Commit Graph

3694 Commits

Author SHA1 Message Date
Anand Shukla
0df7ddbd31 A pass to combine multiple backedges that go to same target
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7201 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-18 16:08:32 +00:00
Brian Gaeke
bfedb912cd Use getClassB for load and store; we don't want to abort when we
try to load or store through a bool*.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7195 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-17 21:30:06 +00:00
Brian Gaeke
1d2ba44224 Fix typo in call to isUnresolvableFunc, which was breaking the build.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7194 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-17 19:07:46 +00:00
John Criswell
a676b78791 Dinakar and I fixed a bug where we were trying to get the initializer of
an external constant.  Since external constants don't have initializers, we
were failing on an assert() call in llvm/GlobalVariable.h.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7193 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-17 19:06:55 +00:00
Vikram S. Adve
1da1d32fc7 Rematerialize nodes from the globals graph into the current graph
after all callees are inlined into the current graph.

NOTE: There's also a major bug fix for the BU pass in DataStructure.cpp,
which ensures that resolvable indirect calls are not moved out to the
globals graph, so that they are eventually inlined (if possible).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7189 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-16 21:42:03 +00:00
Vikram S. Adve
03e19ddaf7 (1) Rematerialize nodes from the globals graph into the current graph
after all callers are inlined into the current graph.
(2) Optimize the way a graph is inlined into its callees in the TD phase:
    (a) Use DSGraph::cloneReachableSubgraph to clone only a subgraph at
        each call site, for faster inlining.
    (b) Clone separately for the same callee at different call sites,
        since only the reachable subgraph is being cloned, not the entire
        caller graph.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7188 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-16 21:40:28 +00:00
Vikram S. Adve
78bbec7fcb Implement 2 important changes: (1) rematerialization from the globals graph,
and (2) faster inlining by cloning only reachable nodes.  In particular:
(1) Added DSGraph::cloneReachableSubgraph and DSGraph::cloneReachableNodes
    to clone the subgraph reachable from a set of root nodes, into the
    current graph, merging the global nodes into thos in the current graph.
    The TD pass now uses this for faster inlining, and so does the
    next function.
(2) Added DSGraph::updateFromGlobalGraph() to rematerialize nodes from the
    globals graph into the current graph in both BU and TD passes.
(3) `I' flags are removed from all nodes in the globals graph, because they
    are difficult to maintain correctly and are not needed anyway.
(4) Aux. function calls are only removed to the globals graph if they
    will never be resovled.  (This is what fixed gap.)  The immediate
    reason is that if we took these out of a function (and moved them to
    the globals graph) we would need to rematerialize these nodes into the
    function graph for every function in the BU pass.  The longer term
    problem is that we would need to find a way to remove them from the
    globals graph iff they have been resolved on all paths through the
    call graph.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7187 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-16 21:36:31 +00:00
Vikram S. Adve
2e1de5ed59 Factor out the test for unresolvable external functions into
isUnresolvableFunc() (I thought I needed this externally.
I don't, but it's still nicer this way.)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7186 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-16 21:25:17 +00:00
Misha Brukman
dfbfc57d78 Fixed the number translation scheme for the integer condition code registers: it
now works in instructions which require a 2-bit or 3-bit INTcc code.

Incidentally, that means that the representation of INTcc registers is now the
same in both integer and FP instructions. Thus, code became much simpler and
cleaner.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7185 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-16 20:30:40 +00:00
Misha Brukman
cf7ec875ba The name should really be `simm11' to follow the naming convention, but this has
no change in functionality.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7184 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-16 20:27:44 +00:00
Misha Brukman
2c4fe5813c No need for a second immediate field if the class already inherits one.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7182 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-15 21:27:14 +00:00
Misha Brukman
a34b619f67 Encode predict = 1 by default, because the Sparc assembler does this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7181 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-15 21:26:49 +00:00
Misha Brukman
07d4516f92 Correctly handle calls to functions which are further away than 2**32 bits will
allow, i.e. make a sequence of instructions to enable an indirect call using
jump-and-link and 2 temporary registers (which we save and ultimately restore).

