[ODE] linux amd64 compile error - 0.6-rc1

[Bram Stolk]

Chris, and others,


This is bad news... we need to figure this one out for 0.6

Chris, I suspect your crash is not related to the asm code for
serialize().

Could you please see if drawstuff actually works for you?

$ cd drawstuff/dstest
$ ./dstest

If that one crashes as well, it may be something related to OpenGL.
Do you have permision to open windows on you x11 server?


About the missing dMassSetCapsule :
This is really weird.. the symbol is defined in mass.cpp
unconditionally.
And mass.cpp is used unconditionally in ode/src/Makefile.am

About the FAILs in ode_test : I do not worry much about those.
As I understand it, that file measures the difference between
the results from optimized code from that of slow but more
mathimatically correct code.
ode_test fails for me as well.

   Bram



Chris M. wrote:
> 
> I added a % to push and pop to get this
> 
> 
>  asm volatile (
>     "mov $0,%%rax\n"
>     "push %%rbx\n"
>     "cpuid\n"
>     "pop %%rbx\n"
>     : : : "%rax","%rcx","%rdx","cc","memory");
> 
> and it compiled , however the test programs that I have run all crash. 
> test_ode gives this
> testRandomNumberGenerator()
>        passed
> testInfinity()
>        passed
> testPad()
>        passed
> testCrossProduct()
>        passed
> testSetZero()
>        FAILED
> testNormalize3()
>        FAILED (code=3)
> testPlaneSpace()
>        FAILED
> testMatrixMultiply()
>        FAILED (1)
>        FAILED (2)
>        FAILED (3)
> testSmallMatrixMultiply()
>        passed (1)
>        passed (2)
>        passed (3)
>        passed (4)
>        passed (5)
>        passed (6)
> testCholeskyFactorization()
>        FAILED (1)
>        maximum difference = 0.000000e+00 - passed (2)
> testCholeskySolve()
>        FAILED (1)
>        maximum difference = 0.000000e+00 - passed (2)
> testInvertPDMatrix()
>        FAILED (1)
>        FAILED (2)
> testIsPositiveDefinite()
>        passed (1)
>        FAILED (2)
> testFastLDLTFactorization()
>        maximum difference = 5.336397e-315 - passed
> testSolveLDLT()
>        maximum difference = 0.000000e+00 - passed
> testLDLTAddTL()
>        maximum difference = 5.336397e-315 - passed
> testLDLTRemove()
>        row/col removal FAILED for row 0
>        row/col removal FAILED for row 1
>        row/col removal FAILED for row 2
>        row/col removal FAILED for row 3
>        row/col removal FAILED for row 4
>        row/col removal FAILED for row 5
>        row/col removal FAILED for row 6
>        row/col removal FAILED for row 7
>        row/col removal FAILED for row 8
>        row/col removal FAILED for row 9
>        row/col removal FAILED for row 10
>        row/col removal FAILED for row 11
>        row/col removal FAILED for row 12
>        row/col removal FAILED for row 13
>        row/col removal FAILED for row 14
>        row/col removal FAILED for row 15
>        row/col removal FAILED for row 16
>        row/col removal FAILED for row 17
>        row/col removal FAILED for row 18
>        row/col removal FAILED for row 19
>        maximum difference = 5.336397e-315 - passed
> testMassFunctions()
>        (Message 2: mass must be > 0 in checkMass()) passed (1)
>        (Message 2: mass must be > 0 in checkMass()) FAILED (2)
>        FAILED (3)
> 
> ODE Message 2: mass must be > 0 in checkMass()
>        FAILED (4)
> ./test_ode: symbol lookup error: ./test_ode: undefined symbol: 
> dMassSetCapsule
> 
> ------------------
> 
> This is with double precision.
> If I rebuild without double precision some things do run and test_ode 
> gives this
> 
> -----------------------
> 
> testRandomNumberGenerator()
>        passed
> testInfinity()
>        passed
> testPad()
>        passed
> testCrossProduct()
>        passed
> testSetZero()
>        passed
> testNormalize3()
>        passed
> testPlaneSpace()
>        passed
> testMatrixMultiply()
>        passed (1)
>        passed (2)
>        passed (3)
> testSmallMatrixMultiply()
>        passed (1)
>        passed (2)
>        passed (3)
>        passed (4)
>        passed (5)
>        passed (6)
> testCholeskyFactorization()
>        passed (1)
>        maximum difference = 9.536743e-07 - passed (2)
> testCholeskySolve()
>        passed (1)
>        maximum difference = 3.325939e-05 - FAILED (2)
> testInvertPDMatrix()
>        passed (1)
>        FAILED (2)
> testIsPositiveDefinite()
>        passed (1)
>        passed (2)
> testFastLDLTFactorization()
>        maximum difference = 4.768372e-07 - passed
> testSolveLDLT()
>        maximum difference = 2.095699e-04 - FAILED
> testLDLTAddTL()
>        maximum difference = 9.536743e-07 - passed
> testLDLTRemove()
>        maximum difference = 4.768372e-07 - passed
> testMassFunctions()
>        (Message 2: inertia must be positive definite in checkMass()) 
> passed (1)
>         passed (2)
>        passed (3)
>        passed (4)
> ./test_ode: symbol lookup error: ./test_ode: undefined symbol: 
> dMassSetCapsule
> 
> 
> 
>> From: Bram Stolk <bram at sara.nl>
>> To: "Chris M." <czb182 at hotmail.com>
>> CC: ode at q12.org
>> Subject: Re: [ODE] linux amd64 compile error - 0.6-rc1
>> Date: Thu, 25 May 2006 10:22:41 +0200
>>
>> Chris M. wrote:
>>
>>> Hello, I just downloaded the 0.6-rc1 to give it a test of my gentoo 
>>> linux athlon64 box.
>>> Im getting some compilation errors.  I'm wasn't sure where to file a 
>>> bug report, hopefully this is fine.
>>
>>
>> This is the perfect place, we need to iron out matters like these for
>> 0.6, just here.
>>
>>> timer.cpp: In function `void dStopwatchStart(dStopwatch*)':
>>> timer.cpp:139: error: invalid `asm': operand number missing after 
>>> %-letter
>>> timer.cpp:139: error: invalid `asm': operand number missing after 
>>> %-letter
>>
>>
>> The x64 code looks suspect to me:
>>   asm volatile (
>>         "mov $0,%%rax\n"
>>         "push %rbx\n"
>>         "cpuid\n"
>>         "pop %rbx\n"
>>         : : : "%rax","%rcx","%rdx","cc","memory");
>>
>> Why is rbx prefixed with one % char, and rax with two?
>>
>> Could you please try adding a '%' to the push and to the pop command?
>>
>> Thanks you for testing,
>>
>>    Bram
>>
>>
>>>
>>> This is with gcc 3.4.5
>>>
>>>
>>> _______________________________________________
>>> ODE mailing list
>>> ODE at q12.org
>>> http://q12.org/mailman/listinfo/ode
>>
>>
> 
>