Re: [squeak-dev] Re: [4.2] tests status

 
On Sun, Dec 26, 2010 at 08:16:22AM +0100, Andreas Raab wrote:
(cc to vm-dev list)

I have found an additional issue with the FloatMath updates. When running
on a VM compiled for 64 bits, an updated trunk image hangs up with no input
event processing and no display updates.

This is on a Linux VM compiled with no optimization (CFLAGS='-g -m64 -O0).

The problem is related to Float>>fractionPart. If this method is reverted
to the prior version, or if the FloatMathPlugin is not present, the image
no longer hangs.

The primitive in question is FloatMathPlugin>>primitiveFractionalPart which
presumably has a bug related to pointer size. The bug is exposed by compiling
with 64 bit pointers, and the updated trunk image will hang up if the primitive
is being called by Float>>fractionPart.

The FloatMathPlugin>>primitiveFractionalPart method looks fine (variables
all declared correctly). The error occurs within the __ieee754_modf(rcvr, &trunc)
function call, which produces completely bogus results when compiled -m64.
This function is implemented as modf() in fdlibm. A stand-alone C test
program confirms that the bug is in the library function itself.

Presumably this bug would affect Mac VMs compiled for 64-bits, and I think
there are a number of these in circulation (though I do not know if they
include the FloatMathPlugin).

Dave

 
I fail to see how modf() could possibly be inexact, it just has to cut
the mantissa in two parts...
Therefore, primitive 52 should lead to bit identical results across platforms.
The only thing to change in order to be FloatMathPlugin compatible
would be to fail the primitive for a nan receiver.

The fdlibm do define a replacement __ieee754_modf() because it has to
(it is a replacement of libm and cannot use default libm).
Maybe we don't have to use it, and can safely revert to modf().

Nicolas

2010/12/29 David T. Lewis <[hidden email]>:

 
On Wed, Dec 29, 2010 at 07:13:25PM +0100, Nicolas Cellier wrote:
Judging by the results of FloatConsistencyTests>>testFractionStd,
primitive 52 does in fact produce different results from
FloatMathPlugin>>primitiveFractionalPart. Whether primitive 52
would be consistent across platforms I cannot say.

But more importantly, I'm afraid that the FloatMathPlugin is
completely broken when compiled for 64 bits. I reverted the
Float>>fractionPart method to use primitive 52 in order to be
able to run the image, then ran the full suite of KernelTests-Numbers
tests. The results are not encouraging:

