dcddd64424
Fix scalarrepl to not 'miscompile' undefined code, part #2 .
...
This fixes the store case, my previous patch just fixed the load
case. rdar://5707076.
llvm-svn: 46932
2008-02-10 19:05:37 +00:00
26f75e2576
Fix PR 1995.
...
llvm-svn: 46898
2008-02-08 22:49:13 +00:00
c676a0329c
Temporarily reverting:
...
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20080128/057882.html
This is causing a miscompilation on PPC G5 and just now seeing it on iMac x86-64.
llvm-svn: 46822
2008-02-06 20:03:07 +00:00
682a7dc653
Fix a bug compiling PR1978 (perhaps not the only one though) which
...
was incorrectly simplifying "x == (gep x, 1, i)" into false, even
though i could be negative. As it turns out, all the code to
handle this already existed, we just need to disable the incorrect
optimization case and let the general case handle it.
llvm-svn: 46739
2008-02-05 04:45:32 +00:00
a82c9931f4
Be more precise when eliminating pointers bue to memcpy's. This allows more
...
stores to be deleted in some cases.
llvm-svn: 46694
2008-02-04 04:53:00 +00:00
c4a7c41869
Allow GVN to hack on memcpy's, making them open to further optimization.
...
llvm-svn: 46693
2008-02-04 02:59:58 +00:00
3b59214320
There are some cases where icmp(add) can be folded into a new icmp. Handle them.
...
llvm-svn: 46687
2008-02-03 16:33:09 +00:00
c7a4ba044b
Hack on vectors too.
...
llvm-svn: 46684
2008-02-03 08:19:11 +00:00
e6e3a7f6ea
Fold away one multiply in instcombine. This would normally be caught in
...
reassociate anyways, but they could be generated during instcombine's run.
llvm-svn: 46683
2008-02-03 07:42:09 +00:00
6b016703a9
Remove a couple more cases of "getNumUses() == 0". No need to walk the linked
...
list just to see if whether the list is empty.
llvm-svn: 46555
2008-01-30 08:01:28 +00:00
625e89c8b0
Use empty() instead of comparing size() with zero.
...
llvm-svn: 46554
2008-01-30 07:54:16 +00:00
0e3a5cb14a
Don't DCE FreeInst's. We were using those! Patch from Owen Anderson.
...
llvm-svn: 46553
2008-01-30 07:45:37 +00:00
4e4b116750
Make DSE much more aggressive by performing DCE earlier. Update a testcase to reflect this increased aggressiveness.
...
llvm-svn: 46542
2008-01-30 01:24:47 +00:00
b9e5b8fb9e
Fix a bug where scalarrepl would discard offset if type would match.
...
In practice this can only happen on code with already undefined behavior,
but this is still a good thing to handle correctly.
llvm-svn: 46539
2008-01-30 00:39:15 +00:00
70de4cb1cd
Use empty() instead of comparing size() with zero.
...
llvm-svn: 46514
2008-01-29 13:02:09 +00:00
17819d971e
eliminate additions of 0.0 when they are obviously dead. This has to be careful to
...
avoid turning -0.0 + 0.0 -> -0.0 which is incorrect.
llvm-svn: 46499
2008-01-29 06:52:45 +00:00
48d37806e2
Add support for eliminating memcpy's at the end of functions. Also fix some errors I noticed in
...
the handling of eliminating stores to byval arguments.
llvm-svn: 46494
2008-01-29 06:18:36 +00:00
67fa0521b6
Filter loops that subtract induction variables.
...
These loops are not yet handled.
Fix PR 1912.
llvm-svn: 46484
2008-01-29 02:20:41 +00:00
8ea81e8ba4
Handle some more combinations of extend and icmp. Fixes PR1940.
...
llvm-svn: 46431
2008-01-28 03:48:02 +00:00
710b441174
Fix PR1932 by disabling an xform invalid for fdiv.
...
llvm-svn: 46429
2008-01-28 00:58:18 +00:00
1b706dd680
Fix PR1938 by forcing the code that uses an undefined value to branch one
...
way or the other. Rewriting the code itself prevents subsequent analysis
passes from making contradictory conclusions about the code that could
cause an infeasible path to be made feasible.
llvm-svn: 46427
2008-01-28 00:32:30 +00:00
efb16f7057
Be more careful modifying the use_list while also iterating through it.
...
llvm-svn: 46417
2008-01-27 18:35:00 +00:00
60361a16f8
The CorrelatedExpressionElimination pass is known to be buggy. Remove it.
...
This fixes PR1769.
llvm-svn: 46408
2008-01-27 06:11:41 +00:00
fa1e7eef30
Fold fptrunc(add (fpextend x), (fpextend y)) -> add(x,y), as GCC does.
...
llvm-svn: 46406
2008-01-27 05:29:54 +00:00
5079483957
If there are no machine instructions emitted for a function, then insert
...
a "nop" instruction so that we don't have the function's label associated
with something that it's not supposed to be associated with.
llvm-svn: 46394
2008-01-26 06:51:24 +00:00
0862e3421d
If we have a function like this:
...
void bork() {
int *address = 0;
*address = 0;
}
It's compiled into LLVM code that looks like this:
define void @bork() noreturn nounwind {
entry:
unreachable
}
This is bad on some platforms (like PPC) because it will generate the label for
the function but no body. The label could end up being associated with some
non-code related stuff, like a section. This places a "trap" instruction if the
SimplifyCFG pass removed all code from the function leaving only one
"unreachable" instruction.
llvm-svn: 46387
2008-01-26 01:43:44 +00:00
6af19fd1e2
DeadStoreElimination can treat byval parameters as if there were alloca's for the purpose of removing end-of-function stores.
...
llvm-svn: 46351
2008-01-25 10:10:33 +00:00
f069264164
Enable the fix I just checked in, silly me.
...
llvm-svn: 46247
2008-01-22 05:42:02 +00:00
78712e5b59
Multiply can be evaluated in a different type, so long as the target type has
...
a smaller bitwidth.
llvm-svn: 46244
2008-01-22 05:08:48 +00:00
afa84da4e0
Make sure the caller doesn't use freed memory.
...
Fixes PR1935.
llvm-svn: 46203
2008-01-20 16:51:46 +00:00
fe3bef091f
Initializing an unsigned with ~0UL causes the compiler
...
to complain on x86-64 (gcc 4.1). Use ~0U instead.
llvm-svn: 46197
2008-01-20 10:49:23 +00:00
b5ca2e9fcb
I noticed that the trampoline straightening transformation could
...
drop attributes on varargs call arguments. Also, it could generate
invalid IR if the transformed call already had the 'nest' attribute
somewhere (this can never happen for code coming from llvm-gcc,
but it's a theoretical possibility). Fix both problems.
llvm-svn: 45973
2008-01-14 19:52:09 +00:00
92bd785323
Turn a memcpy from a double* into a load/store of double instead of
...
a load/store of i64. The later prevents promotion/scalarrepl of the
source and dest in many cases.
This fixes the 300% performance regression of the byval stuff on
stepanov_v1p2.
llvm-svn: 45945
2008-01-14 00:28:35 +00:00
57974c8d51
factor memcpy/memmove simplification out to its own SimplifyMemTransfer
...
method, no functionality change.
llvm-svn: 45944
2008-01-13 23:50:23 +00:00
8c5cdddfb9
simplify some code. If we can infer alignment for source and dest that are
...
greater than memcpy alignment, and if we lower to load/store, use the best
alignment info we have.
llvm-svn: 45943
2008-01-13 22:30:28 +00:00
5a86612d3f
simplify some code by adding a InsertBitCastBefore method,
...
make memmove->memcpy conversion a bit simpler.
llvm-svn: 45942
2008-01-13 22:23:22 +00:00
5bc253c8f2
Fix PR1907, a nasty miscompilation because instcombine didn't
...
realize that ne & sgt was a signed comparison (it was only
looking at whether the left compare was signed).
llvm-svn: 45937
2008-01-13 20:59:02 +00:00
781f6549db
When turning a call to a bitcast function into a direct call,
...
if this becomes a varargs call then deal correctly with any
parameter attributes on the newly vararg call arguments.
llvm-svn: 45931
2008-01-13 08:02:44 +00:00
2940c5c56d
Implement PR1795, an instcombine hack for forming GEPs with integer pointer arithmetic.
