1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 11:42:57 +01:00
Commit Graph

529 Commits

Author SHA1 Message Date
Nick Lewycky
29f7380f34 Implement unionWith.
llvm-svn: 34833
2007-03-02 03:33:05 +00:00
Reid Spencer
39b817765b Add a FIXME
llvm-svn: 34828
2007-03-02 02:01:34 +00:00
Reid Spencer
eabc981a60 Fix a problem where shifting by 64-bits leads to incorrect results on PPC
but not on X86 becuase shift by word size is "undefined".

llvm-svn: 34825
2007-03-02 01:19:42 +00:00
Reid Spencer
f8f8413839 Combine two lines that can be.
llvm-svn: 34818
2007-03-01 23:30:25 +00:00
Reid Spencer
c94231581b Make the static table of results in sqrt const.
llvm-svn: 34791
2007-03-01 17:47:31 +00:00
Reid Spencer
6a95676875 Add methods for bit width modification: sextOrTrunc, zextOrTrunc.
llvm-svn: 34789
2007-03-01 17:15:32 +00:00
Reid Spencer
3f3487918c Remove the "isSigned" parameters from ConstantRange. It turns out they
are not needed as the results are the same with or without it.

Patch by Nicholas Lewycky.

llvm-svn: 34782
2007-03-01 07:54:15 +00:00
Reid Spencer
0d77fe7f06 Use a real table in sqrt to shorten and quicken the code.
Thanks for the idea Chris.

llvm-svn: 34779
2007-03-01 06:23:32 +00:00
Reid Spencer
b6dd32889e Add a square root function.
llvm-svn: 34775
2007-03-01 05:39:56 +00:00
Reid Spencer
f119927a6f Move ConstantRange class to lib/Support from lib/Analysis and make its
interface not depend on Type or ICmpInst.

llvm-svn: 34761
2007-02-28 22:02:48 +00:00
Reid Spencer
d91c898cc0 Make the trunc/sext/zext methods return APInt& so that these operations
can be chained together with other operations.

llvm-svn: 34743
2007-02-28 17:34:32 +00:00
Reid Spencer
b1eaa7bd52 Fix a bug in RoundDoubleToAPInt where it would force the size to 64 bits
instead of honoring the client's requested bit width.

llvm-svn: 34712
2007-02-28 01:30:08 +00:00
Reid Spencer
e306ef6798 Join two lines that can be joined.
llvm-svn: 34705
2007-02-27 23:49:07 +00:00
Reid Spencer
edbe747bab Implement countLeadingOnes() and getMinSignedBits(). This helps to minimize
the bit width of negative numbers by computing the minimum bit width for a
negative value. E.g. 0x1800000000000000 could be just 0x8000000000000000

llvm-svn: 34695
2007-02-27 21:59:26 +00:00
Reid Spencer
9e66d0921a Improve APInt interface:
1. Add unsigned and signed versions of methods so a "bool" argument doesn't
   need to be passed in.
2. Make the various getMin/getMax functions all be inline since they are
   so simple.
3. Simplify sdiv and srem code.

llvm-svn: 34680
2007-02-27 20:24:31 +00:00
Reid Spencer
e0c8670bfe Conform to single person attribution in file comment.
llvm-svn: 34678
2007-02-27 19:31:09 +00:00
Reid Spencer
b61f731ddd 1. Fix three serious bugs in the comparison code.
2. Change RoundDoubleToAPInt to take a bit width parameter. Use that
   parameter to limit the bit width of the result.

llvm-svn: 34673
2007-02-27 18:23:40 +00:00
Reid Spencer
1e498b04f5 Get rid of extraneous // in file comment.
llvm-svn: 34662
2007-02-27 07:35:18 +00:00
Reid Spencer
67498a4072 Simplify and document RoundDoubleToAPInt.
llvm-svn: 34648
2007-02-27 01:28:10 +00:00
Reid Spencer
97d7d18442 1. Make sure all delete operators of arrays use the array form of delete.
2. Rewrite operator=(const APInt& RHS) to allow the RHS to be a different
   bit width than the LHS. This makes it possible to use APInt as the key
   of a DenseMap, as needed for the IntConstants map in Constants.cpp
3. Fix operator=(uint64_t) to clear unused bits in case the client assigns
   a value that has more bits than the APInt allows.
4. Assert that bit widths are equal in operator==
5. Revise getHashValue() to put the bit width in the low order six bits.
   This should help to make i1 0, i2 0, ... i64 0 all distinct in the
   IntConstants DenseMap.

