and add a doxygen comment.
Cache the phi entry to avoid doing tons of
PHINode::getBasicBlockIndex calls in the common case.
On my insane testcase from re2c, this speeds up CGP from
617.4s to 7.9s (78x).
llvm-svn: 96083
to a PHI, avoid it in the common case where the BB occurs
in the same index for multiple phis. This speeds up CGP on
an insane testcase from 8.35 to 3.58s.
llvm-svn: 96080
Functions explicitly marked inline will get an inlining threshold slightly
more aggressive than the default for -O3. This means than -O3 builds are
mostly unaffected while -Os builds will be a bit bigger and faster.
The difference depends entirely on how many 'inline's are sprinkled on the
source.
In the CINT2006 suite, only these tests are significantly affected under -Os:
Size Time
471.omnetpp +1.63% -1.85%
473.astar +4.01% -6.02%
483.xalancbmk +4.60% 0.00%
Note that 483.xalancbmk runs too quickly to give useful timing results.
llvm-svn: 96066
2. don't bother trying to merge globals in non-default sections,
doing so is quite dubious at best anyway.
3. fix a bug reported by Arnaud de Grandmaison where we'd try to
merge two globals in different address spaces.
llvm-svn: 95995
bug fixes, and with improved heuristics for analyzing foreign-loop
addrecs.
This change also flattens IVUsers, eliminating the stride-oriented
groupings, which makes it easier to work with.
llvm-svn: 95975
what it does. Enhance it to return false to optimizing vector
sign extensions from vector comparisions, which is the idiom used
to get a splatted vector for a vector comparison.
Doing this breaks vector-casts.ll, add some compensating
transformations to handle the important case they cover without
depending on this canonicalization.
This fixes rdar://7434900 a serious pessimization of vector compares.
llvm-svn: 95855
block. Other blocks may have pointer cycles that will crash
basicaa and other alias analyses. In any case, there is no
point wasting cycles optimizing dead blocks. This fixes
rdar://7635088
llvm-svn: 95852
Initial skeleton and SCEVUnknown lowering implemented,
the rest should come relatively quickly. Move testcase
to new directory.
Move pass to right before SimplifyLibCalls - which is
moved down a bit so we can take advantage of a few opts.
llvm-svn: 95628
This time it's for real! I am going to hook this up in the frontends as well.
The inliner has some experimental heuristics for dealing with the inline hint.
When given a -respect-inlinehint option, functions marked with the inline
keyword are given a threshold just above the default for -O3.
We need some experiments to determine if that is the right thing to do.
llvm-svn: 95466
xform it is checking to actually pass. There is no need to match
m_SelectCst<0, -1> since instcombine canonicalizes that into not(sext).
Add matches for sext(not(x)) in addition to not(sext(x)).
llvm-svn: 95420
short-circuited conditions to AND/OR expressions, and those expressions
are often converted back to a short-circuited form in code gen. The
original source order may have been optimized to take advantage of the
expected values, and if we reassociate them, we change the order and
subvert that optimization. Radar 7497329.
llvm-svn: 95333
This makes the inliner about as agressive as it was before my changes to the
inliner cost calculations. These levels give the same performance and slightly
smaller code than before.
llvm-svn: 95320
Fix bugs where we would compute out of bounds as in bounds, and where
we couldn't know that the linker could override the size of an array.
Add a few new testcases, change existing testcase to use a private
global array instead of extern.
llvm-svn: 95283
The SRThreshold value makes perfect sense for checking if an entire aggregate
should be promoted to a scalar integer, but it is not so good for splitting
an aggregate into its separate elements. A struct may contain a large embedded
array along with some scalar fields that would benefit from being split apart
by SROA. Even if the total aggregate size is large, it may still be good to
perform SROA. Thus, the most important piece of this patch is simply moving
the aggregate size comparison vs. SRThreshold so that it guards only the
aggregate promotion.
We have also been checking the number of elements to decide if an aggregate
should be split up. The limit of "SRThreshold/4" seemed rather arbitrary,
and I don't think it's very useful to derive this limit from SRThreshold
anyway. I've collected some data showing that the current default limit of
32 (since SRThreshold defaults to 128) is a reasonable cutoff for struct
types. One thing suggested by the data is that distinguishing between structs
and arrays might be useful. There are (obviously) a lot more large arrays
than large structs (as measured by the number of elements and not the total
size -- a large array inside a struct still counts as a single element given
the way we do SROA right now). Out of 8377 arrays where we successfully
performed SROA while compiling a large set of benchmarks, only 16 of them had
more than 8 elements. And, for those 16 arrays, it's not at all clear that
SROA was actually beneficial. So, to offset the compile time cost of
investigating more large structs for SROA, the patch lowers the limit on array
elements to 8.