...
llvm-svn: 45745
2008-01-08 07:23:51 +00:00
b18c30acec
Small cleanup for handling of type/parameter attribute
...
incompatibility.
llvm-svn: 45704
2008-01-07 17:16:06 +00:00
efb08802ec
Deleting an empty file. Thanks, /usr/bin/patch!
...
llvm-svn: 45675
2008-01-07 02:29:04 +00:00
6047b6e140
With this patch, the LowerGC transformation becomes the
...
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 01:30:53 +00:00
404eb05247
The transform that tries to turn calls to bitcast functions into
...
direct calls bails out unless caller and callee have essentially
equivalent parameter attributes. This is illogical - the callee's
attributes should be of no relevance here. Rework the logic, which
incidentally fixes a crash when removed arguments have attributes.
llvm-svn: 45658
2008-01-06 18:27:01 +00:00
55e5090fe8
When transforming a call to a bitcast function into
...
a direct call with cast parameters and cast return
value (if any), instcombine was prepared to cast any
non-void return value into any other, whether castable
or not. Add a new predicate for testing whether casting
is valid, and check it both for the return value and
(as a cleanup) for the parameters.
llvm-svn: 45657
2008-01-06 10:12:28 +00:00
e666bc272d
remove a couple more unsafe xforms in the face of overflow.
...
llvm-svn: 45613
2008-01-05 01:22:42 +00:00
db026d703b
remove the (x-y) < 0 comparison xform, it miscompiles
...
things that are not equality comparisons, for example:
(2147479553+4096)-2147479553 < 0 != (2147479553+4096) < 2147479553
llvm-svn: 45612
2008-01-05 01:18:20 +00:00
30e434569e
fix typo
...
llvm-svn: 45594
2008-01-04 20:02:18 +00:00
f3ebc3f3d2
Remove attribution from file headers, per discussion on llvmdev.
...
llvm-svn: 45418
2007-12-29 20:36:04 +00:00
a087a8d2ce
remove attribution from lib Makefiles.
...
llvm-svn: 45415
2007-12-29 20:09:26 +00:00
b053b80b79
Disable null pointer folding transforms for non-generic address spaces. This should probably be a target-specific predicate based on address space. That way for targets where this isn't applicable the predicate can be optimized away.
...
llvm-svn: 45403
2007-12-29 07:56:53 +00:00
7363914ef7
Repair a transform that Chris noticed a bug in. Thanks to Nicholas for pointing out my stupid mistakes when writing this patch. :-)
...
llvm-svn: 45384
2007-12-28 07:42:12 +00:00
5179819beb
disable this instcombine xform, it miscompiles:
...
define i32 @main() {
entry:
%z = alloca i32 ; <i32*> [#uses=2]
store i32 0, i32* %z
%tmp = load i32* %z ; <i32> [#uses=1]
%sub = sub i32 %tmp, 1 ; <i32> [#uses=1]
%cmp = icmp ult i32 %sub, 0 ; <i1> [#uses=1]
%retval = select i1 %cmp, i32 1, i32 0 ; <i32> [#uses=1]
ret i32 %retval
}
into ret 1, instead of ret 0.
Christopher, please investigate.
llvm-svn: 45383
2007-12-28 06:24:31 +00:00
ef1bbfc762
Don't break critical edges for single-bb loops, this helps with PR1877, though
...
it is only a partial fix. This change is noise for most programs, but
speeds up Shootout-C++/matrix by 20%, Ptrdist/ks by 24%, smg2000 by 8%,
hexxagon by 9%, bzip2 by 9% (not sure I trust this), ackerman by 13%, etc.
OTOH, it slows down Shootout/fib2 by 40% (I'll update PR1877 with this info).
llvm-svn: 45354
2007-12-25 19:06:45 +00:00
62a806d565
add a -backedge-hack llc-beta option to codegenprepare.
...
When specified, don't split backedges of single-bb loops.
This helps address PR1877
llvm-svn: 45344
2007-12-24 19:32:55 +00:00
74b2ab59fd
implement InstCombine/shift-trunc-shift.ll. This allows
...
us to compile:
#include <math.h>
int t1(double d) { return signbit(d); }
into:
_t1:
movd %xmm0, %rax
shrq $63, %rax
ret
instead of:
_t1:
movd %xmm0, %rax
shrq $32, %rax
shrl $31, %eax
ret
on x86-64.
llvm-svn: 45311
2007-12-22 09:07:47 +00:00
7d82bc46b8
Implement review feedback, including additional transforms
...
(icmp slt (sub A B) 1) -> (icmp sle A B)
icmp sgt (sub A B) -1) -> (icmp sge A B)
and add testcase.
llvm-svn: 45256
2007-12-20 07:21:11 +00:00
26ee54eb05
Clean up previous patch: PHI uses should not prevent iv reuse if all other uses are addresses. This trades a constant multiply for one fewer iv.
...
llvm-svn: 45251
2007-12-20 02:20:53 +00:00
16a51da0e2
simplify this code with the new m_Zero() pattern. Make sure the select only
...
has a single use, and generalize it to not require N to be a constant.
llvm-svn: 45250
2007-12-20 01:56:58 +00:00
e2a8ba7fec
Allow iv reuse if the user is a PHI node which is in turn used as addresses.
...
llvm-svn: 45230
2007-12-19 23:33:23 +00:00
aa31b92508
When inlining through an 'nounwind' call, mark inlined
...
calls 'nounwind'. It is important for correct C++
exception handling that nounwind markings do not get
lost, so this transformation is actually needed for
correctness.
llvm-svn: 45218
2007-12-19 21:13:37 +00:00
f00ac6dd93
Fold subtracts into integer compares vs. zero. This improves generate code for this case on X86
...
from
_foo:
movl $99, %ecx
movl 4(%esp), %eax
subl %eax, %ecx
xorl %edx, %edx
testl %ecx, %ecx
cmovs %edx, %eax
ret
to
_foo:
xorl %ecx, %ecx
movl 4(%esp), %eax
cmpl $99, %eax
cmovg %ecx, %eax
ret
llvm-svn: 45173
2007-12-18 21:32:20 +00:00
b7016c53d1
Fix comments
...
llvm-svn: 45170
2007-12-18 20:33:11 +00:00
74dbad9216
Remove an orthogonal transformation of the selection condition from my most recent submission.
...
llvm-svn: 45169
2007-12-18 20:30:28 +00:00
3353ed09ac
Rename isNoReturn to doesNotReturn, and isNoUnwind to
...
doesNotThrow.
llvm-svn: 45160
2007-12-18 09:59:50 +00:00
30291f4a30
Fix typos.
...
llvm-svn: 45159
2007-12-18 09:45:40 +00:00
8b09a464b4
Fold certain additions through selects (and their compares) so as to eliminate subtractions. This code is often produced by the SMAX expansion in SCEV.
...
This implements test/Transforms/InstCombine/2007-12-18-AddSelCmpSub.ll
llvm-svn: 45158
2007-12-18 09:34:41 +00:00
c4a80fb574
Get rid of annoying spaces.
...
llvm-svn: 45100
2007-12-17 17:40:29 +00:00
2a5967b36a
Fix GLIBCXX_DEBUG errors. Erase invalidates std::vector iterators
...
passed the erased element.
llvm-svn: 45099
2007-12-17 17:39:51 +00:00
edf0788758
Change the PointerType api for creating pointer types. The old functionality of PointerType::get() has become PointerType::getUnqual(), which returns a pointer in the generic address space. The new prototype of PointerType::get() requires both a type and an address space.
