Commit Graph

5147 Commits

Author SHA1 Message Date
Evan Cheng
87bb991aa8 Teach the spiller to commute instructions in order to fold a reload. This hits 410 times on 444.namd and 122 times on 252.eon.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52266 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-13 23:58:02 +00:00
Duncan Sands
d4b9c17fb7 Disable some DAG combiner optimizations that may be
wrong for volatile loads and stores.  In fact this
is almost all of them!  There are three types of
problems: (1) it is wrong to change the width of
a volatile memory access.  These may be used to
do memory mapped i/o, in which case a load can have
an effect even if the result is not used.  Consider
loading an i32 but only using the lower 8 bits.  It
is wrong to change this into a load of an i8, because
you are no longer tickling the other three bytes.  It
is also unwise to make a load/store wider.  For
example, changing an i16 load into an i32 load is
wrong no matter how aligned things are, since the
fact of loading an additional 2 bytes can have
i/o side-effects.  (2) it is wrong to change the
number of volatile load/stores: they may be counted
by the hardware.  (3) it is wrong to change a volatile
load/store that requires one memory access into one
that requires several.  For example on x86-32, you
can store a double in one processor operation, but to
store an i64 requires two (two i32 stores).  In a
multi-threaded program you may want to bitcast an i64
to a double and store as a double because that will
occur atomically, and be indivisible to other threads.
So it would be wrong to convert the store-of-double
into a store of an i64, because this will become two
i32 stores - no longer atomic.  My policy here is
to say that the number of processor operations for
an illegal operation is undefined.  So it is alright
to change a store of an i64 (requires at least two
stores; but could be validly lowered to memcpy for
example) into a store of double (one processor op).
In short, if the new store is legal and has the same
size then I say that the transform is ok.  It would
also be possible to say that transforms are always
ok if before they were illegal, whether after they
are illegal or not, but that's more awkward to do
and I doubt it buys us anything much.
However this exposed an interesting thing - on x86-32
a store of i64 is considered legal!  That is because
operations are marked legal by default, regardless of
whether the type is legal or not.  In some ways this
is clever: before type legalization this means that
operations on illegal types are considered legal;
after type legalization there are no illegal types
so now operations are only legal if they really are.
But I consider this to be too cunning for mere mortals.
Better to do things explicitly by testing AfterLegalize.
So I have changed things so that operations with illegal
types are considered illegal - indeed they can never
map to a machine operation.  However this means that
the DAG combiner is more conservative because before
it was "accidentally" performing transforms where the
type was illegal because the operation was nonetheless
marked legal.  So in a few such places I added a check
on AfterLegalize, which I suppose was actually just
forgotten before.  This causes the DAG combiner to do
slightly more than it used to, which resulted in the X86
backend blowing up because it got a slightly surprising
node it wasn't expecting, so I tweaked it.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52254 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-13 19:07:40 +00:00
Duncan Sands
edfcf598fa Sometimes (rarely) nodes held in LegalizeTypes
maps can be deleted.  This happens when RAUW
replaces a node N with another equivalent node
E, deleting the first node.  Solve this by
adding (N, E) to ReplacedNodes, which is already
used to remap nodes to replacements.  This means
that deleted nodes are being allowed in maps,
which can be delicate: the memory may be reused
for a new node which might get confused with the
old deleted node pointer hanging around in the
maps, so detect this and flush out maps if it
occurs (ExpungeNode).  The expunging operation
is expensive, however it never occurs during
a llvm-gcc bootstrap or anywhere in the nightly
testsuite.  It occurs three times in "make check":
Alpha/illegal-element-type.ll,
PowerPC/illegal-element-type.ll and
X86/mmx-shift.ll.  If expunging proves to be too
expensive then there are other more complicated
ways of solving the problem.
In the normal case this patch adds the overhead
of a few more map lookups, which is hopefully
negligable.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52214 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-11 11:42:12 +00:00
Dan Gohman
9ea3f56d07 Teach isGAPlusOffset to respect a GlobalAddressSDNode's offset
value, which is something that apparently isn't used much.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52158 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-09 22:05:52 +00:00
Dan Gohman
f5025cfa68 CodeGen support for aggregate-value function arguments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52156 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-09 21:19:23 +00:00
Duncan Sands
893076354a Various tweaks related to apint codegen. No functionality
change for non-funky-sized integers.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52151 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-09 15:48:25 +00:00
Dan Gohman
8a6ccb5f73 Handle empty aggregate values.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52150 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-09 15:21:47 +00:00
Duncan Sands
8eab8a2798 Remove some DAG combiner assumptions about sizes
of integer types.  Fix the isMask APInt method to
actually work (hopefully) rather than crashing
because it adds apints of different bitwidths.