This fixes Apple Radar 7563690.
llvm-svn: 95224
disabled by default. This divides the existing load PRE code into 2 phases:
first it checks that it is safe to move the load to each of the predecessors
where it is unavailable, and then if it is safe, the code is changed to move
the load. Radar 7571861.
llvm-svn: 95007
of objc message send was getting marked arm_apcscc, but the prototype
isn't. This is fine at runtime because objcmsgsend is implemented in
assembly. Only turn a mismatched caller and callee into 'unreachable'
if the callee is a definition.
llvm-svn: 94986
case, instcombine can't zap the invoke for fear of changing the CFG.
However, we have to do something to prevent the next iteration of
instcombine from inserting another store -> undef before the invoke
thereby getting into infinite iteration between dead store elim and
store insertion.
Just zap the callee to null, which will prevent the next iteration
from doing anything.
llvm-svn: 94985
This bug was exposed by my inliner cost changes in r94615, and caused failures
of lencod on most architectures when building with LTO.
This patch fixes lencod and 464.h264ref on x86-64 (and likely others).
llvm-svn: 94858
create a testcase where this matters. The select+load transformation only
occurs when isSafeToLoadUnconditionally is true, and in those situations,
instcombine also changes the underlying objects to be aligned. This seems
like a good idea regardless, and I've verified that it doesn't pessimize
the subsequent realignment.
llvm-svn: 94850
(via APInt &RHSKnownZero = KnownZero, etc) seems dangerous and confusing to me: it
is easy not to notice this, and then wonder why KnownZero/RHSKnownZero changed
underneath you when you modified RHSKnownZero/KnownZero etc. So get rid of this.
No intended functionality change (tested with "make check" + llvm-gcc bootstrap).
llvm-svn: 94802
This was already being done in SSAUpdater::GetValueAtEndOfBlock so I've
just changed SSAUpdater to check for existing PHIs in both places.
llvm-svn: 94690
Modules and ModuleProviders. Because the "ModuleProvider" simply materializes
GlobalValues now, and doesn't provide modules, it's renamed to
"GVMaterializer". Code that used to need a ModuleProvider to materialize
Functions can now materialize the Functions directly. Functions no longer use a
magic linkage to record that they're materializable; they simply ask the
GVMaterializer.
Because the C ABI must never change, we can't remove LLVMModuleProviderRef or
the functions that refer to it. Instead, because Module now exposes the same
functionality ModuleProvider used to, we store a Module* in any
LLVMModuleProviderRef and translate in the wrapper methods. The bindings to
other languages still use the ModuleProvider concept. It would probably be
worth some time to update them to follow the C++ more closely, but I don't
intend to do it.
Fixes http://llvm.org/PR5737 and http://llvm.org/PR5735.
llvm-svn: 94686
parameter with a default value, instead of just hardcoding it in the
implementation. The limit of MaxLookup = 6 was introduced in r69151 to fix
a performance problem with O(n^2) behavior in instcombine, but the scalarrepl
pass is relying on getUnderlyingObject to go all the way back to an AllocaInst.
Making the limit part of the method signature makes it clear that by default
the result is limited and should help avoid similar problems in the future.
This fixes pr6126.
llvm-svn: 94433
"sext cond" instead of a select. This simplifies some instcombine
code, matches the policy for zext (cond ? 1 : 0 -> zext), and allows
us to generate better code for a testcase on ppc.
llvm-svn: 94339
for arbitrary terminators in predecessors, don't assume
it is a conditional or uncond branch. The testcase shows
an example where they can happen with switches.
llvm-svn: 94323
handle the case when we can infer an input to the xor
from all inputs that agree, instead of going into an
infinite loop. Another part of PR6199
llvm-svn: 94321
missing ones are libsupport, libsystem and libvmcore. libvmcore is
currently blocked on bugpoint, which uses EH. Once it stops using
EH, we can switch it off.
This #if 0's out 3 unit tests, because gtest requires RTTI information.
Suggestions welcome on how to fix this.
llvm-svn: 94164
loop-variant components, adds must be inserted after the increment.
Keep track of the increment position for this case, and insert
these adds in the correct location.
llvm-svn: 94110
ValueMapper.cpp ends up calling an out of line
__ZNK4llvm12PATypeHolder3getEv, which is a template and llvm-config
determines arbitrarily to use the one in libipo. This sucks, but
keeping the #include is a reasonable workaround.
llvm-svn: 94103
This new version is much more aggressive about doing "full" reduction in
cases where it reduces register pressure, and also more aggressive about
rewriting induction variables to count down (or up) to zero when doing so
reduces register pressure.