...
llvm-svn: 45082
2007-12-17 01:12:55 +00:00
8e4847ee95
Make instcombine promote inline asm calls to 'nounwind'
...
calls. Remove special casing of inline asm from the
inliner. There is a potential problem: the verifier
rejects invokes of inline asm (not sure why). If an
asm call is not marked "nounwind" in some .ll, and
instcombine is not run, but the inliner is run, then
an illegal module will be created. This is bad but
I'm not sure what the best approach is. I'm tempted
to remove the check in the verifier...
llvm-svn: 45073
2007-12-16 15:51:49 +00:00
2011df4e39
Fix typo.
...
llvm-svn: 44997
2007-12-13 07:50:36 +00:00
37c36ed79a
Be extra careful with extension use optimation. Now turned on by default.
...
llvm-svn: 44981
2007-12-13 03:32:53 +00:00
309e5a723b
1. "Upgrage" comments.
...
2. Using zero-extended value of Scale and unsigned division is safe provided
that Scale doesn't have the sign bit set.
Previously these 2 instructions:
%p = bitcast [100 x {i8,i8,i8}]* %x to i8*
%q = getelementptr i8* %p, i32 -4
were combined into:
%q = getelementptr [100 x { i8, i8, i8 }]* %x, i32 0,
i32 1431655764, i32 0
what was incorrect.
llvm-svn: 44936
2007-12-12 15:21:32 +00:00
63d33cfd2b
Don't muck with phi nodes; bug fixes.
...
llvm-svn: 44905
2007-12-12 02:53:41 +00:00
7bc8942532
Bug fix. Only safe to perform extension uses optimization if the source of extension is also defined in the same BB as the extension.
...
llvm-svn: 44896
2007-12-12 00:51:06 +00:00
9f76be61d1
Make PruneEH update the nounwind/noreturn attributes
...
on functions as it calculates them.
llvm-svn: 44802
2007-12-10 19:09:40 +00:00
086b2c4537
Fix several cache coherence bugs in MemDep/GVN that were found. Also add some (disabled) debugging code
...
to make such problems easier to diagnose in the future, written by Duncan Sands.
llvm-svn: 44695
2007-12-08 01:37:09 +00:00
d2bbbabbfb
simplify some code.
...
llvm-svn: 44655
2007-12-06 06:25:04 +00:00
0ccb663cca
move some ashr-specific code out of commonShiftTransforms into visitAShr.
...
llvm-svn: 44650
2007-12-06 01:59:46 +00:00
d3d8017bcb
If both result of the {s|z}xt and its source are live out, rewrite all uses of the source with result of extension.
...
llvm-svn: 44643
2007-12-05 23:58:20 +00:00
38ef3a8ec7
Rather than having special rules like "intrinsics cannot
...
throw exceptions", just mark intrinsics with the nounwind
attribute. Likewise, mark intrinsics as readnone/readonly
and get rid of special aliasing logic (which didn't use
anything more than this anyway).
llvm-svn: 44544
2007-12-03 20:06:50 +00:00
19970e62e7
update file comment.
...
llvm-svn: 44543
2007-12-03 19:43:18 +00:00
bd75910fa7
If ExitValue operand is also defined in Loop header then
...
insert new ExitValue after this operand definition.
This fixes PR1828.
llvm-svn: 44539
2007-12-03 19:17:21 +00:00
68b6f50938
Integrate the readonly/readnone logic more deeply
...
into alias analysis. This meant updating the API
which now has versions of the getModRefBehavior,
doesNotAccessMemory and onlyReadsMemory methods
which take a callsite parameter. These should be
used unless the callsite is not known, since in
general they can do a better job than the versions
that take a function. Also, users should no longer
call the version of getModRefBehavior that takes
both a function and a callsite. To reduce the
chance of misuse it is now protected.
llvm-svn: 44487
2007-12-01 07:51:45 +00:00
ffa55112e0
Fix a miscompilation in spiff on PPC.
...
llvm-svn: 44437
2007-11-29 18:02:22 +00:00
5208d1ab4a
Add some convenience methods for querying attributes, and
...
use them.
llvm-svn: 44403
2007-11-28 17:07:01 +00:00
ad0ea2d430
Fix PR1146: parameter attributes are longer part of
...
the function type, instead they belong to functions
and function calls. This is an updated and slightly
corrected version of Reid Spencer's original patch.
The only known problem is that auto-upgrading of
bitcode files doesn't seem to work properly (see
test/Bitcode/AutoUpgradeIntrinsics.ll). Hopefully
a bitcode guru (who might that be? :) ) will fix it.
llvm-svn: 44359
2007-11-27 13:23:08 +00:00
b0dd27ee91
Make LoopInfoBase more generic, in preparation for having MachineLoopInfo. This involves a small interface change.
...
llvm-svn: 44348
2007-11-27 03:43:35 +00:00
9f0b6e9d46
Fix another bug that was causing siod to fail.
...
llvm-svn: 44325
2007-11-26 07:17:19 +00:00
4f833c7610
Allow GVN to eliminate read-only function calls when it can detect that they are redundant.
...
llvm-svn: 44323
2007-11-26 02:26:36 +00:00
2f76e373ae
Remove another leak. Due to some reason AliasSetTracker didn't had any dtor...
...
llvm-svn: 44320
2007-11-25 23:52:02 +00:00
c00e8adfe0
Implement PR1822
...
llvm-svn: 44318
2007-11-25 21:27:53 +00:00
185eeac0f8
Fix PR1816. If a bitcast of a function only exists because of a
...
trivial difference in function attributes, allow calls to it to
be converted to direct calls. Based on a patch by Török Edwin.
While there, move the various lists of mutually incompatible
parameters etc out of the verifier and into ParameterAttributes.h.
llvm-svn: 44315
2007-11-25 14:10:56 +00:00
0cf083815a
add a comment.
...
llvm-svn: 44293
2007-11-23 22:35:18 +00:00
b23192f483
Remove some logic I thoughtlessly copied over
...
from the old ADCE implementation (there it was
correct because the transform was being done
for read-only functions).
llvm-svn: 44287
2007-11-23 09:10:17 +00:00
1985d96dc9
Fix PR1817.
...
llvm-svn: 44284
2007-11-22 23:47:13 +00:00
a915b538d3
Turn invokes of nounwind functions into ordinary calls.
...
llvm-svn: 44280
2007-11-22 22:24:59 +00:00
1c97d752df
Readonly/readnone functions are allowed to throw
...
exceptions, so don't turn invokes of them into
calls.
llvm-svn: 44278
2007-11-22 21:40:06 +00:00
1d6f196245
typo
...
llvm-svn: 44262
2007-11-21 05:21:54 +00:00
d32081002e
Add explicit keywords.
...
llvm-svn: 44234
2007-11-19 15:30:20 +00:00
7d97662467
Remove indeterminism from a loop. We think this will
...
fix an occasional nonrepeatable bootstrap failure we've
been seeing on Darwin.
llvm-svn: 44202
2007-11-17 02:48:01 +00:00
c53b18362a
Fix PR1800 by correcting mistaken logic.
