13017 Commits

Author SHA1 Message Date
Chris Lattner
952bbd726b This patch was contributed by Daniel Berlin!
Speed up SCCP substantially by processing overdefined values quickly.  This
patch speeds up SCCP by about 30-40% on large testcases.

llvm-svn: 14861
2004-07-15 23:36:43 +00:00
Chris Lattner
0193bbef1f Fix PR404 try #2
This version takes about 1s longer than the previous one (down to 2.35s),
but on the positive side, it actually works :)

llvm-svn: 14856
2004-07-15 08:20:22 +00:00
Chris Lattner
1f83f10da7 Give SetVector range support
llvm-svn: 14855
2004-07-15 08:18:31 +00:00
Chris Lattner
36f2a4eb77 Nuke a clearly bogus assertion
llvm-svn: 14854
2004-07-15 07:44:34 +00:00
Chris Lattner
fde0e6bb95 Revert previous patch until I get a bug fixed
llvm-svn: 14853
2004-07-15 05:36:31 +00:00
Chris Lattner
525a4d1e7e QOI issue fixed
llvm-svn: 14852
2004-07-15 04:30:35 +00:00
Chris Lattner
9896dfb8e7 Fix PR404: Loop simplify is really slow on 252.eon
This eliminates an N*N*logN algorithm from the loop simplify pass, replacing
it with a much simpler and faster alternative.  In a debug build, this reduces
gccas time on eon from 85s to 42s.

llvm-svn: 14851
2004-07-15 04:27:04 +00:00
Chris Lattner
6d978d5864 Bug fixed
llvm-svn: 14850
2004-07-15 02:59:24 +00:00
Chris Lattner
c19f2134fc Final fix for PR341: eliminate operator<<(ostream, Value*). Clients should
now send references to ostreams instead of pointers.  Sending pointers to
ostreams will print their addresses.

llvm-svn: 14849
2004-07-15 02:54:36 +00:00
Chris Lattner
d7870e4b1a Fixes for PR341
llvm-svn: 14848
2004-07-15 02:51:32 +00:00
Chris Lattner
94212f7102 Fixes for PR341
llvm-svn: 14847
2004-07-15 02:51:31 +00:00
Chris Lattner
5ab6b23a7d Fix for PR341
llvm-svn: 14846
2004-07-15 02:50:38 +00:00
Chris Lattner
47c6a06211 Fix for PR341
llvm-svn: 14845
2004-07-15 02:40:04 +00:00
Chris Lattner
b9ec0b791c Revert stuff that I didn't mean to checkin
llvm-svn: 14844
2004-07-15 02:33:38 +00:00
Chris Lattner
bb7fe18493 Fixes for PR341
llvm-svn: 14843
2004-07-15 02:31:46 +00:00
Chris Lattner
d7d7bc70fd Fix for PR341
llvm-svn: 14842
2004-07-15 02:26:49 +00:00
Chris Lattner
c4888ccda7 Patches towards fixing PR341
llvm-svn: 14841
2004-07-15 02:14:30 +00:00
Chris Lattner
2db1894038 Progress on PR341
llvm-svn: 14840
2004-07-15 02:06:12 +00:00
Chris Lattner
0f61d55197 Fixes working towards PR341
llvm-svn: 14839
2004-07-15 01:50:47 +00:00
Chris Lattner
ffd5effa5f Bug fixes for PR341
llvm-svn: 14838
2004-07-15 01:29:12 +00:00
Chris Lattner
d1b726d18f Implement folding of expressions like 'uint cast (int* getelementptr (int*
null, uint 1) to uint)' to a constant integer.  We can only do this with
primitive LLVM types, because other types have target-specific sizes.

llvm-svn: 14837
2004-07-15 01:16:59 +00:00
Chris Lattner
26e5829b8f Yeah, how about those 'primative' types huh?
llvm-svn: 14836
2004-07-15 01:15:53 +00:00
Chris Lattner
de4449ef34 Now that we codegen the portable "sizeof" efficiently, we can use it for
malloc lowering.  This means that lowerallocations doesn't need targetdata
anymore.  yaay.

llvm-svn: 14835
2004-07-15 01:08:08 +00:00
Chris Lattner
210ffe4b77 Improve codegen for the LLVM offsetof/sizeof "operator". Before we compiled
this LLVM function:

int %foo() {
        ret int cast (int** getelementptr (int** null, int 1) to int)
}

into:

foo:
        mov %EAX, 0
        lea %EAX, DWORD PTR [%EAX + 4]
        ret

now we compile it into:

foo:
        mov %EAX, 4
        ret

This sequence is frequently generated by the MSIL front-end, and soon the malloc lowering pass and
Java front-ends as well..

-Chris

llvm-svn: 14834
2004-07-15 00:58:53 +00:00
Chris Lattner
626552badd You cannot pass functions by value, what does that even mean??
llvm-svn: 14833
2004-07-14 23:14:07 +00:00
Chris Lattner
a4870e6ca2 this "feature" is gone
llvm-svn: 14832
2004-07-14 23:12:09 +00:00
Chris Lattner
7397a61a7f This is logically part of the last patch. Just more really horrible code
that is made unnecessary by it.

llvm-svn: 14831
2004-07-14 23:07:13 +00:00
Chris Lattner
d81b8323ee ** Finally DeclareNewGlobalValue is dead!
* Simplify a lot of code because type's cannot be in function symbol tables
* Fix memory leaks in handling of redefined function prototypes
* Don't use SymbolTable directly for stuff that we can go through the Module
  for.
* Fix some minor bugs on obscure testcases like:
      test/Feature/globalredefinition.ll
* Do not create GlobalVariable objects for forward referenced Functions!
* When forward referencing a function in a constant expression, do not create
  a placeholder, add a bunch of references to it, then turn around and
  replaceAllUsesOfWith on it with a new global, deleting the placeholder.
  Instead, when we find the real definition of the global, just use the
  placeholder instead of creating a new object.