It looks like isShiftedMask is also broken, but
I'm leaving that one to the APInt people (it is
not used anywhere).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52142 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-09 11:32:28 +00:00
Duncan Sands
8e4eb09b1e Remove comparison methods for MVT. The main cause
of apint codegen failure is the DAG combiner doing
the wrong thing because it was comparing MVT's using
< rather than comparing the number of bits.  Removing
the < method makes this mistake impossible to commit.
Instead, add helper methods for comparing bits and use
them.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52098 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-08 20:54:56 +00:00
Dan Gohman
1d685a46bf CodeGen support for insertvalue and extractvalue, and for loads and
stores of aggregate values.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52069 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-07 02:02:36 +00:00
Owen Anderson
2d389e8ad7 Connect successors before creating the DAG node for the branch. This has
no visible functionality change, but enables a future patch where node creation
will update the CFG if it decides to create an unconditional rather than a conditional branch.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52067 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-07 00:00:23 +00:00
Evan Cheng
877333b99d Enable stack coloring by default.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52057 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-06 19:52:44 +00:00
Duncan Sands
3b3adbb745 Tighten up the abstraction slightly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52045 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-06 12:49:32 +00:00
Duncan Sands
83ec4b6711 Wrap MVT::ValueType in a struct to get type safety
and better control the abstraction.  Rename the type
to MVT.  To update out-of-tree patches, the main
thing to do is to rename MVT::ValueType to MVT, and
rewrite expressions like MVT::getSizeInBits(VT) in
the form VT.getSizeInBits().  Use VT.getSimpleVT()
to extract a MVT::SimpleValueType for use in switch
statements (you will get an assert failure if VT is
an extended value type - these shouldn't exist after
type legalization).
This results in a small speedup of codegen and no
new testsuite failures (x86-64 linux).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52044 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-06 12:08:01 +00:00
Evan Cheng
9c3c221364 Refine stack slot interval weight computation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52040 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-06 07:54:39 +00:00
Owen Anderson
417dc2f595 Remove debugging code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52016 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-05 18:43:34 +00:00
Owen Anderson
a9efb264b0 Use the newly created helper on LiveIntervals.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52013 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-05 17:22:53 +00:00
Owen Anderson
c4dc132c8a Add a helper for constructing new live ranges that ended from an instruction to the end of its MBB.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52012 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-05 17:15:43 +00:00
Evan Cheng
fb4db316d8 Fix a memcpy lowering bug. Even though the memcpy alignment is smaller than the desired alignment, the frame destination alignment may still be larger than the desired alignment. Don't change its alignment to something smaller.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51970 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 23:37:54 +00:00
Evan Cheng
75bb734f1d Oops. Should not be enabled by default.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51953 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 18:09:20 +00:00
Owen Anderson
38b425020b Correctly construct live intervals for the copies we inserted into the predecessors of a block containing a PHI.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51950 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 17:55:58 +00:00
Evan Cheng
7a315e865a Revert this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51949 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 17:21:44 +00:00
Evan Cheng
3f32d65912 Add a stack slot coloring pass. Not yet enabled.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51934 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 09:18:41 +00:00
Evan Cheng
d8a46e3a74 LowerSubregs should not clobber any analysis.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51933 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 09:17:16 +00:00
Evan Cheng
cb74266cdb Move #include to right place.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51932 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 09:16:33 +00:00
Evan Cheng
b3dd264f9d Register if-converter pass for -debug-pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51931 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 09:15:51 +00:00
Duncan Sands
a0fcc08e65 Change packed struct layout so that field sizes
are the same as in unpacked structs, only field
positions differ.  This only matters for structs
containing x86 long double or an apint; it may
cause backwards compatibility problems if someone
has bitcode containing a packed struct with a
field of one of those types.