====
29 December 2010 1:21:48 pm
Results of KernelTests-Numbers tests with 64-bit VM and FloatMathPlugin, no compiler optimization.
Float>>fractionPart is reverted to prior version to permit the image to run, so FloatConsistencyTests>>testFractionStd is an expected failure.
Test errors:
        FloatConsistencyTests>>#testAddArray
        FloatConsistencyTests>>#testArcCos
        FloatConsistencyTests>>#testArcCosH
        FloatConsistencyTests>>#testArcSin
        FloatConsistencyTests>>#testArcSinH
        FloatConsistencyTests>>#testArcTan
        FloatConsistencyTests>>#testArcTan2
        FloatConsistencyTests>>#testArcTanH
        FloatConsistencyTests>>#testCos
        FloatConsistencyTests>>#testCosH
        FloatConsistencyTests>>#testDivArray
        FloatConsistencyTests>>#testDot
        FloatConsistencyTests>>#testExp
        FloatConsistencyTests>>#testFloatAt
        FloatConsistencyTests>>#testFraction
        FloatConsistencyTests>>#testHypot
        FloatConsistencyTests>>#testLog
        FloatConsistencyTests>>#testLog10
        FloatConsistencyTests>>#testMulArray
        FloatConsistencyTests>>#testSin
        FloatConsistencyTests>>#testSinH
        FloatConsistencyTests>>#testSqrt
        FloatConsistencyTests>>#testSubArray
        FloatConsistencyTests>>#testTan
        FloatConsistencyTests>>#testTanH
        FloatConsistencyTests>>#testTimesTwoPower
        FloatTest>>#testFractionAsFloat
        LargePositiveIntegerTest>>#testReciprocalModulo
        RandomTest>>#testNext
        SqNumberParserTest>>#testFloatPrintString
Test failures:
        FloatConsistencyTests>>#testAddScalar
        FloatConsistencyTests>>#testArcCosHStd
        FloatConsistencyTests>>#testArcCosStd
        FloatConsistencyTests>>#testArcSinHStd
        FloatConsistencyTests>>#testArcSinStd
        FloatConsistencyTests>>#testArcTanHStd
        FloatConsistencyTests>>#testArcTanStd
        FloatConsistencyTests>>#testCosHStd
        FloatConsistencyTests>>#testCosStd
        FloatConsistencyTests>>#testDivScalar
        FloatConsistencyTests>>#testExpStd
        FloatConsistencyTests>>#testFractionStd
        FloatConsistencyTests>>#testLength
        FloatConsistencyTests>>#testMulScalar
        FloatConsistencyTests>>#testNormalize
        FloatConsistencyTests>>#testSinHStd
        FloatConsistencyTests>>#testSinStd
        FloatConsistencyTests>>#testSubScalar
        FloatConsistencyTests>>#testSum
        FloatConsistencyTests>>#testTanHStd
        FloatConsistencyTests>>#testTanStd
        FloatTest>>#testArcTan
        FloatTest>>#testCharacterization
        FloatTest>>#testComparison
        FloatTest>>#testDegreeCos
        FloatTest>>#testDegreeSin
        FloatTest>>#testDegreeSinForExceptionalValues
        FloatTest>>#testFloatRounded
        FloatTest>>#testFloatTruncated
        FloatTest>>#testFractionAsFloat2
        FloatTest>>#testInfinity1
        FloatTest>>#testInfinity2
        FloatTest>>#testIntegerAsFloat
        FloatTest>>#testNonSignalingNaN
        FloatTest>>#testSignalingNaN
        FloatTest>>#testStoreBase16
        FloatTest>>#testStoreOn
        FractionTest>>#testDegreeCos
        FractionTest>>#testDegreeSin
        FractionTest>>#testLn
        FractionTest>>#testLog
        IntegerTest>>#testDegreeCos
        IntegerTest>>#testDegreeSin
        IntegerTest>>#testLn
        IntegerTest>>#testLog
        IntegerTest>>#testSqrtFloor
        NumberParsingTest>>#testFloatFromStreamWithExponent
        NumberParsingTest>>#testFloatFromStringWithExponent
        NumberTest>>#testReadFrom
        SqNumberParserTest>>#testFloatFromStreamWithExponent
        SqNumberParserTest>>#testFloatGradualUnderflow
Tests passed:
        RandomTest>>#testClassComment
        RandomTest>>#testCoverage
        RandomTest>>#testNew
        RandomTest>>#testUnCategorizedMethods
        ExtendedNumberParserTest>>#testClassComment
        ExtendedNumberParserTest>>#testCoverage
        ExtendedNumberParserTest>>#testFractionPartWithoutIntegerPart
        ExtendedNumberParserTest>>#testIntegerPartWithoutFraction
        ExtendedNumberParserTest>>#testInvalidExponent
        ExtendedNumberParserTest>>#testInvalidRadix
        ExtendedNumberParserTest>>#testInvalidScale