Warning: if the delay slot of a function call is used to do meaningful work and
not just a NOP, this behavior is incorrect. However, the Sparc backend does not
yet utilize the delay slots effectively, so it is not necessary to make an
overly complicated algorithm for something that's not used.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7178 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-15 19:09:43 +00:00
Misha Brukman
859e09f06e Clean up my last checkin: code is easier to read and explains the differences in
usage of the special file handle RTLD_SELF on Sparc/Solaris vs. 0 on Linux/x86.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7177 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-15 15:58:26 +00:00
Misha Brukman
291c2c59b0 On Sparc/Solaris, the special handle RTLD_SELF is used as a handle referring to
the program's executing image, not 0 as it is on Linux/x86 and possibly other
systems.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7176 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-15 15:55:32 +00:00
Misha Brukman
173e250548 * Added support for the %ccr register
* FP double registers are now coded correctly
* Removed function which converted registers based on register types, it was
  broken (because regTypes are broken)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7175 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-14 23:26:03 +00:00
Misha Brukman
1ba3138cdc The word separate' only has one e'.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7174 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-14 17:26:34 +00:00
Misha Brukman
bc0e998c49 The word separate' only has one e'.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7173 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-14 17:20:40 +00:00
Vikram S. Adve
919fc8c367 Trace loads and stores as they happen (stores were being
remembered in valuesStoredInFunction, but never traced at function return,
and that's too late to be finding the error anyway).
Stores trace both the value and the address being stored to,
but after some experience I think only values should be traced.
The pointer hash table just fills up far too quickly if every
store address were traced.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7169 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-11 21:57:43 +00:00
Brian Gaeke
0517c5ac92 If we have a constant pointer reference to a function, we were printing
out the entire llvm disassembly for the function at global constant-output
time, which caused the assembler to barf in 164.gzip. This fixes that
particular problem (though 164.gzip has other problems with X86 llc.)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7168 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-11 21:57:01 +00:00
Brian Gaeke
1aa476edf0 Fix some serious floating-point bugs (fixes test cases such as Oscar,
Fhourstones, McCat-vor, and many others...)

Printer.cpp: Print implicit uses for AddRegFrm instructions.  Break gas
bug workarounds up into separate stanzas of code for each bug.  Add new
workarounds for fild and fistp.

X86InstrInfo.def: Add O_ST0 implicit uses for more FP instrs where they
obviously apply. Also add PrintImplUses flags for FP instrs where they
are necessary for gas to understand the output.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7165 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-11 18:18:35 +00:00
Anand Shukla
666ff520e6 Added functionality to instrmentation pass
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7161 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 21:55:57 +00:00
Misha Brukman
1115e0483f Fixed grammatical error.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7160 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 21:38:28 +00:00
Vikram S. Adve
951df2b1bd Several important bug fixes:
(1) Cannot use ANDN(ot), ORN, and XORN for boolean ops, only bitwise ops.

(2) Conditional move instructions must distinguish signed and unsigned
    condition codes, e.g., MOVLE vs. MOVLEU.

(3) Conditional-move-on-register was using the cond-move-on-cc opcodes,
    which produces a valid-looking instruction with bogus registers!

(4) Here's a really cute one: dividing-by-2^k for negative numbers needs to
    add 2^k-1 before shifting, not add 1 after shifting.  Sadly, these
    are the same when k=0 so our poor test case worked fine.

(5) Casting between signed and unsigned values was not correct:
    completely reimplemented.

(6) Zero-extension on unsigned values was bogus: I was only doing the
    SRL and not the SLLX before it.  Don't know WHAT I was thinking!

(7) And the most important class of changes: Sign-extensions on signed values.
    Signed values are not sign-extended after ordinary operations,
    so they must be sign-extended before the following cases:
	-- passing to an external or unknown function
	-- returning from a function
	-- using as operand 2 of DIV or REM
	-- using as either operand of condition-code setting operation
           (currently only SUBCC), with smaller than 32-bit operands


Also, a couple of improvements:

(1) Fold cast-to-bool into Not(bool).  Need to do this for And, Or, XOR also.

(2) Convert SetCC-Const into a conditional-move-on-register (case 41)
    if the constant is 0.  This was only being done for branch-on-SetCC-Const
    when the branch is folded with the SetCC-Const.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7159 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 20:07:54 +00:00
Vikram S. Adve
940a3a47b0 Bug fix in creating constants: need 1U << 31, not 1 << 31.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7158 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:48:19 +00:00
Vikram S. Adve
ad80db6d7d Fold cast-to-bool into not. Later, this should also be folded into other
boolean operations: AND, OR, XOR.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7157 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:47:42 +00:00
Vikram S. Adve
4d2faf6307 Values stored in CallArgsDescriptor cannot be const.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7156 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:46:15 +00:00
Vikram S. Adve
24d0410db0 isMarkedForSpill() should be const.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7155 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:45:28 +00:00
Vikram S. Adve
627eb31cd7 Change interface to MachineInstr::substituteValue to specify more precisely
which args can be substituted: defsOnly, defsAndUses or usesOnly.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7154 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:45:07 +00:00
Vikram S. Adve
51bda6fccb Print out all neighbors in interference graph, not just higher-numbered ones.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7153 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:43:33 +00:00
Vikram S. Adve
3bf089227b Several fixes to handling of int CC register:
(1) An int CC live range must be spilled if there are any interferences,
    even if no other "neighbour" in the interf. graph has been allocated
    that reg. yet.  This is actually true of any class with only one reg!

