Commit Graph

20126 Commits

Author SHA1 Message Date
Chris Lattner
263d1e469d Fold two consequtive branches that share a common destination between them.
This implements SimplifyCFG/branch-fold.ll, and is useful on ?:/min/max heavy
code


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23410 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 18:47:20 +00:00
Chris Lattner
9a82a3ccac new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23409 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 18:43:57 +00:00
Chris Lattner
055dc102e9 simplify some logic further
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23408 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 07:23:18 +00:00
Chris Lattner
f58c1a578e pull a bunch of logic out of SimplifyCFG into a helper fn
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23407 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 06:39:30 +00:00
Chris Lattner
7c439929bc speed up Archive::isBytecodeArchive in the case when the archive doesn't have
an llvm-ranlib symtab.  This speeds up gccld -native on an almost empty .o file
from 1.63s to 0.18s.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23406 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 06:22:58 +00:00
Chris Lattner
7b5634d213 Speed up isBytecodeLPath from 20s to .01s in common cases. This makes -native
not completely painful to use.  Once we decide a directory has a bytecode
library, we know it this function returns true, no need to scan entire directories.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23405 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 06:11:24 +00:00
Chris Lattner
f394213021 1. Do not use .c_str() to keep a persistent handle on a temporary string.
2. Concatenate -lfoo and -L/bar options into a single option instead of
   passing "-L /bar" (for example) which doesn't work on Darwin.
3. Send -v output to stderr instead of stdout


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23404 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 06:05:46 +00:00
Chris Lattner
5ae7911b24 Turn (X^C1) == C2 into X == C1^C2 iff X&~C1 = 0 (and move a function)
This happens all the time on PPC for bool values, e.g. eliminating a xori
in inverted-bool-compares.ll.

This should be added to the dag combiner as well.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23403 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 00:55:52 +00:00
Chris Lattner
e8033c00d6 new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23402 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-23 00:53:06 +00:00
Chris Lattner
77f800545e Testcase for PR629
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23401 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-21 06:53:56 +00:00
Chris Lattner
3c3fe462f7 Expose the LiveInterval interfaces as public headers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23400 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-21 04:19:09 +00:00
Chris Lattner
779a6513fe move the live interval headers out of lib/CodeGen/
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23399 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-21 04:18:25 +00:00
Chris Lattner
e9967146ad Recommend what I actually test
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23398 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-21 03:56:26 +00:00
Chris Lattner
e9487f0dc8 Start threading across blocks with code in them, so long as the code does
not define a value that is used outside of it's block.  This catches many
more simplifications, e.g. 854 in 176.gcc, 137 in vpr, etc.

This implements branch-phi-thread.ll:test3.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23397 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-20 01:48:40 +00:00
Chris Lattner
055135d1c1 make this test harder: add a case where instructions are in the bb to be
threaded over


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23396 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-20 01:43:41 +00:00
Chris Lattner
2e42e36698 Implement merging of blocks with the same condition if the block has multiple
predecessors.  This implements branch-phi-thread.ll::test1


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23395 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-20 00:43:16 +00:00
Chris Lattner
ecd7b6d110 new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23394 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-20 00:41:55 +00:00
Chris Lattner
9c88d98162 Reject a case we don't handle yet
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23393 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 23:57:04 +00:00
Chris Lattner
7e1ff8d2d9 remove debugging code :-/
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23392 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 23:50:15 +00:00
Chris Lattner
eaba3a194c Implement SimplifyCFG/branch-phi-thread.ll, the most trivial case of threading
control across branches with determined outcomes.  More generality to follow.
This triggers a couple thousand times in specint.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23391 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 23:49:37 +00:00
Chris Lattner
b6e2b7e4cd new testcase.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23390 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 23:48:04 +00:00
Nate Begeman
44728a7bb6 Stub out the rest of the DAG Combiner. Just need to fill in the
select_cc bits and then wrap it in a convenience function for  use with
regular select.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23389 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 22:34:01 +00:00
Chris Lattner
cea8688ee4 Teach the local spiller to turn stack slot loads into register-register copies
when possible, avoiding the load (and avoiding the copy if the value is already
in the right register).

This patch came about when I noticed code like the following being generated:

  store R17 -> [SS1]
  ...blah...
  R4 = load [SS1]

This was causing an LSU reject on the G5.  This problem was due to the register
allocator folding spill code into a reg-reg copy (producing the load), which
prevented the spiller from being able to rewrite the load into a copy, despite
the fact that the value was already available in a register.  In the case
above, we now rip out the R4 load and replace it with a R4 = R17 copy.