        ExtendedNumberParserTest>>#testNew
        ExtendedNumberParserTest>>#testPositive
        ExtendedNumberParserTest>>#testPositiveExponent
        ExtendedNumberParserTest>>#testUnCategorizedMethods
        NumberTest>>#testClassComment
        NumberTest>>#testCoverage
        NumberTest>>#testFractionPart
        NumberTest>>#testIntegerPart
        NumberTest>>#testNew
        NumberTest>>#testOne
        NumberTest>>#testPrintShowingDecimalPlaces
        NumberTest>>#testPrintShowingDecimalPlaces2
        NumberTest>>#testPrintShowingDecimalPlaces3
        NumberTest>>#testRaisedTo
        NumberTest>>#testRaisedToInteger
        NumberTest>>#testRaisedToIntegerWithFloats
        NumberTest>>#testReciprocal
        NumberTest>>#testUnCategorizedMethods
        FloatConsistencyTests>>#testMD5
        LargePositiveIntegerTest>>#testBitShift
        LargePositiveIntegerTest>>#testClassComment
        LargePositiveIntegerTest>>#testCompactClassIndex
        LargePositiveIntegerTest>>#testCoverage
        LargePositiveIntegerTest>>#testDenormalizedPrintString
        LargePositiveIntegerTest>>#testEmptyTemplate
        LargePositiveIntegerTest>>#testMultDicAddSub
        LargePositiveIntegerTest>>#testNew
        LargePositiveIntegerTest>>#testNormalize
        LargePositiveIntegerTest>>#testUnCategorizedMethods
        LargeNegativeIntegerTest>>#testClassComment
        LargeNegativeIntegerTest>>#testCompactClassIndex
        LargeNegativeIntegerTest>>#testCoverage
        LargeNegativeIntegerTest>>#testDenormalizedPrintString
        LargeNegativeIntegerTest>>#testEmptyTemplate
        LargeNegativeIntegerTest>>#testNew
        LargeNegativeIntegerTest>>#testUnCategorizedMethods
        SmallIntegerTest>>#testBasicNew
        SmallIntegerTest>>#testClassComment
        SmallIntegerTest>>#testCoverage
        SmallIntegerTest>>#testDivide
        SmallIntegerTest>>#testEven
        SmallIntegerTest>>#testMaxVal
        SmallIntegerTest>>#testMinVal
        SmallIntegerTest>>#testNew
        SmallIntegerTest>>#testOdd
        SmallIntegerTest>>#testPrintPaddedWith
        SmallIntegerTest>>#testPrintString
        SmallIntegerTest>>#testUnCategorizedMethods
        IntegerDigitLogicTest>>#testAndSingleBitWithMinusOne
        IntegerDigitLogicTest>>#testLargeShift
        IntegerDigitLogicTest>>#testMixedSignDigitLogic
        IntegerDigitLogicTest>>#testNBitAndNNegatedEqualsN
        IntegerDigitLogicTest>>#testNNegatedEqualsNComplementedPlusOne
        IntegerDigitLogicTest>>#testShiftMinusOne1LeftThenRight
        IntegerDigitLogicTest>>#testShiftOneLeftThenRight
        IntegerTest>>#testBenchFib
        IntegerTest>>#testBitAt
        IntegerTest>>#testBitLogic
        IntegerTest>>#testCreationFromBytes1
        IntegerTest>>#testCreationFromBytes2
        IntegerTest>>#testCreationFromBytes3
        IntegerTest>>#testCrossSumBase
        IntegerTest>>#testDifferentBases
        IntegerTest>>#testEven
        IntegerTest>>#testHighBit
        IntegerTest>>#testHighBitOfMagnitude
        IntegerTest>>#testIntegerByteEncoded
        IntegerTest>>#testIntegerPadding
        IntegerTest>>#testIntegerReadFrom
        IntegerTest>>#testIsInteger
        IntegerTest>>#testIsPowerOfTwo
        IntegerTest>>#testIsPowerOfTwoM6873
        IntegerTest>>#testIsPrime
        IntegerTest>>#testIsProbablyPrime
        IntegerTest>>#testLargePrimesUpTo
        IntegerTest>>#testLowBit
        IntegerTest>>#testNegativeIntegerPrinting
        IntegerTest>>#testNew
        IntegerTest>>#testNumberOfDigits
        IntegerTest>>#testPositiveIntegerPrinting
        IntegerTest>>#testPrimesUpTo
        IntegerTest>>#testPrintOnBaseShowRadix
        IntegerTest>>#testPrintStringBase
        IntegerTest>>#testReadFrom
        IntegerTest>>#testReciprocalModulo
        IntegerTest>>#testRomanPrinting
        IntegerTest>>#testStringAsNumber
        IntegerTest>>#testTwoComplementBitLogicWithCarry
        IntegerTest>>#testTwoComplementRightShift
        SqNumberParserTest>>#testClassComment
        SqNumberParserTest>>#testCoverage
        SqNumberParserTest>>#testFail
        SqNumberParserTest>>#testFloatFromStreamAsNumber
        SqNumberParserTest>>#testFloatReadError