(2) SparcIntCCRegClass::colorIGNode sets the color even if the LR must
    be spilled so that the machine-independent spill code doesn't have to
    make the machine-dependent decision of which CC name to use based on
    operand type: %xcc or %icc.  (These are two halves of the same
register.)

(3) LR->isMarkedForSpill() is no longer the same as LR->hasColor().
    These should never have been the same, and this is necessary now for #2.

(4) All RDCCR and WRCCR instructions are directly generated with the
    phony number for %ccr so that EmitAssembly/EmitBinary doesn't have to
    deal with this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7152 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:42:55 +00:00
Vikram S. Adve
b15f8d446a Several fixes to handling of int CC register:
(1) An int CC live range must be spilled if there are any interferences,
    even if no other "neighbour" in the interf. graph has been allocated
    that reg. yet.  This is actually true of any class with only one reg!

(2) SparcIntCCRegClass::colorIGNode sets the color even if the LR must
    be spilled so that the machine-independent spill code doesn't have to
    make the machine-dependent decision of which CC name to use based on
    operand type: %xcc or %icc.  (These are two halves of the same register.)

(3) LR->isMarkedForSpill() is no longer the same as LR->hasColor().
    These should never have been the same, and this is necessary now for #2.

(4) All RDCCR and WRCCR instructions are directly generated with the
    phony number for %ccr so that EmitAssembly/EmitBinary doesn't have to
    deal with this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7151 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 19:42:11 +00:00
Misha Brukman
b5c520bfb6 Lowercase versions of `occurrence' need to be spelled correctly, too.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7142 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 17:05:26 +00:00
Misha Brukman
dd6cb6a43b Occurrence' has no a' and the `r' is doubled.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7140 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-10 16:49:51 +00:00
Misha Brukman
ac25de3bf0 Some beautification changes (tabs to spaces, removed extra blank lines);
no functionality changed.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7138 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-09 17:33:50 +00:00
John Criswell
cf6d7b0353 Removed size_t from the parameter list of the malloc() prototype in generated C
code.  This prevents a compiler warning on Sparc that causes the tests to fail.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7125 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-08 14:52:09 +00:00
Misha Brukman
2988612f1b Elaborated assembly syntax of instructions in the comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7120 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-07 22:18:42 +00:00
Misha Brukman
35f19cc7b2 Removed unnecessary assignment (it was taken care by a superclass) and clarified
some comments.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7119 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-07 22:18:06 +00:00
Brian Gaeke
b44210d768 Insert workaround for GAS bug in assembling FLD/FSTP XWORD PTR [...]
instructions, by outputting them as bytes.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7115 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-07 18:34:20 +00:00
Misha Brukman
de11f2d038 Moved RegClassIDs enum to be next to the RegTypes enum.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7114 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-07 16:52:39 +00:00
Vikram S. Adve
9c570ee129 Correction to last fix: Pointer types do not return true in Type::IsIntegral().
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7113 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-06 22:50:31 +00:00
Vikram S. Adve
97539fc4b6 Choose register instead of immediate for ConstantExpr in ChooseRegOrImmed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7112 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-06 20:33:21 +00:00
Vikram S. Adve
786833ad34 Major bug fix though it happened rarely (only on a compare after an
integer overflow):
We need to use %icc and not %xcc for comparisons on 32-bit or smaller
integer values.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7111 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-06 20:13:59 +00:00
Vikram S. Adve
46d3f8e34c Make the RegClassID values public -- there is no other way to get them.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7109 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-06 19:53:59 +00:00
Misha Brukman
9cedd433eb Apparently, the "regType" and "regClass" used in the Sparc backend are not both
correct: empirically, "regType" is wrong for a number of registers. Thus, one
can only rely on the "regClass" to figure out what kind of register one is
dealing with.

This change switches to using only "regClass" and adds a few extra DEBUG() print
statements and a few clean-ups in comments and code, mostly minor.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7103 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-03 18:36:47 +00:00
Chris Lattner
5bfccb9def Reuse the values if they are constants: this is important so that we index into the right structure field
This fixes bug: BasicAA/2003-07-03-BasicAACrash.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7093 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-03 06:42:38 +00:00
Chris Lattner
85cfe01a5e Remove globals more aggressively from graphs.
Fix a bug where we removed nodes that were marked U.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7090 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-03 02:03:53 +00:00
Chris Lattner
7079386e02 INCLUDE_PARENT_GRAPH is required!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7088 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-02 23:57:05 +00:00
Chris Lattner
10c45d6cd5 Disable incorrect mustalias code
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7087 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-02 23:56:51 +00:00
Chris Lattner
61691c5d85 Remove space at end of line
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7084 91177308-0d34-0410-b5e6-96231b3b80d8
2003-07-02 23:44:15 +00:00