Make it always return APInts with the same bitwidth for the same ConstantRange bitwidth to simply clients

llvm-svn: 160294

llvm-svn: 159352

llvm-svn: 159319

this fixes the missed optimization I was seeing in the CorrelatedValuePropagation pass

llvm-svn: 157032

llvm-svn: 133648

zextOrTrunc(), and APSInt methods extend(), extOrTrunc() and new method
trunc(), to be const and to return a new value instead of modifying the
object in place.

llvm-svn: 121120

also fixes PR8250.

llvm-svn: 114972

Fix zeroExtend and signExtend to support empty sets, and to return the smallest
possible result set which contains the extension of each element in their
inputs. For example zext i8 [100, 10) to i16 is now [0, 256), not i16 [100, 10)
which contains 63446 members.

llvm-svn: 113187

 - remove ashr which never worked.
 - fix lshr and shl and add tests.
 - remove dead function "intersect1Wrapped".
 - add a new sub method to subtract ranges, with test.

llvm-svn: 110861

llvm-svn: 110504

llvm-svn: 81102

equality. Prefer EXPECT_EQ(foo, Full) over EXPECT_TRUE(foo.isFullSet()) because
the former will print out the contents of the constant range that failed.

llvm-svn: 81094

This can break when there are implicit conversions from types raw_ostream
understands but std::ostream doesn't, but it increases the number of cases that
Just Work.

llvm-svn: 81093

means that raw_ostream no longer has to #include <iosfwd>.  Nothing in llvm
should use raw_os_ostream.h, but llvm-gcc and some unit tests do.

llvm-svn: 79886

llvm-svn: 79879

ConstantRange doesn't have an std::ostream inserter anymore.

llvm-svn: 79831

llvm-svn: 76744

return the smallest union of two ranges instead of just any range that happens
to contain the union.

llvm-svn: 76360

all values belonging to the intersection will belong to the resulting range.
The former was inconsistent about that point (either way is fine, just pick
one.) This is part of PR4545.

llvm-svn: 76289

again to Daniel Dunbar and KLEE!

llvm-svn: 75449

Found by Daniel Dunbar and KLEE.

llvm-svn: 75448

generally return Full on all wrapped inputs. "Fixes" PR4545.

llvm-svn: 75444

large as the full set, only those one size smaller. Thanks to Daniel Dunbar
who found this bug using Klee!

llvm-svn: 75443

llvm-svn: 75413

Some = [10, 2730). A subset of that is [1024..2048) which covers every possible
10-bit pattern.

llvm-svn: 75411

llvm-svn: 75410

A full set is a constant range that represents any number. If you take the
umax of that and [5, 10) you end up with [5, INT_MAX] because the values less
than 5 would be umax's against a value which is at least 5.

llvm-svn: 75372

changes.

llvm-svn: 75369

merge the new functionality and unittests into ConstantRange. Thanks to
Nick Lewycky for pointing out that it isn't necessary to have two separate
classes here.

llvm-svn: 75191

what ConstantRange does for unsigned integers. Factor out a
common base class for common functionality.

Add some new functions for performing arithmetic on constant
ranges. Some of these are currently just stubbed out with
conservative implementations.

Add unittests for ConstantRange and ConstantSignedRange.

llvm-svn: 75177