It currently uses fairly simplistic algorithms for finding reuse
opportunities, but it introduces a new framework allows it to combine
multiple strategies at once to form hybrid solutions, instead of doing
all full-reduction or all base+index.
llvm-svn: 94061
than the scaled register. This makes it more likely that subsequent
AddrModeMatcher queries will match the new address the same way as the
old, instead of accidentally matching what had been the base register
as the new scaled register, and then failing to match the scaled register.
This fixes some problems with address-mode sinking multiple muls into a
block, which will be a lot more common with some upcoming
LoopStrengthReduction changes.
llvm-svn: 93935
are the same. I had already fixed a similar problem where the source and
destination were different bitcasts derived from the same alloca, but the
previous fix still did not handle the case where both operands are exactly
the same value. Radar 7552893.
llvm-svn: 93848
aggressive changed the canonical form from sext(trunc(x)) to ashr(lshr(x)),
make sure to transform a couple more things into that canonical form,
and catch a case where we missed turning zext/shl/ashr into a single sext.
llvm-svn: 93787
added to the FSub version. However, the original version of this xform guarded
against doing this for floating point (!Op0->getType()->isFPOrFPVector()).
This is causing LLVM to perform incorrect xforms for code like:
void func(double *rhi, double *rlo, double xh, double xl, double yh, double yl){
double mh, ml;
double c = 134217729.0;
double up, u1, u2, vp, v1, v2;
up = xh*c;
u1 = (xh - up) + up;
u2 = xh - u1;
vp = yh*c;
v1 = (yh - vp) + vp;
v2 = yh - v1;
mh = xh*yh;
ml = (((u1*v1 - mh) + (u1*v2)) + (u2*v1)) + (u2*v2);
ml += xh*yl + xl*yh;
*rhi = mh + ml;
*rlo = (mh - (*rhi)) + ml;
}
The last line was optimized away, but rl is intended to be the difference
between the infinitely precise result of mh + ml and after it has been rounded
to double precision.
llvm-svn: 93369
in JT.
2) When cloning blocks for PHI or xor conditions, use
instsimplify to simplify the code as we go. This allows us to
squish common cases early in JT which opens up opportunities for
subsequent iterations, and allows it to completely simplify the
testcase.
llvm-svn: 93253
condition is a xor with a phi node. This eliminates nonsense
like this from 176.gcc in several places:
LBB166_84:
testl %eax, %eax
- setne %al
- xorb %cl, %al
- notb %al
- testb $1, %al
- je LBB166_85
+ je LBB166_69
+ jmp LBB166_85
This is rdar://7391699
llvm-svn: 93221
trunc has multiple uses. Codegen is not able to coalesce the subreg case
correctly and so this leads to higher register pressure and spilling (see PR5997).
This speeds up 256.bzip2 from 8.60 -> 8.04s on my machine, ~7%.
llvm-svn: 93200
BitsToClear case. This allows it to promote expressions which have an
and/or/xor after the lshr, promoting cases like test2 (from PR4216)
and test3 (random extample extracted from a spec benchmark).
clang now compiles the code in PR4216 into:
_test_bitfield: ## @test_bitfield
movl %edi, %eax
orl $194, %eax
movl $4294902010, %ecx
andq %rax, %rcx
orl $32768, %edi
andq $39936, %rdi
movq %rdi, %rax
orq %rcx, %rax
ret
instead of:
_test_bitfield: ## @test_bitfield
movl %edi, %eax
orl $194, %eax
movl $4294902010, %ecx
andq %rax, %rcx
shrl $8, %edi
orl $128, %edi
shlq $8, %rdi
andq $39936, %rdi
movq %rdi, %rax
orq %rcx, %rax
ret
which is still not great, but is progress.
llvm-svn: 93145
new BitsToClear result which allows us to start promoting
expressions that end with a lshr-by-constant. This is
conservatively correct and better than what we had before
(see testcases) but still needs to be extended further.
llvm-svn: 93144
the zext dest type. This allows us to handle test52/53 in cast.ll,
and allows llvm-gcc to generate much better code for PR4216 in -m64
mode:
_test_bitfield: ## @test_bitfield
orl $32962, %edi
movl %edi, %eax
andl $-25350, %eax
ret
This also fixes a bug handling vector extends, ensuring that the
mask produced is a vector constant, not an integer constant.
llvm-svn: 93127
elimination of a sign extend to be a win, which simplifies
the client of CanEvaluateSExtd, and allows us to eliminate
more casts (examples taken from real code).
llvm-svn: 93109
lshr+ashr instead of trunc+sext. We want to avoid type
conversions whenever possible, it is easier to codegen expressions
without truncates and extensions.
llvm-svn: 93107
1) don't try to optimize a sext or zext that is only used by a trunc, let
the trunc get optimized first. This avoids some pointless effort in
some common cases since instcombine scans down a block in the first pass.
