llvm-svn: 11892

Functions with linkonce linkage are declared with weak linkage.
Global floating point constants used to represent unprintable values
(such as NaN and infinity) are declared static so that they don't interfere
with other CBE generated translation units.

llvm-svn: 11884

are beastly ConstantPointerRefs in the way...

llvm-svn: 11883

MRegisterInfo::is{Physical,Virtual}Register. Apply appropriate fixes
to relevant files.

llvm-svn: 11882

multiple type names for the same structural type.  Make DTE eliminate all
but one of the type names

llvm-svn: 11879

llvm-svn: 11875

llvm-svn: 11872

having the compiler emit RTTI and vtables to EVERY translation unit.

llvm-svn: 11871

   if (X == 0 || X == 2)

...where the comparisons and branches are in different blocks... into a switch
instruction.  This comes up a lot in various programs, and works well with
the switch/switch merging code I checked earlier.  For example, this testcase:

int switchtest(int C) {
  return C == 0 ? f(123) :
         C == 1 ? f(3123) :
         C == 4 ? f(312) :
         C == 5 ? f(1234): f(444);
}

is converted into this:
        switch int %C, label %cond_false.3 [
                 int 0, label %cond_true.0
                 int 1, label %cond_true.1
                 int 4, label %cond_true.2
                 int 5, label %cond_true.3
        ]

instead of a whole bunch of conditional branches.

Admittedly the code is ugly, and incomplete.  To be complete, we need to add
br -> switch merging and switch -> br merging.  For example, this testcase:

struct foo { int Q, R, Z; };
#define A (X->Q+X->R * 123)
int test(struct foo *X) {
  return A  == 123 ? X1() :
        A == 12321 ? X2():
        (A == 111 || A == 222) ? X3() :
        A == 875 ? X4() : X5();
}

Gets compiled to this:
        switch int %tmp.7, label %cond_false.2 [
                 int 123, label %cond_true.0
                 int 12321, label %cond_true.1
                 int 111, label %cond_true.2
                 int 222, label %cond_true.2
        ]
...
cond_false.2:           ; preds = %entry
        %tmp.52 = seteq int %tmp.7, 875         ; <bool> [#uses=1]
        br bool %tmp.52, label %cond_true.3, label %cond_false.3

where the branch could be folded into the switch.

This kind of thing occurs *ALL OF THE TIME*, especially in programs like
176.gcc, which is a horrible mess of code.  It contains stuff like *shudder*:

#define SWITCH_TAKES_ARG(CHAR) \
  (   (CHAR) == 'D' \
   || (CHAR) == 'U' \
   || (CHAR) == 'o' \
   || (CHAR) == 'e' \
   || (CHAR) == 'u' \
   || (CHAR) == 'I' \
   || (CHAR) == 'm' \
   || (CHAR) == 'L' \
   || (CHAR) == 'A' \
   || (CHAR) == 'h' \
   || (CHAR) == 'z')

and

#define CONST_OK_FOR_LETTER_P(VALUE, C)                 \
  ((C) == 'I' ? SMALL_INTVAL (VALUE)                    \
   : (C) == 'J' ? SMALL_INTVAL (-(VALUE))               \
   : (C) == 'K' ? (unsigned)(VALUE) < 32                \
   : (C) == 'L' ? ((VALUE) & 0xffff) == 0               \
   : (C) == 'M' ? integer_ok_for_set (VALUE)            \
   : (C) == 'N' ? (VALUE) < 0                           \
   : (C) == 'O' ? (VALUE) == 0                          \
   : (C) == 'P' ? (VALUE) >= 0                          \
   : 0)