...
llvm-svn: 44188
2007-11-16 06:04:17 +00:00
a77e74edba
Implement PR1796 and Transforms/SimplifyCFG/noreturn-call.ll
...
by inserting unreachable after no-return calls.
llvm-svn: 44099
2007-11-14 06:19:25 +00:00
61ce4dff7a
Implement PR1786 by iterating between dead cycle elimination
...
and simplifycfg in the rare cases when it is needed.
llvm-svn: 44044
2007-11-13 07:32:38 +00:00
19ca5c7021
Better check
...
llvm-svn: 43897
2007-11-08 18:45:15 +00:00
8cf11aa330
Fix PR1780
...
llvm-svn: 43893
2007-11-08 17:39:28 +00:00
8201a9bca0
fix const correctness, BB is const, so its predecessors are too
...
llvm-svn: 43780
2007-11-06 22:07:40 +00:00
cbad11e027
don't put erase or query for non-allocainst pointers in an set of allocainsts*'s
...
llvm-svn: 43779
2007-11-06 22:07:22 +00:00
d8515f8e80
Implement PR1777 by detecting dependent phis that
...
all compute the same value.
llvm-svn: 43777
2007-11-06 21:52:06 +00:00
f042e862fd
At the point of calculating the shift amount, the
...
type of SV has changed from what it originally was.
However we need the store width of the original.
llvm-svn: 43775
2007-11-06 20:39:11 +00:00
362709dff1
wrap long lines
...
llvm-svn: 43745
2007-11-06 01:15:27 +00:00
4decbc5002
Fix an abort in instcombine when folding creates a vector rem instruction.
...
llvm-svn: 43743
2007-11-05 23:16:33 +00:00
b98d2050a2
If a value is incoming from outside the loop then the value does not need remapping and the value is never tracked through LastValueMap.
...
llvm-svn: 43728
2007-11-05 19:32:30 +00:00
f07fa24289
If a long double is in a packed struct, it may be
...
that there is no padding.
llvm-svn: 43691
2007-11-05 00:35:07 +00:00
d568767ecb
Finishing initial docs for all transformations in Passes.html.
...
Also cleaned up some comments in source files.
llvm-svn: 43674
2007-11-04 16:15:04 +00:00
399d97987b
Change uses of getTypeSize to getABITypeSize, getTypeStoreSize
...
or getTypeSizeInBits as appropriate in ScalarReplAggregates.
The right change to make was not always obvious, so it would
be good to have an sroa guru review this. While there I noticed
some bugs, and fixed them: (1) arrays of x86 long double have
holes due to alignment padding, but this wasn't being spotted
by HasStructPadding (renamed to HasPadding). The same goes
for arrays of oddly sized ints. Vectors also suffer from this,
in fact the problem for vectors is much worse because basic
vector assumptions seem to be broken by vectors of type with
alignment padding. I didn't try to fix any of these vector
problems. (2) The code for extracting smaller integers from
larger ones (in the "int union" case) was wrong on big-endian
machines for integers with size not a multiple of 8, like i1.
Probably this is impossible to hit via llvm-gcc, but I fixed
it anyway while there and added a testcase. I also got rid of
some trailing whitespace and changed a function name which
had an obvious typo in it.
llvm-svn: 43672
2007-11-04 14:43:57 +00:00
ce8c62665a
Disable tail duplication of call instructions. The cost
...
metric is way off for these in general, and this works around
buggy code like that in PR1764. we'll see if there is a big
performance impact of this. If so, I'll revert it tomorrow.
llvm-svn: 43668
2007-11-04 06:37:55 +00:00
44b8721de8
Executive summary: getTypeSize -> getTypeStoreSize / getABITypeSize.
...
The meaning of getTypeSize was not clear - clarifying it is important
now that we have x86 long double and arbitrary precision integers.
The issue with long double is that it requires 80 bits, and this is
not a multiple of its alignment. This gives a primitive type for
which getTypeSize differed from getABITypeSize. For arbitrary precision
integers it is even worse: there is the minimum number of bits needed to
hold the type (eg: 36 for an i36), the maximum number of bits that will
be overwriten when storing the type (40 bits for i36) and the ABI size
(i.e. the storage size rounded up to a multiple of the alignment; 64 bits
for i36).
This patch removes getTypeSize (not really - it is still there but
deprecated to allow for a gradual transition). Instead there is:
(1) getTypeSizeInBits - a number of bits that suffices to hold all
values of the type. For a primitive type, this is the minimum number
of bits. For an i36 this is 36 bits. For x86 long double it is 80.
This corresponds to gcc's TYPE_PRECISION.
(2) getTypeStoreSizeInBits - the maximum number of bits that is
written when storing the type (or read when reading it). For an
i36 this is 40 bits, for an x86 long double it is 80 bits. This
is the size alias analysis is interested in (getTypeStoreSize
returns the number of bytes). There doesn't seem to be anything
corresponding to this in gcc.
(3) getABITypeSizeInBits - this is getTypeStoreSizeInBits rounded
up to a multiple of the alignment. For an i36 this is 64, for an
x86 long double this is 96 or 128 depending on the OS. This is the
spacing between consecutive elements when you form an array out of
this type (getABITypeSize returns the number of bytes). This is
TYPE_SIZE in gcc.
Since successive elements in a SequentialType (arrays, pointers
and vectors) need to be aligned, the spacing between them will be
given by getABITypeSize. This means that the size of an array
is the length times the getABITypeSize. It also means that GEP
computations need to use getABITypeSize when computing offsets.
Furthermore, if an alloca allocates several elements at once then
these too need to be aligned, so the size of the alloca has to be
the number of elements multiplied by getABITypeSize. Logically
speaking this doesn't have to be the case when allocating just
one element, but it is simpler to also use getABITypeSize in this
case. So alloca's and mallocs should use getABITypeSize. Finally,
since gcc's only notion of size is that given by getABITypeSize, if
you want to output assembler etc the same as gcc then getABITypeSize
is the size you want.
Since a store will overwrite no more than getTypeStoreSize bytes,
and a read will read no more than that many bytes, this is the
notion of size appropriate for alias analysis calculations.
In this patch I have corrected all type size uses except some of
those in ScalarReplAggregates, lib/Codegen, lib/Target (the hard
cases). I will get around to auditing these too at some point,
but I could do with some help.
Finally, I made one change which I think wise but others might
consider pointless and suboptimal: in an unpacked struct the
amount of space allocated for a field is now given by the ABI
size rather than getTypeStoreSize. I did this because every
other place that reserves memory for a type (eg: alloca) now
uses getABITypeSize, and I didn't want to make an exception
for unpacked structs, i.e. I did it to make things more uniform.
This only effects structs containing long doubles and arbitrary
precision integers. If someone wants to pack these types more
tightly they can always use a packed struct.
llvm-svn: 43620
2007-11-01 20:53:16 +00:00
2ed651ace7
Fix test/Transforms/DeadStoreElimination/PartialStore.ll, which had been
...
silently failing because of an incorrect run line for some time.
llvm-svn: 43605
2007-11-01 05:29:16 +00:00
74709473ed
Fix InstCombine/2007-10-31-RangeCrash.ll
...
llvm-svn: 43596
2007-11-01 02:18:41 +00:00
54048ec9e0
Fix a typo in a comment.
...
llvm-svn: 43553
2007-10-31 14:35:39 +00:00
240c1adade
At end of LSR, replace uses of now constant (as result of SplitCriticalEdge) PHI node with the constant value.
...
llvm-svn: 43533
2007-10-30 23:45:15 +00:00
c2dbfee43f
It's not safe to tell SplitCriticalEdge to merge identical edges. It may delete the phi instruction that's being processed.
...
llvm-svn: 43524
2007-10-30 22:27:26 +00:00
b024c4c81d
- Bug fixes.
...
- Allow icmp rewrite using an iv / stride of a smaller integer type.
llvm-svn: 43480
2007-10-29 22:07:18 +00:00
2aec186dd0
Don't bitcast from pointer-to-vector to pointer-to-array when
...
lowering load and store instructions.
llvm-svn: 43468
2007-10-29 20:34:35 +00:00
3bcd5fe9f1
Use an array instead of a fixed-length std::vector.
...
llvm-svn: 43467
2007-10-29 20:24:00 +00:00
d9911e21df
Do a real assert if there is an unhandled vector instruction instead
...
of just printing to cerr.
llvm-svn: 43466
2007-10-29 20:14:29 +00:00
7414e21ec0
Update a comment to reflect the current code.