llvm-svn: 34646
2007-02-26 23:38:21 +00:00
Reid Spencer
2dc808284e Implement the getHashValue method.
Fix toString use of getValue to use getZExtValue()

llvm-svn: 34642
2007-02-26 21:02:27 +00:00
Reid Spencer
728cf2d5cf Re-enable this. The header was committed.
llvm-svn: 34634
2007-02-26 08:10:54 +00:00
Chris Lattner
4b7b0a145a another missing header :( :( :(
llvm-svn: 34632
2007-02-26 07:51:00 +00:00
Reid Spencer
050bd4479e 1. Remove redundant calls to clearUsedBits().
2. Fix countTrailingZeros to use a faster algorithm.
3. Simplify sext() slightly by using isNegative().
4. Implement ashr using word-at-a-time logic instead of bit-at-a-time
5. Rename locals named isNegative so they don't clash with method name.
6. Fix fromString to compute negated value correctly.

llvm-svn: 34629
2007-02-26 07:44:38 +00:00
Reid Spencer
6177d67365 Rewrite lshr to not do bit by bit copy but to copy and shift whole words.
This makes it much more efficient.

llvm-svn: 34618
2007-02-26 01:19:48 +00:00
Reid Spencer
82a5b300ad Fix sext operation. Shifting by zero would leave an incorrect mask.
llvm-svn: 34617
2007-02-25 23:54:00 +00:00
Reid Spencer
c0d07ad61a 1. Fix the flip() method to correctly flip all words of the APInt.
2. Implement the trunc, sext, and zext operations.
3. Improve fromString to accept negative values as input.

llvm-svn: 34616
2007-02-25 23:44:53 +00:00
Reid Spencer
e6f7201099 1. Provide more detail in file comment.
2. Move comments for methods to .h file, delete them in .cpp file.
3. All places that were doing manual clear of high order bits now call the
   clearUnusedBits() method in order to not depend on undefined behavior
   of the >> operator when the number of bits shifted equals the word size.
4. Reduced # of loc by using the new result of clearUnusedBits() method.
5. Simplified logic (decreased indentation) in a few places.
6. Added code comments to larger functions that needed them.
7. Added FIXME notes about weak implementations of things (e.g. bit-by-bit
   shift right is sub-optimal).

llvm-svn: 34603
2007-02-25 19:32:03 +00:00
Reid Spencer
ec6159bbea Allow this to compile now that the header file is checked in.
llvm-svn: 34581
2007-02-25 07:30:14 +00:00
Chris Lattner
5adebeb207 this doesn't compile, disable it
llvm-svn: 34571
2007-02-25 02:32:44 +00:00
Reid Spencer
0fd90c3447 Clean up lshr and ashr to coding standards.
Handle the single word cases for shiftAmt == BitWidth.

llvm-svn: 34569
2007-02-25 01:56:07 +00:00
Reid Spencer
66e162cfcd Whoops, last word with bits in large shift left wasn't correct.
llvm-svn: 34565
2007-02-25 01:08:58 +00:00
Reid Spencer
4dcbccdc99 Fix the > 64 bits case for left shift.
llvm-svn: 34564
2007-02-25 00:56:44 +00:00
Reid Spencer
ecc067a698 Fix the remainder shifting in KnuthDiv.
llvm-svn: 34562
2007-02-24 20:38:01 +00:00
Reid Spencer
d6956a0e8c 1. Fix a bug in fromString for the <= 64bits case
2. Fix shl when shiftAmount == BitWidth.

llvm-svn: 34560
2007-02-24 20:19:37 +00:00
Reid Spencer
e2bb0cf83e 1. Fix last bug in KnuthDiv. All divide tests pass up to 1024 bits now.
2. Clean up comments, style, coding standards, etc.
3. Simplify a constructor.

