Commit Graph

13002 Commits

Author SHA1 Message Date
Chris Lattner 558cebc775 Fix for PR341
llvm-svn: 14842
2004-07-15 02:26:49 +00:00
Chris Lattner 7dc9de5f43 Patches towards fixing PR341
llvm-svn: 14841
2004-07-15 02:14:30 +00:00
Chris Lattner 32c518e526 Progress on PR341
llvm-svn: 14840
2004-07-15 02:06:12 +00:00
Chris Lattner 9a63520b1a Fixes working towards PR341
llvm-svn: 14839
2004-07-15 01:50:47 +00:00
Chris Lattner cae777bb4f Bug fixes for PR341
llvm-svn: 14838
2004-07-15 01:29:12 +00:00
Chris Lattner 4bbd409749 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 030f84f215 Yeah, how about those 'primative' types huh?
llvm-svn: 14836
2004-07-15 01:15:53 +00:00
Chris Lattner ba7aef39fd 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 33930ad7bd 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 969d6fbcee You cannot pass functions by value, what does that even mean??
llvm-svn: 14833
2004-07-14 23:14:07 +00:00
Chris Lattner 1e780d3d26 this "feature" is gone
llvm-svn: 14832
2004-07-14 23:12:09 +00:00
Chris Lattner 4babbe04e1 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 8abe1a11b4 ** 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 7e40f4f9c5 New, horrible, testcase
llvm-svn: 14829
2004-07-14 22:59:47 +00:00
Chris Lattner 4c9210ed64 * 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 e875d48a92 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 def1a51232 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 cbdf371d30 Simplify logic.
llvm-svn: 14825
2004-07-14 20:27:12 +00:00
Chris Lattner 47fb34c00a 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 91393ee84f 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 2661c50173 Make sure MTSPR instruction is inserted into the BasicBlock
llvm-svn: 14822
2004-07-14 18:26:31 +00:00
Chris Lattner 92aecc9e3c Implement new helper methods for creating two-index GEP instructions
llvm-svn: 14821
2004-07-14 18:14:33 +00:00
Chris Lattner 29f231ce4e * 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 5193410cd9 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 d4cbe67ce6 * 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 42dd47437e 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 70ef57d001 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 6652224c2a 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 88357937b6 Finegrainify namespacification
llvm-svn: 14808
2004-07-14 06:39:48 +00:00
Chris Lattner 4accae9c3e 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 590b5f12db Add Machine-CFG edges to SparcV9 MachineBasicBlocks.
llvm-svn: 14806
2004-07-14 05:40:50 +00:00
Chris Lattner 95230b0cbc Split the basic block handling case out of getVal into getBBVal.
llvm-svn: 14805
2004-07-14 01:33:11 +00:00
Chris Lattner e52ed45890 Make Argument::print more resilient to non-verifiable IR
llvm-svn: 14801
2004-07-13 23:14:34 +00:00
Chris Lattner 8c54e58f7c Catch aggregates passed by value sooner rather than later.
llvm-svn: 14800
2004-07-13 20:09:51 +00:00
Misha Brukman aa678b5ed4 * 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
Chris Lattner ab1aedab13 Actually, use an exception to stop verification. This gives us much better
error messages because verification stops at the first error.

llvm-svn: 14794
2004-07-13 08:48:04 +00:00
Chris Lattner 2c1bc9e7e2 There is no reason to abort and print a stack trace if there is a verification
error.  Just print the message like a good little tool.

llvm-svn: 14793
2004-07-13 08:45:41 +00:00
Chris Lattner d25cad9f38 Fine-grainify namespacification, prune #include
llvm-svn: 14792
2004-07-13 08:42:12 +00:00
Chris Lattner e84a2ba99f Fix typeo and refactor bb productions to make it possible for us to reuse any
forward reference blocks if they have been created (instead of creating a new
block, replaceAllUsesOfWith, then nuking the placeholder).   This is not yet
implemented.

llvm-svn: 14791
2004-07-13 08:39:15 +00:00
Chris Lattner d9c9c496e6 Eliminate some mega-cruft here. There is no reason to DERIVE FROM IR CLASSES
just to keep track of some per-object state!  Gaah!  Whoever wrote this stuff...
oh wait, that would be me.  Never mind.