...
llvm-svn: 43463
2007-10-29 19:32:39 +00:00
f5feb01056
Remove an unused function argument.
...
llvm-svn: 43462
2007-10-29 19:31:25 +00:00
50d42224d0
Fix a typo in a comment.
...
llvm-svn: 43461
2007-10-29 19:26:14 +00:00
8e8adada83
Avoid calling ValidStride when not all uses are addresses.
...
llvm-svn: 43460
2007-10-29 19:23:53 +00:00
9dbe99dcd6
A number of LSR fixes:
...
- ChangeCompareStride only reuse stride that is larger than current stride. It
will let the general reuse mechanism to try to reuse a smaller stride.
- Watch out for multiplication overflow in ChangeCompareStride.
- Replace std::set with SmallPtrSet.
llvm-svn: 43408
2007-10-26 23:08:19 +00:00
d78a3e5555
Fix a crash. Make sure TLI is not null.
...
llvm-svn: 43384
2007-10-26 17:24:46 +00:00
7f3d02471d
Loosen up iv reuse to allow reuse of the same stride but a larger type when truncating from the larger type to smaller type is free.
...
e.g.
Turns this loop:
LBB1_1: # entry.bb_crit_edge
xorl %ecx, %ecx
xorw %dx, %dx
movw %dx, %si
LBB1_2: # bb
movl L_X$non_lazy_ptr, %edi
movw %si, (%edi)
movl L_Y$non_lazy_ptr, %edi
movw %dx, (%edi)
addw $4, %dx
incw %si
incl %ecx
cmpl %eax, %ecx
jne LBB1_2 # bb
into
LBB1_1: # entry.bb_crit_edge
xorl %ecx, %ecx
xorw %dx, %dx
LBB1_2: # bb
movl L_X$non_lazy_ptr, %esi
movw %cx, (%esi)
movl L_Y$non_lazy_ptr, %esi
movw %dx, (%esi)
addw $4, %dx
incl %ecx
cmpl %eax, %ecx
jne LBB1_2 # bb
llvm-svn: 43375
2007-10-26 01:56:11 +00:00
29e29e63bd
Do not rewrite compare instruction using iv of a different stride if the new
...
stride may be rewritten using the stride of the compare instruction.
llvm-svn: 43367
2007-10-25 22:45:20 +00:00
5a38108374
Remove code that's commented out.
...
llvm-svn: 43356
2007-10-25 18:38:24 +00:00
133694db06
If a loop termination compare instruction is the only use of its stride,
...
and the compaison is against a constant value, try eliminate the stride
by moving the compare instruction to another stride and change its
constant operand accordingly. e.g.
loop:
...
v1 = v1 + 3
v2 = v2 + 1
if (v2 < 10) goto loop
=>
loop:
...
v1 = v1 + 3
if (v1 < 30) goto loop
llvm-svn: 43336
2007-10-25 09:11:16 +00:00
55b8302dfe
simplify some code by using the new isNaN predicate
...
llvm-svn: 43305
2007-10-24 18:54:45 +00:00
c62877e9da
Implement a couple of foldings for ordered and unordered comparisons,
...
implementing cases related to PR1738.
llvm-svn: 43289
2007-10-24 05:38:08 +00:00
e0c3d9f338
Strength reduction improvements.
...
- Avoid attempting stride-reuse in the case that there are users that
aren't addresses. In that case, there will be places where the
multiplications won't be folded away, so it's better to try to
strength-reduce them.
- Several SSE intrinsics have operands that strength-reduction can
treat as addresses. The previous item makes this more visible, as
any non-address use of an IV can inhibit stride-reuse.
- Make ValidStride aware of whether there's likely to be a base
register in the address computation. This prevents it from thinking
that things like stride 9 are valid on x86 when the base register is
already occupied.
Also, XFAIL the 2007-08-10-LEA16Use32.ll test; the new logic to avoid
stride-reuse elimintes the LEA in the loop, so the test is no longer
testing what it was intended to test.
llvm-svn: 43231
2007-10-22 20:40:42 +00:00
a37eaf2bf9
Move the SCEV object factors from being static members of the individual
...
SCEV subclasses to being non-static member functions of the ScalarEvolution
class.
llvm-svn: 43224
2007-10-22 18:31:58 +00:00
7499a3b092
Reg2Mem cleanup and optimizations:
...
- enable phi instructions demotion to stack
- create alloca instructions in the entry block
llvm-svn: 43208
2007-10-21 23:05:16 +00:00
df49cf52e2
Try again.
...
Instead of loading small global string from memory, use
integer constant.
llvm-svn: 43148
2007-10-18 19:52:32 +00:00
09b83ba6f1
Allow GVN to eliminate redundant calls to functions without side effects.
...
llvm-svn: 43147
2007-10-18 19:39:33 +00:00
cdcc1d0444
Reverting r43070 for now. It's causing llc test failures.
...
llvm-svn: 43103
2007-10-17 23:51:13 +00:00
91ff13edcc
Apply "Instead of loading small c string constant, use integer constant directly" transformation while processing load instruction.
...
llvm-svn: 43070
2007-10-17 07:24:40 +00:00
8d818f5e80
Use immediate stores.
...
llvm-svn: 43055
2007-10-16 23:44:18 +00:00
bff4aea328
Achieve same result but use fewer lines of code.
...
llvm-svn: 42985
2007-10-15 15:31:35 +00:00
371e6ca690
Dest type is always i8 *. This allows some simplification.
...
Do not filter memmove.
llvm-svn: 42930
2007-10-12 20:10:21 +00:00
ad618f66e6
Fix a bug in my patch last night that broke InstCombine/2007-10-12-Crash.ll
...
llvm-svn: 42920
2007-10-12 18:05:47 +00:00
5d8f7e0cc7
eliminate warning
...
llvm-svn: 42892
2007-10-12 07:44:54 +00:00
d8675e4915
Fix some 80 column violations.
...
Fix DecomposeSimpleLinearExpr to handle simple constants better.
Don't nuke gep(bitcast(allocation)) if the bitcast(allocation) will
fold the allocation. This fixes PR1728 and Instcombine/malloc3.ll
llvm-svn: 42891
2007-10-12 05:30:59 +00:00
899cc56612
Lower memcpy if it makes sense.
...
llvm-svn: 42864
2007-10-11 17:21:57 +00:00
2af23f976b
Do not walk invalid iterator.
...
llvm-svn: 42812
2007-10-09 21:31:36 +00:00
a69f987b66
Fix bug in updating dominance frontier after loop
...
unswitch when frontier includes basic blocks that
are not inside loop.
llvm-svn: 42654
2007-10-05 22:29:34 +00:00
3574759d85
Fix 80 col violation.
...
llvm-svn: 42591
2007-10-03 21:17:43 +00:00
e192e32577
Refactor code in a separate method.
...
llvm-svn: 42590
2007-10-03 21:16:08 +00:00
c731c97fac
Use empty() member functions when that's what's being tested for instead
...
of comparing begin() and end().
llvm-svn: 42585
2007-10-03 19:26:29 +00:00
9d559cfff5
Tone down an overzealous optimization.
...
llvm-svn: 42582
2007-10-03 17:45:27 +00:00
b6c05b1f90
Fix stride computations for long double arrays.
...
llvm-svn: 42508
2007-10-01 23:08:35 +00:00
2a60ff1aeb
Relax unsafe use check. If there is one unconditional use inside the loop then it is safe to promote value even if there is another conditional use inside the loop.
...
llvm-svn: 42493
2007-10-01 18:12:58 +00:00
1d1d0e7735
Don't do SRA for unions with long double fields.
...
Fixes a SWB crash.
llvm-svn: 42422
2007-09-28 00:21:38 +00:00
7bba386f72
Handle multiple induction variables.