Extended testing revealed some additional bugs in shifting. I'll fix these
tomorrow.

llvm-svn: 34559
2007-02-24 10:01:42 +00:00
Reid Spencer
b45646b14e 1. Make internal functions take const arguments where they should, just
to be safe.
2. Make internal functions that return a carry/borrow return bool instead
   of uint64_t because the carry/borrow can only be in range [0,1].
3. Assert that the pointers to KnuthDiv are all different so that the
   result and operands can't overlap.
4. Add debug output to KnuthDiv function.
5. Fix a problem with KnuthDiv by separating the b's complement operation
   from the subtraction borrow operation. This fixes a wide range of
   division problems, but alas, not all of them.

llvm-svn: 34554
2007-02-24 03:58:46 +00:00
Chris Lattner
e37ee33c4e remove folding set debug output
llvm-svn: 34549
2007-02-24 01:41:48 +00:00
Chris Lattner
042d3195c9 Avoid TBAA issue.
llvm-svn: 34539
2007-02-23 22:31:24 +00:00
Reid Spencer
3573b012d3 1. Fix a carry out problem in add if destination and x point to the same
memory (as done in fromString).
2. Implement Knuth divide more closely to what is recommended in his book.
3. Fix computation of the remainder for Knuth Divide (bad shifting).
4. Remove some cruft from the file

llvm-svn: 34518
2007-02-23 01:57:13 +00:00
Reid Spencer
cbff120eeb When converting from 64 to 32-bits, use the actual number of words to
extract the value, not the number of words implied by the active bits.
This fixes numerous, but not all divide bugs.

llvm-svn: 34484
2007-02-22 00:58:45 +00:00
Reid Spencer
d51b03d452 Fix countLeadingZeros in the case that the bitwidth evenly divides the
word size. This fixes all reads of uninitialized data (buffer over read)
and makes APInt.cpp memory clean, per valgrind. The only remaining
problem is division in a few cases.

llvm-svn: 34483
2007-02-22 00:22:00 +00:00
Reid Spencer
df5fe8163d Reorganize some code to make it clearer, avoid a few uninitialized memory
reads, and reduce the number of temporary APInt instances we construct.

llvm-svn: 34467
2007-02-21 08:21:52 +00:00
Reid Spencer
95874f6bc9 Fix the carry in addition.
llvm-svn: 34465
2007-02-21 05:44:56 +00:00
Reid Spencer
c4662ad089 1. Add a dump() method for faster debugging.
2. Change 0 initialization of union to larger component so all is zeroed.
3. Fix the borrow logic in subtraction so it works for > 128 bits.
4. Rewrite fromString to use a simpler but correct algorithm and also to
   not set the bit width contrary to the user's request.
5. Optimize toString a bit by making it only do one Knuth divide per
   iteration instead of two.

With these changes, all arithmetic passes (verified by pari/GP) up to
1024 bits except for certain division cases.

llvm-svn: 34463
2007-02-21 03:55:44 +00:00
Reid Spencer
1fbf856864 Fix countLeadingZeros to actually return the correct number.
Fix toString to correctly return "0" for zero valued APInts over 128 bits.

llvm-svn: 34459
2007-02-21 00:29:48 +00:00
Reid Spencer
17d47c5062 Make long addition and subtraction work. Speed things up by using internal
functions more.

llvm-svn: 34458
2007-02-20 23:40:25 +00:00
Reid Spencer
b27fcd8cb6 Clean up variable names in operator*.
Attempt #3 for getting a portable INFINITY value.

llvm-svn: 34454
2007-02-20 20:42:10 +00:00
Reid Spencer
4caebb1af8 Use INFINITY macro from math.h instead of constructing hex floating point
constants (avoids warnings).

llvm-svn: 34452
2007-02-20 18:29:12 +00:00
Reid Spencer
80b41ab5f7 First version that can process arith.cpp test case up to 1024 bits:
1. Ensure pVal is set to 0 in each constructor.
2. Fix roundToDouble to make correct calculations and not read beyond the
   end of allocated memory.
3. Implement Knuth's "classical algorithm" for division from scratch and
   eliminate buffer overflows and uninitialized mememory reads. Document
   it properly too.
4. Implement a wrapper function for KnuthDiv which handles the 64-bit to
   32-bit conversion and back. It also implement short division for the
   n == 1 case that Knuth's algorithm can't handle.
5. Simplify the logic of udiv and urem a little, make them exit early, and
   have them use the "divide" wrapper function to perform the division
   or remainder operation.
6. Move the toString function to the end of the file, closer to where
   the division functions are located.

Note: division is still broken for some > 64 bit values, but at least it
      doesn't crash any more.
llvm-svn: 34449
2007-02-20 08:51:03 +00:00