This speeds up several programs on X86 (which spills a lot :) ), e.g.
smg2k from 22.39->20.60s, povray from 12.93->12.66s, 168.wupwise from
68.54->53.83s (!), 197.parser from 7.33->6.62s (!), etc.  This may have a larger
impact in some cases on the G5 (by avoiding LSU rejects), though it probably
won't trigger as often (less spilling in general).

Targets that implement folding of loads/stores into copies should implement
the isLoadFromStackSlot hook to get this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23388 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 06:56:21 +00:00
Chris Lattner
a92aab74dd Implement the isLoadFromStackSlot interface
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23387 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-19 05:23:44 +00:00
Chris Lattner
7203e158da Refactor this code a bit and make it more general. This now compiles:
struct S { unsigned int i : 6, j : 11, k : 15; } b;
void plus2 (unsigned int x) { b.j += x; }

To:

_plus2:
        lis r2, ha16(L_b$non_lazy_ptr)
        lwz r2, lo16(L_b$non_lazy_ptr)(r2)
        lwz r4, 0(r2)
        slwi r3, r3, 6
        add r3, r4, r3
        rlwimi r3, r4, 0, 26, 14
        stw r3, 0(r2)
        blr


instead of:

_plus2:
        lis r2, ha16(L_b$non_lazy_ptr)
        lwz r2, lo16(L_b$non_lazy_ptr)(r2)
        lwz r4, 0(r2)
        rlwinm r5, r4, 26, 21, 31
        add r3, r5, r3
        rlwimi r4, r3, 6, 15, 25
        stw r4, 0(r2)
        blr

by eliminating an 'and'.

I'm pretty sure this is as small as we can go :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23386 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 07:22:02 +00:00
Chris Lattner
150f12af7f Compile
struct S { unsigned int i : 6, j : 11, k : 15; } b;
void plus2 (unsigned int x) {
  b.j += x;
}

to:

plus2:
        mov %EAX, DWORD PTR [b]
        mov %ECX, %EAX
        and %ECX, 131008
        mov %EDX, DWORD PTR [%ESP + 4]
        shl %EDX, 6
        add %EDX, %ECX
        and %EDX, 131008
        and %EAX, -131009
        or %EDX, %EAX
        mov DWORD PTR [b], %EDX
        ret

instead of:

plus2:
        mov %EAX, DWORD PTR [b]
        mov %ECX, %EAX
        shr %ECX, 6
        and %ECX, 2047
        add %ECX, DWORD PTR [%ESP + 4]
        shl %ECX, 6
        and %ECX, 131008
        and %EAX, -131009
        or %ECX, %EAX
        mov DWORD PTR [b], %ECX
        ret


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23385 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 06:30:59 +00:00
Chris Lattner
0b7c0bf249 Generalize this transform, using MaskedValueIsZero, allowing us to compile:
struct S { unsigned int i : 6, j : 11, k : 15; } b;
void plus3 (unsigned int x) { b.k += x; }

To:

plus3:
        mov %EAX, DWORD PTR [%ESP + 4]
        shl %EAX, 17
        add DWORD PTR [b], %EAX
        ret

instead of:

plus3:
        mov %EAX, DWORD PTR [%ESP + 4]
        shl %EAX, 17
        mov %ECX, DWORD PTR [b]
        add %EAX, %ECX
        and %EAX, -131072
        and %ECX, 131071
        or %ECX, %EAX
        mov DWORD PTR [b], %ECX
        ret


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23384 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 06:02:59 +00:00
Chris Lattner
5aa7666ebe fix typeo
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23383 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 05:25:20 +00:00
Chris Lattner
0d947ea943 Remove unintentionally committed code
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23382 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 05:12:51 +00:00
Chris Lattner
11021cb988 implement shift.ll:test25. This compiles:
struct S { unsigned int i : 6, j : 11, k : 15; } b;
void plus3 (unsigned int x) {
  b.k += x;
}

to:

_plus3:
        lis r2, ha16(L_b$non_lazy_ptr)
        lwz r2, lo16(L_b$non_lazy_ptr)(r2)
        lwz r3, 0(r2)
        rlwinm r4, r3, 0, 0, 14
        add r4, r4, r3
        rlwimi r4, r3, 0, 15, 31
        stw r4, 0(r2)
        blr

instead of:

_plus3:
        lis r2, ha16(L_b$non_lazy_ptr)
        lwz r2, lo16(L_b$non_lazy_ptr)(r2)
        lwz r4, 0(r2)
        srwi r5, r4, 17
        add r3, r5, r3
        slwi r3, r3, 17
        rlwimi r3, r4, 0, 15, 31
        stw r3, 0(r2)
        blr


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23381 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 05:12:10 +00:00
Chris Lattner
ab4ea6e980 new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23380 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 05:10:39 +00:00
Chris Lattner
c8e7756791 Implement add.ll:test29. Codegening:
struct S { unsigned int i : 6, j : 11, k : 15; } b;
void plus1 (unsigned int x) {
  b.i += x;
}

as:
_plus1:
        lis r2, ha16(L_b$non_lazy_ptr)
        lwz r2, lo16(L_b$non_lazy_ptr)(r2)
        lwz r4, 0(r2)
        add r3, r4, r3
        rlwimi r3, r4, 0, 0, 25
        stw r3, 0(r2)
        blr

instead of:

_plus1:
        lis r2, ha16(L_b$non_lazy_ptr)
        lwz r2, lo16(L_b$non_lazy_ptr)(r2)
        lwz r4, 0(r2)
        rlwinm r5, r4, 0, 26, 31
        add r3, r5, r3
        rlwimi r3, r4, 0, 0, 25
        stw r3, 0(r2)
        blr


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23379 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 04:24:45 +00:00
Chris Lattner
dc0ffccb19 new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23378 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 04:22:59 +00:00
Chris Lattner
3255bd101d remove debug output
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23377 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 03:50:25 +00:00
Chris Lattner
e9bed7d107 Implement or.ll:test21. This teaches instcombine to be able to turn this:
struct {
   unsigned int bit0:1;
   unsigned int ubyte:31;
} sdata;

void foo() {
  sdata.ubyte++;
}

into this:

foo:
        add DWORD PTR [sdata], 2
        ret

instead of this:

foo:
        mov %EAX, DWORD PTR [sdata]
        mov %ECX, %EAX
        add %ECX, 2
        and %ECX, -2
        and %EAX, 1
        or %EAX, %ECX
        mov DWORD PTR [sdata], %EAX
        ret


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23376 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 03:42:07 +00:00
Chris Lattner
1e122d1dc2 new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23375 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-18 03:39:02 +00:00
Chris Lattner
6a78c2157a Implement hook for ppc
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23374 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-17 01:03:26 +00:00
Chris Lattner
7d3041e060 add a new callback
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23373 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-17 01:02:45 +00:00
Nate Begeman
452d7bebaa More DAG combining. Still need the branch instructions, and select_cc
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23371 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-16 00:54:12 +00:00
Chris Lattner
b9f01eb7bc Fix a minor bug, add comments
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23370 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-16 00:29:46 +00:00
Chris Lattner
a28aec1da7 teach the type inference code how to infer types for instructions and node
xforms.  Run type inference on result patterns, so we always have fully typed
results (and to catch errors in .td files).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23369 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-15 22:23:50 +00:00
Chris Lattner
ae5b350b3b put instructions into a map instead of a vector for quick lookup
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23368 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-15 21:57:35 +00:00
Chris Lattner
ae6d828de8 when parsing instructions remember information about the types taken and
returned.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23367 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-15 21:51:12 +00:00
Chris Lattner
4ac85b3e94 disable this for now
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23366 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-15 21:44:00 +00:00
Chris Lattner
abbb605d80 Start parsing "Pattern" nodes
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23365 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-15 21:42:00 +00:00
Chris Lattner
b39e4be1e1 rename a couple of methods, add structure for pattern parsing
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23364 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-15 02:38:02 +00:00
Chris Lattner
f1311843d5 Verify that xform functions only occur in logical places
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23363 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-14 23:05:13 +00:00
Chris Lattner
2175c18c4f Promote xform fns to be explicit nodes in result patterns, and clean off
predicates since they will have already matched at this point.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23362 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-14 23:01:59 +00:00
Chris Lattner
b0276200e6 start building the instruction dest pattern correctly. Change the xform
functions to preserve the Record for the xform instead of making it into a
function name.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23361 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-14 22:55:26 +00:00
Chris Lattner
7da852fbab catch unnamed inputs
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@23360 91177308-0d34-0410-b5e6-96231b3b80d8
2005-09-14 22:06:36 +00:00