...
This fixes PR714.
llvm-svn: 42309
2007-09-25 18:24:48 +00:00
440d13b55b
Do not reserve DOM check for GetElementPtrInst.
...
llvm-svn: 42306
2007-09-25 17:55:50 +00:00
5e1651d270
doh..
...
llvm-svn: 42300
2007-09-25 17:43:08 +00:00
87d7e8ebcb
Add transformation to update loop interation space. Now,
...
for (i=A; i<N; i++) {
if (i < X && i > Y)
do_something();
}
is transformed into
U=min(N,X); L=max(A,Y);
for (i=L;i<U;i++)
do_somethihg();
llvm-svn: 42299
2007-09-25 17:31:19 +00:00
9e30e1a3be
Do not promote null values because it may be unsafe to do so.
...
llvm-svn: 42270
2007-09-24 20:02:42 +00:00
75470c3bf1
explicit keywords.
...
llvm-svn: 42262
2007-09-24 15:48:49 +00:00
361e52f39c
Fix PR1692
...
llvm-svn: 42209
2007-09-21 21:18:19 +00:00
46da2a6262
Add partial caching of non-local memory dependence queries. This provides a modest
...
speedup for GVN.
llvm-svn: 42185
2007-09-21 03:53:52 +00:00
83cc3f8f51
Update aux. info associated with an instruction before erasing instruction.
...
llvm-svn: 42180
2007-09-20 23:45:50 +00:00
6117a3b696
Don't increment invalid iterator.
...
llvm-svn: 42178
2007-09-20 23:01:50 +00:00
eae7e7d00b
Fix optimization. %x = sub %x, %y does not imply that %y is zero.
...
llvm-svn: 42157
2007-09-20 00:48:36 +00:00
464276f831
Avoid unsafe promotion.
...
llvm-svn: 42149
2007-09-19 20:18:51 +00:00
d31649bc59
Improve comment.
...
llvm-svn: 42132
2007-09-19 10:25:38 +00:00
56df7dec2b
A global variable with external weak linkage can be null, while
...
an alias could alias such a global variable.
llvm-svn: 42130
2007-09-19 10:10:31 +00:00
69a55a38ed
Relax loop ExitCondition predicate restriction.
...
llvm-svn: 42122
2007-09-19 00:28:47 +00:00
455a53b7db
Filter loops where split condition's false branch is not empty. For example
...
for (int i = 0; i < N; ++i) {
if (i == somevalue)
dosomething();
else
dosomethingelse();
}
llvm-svn: 42121
2007-09-19 00:15:16 +00:00
4c238c451f
Bail out early, before modifying anything.
...
llvm-svn: 42120
2007-09-19 00:11:01 +00:00
31f2c8592c
Work is incomplete. Loop is not modified at all right now.
...
llvm-svn: 42119
2007-09-19 00:08:13 +00:00
fcda998ab2
Fix PR1657
...
llvm-svn: 42075
2007-09-18 01:54:42 +00:00
267c07b51f
Do not eliminate loop when it is invalid to do so. For example,
...
for(int i = 0; i < N; i++) {
if ( i == XYZ) {
A;
else
B;
}
C;
D;
}
llvm-svn: 42058
2007-09-17 21:01:05 +00:00
712dbe9d13
Skeleton for transformations to truncate loop's iteration space.
...
llvm-svn: 42054
2007-09-17 20:39:48 +00:00
0625bd6472
Merge DenseMapKeyInfo & DenseMapValueInfo into DenseMapInfo
...
Add a new DenseMapInfo::isEqual method to allow clients to redefine
the equality predicate used when probing the hash table.
llvm-svn: 42042
2007-09-17 18:34:04 +00:00
2ac2652779
Instcombine x-((x/y)*y) into a remainder operator.
...
llvm-svn: 42035
2007-09-17 17:31:57 +00:00
6d5da71288
Factor the trampoline transformation into a subroutine.
...
llvm-svn: 42021
2007-09-17 10:26:40 +00:00
4cd516b50b
Be more careful when constant-folding PHI nodes.
...
llvm-svn: 41998
2007-09-16 08:04:16 +00:00
8d0cb881e5
Remove RLE. It is subsumed by GVN.
...
llvm-svn: 41968
2007-09-14 22:33:52 +00:00
98d3a08d8f
Remove the assumption that FP's are either float or
...
double from some of the many places in the optimizers
it appears, and do something reasonable with x86
long double.
Make APInt::dump() public, remove newline, use it to
dump ConstantSDNode's.
Allow APFloats in FoldingSet.
Expand X86 backend handling of long doubles (conversions
to/from int, mostly).
llvm-svn: 41967
2007-09-14 22:26:36 +00:00
d9111b88d1
silence a bogus gcc warning.
...
llvm-svn: 41949
2007-09-14 03:07:24 +00:00
264d4813c7
Temporary reverting r41817
...
(http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070910/053370.html ). It's
causing SPASS to fail.
llvm-svn: 41938
2007-09-14 01:13:55 +00:00
7b412cb823
Change llvm.gcroot to not init the root to null at runtime, this prevents
...
using it for live-in values etc.
llvm-svn: 41879
2007-09-12 17:53:10 +00:00
9204663bcb
Turn calls to trampolines into calls to the underlying
...
nested function.
llvm-svn: 41844
2007-09-11 14:35:41 +00:00
7ed6eb8992
Avoid negative logic.
...
llvm-svn: 41829
2007-09-11 01:10:45 +00:00
8c95373ced
Refactor code into a separate method.
...
llvm-svn: 41826
2007-09-11 00:42:56 +00:00
d67479b6ee
Clear split info object.
...
llvm-svn: 41823
2007-09-11 00:23:56 +00:00
a28a7f1b2d
Split condition does not have to be ICmpInst in all cases.
...
llvm-svn: 41822
2007-09-11 00:12:56 +00:00
f4202e91f8
Check all terminators inside loop.
...
llvm-svn: 41821
2007-09-10 23:57:58 +00:00
e804567cd8
remove some dead code, this is handled by constant folding.
...
llvm-svn: 41819
2007-09-10 23:46:29 +00:00
2181b8e86a
Swap exit condition operands if it works.
...
llvm-svn: 41817
2007-09-10 23:34:06 +00:00
c75cbe6473
Prevent tailcallelim from breaking "recursive" calls to builtins.
...
llvm-svn: 41804
2007-09-10 20:58:55 +00:00
f8ab0a9acc
Filter exit conditions which are not yet handled.
...
llvm-svn: 41800
2007-09-10 18:33:42 +00:00
d7409fdce5
Require SCEV before LCSSA.
...
llvm-svn: 41798
2007-09-10 18:08:23 +00:00
85a51e0060
Don't zap back to back volatile load/stores
...
llvm-svn: 41759
2007-09-07 05:33:03 +00:00
bed9dc423c
Next round of APFloat changes.
...
Use APFloat in UpgradeParser and AsmParser.
Change all references to ConstantFP to use the
APFloat interface rather than double. Remove
the ConstantFP double interfaces.
Use APFloat functions for constant folding arithmetic
and comparisons.
(There are still way too many places APFloat is
just a wrapper around host float/double, but we're
getting there.)
llvm-svn: 41747
2007-09-06 18:13:44 +00:00
0c5c47944a
Use isTrueWhenEqual. Thanks Chris!
...
llvm-svn: 41741
2007-09-06 02:40:25 +00:00
b0b066eaaa
When the two operands of an icmp are equal, there are five possible predicates
...
that would make the icmp true. Fixes PR1637.
llvm-svn: 41740
2007-09-06 01:10:22 +00:00
2320323647
Forgot to obey 80 column rule. Fixing that.
...
llvm-svn: 41725
2007-09-05 20:36:41 +00:00
e58572233d
Added default parameters to GetElementPtrInstr constructor call. Visual Studio 2k5 was getting confused and was unable to compile it. Suspected compiler error.