llvm-svn: 14790
2004-07-13 08:28:21 +00:00
Chris Lattner 30b35cf3f7 Inline the now trivial setValueNameInternal function into both callers
llvm-svn: 14789
2004-07-13 08:12:39 +00:00
Chris Lattner 85a351f78d Now that basic blocks are eagerly inserted into the Function, we can use
the funciton symbol table to check for conflicts instead of having to
keep a shadow named LocalSymtab.  Totally eliminate LocalSymtab.  Verified
that this did not cause a regression on the testcase for PR107.

llvm-svn: 14788
2004-07-13 08:10:10 +00:00
Chris Lattner 2ed776bb88 A couple of substantial cleanup fixes:
1. Split setValueName into two separate functions, one that is only used
     at function scope and doesn't have to deal with duplicates, and one
     that can be used either at global or function scope but that does deal
     with conflicts.  Conflicts were only in there because of the crappy old
     CFE and probably should be entirely eliminated.
  2. Insert BasicBlock's into the parent functions when they are created
     instead of when they are complete.  This effects name lookup (for the
     better), which will be exploited in the next patch.

llvm-svn: 14787
2004-07-13 07:59:27 +00:00
Brian Gaeke b88d2a44f9 Fix warning on SparcV9, where sizeof (int) != sizeof (void *).
llvm-svn: 14786
2004-07-13 07:37:43 +00:00
Chris Lattner cc459c9d66 Replace a bunch of complex ConstantPointerRef referring code with simple
code.

llvm-svn: 14785
2004-07-13 06:58:07 +00:00
Chris Lattner 8e30a4dfc5 Make tblgen not try to be smart. This is better handled in makefiles if
at all.  Patch contributed by Vladimir Prus!

llvm-svn: 14784
2004-07-13 06:11:46 +00:00
Chris Lattner 35e24774eb Factor some code to handle "load (constantexpr cast foo)" just like
"load (cast foo)".  This allows us to compile C++ code like this:

class Bclass {
  public: virtual int operator()() { return 666; }
};

class Dclass: public Bclass {
  public: virtual int operator()() { return 667; }
} ;

int main(int argc, char** argv) {
  Dclass x;
  return x();
}

Into this:

int %main(int %argc, sbyte** %argv) {
entry:
        call void %__main( )
        ret int 667
}

Instead of this:

int %main(int %argc, sbyte** %argv) {
entry:
        %x = alloca "struct.std::bad_typeid"            ; <"struct.std::bad_typeid"*> [#uses=3]
        call void %__main( )
        %tmp.1.i.i = getelementptr "struct.std::bad_typeid"* %x, uint 0, uint 0, uint 0         ; <int (...)***> [#uses=1]
        store int (...)** getelementptr ([3 x int (...)*]*  %vtable for Bclass, int 0, long 2), int (...)*** %tmp.1.i.i
        %tmp.3.i = getelementptr "struct.std::bad_typeid"* %x, int 0, uint 0, uint 0            ; <int (...)***> [#uses=1]
        store int (...)** getelementptr ([3 x int (...)*]*  %vtable for Dclass, int 0, long 2), int (...)*** %tmp.3.i
        %tmp.5 = load int ("struct.std::bad_typeid"*)** cast (int (...)** getelementptr ([3 x int (...)*]*  %vtable for Dclass, int 0, long 2) to int
("struct.std::bad_typeid"*)**)          ; <int ("struct.std::bad_typeid"*)*> [#uses=1]
        %tmp.6 = call int %tmp.5( "struct.std::bad_typeid"* %x )                ; <int> [#uses=1]
	ret int %tmp.6
        ret int 0
}

In order words, we now resolve the virtual function call.

llvm-svn: 14783
2004-07-13 01:49:43 +00:00
Misha Brukman be1b680f54 Correctly load FP constants out of the constant pool.
llvm-svn: 14782
2004-07-12 23:49:47 +00:00
Misha Brukman 9dacda651f Apple's MacOS X is another OS which does not provide alloca() via <alloca.h>
llvm-svn: 14781
2004-07-12 23:37:18 +00:00
Misha Brukman dcb0eec80d Implement getModuleMatchQuality and getJITMatchQuality() for PowerPC
llvm-svn: 14780
2004-07-12 23:36:12 +00:00