This substantially simplifies the asmwriter and should even speed it up on
cases heavy in constantexprs (like C++, Java, MSIL)...

llvm-svn: 14830
2004-07-14 23:03:46 +00:00
Chris Lattner
17ec8fde04 New, horrible, testcase
llvm-svn: 14829
2004-07-14 22:59:47 +00:00
Chris Lattner
7bef7be4c4 * Fairly substantial change. Instead of creating new globalvariables, then
replaceAllUsesWith'ing any forward references, just use the forward
  reference if it exists.

This introduces GetForwardRefForGlobal, which will eventually completely
replace the horrible DeclareNewGlobalValue function.

llvm-svn: 14828
2004-07-14 21:44:00 +00:00
Chris Lattner
5b80b7157a Fold setValueNameMergingDuplicates into ParseGlobalVariable, allowing us
to substantially simplify the result.  In particular, we no longer create
GlobalVariables and then immediately destroy them when they are duplciate
definitions.

The real point of this patch though is that it gets us closer to the
DeclareNewGlobalValue calls...

llvm-svn: 14827
2004-07-14 20:42:57 +00:00
Brian Gaeke
d8746a7ab6 Actually set the endian/pointersize flags on the module being read in!
llvm-svn: 14826
2004-07-14 20:33:13 +00:00
Chris Lattner
fb6e1c2007 Simplify logic.
llvm-svn: 14825
2004-07-14 20:27:12 +00:00
Chris Lattner
9e4dad1400 Make PATypeHolder and friends return non-const pointers to the types they
hold.  Because types are basically immutable anyway, they should not be
referenced as "const Type*" everywhere.  Just "Type*" should suffice!

llvm-svn: 14824
2004-07-14 20:10:26 +00:00
Chris Lattner
7044b0d4e1 Fix a regression from last night. Apparently the CFE is broken and outputs
functions multiple times, expecting them to be merged.  This should be fixed
in the CFE, then here.

llvm-svn: 14823
2004-07-14 19:33:47 +00:00
Misha Brukman
d6a9646f29 Make sure MTSPR instruction is inserted into the BasicBlock
llvm-svn: 14822
2004-07-14 18:26:31 +00:00
Chris Lattner
74bd192fba Implement new helper methods for creating two-index GEP instructions
llvm-svn: 14821
2004-07-14 18:14:33 +00:00
Chris Lattner
91169a26b5 * Change code to use a static_cast instead of reinterpret_cast
* Add comments
* Add two new gep ctors for the very common case of creating a two operand
  GEP instruction.

llvm-svn: 14820
2004-07-14 18:13:54 +00:00
Misha Brukman
0f42826f30 Don't define the same register twice when loading a ConstantPointerRef to a reg
llvm-svn: 14819
2004-07-14 17:57:04 +00:00
Misha Brukman
73d729838f * Fix multiplication by powers of two and otherwise
* Clarify variable name (StoreInst SI instead of LI)

llvm-svn: 14818
2004-07-14 15:29:51 +00:00
Chris Lattner
6472a6b770 Pull out code shared between GV forward-decl and definition processing.
This gives us only a single call site for setValueNameMergingDuplicates.
The next stage is the start merging them together.

llvm-svn: 14811
2004-07-14 08:23:52 +00:00
Chris Lattner
9827ad9f08 Simplify code. Do not allow functions to be redefined more than once.
Since the stupid '%X = const int 4' thing is gone, we can now simplify
setValueNameMergingDuplicates a bit more.

llvm-svn: 14810
2004-07-14 07:12:48 +00:00
Chris Lattner
fafef8ecd1 Remove a gross and crufty "feature" that was never documented and doesn't work.
llvm-svn: 14809
2004-07-14 06:44:56 +00:00
Chris Lattner
633e78c5c6 Finegrainify namespacification
llvm-svn: 14808
2004-07-14 06:39:48 +00:00
Chris Lattner
a262a55378 Revamp handling of labels. In particular, if we create a forward reference
for a basic block, use it when the block is defined instead of deleting it
and creating a new one.  Also, only create at most ONE forward reference
for any block, instead of one for each forward reference.

llvm-svn: 14807
2004-07-14 06:28:35 +00:00
Brian Gaeke
4dc7caee0e Add Machine-CFG edges to SparcV9 MachineBasicBlocks.
llvm-svn: 14806
2004-07-14 05:40:50 +00:00
Chris Lattner
483a35c547 Split the basic block handling case out of getVal into getBBVal.
llvm-svn: 14805
2004-07-14 01:33:11 +00:00
Chris Lattner
183766f21e Make Argument::print more resilient to non-verifiable IR
llvm-svn: 14801
2004-07-13 23:14:34 +00:00
Chris Lattner
1a1afa3566 Catch aggregates passed by value sooner rather than later.
llvm-svn: 14800
2004-07-13 20:09:51 +00:00
Misha Brukman
79b089abff * Specify that FP arith options have 3 operands
* Correctly load FP constants from the constant pool, should be refactored

llvm-svn: 14799
2004-07-13 15:35:45 +00:00