...
llvm-svn: 41721
2007-09-05 16:54:38 +00:00
c656cbb8c2
Update GEP constructors to use an iterator interface to fix
...
GLIBCXX_DEBUG issues.
llvm-svn: 41697
2007-09-04 15:46:09 +00:00
ffac17a223
Fix a gcroot lowering bug.
...
llvm-svn: 41668
2007-09-01 02:00:51 +00:00
0e258b8518
Cut off crazy computation. This helps PR1622 slightly.
...
llvm-svn: 41522
2007-08-28 04:23:55 +00:00
d2456a171d
Use simpler test to filter loops.
...
llvm-svn: 41516
2007-08-27 21:34:31 +00:00
703623d571
Update InvokeInst to work like CallInst
...
llvm-svn: 41506
2007-08-27 19:04:21 +00:00
2b9ec7ff33
Don't DSe volatile stores.
...
llvm-svn: 41456
2007-08-26 21:14:47 +00:00
6114751544
Move exit condition and exit branch from exiting block into loop header and dominator info. This avoid execution of dead iteration. Loop is already filter in the beginning such that this change is safe.
...
llvm-svn: 41394
2007-08-25 02:39:24 +00:00
c1ef32ef3d
Constant split values needs upper bound and lower bound check, just like any other split value.
...
llvm-svn: 41389
2007-08-25 01:09:14 +00:00
4e63e1f5b5
While calculating upper loop bound for first loop and lower loop bound for second loop, take care of edge cases.
...
llvm-svn: 41387
2007-08-25 00:56:38 +00:00
f5a01bf025
Fix regression that I caused yesterday night while adding logic to select appropriate split condition branch.
...
llvm-svn: 41365
2007-08-24 19:32:26 +00:00
4bc9298f2a
It is not safe to execute split condition's true branch first all the time. If split
...
condition predicate is GT or GE then execute false branch first.
llvm-svn: 41358
2007-08-24 06:17:19 +00:00
4be56a5d12
Reject ICMP_NE as index split condition.
...
llvm-svn: 41357
2007-08-24 06:02:25 +00:00
5e46fac6de
Tightenup loop filter.
...
llvm-svn: 41356
2007-08-24 05:36:56 +00:00
504dc0aaed
Remove incomplete cost analysis.
...
llvm-svn: 41354
2007-08-24 05:21:13 +00:00
887db2d832
Remove dead code.
...
llvm-svn: 41295
2007-08-22 21:07:41 +00:00
6f4f23320d
Fix typo.
...
llvm-svn: 41292
2007-08-22 20:55:18 +00:00
31206b56d5
Cosmetic change
...
"True Loop" and "False Loop" naming terminology to refer two loops
after loop cloning is confusing. Instead just use A_Loop and B_Loop.
llvm-svn: 41287
2007-08-22 19:33:29 +00:00
90da534987
Refactor loop condition check in a separate function.
...
llvm-svn: 41282
2007-08-22 18:27:01 +00:00
cd8beb7645
Fix thinko.
...
Starting value of second loop's induction variable can not be lower
then starting value of original loop's induction variable.
llvm-svn: 41280
2007-08-22 18:07:47 +00:00
a12000d572
Rename bunch of variables.
...
llvm-svn: 41250
2007-08-21 21:12:02 +00:00
f98db5e62a
Preserve LCSSA.
...
llvm-svn: 41246
2007-08-21 19:47:46 +00:00
b5933bbbd5
Use SmallVector instead of std::vector.
...
llvm-svn: 41207
2007-08-21 00:31:24 +00:00
8f4228d619
s/ExitBlock/ExitingBlock/g
...
llvm-svn: 41204
2007-08-20 23:51:18 +00:00
49c4f9a889
Replace indunction variable with split value in loop body.
...
This fixes art miscompile.
llvm-svn: 41195
2007-08-20 20:49:01 +00:00
c2e2d15f45
Do not split loops rejected by processOneIterationLoop().
...
llvm-svn: 41194
2007-08-20 20:24:15 +00:00
bfa9499a88
Oops, remove assert that wasn't meant to be committed.
...
llvm-svn: 41170
2007-08-18 23:21:28 +00:00
5b5b1ab9e0
Never insert duplicate edges.
...
llvm-svn: 41169
2007-08-18 23:18:03 +00:00
a0d49dac26
Typo.
...
llvm-svn: 41168
2007-08-18 15:08:56 +00:00
1282b6e181
Avoid spliting loops where two split condition branches are not independent.
...
llvm-svn: 41148
2007-08-18 00:00:32 +00:00
d1fcfcc76c
When one branch of condition is eliminated then head of the other
...
branch is not necessary immediate dominators of merge blcok in all cases.
llvm-svn: 41144
2007-08-17 21:59:16 +00:00
f5023a7a84
Factor out some code into a helper function.
...
llvm-svn: 41131
2007-08-16 22:51:56 +00:00
221a43604e
Add some more comments to GVN.
...
llvm-svn: 41129
2007-08-16 22:02:55 +00:00
3640e78057
Dominance frontier is now required.
...
llvm-svn: 41096
2007-08-15 03:34:53 +00:00
b81bcbde09
Cleanup removeBlocks.
...
Use dominance frontier to fixup incoming edges of successor blocks not domianted by DeadBB.
Use df_iterator to walk and delete basic blocks dominated by DeadBB.
llvm-svn: 41095
2007-08-15 03:31:47 +00:00
0db035567c
Remove unneeded header file.
...
llvm-svn: 41094
2007-08-15 03:01:04 +00:00
f55b79fa71
Avoid triangle loops.
...
llvm-svn: 41093
2007-08-15 02:14:55 +00:00
7cad917160
Avoid nested loops at the moment.
...
llvm-svn: 41090
2007-08-14 23:53:57 +00:00
33ba97d747
Fix dominance frontier update while removing blocks.
...
llvm-svn: 41082
2007-08-14 18:35:57 +00:00
bc271a02fd
Eliminate PHI nodes with constant values during normal GVN processing, even when
...
they're not related to eliminating a load.
llvm-svn: 41081
2007-08-14 18:33:27 +00:00
398602a6eb
Be more aggressive in pruning unnecessary PHI nodes when doing PHI construction.
...
llvm-svn: 41080
2007-08-14 18:16:29 +00:00
676070d503
Make GVN iterative.
...
llvm-svn: 41078
2007-08-14 18:04:11 +00:00
a7b220f23a
Fix a case where GVN was failing to return true when it had, in fact, modified
...
the function.
llvm-svn: 41077
2007-08-14 17:59:48 +00:00
dbe8497d45
Handle last value assignments.
...
llvm-svn: 41063
2007-08-14 01:30:57 +00:00
f74ccbb4e8
StartValue is already calculated.
...
llvm-svn: 41062
2007-08-14 00:15:45 +00:00
948653915f
Preserve simple analysis.
...
llvm-svn: 41054
2007-08-13 22:22:13 +00:00
b8a41bb4f1
Preserve dominator info.
...
llvm-svn: 41053
2007-08-13 22:13:24 +00:00
f258578206
Split loops and do CFG cleanup.
...
llvm-svn: 41029
2007-08-12 07:02:51 +00:00
9f90f965de
Remove unused variables.
...
llvm-svn: 41028
2007-08-12 04:45:36 +00:00
99c8ee2977
Transform a load from an undef/zero global into an undef/global even if we
...
have complex pointer manipulation going on. This allows us to compile
stuff like this:
__m128i foo(__m128i x){
static const unsigned int c_0[4] = { 0, 0, 0, 0 };
__m128i v_Zero = _mm_loadu_si128((__m128i*)c_0);
x = _mm_unpacklo_epi8(x, v_Zero);
return x;
}
into:
_foo:
xorps %xmm1, %xmm1
punpcklbw %xmm1, %xmm0
ret
llvm-svn: 41022
2007-08-11 18:48:48 +00:00
f417c2cc34
Clone loop.