The issue is that only 10 bytes are needed to
hold an x86 long double: the store size is 10
bytes, but the ABI size is 12 or 16 bytes (linux/
darwin) which comes from rounding the store size
up by the alignment.  Because it seemed silly not
to pack an x86 long double into 10 bytes in a
packed struct, this is what was done.  I now
think this was a mistake.  Reserving the ABI size
for an x86 long double field even in a packed
struct makes things more uniform: the ABI size is
now always used when reserving space for a type.
This means that developers are less likely to
make mistakes.  It also makes life easier for the
CBE which otherwise could not represent all LLVM
packed structs (PR2402).
Front-end people might need to adjust the way
they create LLVM structs - see following change
to llvm-gcc.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51928 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 08:21:45 +00:00
Owen Anderson
3094a4c97a We need to subtract one from this index because live ranges are open at the end.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51922 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 00:38:56 +00:00
Scott Michel
ba12f57ebe Fix spellnig error
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51917 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-03 19:13:20 +00:00
Scott Michel
eefc845b5f Find a better place to output hex constants corresponding to integers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51904 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-03 15:39:51 +00:00
Bruno Cardoso Lopes
a8c63f0fc9 Fixed bug in bad behavior in calculateFrameObjectOffsets,
the solution commited is different from the previous patch to
avoid int and unsigned comparison


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51899 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-03 08:46:59 +00:00
Evan Cheng
f73ff3aaf6 Do not run loop-aligner at -fast (e.g. -O0).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51898 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-03 06:56:08 +00:00
Scott Michel
4315eee376 Revert this patch
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51897 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-03 06:18:19 +00:00
Dan Gohman
33b625bee4 Fold adds and subtracts of zero immediately, instead of waiting
for dagcombine to do this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51886 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-02 22:27:05 +00:00
Scott Michel
83ccc0a50c Minor cosmetic patch so that the hex equivalent of a decimal
constant shows up in the assembly language output. Helps with
debugging without a HP calculator having to be handy.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51885 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-02 22:19:12 +00:00
Scott Michel
8bf61e8c2a Add necessary 64-bit support so that gcc frontend compiles (mostly). Current
issue is operand promotion for setcc/select... but looks like the fundamental
stuff is implemented for CellSPU.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51884 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-02 22:18:03 +00:00
Owen Anderson
e3abb0a858 Correctly handle removed instructions at the beginning of MBBs when renumbering.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51876 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-02 17:36:36 +00:00
Dan Gohman
2830741866 Remove an unused variable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51807 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-31 01:44:25 +00:00
Evan Cheng
7645e574d2 Fix indentation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51793 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-30 22:39:32 +00:00
Owen Anderson
dbb8137c11 The coalescer doesn't need LiveVariables now that we have register use iterators.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51790 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-30 22:37:27 +00:00
Owen Anderson
3947e4d142 Preserve the register coallescer, and update live intervals more correctly by triggering a renumbering after phi elimination.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51780 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-30 18:38:26 +00:00
Dan Gohman
493e03851d Remove an unused variable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51721 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-30 00:56:36 +00:00
Owen Anderson
7eec0c2433 Make the renumbering correct in the face of deleted instructions that have been removed from the LiveIntervals maps.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51714 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 23:01:22 +00:00
Bill Wendling
ec58498dc9 Remove <iostream>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51704 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 21:29:39 +00:00
Dan Gohman
21323f3a82 Expand small memmovs using inline code. Set the X86 threshold for expanding
memmove to a more plausible value, now that it's actually being used.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51696 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 19:42:22 +00:00
Owen Anderson
fcc6350ac9 Revert part of my last patch that I didn't intend to commit yet.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51694 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 18:35:21 +00:00
Owen Anderson
4b5b209679 Renumbering needs to account for instruction slot offsets when performing lookups in the index maps.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51691 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 18:15:49 +00:00
Evan Cheng
f26ffe987c Implement vector shift up / down and insert zero with ps{rl}lq / ps{rl}ldq.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51667 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 08:22:04 +00:00
Bill Wendling
bd0879d7d9 Implement "AsCheapAsAMove" for some obviously cheap instructions: xor and the
like.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51662 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-29 01:02:09 +00:00
Bill Wendling
8370d38ade Add a flag to indicate that an instruction is as cheap (or cheaper) than a move
instruction to execute. This can be used for transformations (like two-address
conversion) to remat an instruction instead of generating a "move"
instruction. The idea is to decrease the live ranges and register pressure and
all that jazz.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51660 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-28 22:54:52 +00:00