2) Change the cost model for zext elimination to consider an 'and' cheaper
than a zext. This allows us to do it more aggressively, and for the next
patch to simplify the code quite a bit.
llvm-svn: 93097
commonIntCastTransforms into the callers, eliminating a switch,
and allowing the static predicate methods to be moved down to
live next to the corresponding function. No functionality
change.
llvm-svn: 93089
to an element of a vector in a static ctor) which occurs with an
unrelated patch I'm testing. Annoyingly, EvaluateStoreInto basically
does exactly the same stuff as InsertElement constant folding, but it
now handles vectors, and you can't insertelement into a vector. It
would be 'really nice' if GEP into a vector were not legal.
llvm-svn: 92889
phi nodes when deciding which pointers point to local memory.
I actually checked long ago how useful this is, and it isn't
very: it hardly ever fires in the testsuite, but since Chris
wants it here it is!
llvm-svn: 92836
memcpy, memset and other intrinsics that only access their arguments
to be readnone if the intrinsic's arguments all point to local memory.
This improves the testcase in the README to readonly, but it could in
theory be made readnone, however this would involve more sophisticated
analysis that looks through the memcpy.
llvm-svn: 92829
Previously, instcombine would only promote an expression tree to
the larger type if doing so eliminated two casts. This is because
a need to manually do the sign extend after the promoted expression
tree with two shifts. Now, we keep track of whether the result of
the computation is going to be properly sign extended already. If
so, we can unconditionally promote the expression, which allows us
to zap more sext's.
This implements rdar://6598839 (aka gcc pr38751)
llvm-svn: 92815
Eliminate the 'AddMaskingAnd' transformation, it is redundant with this
more general code right below it:
// A+B --> A|B iff A and B have no bits set in common.
llvm-svn: 92693
that got instantiated. There is no reason for instcombine
to try this hard for simple associative optimizations. Next
up, eliminate the template completely.
llvm-svn: 92692
when doing this transform if the GEP is not inbounds. No testcase because
it is very difficult to trigger this: instcombine already canonicalizes
GEP indices to pointer size, so it relies specific permutations of the
instcombine worklist.
Thanks to Duncan for pointing this possible problem out.
llvm-svn: 92495
on the example in PR4216. This doesn't trigger in the testsuite,
so I'd really appreciate someone scrutinizing the logic for
correctness.
llvm-svn: 92458
when a consequtive sequence of elements all satisfies the
predicate. Like the double compare case, this generates better
code than the magic constant case and generalizes to more than
32/64 element array lookups.
Here are some examples where it triggers. From 403.gcc, most
accesses to the rtx_class array are handled, e.g.:
@rtx_class = constant [153 x i8] c"xxxxxmmmmmmmmxxxxxxxxxxxxmxxxxxxiiixxxxxxxxxxxxxxxxxxxooxooooooxxoooooox3x2c21c2222ccc122222ccccaaaaaa<<<<<<<<<<<<<<<<<<111111111111bbooxxxxxxxxxxcc2211x", align 32 ; <[153 x i8]*> [#uses=547]
%142 = icmp eq i8 %141, 105
@rtx_class = constant [153 x i8] c"xxxxxmmmmmmmmxxxxxxxxxxxxmxxxxxxiiixxxxxxxxxxxxxxxxxxxooxooooooxxoooooox3x2c21c2222ccc122222ccccaaaaaa<<<<<<<<<<<<<<<<<<111111111111bbooxxxxxxxxxxcc2211x", align 32 ; <[153 x i8]*> [#uses=543]
%165 = icmp eq i8 %164, 60
Also, most of the 59-element arrays (mode_class/rid_to_yy, etc)
optimized before are actually range compares. This lets 32-bit
machines optimize them.
