Commit Graph

45 Commits

Author SHA1 Message Date
Duncan Sands
dff6710717 Integrate the readonly/readnone logic more deeply
into alias analysis.  This meant updating the API
which now has versions of the getModRefBehavior,
doesNotAccessMemory and onlyReadsMemory methods
which take a callsite parameter.  These should be
used unless the callsite is not known, since in
general they can do a better job than the versions
that take a function.  Also, users should no longer
call the version of getModRefBehavior that takes
both a function and a callsite.  To reduce the
chance of misuse it is now protected.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@44487 91177308-0d34-0410-b5e6-96231b3b80d8
2007-12-01 07:51:45 +00:00
Duncan Sands
514ab348fd Executive summary: getTypeSize -> getTypeStoreSize / getABITypeSize.
The meaning of getTypeSize was not clear - clarifying it is important
now that we have x86 long double and arbitrary precision integers.
The issue with long double is that it requires 80 bits, and this is
not a multiple of its alignment.  This gives a primitive type for
which getTypeSize differed from getABITypeSize.  For arbitrary precision
integers it is even worse: there is the minimum number of bits needed to
hold the type (eg: 36 for an i36), the maximum number of bits that will
be overwriten when storing the type (40 bits for i36) and the ABI size
(i.e. the storage size rounded up to a multiple of the alignment; 64 bits
for i36).

This patch removes getTypeSize (not really - it is still there but
deprecated to allow for a gradual transition).  Instead there is:

(1) getTypeSizeInBits - a number of bits that suffices to hold all
values of the type.  For a primitive type, this is the minimum number
of bits.  For an i36 this is 36 bits.  For x86 long double it is 80.
This corresponds to gcc's TYPE_PRECISION.

(2) getTypeStoreSizeInBits - the maximum number of bits that is
written when storing the type (or read when reading it).  For an
i36 this is 40 bits, for an x86 long double it is 80 bits.  This
is the size alias analysis is interested in (getTypeStoreSize
returns the number of bytes).  There doesn't seem to be anything
corresponding to this in gcc.

(3) getABITypeSizeInBits - this is getTypeStoreSizeInBits rounded
up to a multiple of the alignment.  For an i36 this is 64, for an
x86 long double this is 96 or 128 depending on the OS.  This is the
spacing between consecutive elements when you form an array out of
this type (getABITypeSize returns the number of bytes).  This is
TYPE_SIZE in gcc.

Since successive elements in a SequentialType (arrays, pointers
and vectors) need to be aligned, the spacing between them will be
given by getABITypeSize.  This means that the size of an array
is the length times the getABITypeSize.  It also means that GEP
computations need to use getABITypeSize when computing offsets.
Furthermore, if an alloca allocates several elements at once then
these too need to be aligned, so the size of the alloca has to be
the number of elements multiplied by getABITypeSize.  Logically
speaking this doesn't have to be the case when allocating just
one element, but it is simpler to also use getABITypeSize in this
case.  So alloca's and mallocs should use getABITypeSize.  Finally,
since gcc's only notion of size is that given by getABITypeSize, if
you want to output assembler etc the same as gcc then getABITypeSize
is the size you want.

Since a store will overwrite no more than getTypeStoreSize bytes,
and a read will read no more than that many bytes, this is the
notion of size appropriate for alias analysis calculations.

In this patch I have corrected all type size uses except some of
those in ScalarReplAggregates, lib/Codegen, lib/Target (the hard
cases).  I will get around to auditing these too at some point,
but I could do with some help.