and

#define LEGITIMIZE_ADDRESS(X,OLDX,MODE,WIN)                     \
{                                                               \
  if (GET_CODE (X) == PLUS && CONSTANT_ADDRESS_P (XEXP (X, 1))) \
    (X) = gen_rtx (PLUS, SImode, XEXP (X, 0),                   \
                   copy_to_mode_reg (SImode, XEXP (X, 1)));     \
  if (GET_CODE (X) == PLUS && CONSTANT_ADDRESS_P (XEXP (X, 0))) \
    (X) = gen_rtx (PLUS, SImode, XEXP (X, 1),                   \
                   copy_to_mode_reg (SImode, XEXP (X, 0)));     \
  if (GET_CODE (X) == PLUS && GET_CODE (XEXP (X, 0)) == MULT)   \
    (X) = gen_rtx (PLUS, SImode, XEXP (X, 1),                   \
                   force_operand (XEXP (X, 0), 0));             \
  if (GET_CODE (X) == PLUS && GET_CODE (XEXP (X, 1)) == MULT)   \
    (X) = gen_rtx (PLUS, SImode, XEXP (X, 0),                   \
                   force_operand (XEXP (X, 1), 0));             \
  if (GET_CODE (X) == PLUS && GET_CODE (XEXP (X, 0)) == PLUS)   \
    (X) = gen_rtx (PLUS, Pmode, force_operand (XEXP (X, 0), NULL_RTX),\
                   XEXP (X, 1));                                \
  if (GET_CODE (X) == PLUS && GET_CODE (XEXP (X, 1)) == PLUS)   \
    (X) = gen_rtx (PLUS, Pmode, XEXP (X, 0),                    \
                   force_operand (XEXP (X, 1), NULL_RTX));      \
  if (GET_CODE (X) == SYMBOL_REF || GET_CODE (X) == CONST       \
           || GET_CODE (X) == LABEL_REF)                        \
    (X) = legitimize_address (flag_pic, X, 0, 0);               \
  if (memory_address_p (MODE, X))                               \
    goto WIN; }

and others.  These macros get used multiple times of course.  These are such
lovely candidates for macros, aren't they?  :)

This code also nicely handles LLVM constructs that look like this:

  if (isa<CastInst>(I))
   ...
  else if (isa<BranchInst>(I))
   ...
  else if (isa<SetCondInst>(I))
   ...
  else if (isa<UnwindInst>(I))
   ...
  else if (isa<VAArgInst>(I))
   ...

where the isa can obviously be a dyn_cast as well.  Switch instructions are a
good thing.

llvm-svn: 11870

llvm-svn: 11868

llvm-svn: 11864

not have any globals.

llvm-svn: 11863

llvm-svn: 11862

bugs.  Thanks Brian!

llvm-svn: 11859

llvm-svn: 11858

 1. Functions do not make things incomplete, only variables
 2. Constant global variables no longer need to be marked incomplete, because
    we are guaranteed that the initializer for the global will be in the
    graph we are hacking on now.  This makes resolution of indirect calls happen
    a lot more in the bu pass, supports things like vtables and the C counterparts
    (giant constant arrays of function pointers), etc...

Testcase here: test/Regression/Analysis/DSGraph/constant_globals.ll

llvm-svn: 11852

local graph that uses the global.

llvm-svn: 11850

MRegisterInfo::is{Physical,Virtual}Register.

llvm-svn: 11849

Make the incompleteness marker faster by looping directly over the globals
instead of over the scalars to find the globals

Fix a bug where we didn't mark a global incomplete if it didn't have any
outgoing edges.  This wouldn't break any current clients but is still wrong.

llvm-svn: 11848

llvm-svn: 11847

llvm-svn: 11846

pair, and look up varargs in the execution stack every time, instead of
just pushing iterators (which can be invalidated during callFunction())
around.  (union GenericValue now has a "pair of uints" member, to support
this mechanism.) Fixes Bug 234.

llvm-svn: 11845

llvm-svn: 11844

llvm-svn: 11841

to objects.

llvm-svn: 11840

assume that if they don't intend to write to a global variable, that they
would mark it as constant.  However, there are people that don't understand
that the compiler can do nice things for them if they give it the information
it needs.

This pass looks for blatently obvious globals that are only ever read from.
Though it uses a trivially simple "alias analysis" of sorts, it is still able
to do amazing things to important benchmarks.  253.perlbmk, for example,
contains several ***GIANT*** function pointer tables that are not marked
constant and should be.  Marking them constant allows the optimizer to turn
a whole bunch of indirect calls into direct calls.  Note that only a link-time
optimizer can do this transformation, but perlbmk does have several strings
and other minor globals that can be marked constant by this pass when run
from GCCAS.

176.gcc has a ton of strings and large tables that are marked constant, both
at compile time (38 of them) and at link time (48 more).  Other benchmarks
give similar results, though it seems like big ones have disproportionally
more than small ones.

