410-b5e6-96231b3b80d8

Adjust to changes in ilist

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19922 91177308-0d34-0410-b5e6-96231b3b80d8

Rename createNode -> createSentinal.

Add a new method, destroySentinal, that is used to delete it (instead of
requiring use of delete.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19921 91177308-0d34-0410-b5e6-96231b3b80d8

clean up comments

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19920 91177308-0d34-0410-b5e6-96231b3b80d8

This has been deprecated for long enough, nuke it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19919 91177308-0d34-0410-b5e6-96231b3b80d8

This file was schizophrenic when it came to representing sizes.  In some
cases it represented them as 'unsigned's, which are not enough for 64-bit
hosts.  In other cases, it represented them as uint64_t's, which are
inefficient for 32-bit hosts.

This patch unifies all of the sizes to use size_t instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19918 91177308-0d34-0410-b5e6-96231b3b80d8

There is no reason to include ostream here, include iosfwd instead.

This file was schizophrenic when it came to representing sizes.  In some
cases it represented them as 'unsigneds', which are not enough for 64-bit
hosts.  In other cases, it represented them as uint64_t's, which are
inefficient for 32-bit hosts.

This patch unifies all of the sizes to use size_t instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19917 91177308-0d34-0410-b5e6-96231b3b80d8

After reading in a bc file, trim the resultant buffer down to what we
really need.  This reduces 4M of memory consumption reading 176.gcc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19916 91177308-0d34-0410-b5e6-96231b3b80d8

Finegrainify namespacification

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19915 91177308-0d34-0410-b5e6-96231b3b80d8

first step towards a correct and complete stack.  also add some forms for things that were getting stuck in the nightly tester.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19914 91177308-0d34-0410-b5e6-96231b3b80d8

Due to previous simplifications, we can simplify the data structures being
used here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19913 91177308-0d34-0410-b5e6-96231b3b80d8

Properly handle volatile.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19912 91177308-0d34-0410-b5e6-96231b3b80d8

Remove some useless map operations.  Loads/stores that are in the same
BB as the load are not included in the Cand* sets at all.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19911 91177308-0d34-0410-b5e6-96231b3b80d8

Before doing expensive global analysis, check to make sure the pointer is
not invalidated on entry and on exit of the block.  This fixes some N^2
behavior in common cases, and speeds up gcc another 5% to 22.35s.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19910 91177308-0d34-0410-b5e6-96231b3b80d8

Fix quotes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19909 91177308-0d34-0410-b5e6-96231b3b80d8

Unbreak VC++ build

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19908 91177308-0d34-0410-b5e6-96231b3b80d8

Minor simplification/speedup.  Replaces a set lookup with a pointer comparison.
This speeds up 176.gcc from 25.73s to 23.48s, which is 9.5%

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19907 91177308-0d34-0410-b5e6-96231b3b80d8

Eliminate generality that is not buying us anything.  In particular, this
will cause us to miss cases where the input pointer to a load could be value
numbered to another load.  Something like this:

  %X = load int* %P1
  %Y = load int* %P2

Those are obviously the same if P1/P2 are the same.  The code this patch
removes attempts to handle that.  However, since GCSE iterates, this doesn't
actually buy us anything: GCSE will first replace P1 or P2 with the other
one, then the load can be value numbered as equal.

Removing this code speeds up gcse a lot.  On 176.gcc in debug mode, this
speeds up gcse from 29.08s -> 25.73s, a 13% savings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19906 91177308-0d34-0410-b5e6-96231b3b80d8