Finally, I made one change which I think wise but others might
consider pointless and suboptimal: in an unpacked struct the
amount of space allocated for a field is now given by the ABI
size rather than getTypeStoreSize.  I did this because every
other place that reserves memory for a type (eg: alloca) now
uses getABITypeSize, and I didn't want to make an exception
for unpacked structs, i.e. I did it to make things more uniform.
This only effects structs containing long doubles and arbitrary
precision integers.  If someone wants to pack these types more
tightly they can always use a packed struct.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43620 91177308-0d34-0410-b5e6-96231b3b80d8
2007-11-01 20:53:16 +00:00
Devang Patel
6ea7ac6a1e Use DominatorTree instead of ETForest.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37505 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-07 23:53:38 +00:00
Devang Patel
1997473cf7 Drop 'const'
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@36662 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-03 01:11:54 +00:00
Devang Patel
3e15bf33e0 Use 'static const char' instead of 'static const int'.
Due to darwin gcc bug, one version of darwin linker coalesces
static const int, which defauts PassID based pass identification.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@36652 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-02 21:39:20 +00:00
Devang Patel
794fd75c67 Do not use typeinfo to identify pass in pass manager.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@36632 91177308-0d34-0410-b5e6-96231b3b80d8
2007-05-01 21:15:47 +00:00
Chris Lattner
8e8f865227 Remove use of Instruction::getNext
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@36201 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-17 17:52:45 +00:00
Owen Anderson
46b58f7869 Expunge DomSet from LoadValueNumbering. This is part of the continuing
work on PR1171.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35723 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-07 04:43:07 +00:00
Reid Spencer
d7d83db5f2 Make classes in anonymous namespaces use VISIBILITY_HIDDEN to help reduce
LLVM's footprint and speed up linking.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@33941 91177308-0d34-0410-b5e6-96231b3b80d8
2007-02-05 23:42:17 +00:00
Chris Lattner
a5370172b6 simplify AnalysisGroup registration, eliminating one typeid call.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29932 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-28 00:42:29 +00:00
Chris Lattner
7f8897f22e eliminate RegisterOpt. It does the same thing as RegisterPass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29925 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-27 22:42:52 +00:00
Chris Lattner
10673b63b0 Fix a problem exposed by the et-forest work. Load-vn needs these passes live
whenever it is live, not just when load-vn is computed initially


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25146 91177308-0d34-0410-b5e6-96231b3b80d8
2006-01-08 09:10:04 +00:00
Andrew Lenharth
c826aea982 the correct fix was to fix AliasAnalysis.getModRefInfo
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@22268 91177308-0d34-0410-b5e6-96231b3b80d8
2005-06-20 15:25:22 +00:00
Andrew Lenharth
e3e1739b7d prevent GCSE from forwarding stores to loads around vaarg. This is uggly, and I am trying to fix the AliasInfo, as it should catch the problem instead.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@22266 91177308-0d34-0410-b5e6-96231b3b80d8
2005-06-20 15:02:05 +00:00
Misha Brukman
2b37d7cf28 Remove trailing whitespace
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@21416 91177308-0d34-0410-b5e6-96231b3b80d8
2005-04-21 21:13:18 +00:00
Chris Lattner
3725c12b26 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
2005-01-29 07:04:10 +00:00
Chris Lattner
ee379a16ee Properly handle volatile.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19912 91177308-0d34-0410-b5e6-96231b3b80d8
2005-01-29 06:42:34 +00:00
Chris Lattner
4f16bd7a74 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
2005-01-29 06:39:25 +00:00
Chris Lattner
15774df4d3 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
2005-01-29 06:31:53 +00:00
Chris Lattner
e212d62a34 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
2005-01-29 06:20:55 +00:00
Chris Lattner
e233b8c64b 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
2005-01-29 06:11:16 +00:00
Chris Lattner
2652da6af9 If we see:
%A = alloca int
  %V = load int* %A

value number %V to undef, not 0.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@19905 91177308-0d34-0410-b5e6-96231b3b80d8
2005-01-29 05:57:01 +00:00
Chris Lattner
fd4b3c4774 Eliminate a virtual method call
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@18964 91177308-0d34-0410-b5e6-96231b3b80d8
2004-12-15 18:14:04 +00:00
Alkis Evlogimenos
20aa474f8f Fixes to make LLVM compile with vc7.1.
Patch contributed by Paolo Invernizzi!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@16152 91177308-0d34-0410-b5e6-96231b3b80d8
2004-09-03 18:19:51 +00:00
Misha Brukman
47b14a4a6a Fix #includes of i*.h => Instructions.h as per PR403.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@15334 91177308-0d34-0410-b5e6-96231b3b80d8
2004-07-29 17:30:56 +00:00
Brian Gaeke
96d4bf7aee Make the create...() functions for some of these passes return a FunctionPass *.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@15276 91177308-0d34-0410-b5e6-96231b3b80d8
2004-07-27 17:43:21 +00:00
Chris Lattner
0f312d6998 Implement the interfaces to update value numbering information. Add an
assert.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13679 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-23 21:13:24 +00:00
Chris Lattner
5da80974c0 Implement test/Regression/Transforms/GCSE/undefined_load.ll
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@12641 91177308-0d34-0410-b5e6-96231b3b80d8
2004-04-03 00:45:16 +00:00
Chris Lattner
002be76733 Add some missing functions. Make sure to handle calls together in case the
client has another VN implementation that can VN calls.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@12427 91177308-0d34-0410-b5e6-96231b3b80d8
2004-03-16 03:41:35 +00:00
Chris Lattner
5a6e947f1d Implement CSE of call instructions in the most trivial case. This implements
GCSE/call_cse.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@12419 91177308-0d34-0410-b5e6-96231b3b80d8
2004-03-15 05:44:59 +00:00
Chris Lattner
3b303d91d7 In a "seeing the forest through the trees" kinda situation, I realized that a
complete rewrite of load-vn will make it a bit faster.  This changes speeds up
the gcse pass (which uses load-vn) from 25.45s to 0.42s on the testcase in
PR209.