400.perlbmk has stuff like this:
400.perlbmk: PL_regkind, even for 32-bit:
@PL_regkind = constant [62 x i8] c"\00\00\02\02\02\06\06\06\06\09\09\0B\0B\0D\0E\0E\0E\11\12\12\14\14\16\16\18\18\1A\1A\1C\1C\1E\1F !!!$$&'((((,-.///88886789:;8$", align 32 ; <[62 x i8]*> [#uses=4]
%811 = icmp ne i8 %810, 33
@PL_utf8skip = constant [256 x i8] c"\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\01\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\02\03\03\03\03\03\03\03\03\03\03\03\03\03\03\03\03\04\04\04\04\04\04\04\04\05\05\05\05\06\06\07\0D", align 32 ; <[256 x i8]*> [#uses=94]
%12 = icmp ult i8 %10, 2
etc.
llvm-svn: 92426
two elements match or don't match with two comparisons. For
example, the testcase compiles into:
define i1 @test5(i32 %X) {
%1 = icmp eq i32 %X, 2 ; <i1> [#uses=1]
%2 = icmp eq i32 %X, 7 ; <i1> [#uses=1]
%R = or i1 %1, %2 ; <i1> [#uses=1]
ret i1 %R
}
This generalizes the previous xforms when the array is larger than
64 elements (and this case matches) and generates better code for
cases where it overlaps with the magic bitshift case.
This generalizes more cases than you might expect. For example,
400.perlbmk has:
@PL_utf8skip = constant [256 x i8] c"\01\01\01\...
%15 = icmp ult i8 %7, 7
403.gcc has:
@rid_to_yy = internal constant [114 x i16] [i16 259, i16 260, ...
%18 = icmp eq i16 %16, 295
and xalancbmk has a bunch of examples, such as
_ZN11xercesc_2_5L15gCombiningCharsE and _ZN11xercesc_2_5L10gBaseCharsE.
llvm-svn: 92417
arrays with variable indices into a comparison of the index
with a constant. The most common occurrence of this that
I see by far is stuff like:
if ("foobar"[i] == '\0') ...
which we compile into: if (i == 6), saving a load and
materialization of the global address. This also exposes
loop trip count information to later passes in many cases.
This triggers hundreds of times in xalancbmk, which is where I first
noticed it, but it also triggers in many other apps. Here are a few
interesting ones from various apps:
@must_be_connected_without = internal constant [8 x i8*] [i8* getelementptr inbounds ([3 x i8]* @.str64320, i64 0, i64 0), i8* getelementptr inbounds ([3 x i8]* @.str27283, i64 0, i64 0), i8* getelementptr inbounds ([4 x i8]* @.str71327, i64 0, i64 0), i8* getelementptr inbounds ([4 x i8]* @.str72328, i64 0, i64 0), i8* getelementptr inbounds ([3 x i8]* @.str18274, i64 0, i64 0), i8* getelementptr inbounds ([6 x i8]* @.str11267, i64 0, i64 0), i8* getelementptr inbounds ([3 x i8]* @.str32288, i64 0, i64 0), i8* null], align 32 ; <[8 x i8*]*> [#uses=2]
%scevgep.i = getelementptr [8 x i8*]* @must_be_connected_without, i64 0, i64 %indvar.i ; <i8**> [#uses=1]
%17 = load ...
%18 = icmp eq i8* %17, null ; <i1> [#uses=1]
-> icmp eq i64 %indvar.i, 7
@yytable1095 = internal constant [84 x i8] c"\12\01(\05\06\07\08\09\0A\0B\0C\0D\0E1\0F\10\11266\1D: \10\11,-,0\03'\10\11B6\04\17&\18\1945\05\06\07\08\09\0A\0B\0C\0D\0E\1E\0F\10\11*\1A\1B\1C$3+>#%;<IJ=ADFEGH9KL\00\00\00C", align 32 ; <[84 x i8]*> [#uses=2]
%57 = getelementptr inbounds [84 x i8]* @yytable1095, i64 0, i64 %56 ; <i8*> [#uses=1]
%mode.0.in = getelementptr inbounds [9 x i32]* @mb_mode_table, i64 0, i64 %.pn ; <i32*> [#uses=1]
load ...
%64 = icmp eq i8 %58, 4 ; <i1> [#uses=1]
-> icmp eq i64 %.pn, 35 ; <i1> [#uses=0]
@gsm_DLB = internal constant [4 x i16] [i16 6554, i16 16384, i16 26214, i16 32767]
%scevgep.i = getelementptr [4 x i16]* @gsm_DLB, i64 0, i64 %indvar.i ; <i16*> [#uses=1]
%425 = load %scevgep.i
%426 = icmp eq i16 %425, -32768 ; <i1> [#uses=0]
-> false
llvm-svn: 92411
pointer to int casts that confuse later optimizations. See PR3351
for details.
This improves but doesn't complete fix 483.xalancbmk because llvm-gcc
does this xform in GCC's "fold" routine as well. Clang++ will do
better I guess.
llvm-svn: 92408
Dan Gohman