        SqNumberParserTest>>#testFloatReadWithRadix
        SqNumberParserTest>>#testIntegerReadFrom
        SqNumberParserTest>>#testIntegerReadWithRadix
        SqNumberParserTest>>#testNew
        SqNumberParserTest>>#testScaledDecimalWithTrailingZeroes
        SqNumberParserTest>>#testUnCategorizedMethods
        ScaledDecimalTest>>#testAsNumber
        ScaledDecimalTest>>#testAsNumberNegatedWithoutDecimalPoint
        ScaledDecimalTest>>#testAsNumberNegatedWithoutDecimalPoint2
        ScaledDecimalTest>>#testAsNumberWithExtendedScale
        ScaledDecimalTest>>#testAsNumberWithRadix
        ScaledDecimalTest>>#testAsNumberWithSuperfluousDecimalPoint
        ScaledDecimalTest>>#testAsNumberWithoutDecimalPoint
        ScaledDecimalTest>>#testAsNumberWithoutDecimalPoint2
        ScaledDecimalTest>>#testClassComment
        ScaledDecimalTest>>#testCoercion
        ScaledDecimalTest>>#testConvertFromFloat
        ScaledDecimalTest>>#testConvertFromFraction
        ScaledDecimalTest>>#testConvertFromFractionWithScale
        ScaledDecimalTest>>#testConvertFromInteger
        ScaledDecimalTest>>#testConvertFromIntegerWithScale
        ScaledDecimalTest>>#testCoverage
        ScaledDecimalTest>>#testIsLiteral
        ScaledDecimalTest>>#testLiteral
        ScaledDecimalTest>>#testNew
        ScaledDecimalTest>>#testOneRaisedToInteger
        ScaledDecimalTest>>#testPrintString
        ScaledDecimalTest>>#testRaisedToInteger
        ScaledDecimalTest>>#testReadFrom
        ScaledDecimalTest>>#testScaleExtension
        ScaledDecimalTest>>#testStoreOn
        ScaledDecimalTest>>#testUnCategorizedMethods
        ScaledDecimalTest>>#testZeroRaisedToInteger
        FractionTest>>#testClassComment
        FractionTest>>#testCoverage
        FractionTest>>#testFractionPrinting
        FractionTest>>#testNew
        FractionTest>>#testReciprocal
        FractionTest>>#testUnCategorizedMethods
        NumberParsingTest>>#testFloatFromStreamAsNumber
        NumberParsingTest>>#testFloatFromStringAsNumber
        NumberParsingTest>>#testFloatReadWithRadix
        NumberParsingTest>>#testIntegerFromString
        NumberParsingTest>>#testIntegerReadFrom
        NumberParsingTest>>#testIntegerReadWithRadix
        NumberParsingTest>>#testNegativeZero
        NumberParsingTest>>#testScaledDecimalWithTrailingZeroes
        FloatTest>>#test32bitGradualUnderflow
        FloatTest>>#test32bitRoundingMode
        FloatTest>>#testClassComment
        FloatTest>>#testCloseTo
        FloatTest>>#testComparisonWhenPrimitiveFails
        FloatTest>>#testContinuedFractions
        FloatTest>>#testCopy
        FloatTest>>#testCoverage
        FloatTest>>#testDegreeCosForExceptionalValues
        FloatTest>>#testDivide
        FloatTest>>#testHugeIntegerCloseTo
        FloatTest>>#testInfinity3
        FloatTest>>#testInfinityCloseTo
        FloatTest>>#testIsZero
        FloatTest>>#testNaN1
        FloatTest>>#testNaN2
        FloatTest>>#testNaN3
        FloatTest>>#testNaN4
        FloatTest>>#testNaN5
        FloatTest>>#testNaNCompare
        FloatTest>>#testNaNisLiteral
        FloatTest>>#testNegativeZeroAbs
        FloatTest>>#testNegativeZeroSign
        FloatTest>>#testNew
        FloatTest>>#testRaisedTo
        FloatTest>>#testReadFromManyDigits
        FloatTest>>#testReciprocal
        FloatTest>>#testSetOfFloat
        FloatTest>>#testSign
        FloatTest>>#testStringAsNumber
        FloatTest>>#testUnCategorizedMethods
        FloatTest>>#testZero1
        FloatTest>>#testZero2
        FloatTest>>#testZeroSignificandAsInteger

 

 
2010/12/29 David T. Lewis <[hidden email]>:
Yes, they differ for NaN...
for fdlibm, NaN fractionPart is -/+ 0
for primitive 52, it is NaN

I'm a bit surprised that the FloatMathPlugin primitive doesn't fail in
case of NaN...
What is the use of SignalNaN class var then ?

> But more importantly, I'm afraid that the FloatMathPlugin is
> completely broken when compiled for 64 bits. I reverted the
> Float>>fractionPart method to use primitive 52 in order to be
> able to run the image, then ran the full suite of KernelTests-Numbers
> tests. The results are not encouraging:
>

Invalid C99 definition of __HI() __LO() macros and co ?

Nicolas