I've also verified that this gives the exact same results as the old one.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@11132 91177308-0d34-0410-b5e6-96231b3b80d8
2004-02-05 17:20:00 +00:00
Chris Lattner
57ef9a2392 This is a big diff with no functionality change. We just reorder some code,
which causes big reindentation.  While I'm at it, I fix the fixme by removing
some dead code.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@11131 91177308-0d34-0410-b5e6-96231b3b80d8
2004-02-05 05:56:23 +00:00
Chris Lattner
270db367e0 finegrainify namespacification
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@11130 91177308-0d34-0410-b5e6-96231b3b80d8
2004-02-05 05:51:40 +00:00
Tanya Lattner
1ad5bc5616 Fixed Chris' typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@11128 91177308-0d34-0410-b5e6-96231b3b80d8
2004-02-05 04:45:21 +00:00
Chris Lattner
adf9b90411 Implement optimizations for handling large basic blocks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@11126 91177308-0d34-0410-b5e6-96231b3b80d8
2004-02-05 00:36:43 +00:00
Brian Gaeke
d0fde30ce8 Put all LLVM code into the llvm namespace, as per bug 109.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@9903 91177308-0d34-0410-b5e6-96231b3b80d8
2003-11-11 22:41:34 +00:00
John Criswell
b576c94c15 Added LLVM project notice to the top of every C++ source file.
Header files will be on the way.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@9298 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-20 19:43:21 +00:00
Misha Brukman
2f2d06506c Fixed spelling and grammar.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8478 91177308-0d34-0410-b5e6-96231b3b80d8
2003-09-11 18:14:24 +00:00
Misha Brukman
7bc439a4b6 Spell `definitely' correctly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8466 91177308-0d34-0410-b5e6-96231b3b80d8
2003-09-11 15:31:17 +00:00
Chris Lattner
bd70a31c51 Fix bug: RLE-Preserve-Volatile.ll
Volatile loads and stores must not be value numbered


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8398 91177308-0d34-0410-b5e6-96231b3b80d8
2003-09-08 18:13:58 +00:00
Misha Brukman
5560c9d49c Spell `necessary' correctly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7944 91177308-0d34-0410-b5e6-96231b3b80d8
2003-08-18 14:43:39 +00:00
Chris Lattner
aed2c6d2cb If the alias analysis algorithm we are using can provide MUST alias information,
expose it directly as value numbering information


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6972 91177308-0d34-0410-b5e6-96231b3b80d8
2003-06-29 00:53:34 +00:00
Chris Lattner
28c6cf22de Implement forwarding from stores to loads of must-aliased pointers.
This implements: GCSE/2003-06-13-LoadStoreEliminate.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6694 91177308-0d34-0410-b5e6-96231b3b80d8
2003-06-16 12:06:41 +00:00
Chris Lattner
f98d8d8611 Calculate and pass load sizes to the alias analysis infrastructure
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@5635 91177308-0d34-0410-b5e6-96231b3b80d8
2003-02-26 19:27:35 +00:00
Chris Lattner
71c7ec9b18 Initial implementation of value numbering for load instructions
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@3540 91177308-0d34-0410-b5e6-96231b3b80d8
2002-08-30 20:28:10 +00:00