This pass is extremely quick and does good things.  I'm going to enable it
in gccas & gccld.  Not bad for 50 SLOC.

llvm-svn: 11836

llvm-svn: 11835

llvm-svn: 11834

llvm-svn: 11833

llvm-svn: 11832

llvm-svn: 11830

where there did not used to be any before

llvm-svn: 11829

llvm-svn: 11828

llvm-svn: 11827

llvm-svn: 11826

llvm-svn: 11824

llvm-svn: 11821

scaled indexes.  This allows us to compile GEP's like this:

int* %test([10 x { int, { int } }]* %X, int %Idx) {
        %Idx = cast int %Idx to long
        %X = getelementptr [10 x { int, { int } }]* %X, long 0, long %Idx, ubyte 1, ubyte 0
        ret int* %X
}

Into a single address computation:

test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        lea %EAX, DWORD PTR [%EAX + 8*%ECX + 4]
        ret

Before it generated:
test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        shl %ECX, 3
        add %EAX, %ECX
        lea %EAX, DWORD PTR [%EAX + 4]
        ret

This is useful for things like int/float/double arrays, as the indexing can be folded into
the loads&stores, reducing register pressure and decreasing the pressure on the decode unit.
With these changes, I expect our performance on 256.bzip2 and gzip to improve a lot.  On
bzip2 for example, we go from this:

10665 asm-printer           - Number of machine instrs printed
   40 ra-local              - Number of loads/stores folded into instructions
 1708 ra-local              - Number of loads added
 1532 ra-local              - Number of stores added
 1354 twoaddressinstruction - Number of instructions added
 1354 twoaddressinstruction - Number of two-address instructions
 2794 x86-peephole          - Number of peephole optimization performed

to this:
9873 asm-printer           - Number of machine instrs printed
  41 ra-local              - Number of loads/stores folded into instructions
1710 ra-local              - Number of loads added
1521 ra-local              - Number of stores added
 789 twoaddressinstruction - Number of instructions added
 789 twoaddressinstruction - Number of two-address instructions
2142 x86-peephole          - Number of peephole optimization performed

... and these types of instructions are often in tight loops.

Linear scan is also helped, but not as much.  It goes from:

8787 asm-printer           - Number of machine instrs printed
2389 liveintervals         - Number of identity moves eliminated after coalescing
2288 liveintervals         - Number of interval joins performed
3522 liveintervals         - Number of intervals after coalescing
5810 liveintervals         - Number of original intervals
 700 spiller               - Number of loads added
 487 spiller               - Number of stores added
 303 spiller               - Number of register spills
1354 twoaddressinstruction - Number of instructions added
1354 twoaddressinstruction - Number of two-address instructions
 363 x86-peephole          - Number of peephole optimization performed

to:

7982 asm-printer           - Number of machine instrs printed
1759 liveintervals         - Number of identity moves eliminated after coalescing
1658 liveintervals         - Number of interval joins performed
3282 liveintervals         - Number of intervals after coalescing
4940 liveintervals         - Number of original intervals
 635 spiller               - Number of loads added
 452 spiller               - Number of stores added
 288 spiller               - Number of register spills
 789 twoaddressinstruction - Number of instructions added
 789 twoaddressinstruction - Number of two-address instructions
 258 x86-peephole          - Number of peephole optimization performed

Though I'm not complaining about the drop in the number of intervals.  :)

llvm-svn: 11820

  to do analysis.

*** FOLD getelementptr instructions into loads and stores when possible,
    making use of some of the crazy X86 addressing modes.

For example, the following C++ program fragment:

struct complex {
    double re, im;
    complex(double r, double i) : re(r), im(i) {}
};
inline complex operator+(const complex& a, const complex& b) {
    return complex(a.re+b.re, a.im+b.im);
}
complex addone(const complex& arg) {
    return arg + complex(1,0);
}

Used to be compiled to:
_Z6addoneRK7complex:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
***     mov %EDX, %ECX
        fld QWORD PTR [%EDX]
        fld1
        faddp %ST(1)
***     add %ECX, 8
        fld QWORD PTR [%ECX]
        fldz
        faddp %ST(1)
***     mov %ECX, %EAX
        fxch %ST(1)
        fstp QWORD PTR [%ECX]
***     add %EAX, 8
        fstp QWORD PTR [%EAX]
        ret

Now it is compiled to:
_Z6addoneRK7complex:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        fld QWORD PTR [%ECX]
        fld1
        faddp %ST(1)
        fld QWORD PTR [%ECX + 8]
        fldz
        faddp %ST(1)
        fxch %ST(1)
        fstp QWORD PTR [%EAX]
        fstp QWORD PTR [%EAX + 8]
        ret

Other programs should see similar improvements, across the board.  Note that
in addition to reducing instruction count, this also reduces register pressure
a lot, always a good thing on X86.  :)

llvm-svn: 11819

llvm-svn: 11818