...
llvm-svn: 40998
2007-08-10 18:07:13 +00:00
9a4761464f
Remove unncessary duplication.
...
llvm-svn: 40979
2007-08-10 00:59:03 +00:00
7bdf4531bb
Calculate exit and start value of true loop and false loop respectively.
...
llvm-svn: 40978
2007-08-10 00:53:35 +00:00
67af6cd7ea
ExitCondition and Induction variable are loop constraints
...
not split condition constraints.
llvm-svn: 40977
2007-08-10 00:33:50 +00:00
a8e4b4bc7b
when we see a unaligned load from an insufficiently aligned global or
...
alloca, increase the alignment of the load, turning it into an aligned load.
This allows us to compile:
#include <xmmintrin.h>
__m128i foo(__m128i x){
static const unsigned int c_0[4] = { 0, 0, 0, 0 };
__m128i v_Zero = _mm_loadu_si128((__m128i*)c_0);
x = _mm_unpacklo_epi8(x, v_Zero);
return x;
}
into:
_foo:
punpcklbw _c_0.5944, %xmm0
ret
.data
.lcomm _c_0.5944,16,4 # c_0.5944
instead of:
_foo:
movdqu _c_0.5944, %xmm1
punpcklbw %xmm1, %xmm0
ret
.data
.lcomm _c_0.5944,16,2 # c_0.5944
llvm-svn: 40971
2007-08-09 19:05:49 +00:00
9b1cc8cac0
Make NonLocal and None const in the right way. :-)
...
llvm-svn: 40961
2007-08-09 04:42:44 +00:00
42e3e5bec1
Traverse loop blocks' terminators to find split candidates.
...
llvm-svn: 40960
2007-08-09 01:39:01 +00:00
0183c797c4
Add cost analysis.
...
llvm-svn: 40952
2007-08-08 22:25:28 +00:00
0e34ee25ab
Preserve dom info while processing one iteration loop.
...
llvm-svn: 40947
2007-08-08 21:39:47 +00:00
b84d3b1c92
Change the None and NonLocal markers in memdep to be const.
...
llvm-svn: 40946
2007-08-08 21:39:39 +00:00
8abc5c82b7
Clear split info.
...
llvm-svn: 40944
2007-08-08 21:18:27 +00:00
593bf9ceb3
Handle multiple split conditions.
...
llvm-svn: 40941
2007-08-08 21:02:17 +00:00
680862880d
Global values also don't undead-ify pointers in our dead alloca's set.
...
llvm-svn: 40936
2007-08-08 19:12:31 +00:00
ddf4aee543
Make handleEndBlock significantly faster with one trivial improvement,
...
and one hack to avoid hitting a bad case when the alias analysis is imprecise.
llvm-svn: 40935
2007-08-08 18:38:28 +00:00
50df9685b0
Small improvement: if a function doesn't access memory, we don't need to scan
...
it for potentially undeading pointers.
llvm-svn: 40933
2007-08-08 17:58:56 +00:00
52aaabf74d
Add some comments, remove a dead argument, and simplify some control flow.
...
No functionality change.
llvm-svn: 40932
2007-08-08 17:50:09 +00:00
b17ab03081
A few more small cleanups.
...
llvm-svn: 40922
2007-08-08 06:06:02 +00:00
0aecf0ebef
First round of cleanups from Chris' feedback.
...
llvm-svn: 40919
2007-08-08 04:52:29 +00:00
68de1ae816
Embrace patch review feedback.
...
llvm-svn: 40915
2007-08-08 01:51:27 +00:00
c7e53bdcfd
Fix new compare instruction's signness. Caught by Chris during review.
...
llvm-svn: 40912
2007-08-07 23:17:52 +00:00
0cc1a76283
Don't insert nearly as many redundant phi nodes.
...
llvm-svn: 40909
2007-08-07 23:12:31 +00:00
19211b6528
Use eraseFromParent().
...
llvm-svn: 40903
2007-08-07 17:45:35 +00:00
bacdbaa0da
Fix comment typo
...
llvm-svn: 40898
2007-08-07 16:52:03 +00:00
816a190cdf
Fix GLIBCXX_DEBUG error triggered by incrementing erased iterator.
...
llvm-svn: 40897
2007-08-07 16:44:38 +00:00
c70106cb30
Begin loop index split pass.
...
llvm-svn: 40883
2007-08-07 00:25:56 +00:00
8052019a20
It's safe to fold not of fcmp.
...
llvm-svn: 40870
2007-08-06 20:04:16 +00:00
77b2accbca
Make this code more efficient.
...
llvm-svn: 40861
2007-08-06 15:09:17 +00:00
d959cfc882
Silence some warnings from doxygen about @param argument name not matching the
...
actual argument name of the documented function.
llvm-svn: 40851
2007-08-05 19:35:22 +00:00
f0da7975ea
at the end of instcombine, explicitly clear WorklistMap.
...
This shrinks it down to something small. On the testcase
from PR1432, this speeds up instcombine from 0.7959s to 0.5000s,
(59%)
llvm-svn: 40840
2007-08-05 08:47:58 +00:00
20f0811fc0
Clean up comments, fix up some confusing code logic.
...
Predsimplify fails llvm-gcc bootstrap.
llvm-svn: 40815
2007-08-04 18:45:32 +00:00
7132e00de7
This is the patch to provide clean intrinsic function overloading support in LLVM. It cleans up the intrinsic definitions and generally smooths the process for more complicated intrinsic writing. It will be used by the upcoming atomic intrinsics as well as vector and float intrinsics in the future.
...
This also changes the syntax for llvm.bswap, llvm.part.set, llvm.part.select, and llvm.ct* intrinsics. They are automatically upgraded by both the LLVM ASM reader and the bitcode reader. The test cases have been updated, with special tests added to ensure the automatic upgrading is supported.
llvm-svn: 40807
2007-08-04 01:51:18 +00:00
2d19aae4ca
Fix a subtle miscompilation. This allows 197.parser to be compiled correctly.
...
llvm-svn: 40791
2007-08-03 19:59:35 +00:00
774761c503
Fix a subtle iterator invalidation bug in a recursive algorithm.
...
llvm-svn: 40776
2007-08-03 11:03:26 +00:00
1f70816c73
Fix an accidental commit.
...
llvm-svn: 40758
2007-08-02 21:33:36 +00:00
a8ba659976
Fix 80 col. violations.
...
llvm-svn: 40751
2007-08-02 18:20:52 +00:00
9699a6ea03
Fix 80 col. violations.
...
llvm-svn: 40750
2007-08-02 18:16:06 +00:00
e3590584b9
Fix 80 col. violations.
...
llvm-svn: 40749
2007-08-02 18:11:11 +00:00
0ac1fc8ac1
Fix a bug that was causing several miscompilations on SPEC.
...
llvm-svn: 40746
2007-08-02 17:56:05 +00:00
dc2cf228ce
Replacing a cast with another one does not reduce the number of
...
casts in the input.
llvm-svn: 40741
2007-08-02 17:23:38 +00:00
222b214be7
Disable an xform that causes an infinite loop. This fixes PR1594
...
llvm-svn: 40739
2007-08-02 16:56:32 +00:00
2740694450
wrap some long lines. Major offenders that are left include
...
gvn, gvnpre, dse, and predsimplify. To see these, use:
make check-line-length
llvm-svn: 40738
2007-08-02 16:53:43 +00:00
a882328e61
Update dominator info for the middle blocks created while spliting
...
exit edge to preserve LCSSA.
Fix dominance frontier update during loop unswitch. This fixes PR 1589, again
llvm-svn: 40737
2007-08-02 15:25:57 +00:00
Chris Lattner