Wrong result for complex matrix multiplication on some Windows systems

[jasax]

The issue appears in Windows7-64 bit, both in 0.2.1 and 3.0 dev Julia versions. below I send only the log from 0.2.1 which is more readable. It is observed in several circuits (its related with a basic circuit simulator).

I'm implementing a linear AC circuit solver using complex matrices and vectors.
When solving a resistive circuit, all the data and the solution are real. However, in one example I got (surprisingly) a complex solution. That triggered a detailed analysis of the results.

The workflow is:
I want to solve M*X=b for X in complex values.
We have a Complex{Float64} matrix M=MR+J*W*MI where MR and MI are Float64 matrices, J=sqrt(-1) and W is a real frequency value.
In the example below, however, MI is zero and thus M, X and b should all be real.

The observed error (or misbehavior) is the following:

I factorize M with lu() such that L,U,p=lu(M)
Then I calculate, using regular matrix multiplication, LU=L*U

On the other hand I build a triple loop to calculate the product L*U element by element which leads to the LoopLU matrix (code below)...

LoopLU=zeros(Complex{Float64},N,N)
# Product of L*U is looped
for i in 1:N
    for j in 1:N
        s=0.0+0.0im
        for k in 1:N s += L[i,k]*U[k,j] end
        LoopLU[i,j]=s
    end
end

Then, the difference LU-LoopLU should be zero, but it is not!!! The elements [1,5] and [2,5] are approximately 0+0.3im

Difference LoopLU-L*U

0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.30000000000000004im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 - 0.30000000000000004im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im

By examinng the log in the end, it is seen that despite L and U being real matrices its product is complex!!! ( elements [1,5] and [2,5] again)

L*U
0.001 + 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im -1.0 - 0.30000000000000004im
0.0 + 0.0im 0.00030303030303030303 + 0.0im  0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.30000000000000004im
0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
-0.0002 + 0.0im 0.0 + 0.0im 0.0002 + 0.0im  1.0 + 0.0im 0.0 + 0.0im
0.00030000000000000003 + 0.0im  0.0 + 0.0im -0.0002 + 0.0im 0.0 + 0.0im 0.0 + 0.0im

REMARKS

• the error is the same in Julia 0.2.1 and 0.3-dev, in windows-7 64 bits
• the error occurs in many circuits
• creating random completely filled Float64 matrices and factorizing, the error doesn't show.
• creating random completely filled Complex{Float64 matrices and factorizing, * the error shows up!!!!!!!*
• It seems that the error is somewhere in the routines for performing product of complex numbers.
• TEST PROGRAM WITH RANDOM MATRICES BELOW: HIGHLIGHTS THE ERROR

## RANDOM MATRICES HIGHLIGHTING THE ERROR

###### COMPLETE PROGRAM

N=4
M=rand(Float64,N,N)  
L,U,p=lu(M)
LU=L*U
println(p)

LoopLU=reshape(zeros(Float64,N*N),N,N)
# Product of L*U is looped
for i in 1:N
    for j in 1:N
    s=0.0+0.0im
    for k in 1:N s += L[i,k]*U[k,j] end
    LoopLU[i,j]=s
    end
end
println(LU-LoopLU)    # IT IS ZERO

MC=zeros(Complex{Float64},N,N) 
MC = rand(Float64,N,N)+(0.0+1.0im)*rand(Float64,N,N)     

L,U,p=lu(MC)
LU=L*U
println(p)
LoopLU=zeros(Complex{Float64},N,N) 
# Product of L*U is looped
for i in 1:N
    for j in 1:N
    s=0.0+0.0im
    for k in 1:N s += L[i,k]*U[k,j] end
    LoopLU[i,j]=s
    end
end
println(LU-LoopLU)    # GIVES != ZERO!!!

## COMPLETE LOG OF CIRCUIT OUTPUT in Julia 0.2.1

circuit has 5 vars

MR
.00030000000000000003   0   -2e-4   0   0
0   .00030303030303030303   0   0   1
-2e-4   0   2e-4    1   0
0   0   1   0   0
.001    0   0   0   -1

MI
0   0   0   0   0
0   0   0   0   0
0   0   0   0   0
0   0   0   0   0
0   0   0   0   0

M
0.00030000000000000003 + 0.0im  0.0 + 0.0im -0.0002 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.00030303030303030303 + 0.0im  0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im
-0.0002 + 0.0im 0.0 + 0.0im 0.0002 + 0.0im  1.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.001 + 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im -1.0 + 0.0im

b
0.0 + 0.0im
0.0 + 0.0im
0.0 + 0.0im
10.0 + 0.0im
0.0 + 0.0im

Results from L,U,p=lu(M)

L
1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
-0.2 + 0.0im    0.0 + 0.0im 0.0002 + 0.0im  1.0 + 0.0im 0.0 + 0.0im
0.30000000000000004 + 0.0im 0.0 + 0.0im -0.0002 + 0.0im 0.0 + 0.0im 1.0 + 0.0im

U
0.001 + 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im -1.0 + 0.0im
0.0 + 0.0im 0.00030303030303030303 + 0.0im  0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im -0.2 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.30000000000000004 + 0.0im

p (reordering vector)
5
2
4
3
1

b[p,1] (reordered b vector to use LU factors)
0.0 + 0.0im
0.0 + 0.0im
10.0 + 0.0im
0.0 + 0.0im
0.0 + 0.0im

invM=inv(M)
3333.333333333333 - 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.6666666666666666 + 0.0im  -1.1368683772161603e-13 + 0.0im
-10999.999999999998 + 0.0im 3300.0 + 0.0im  0.0 + 0.0im -2.1999999999999997 + 0.0im 3300.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im
0.6666666666666666 - 0.0im  0.0 + 0.0im 1.0 + 0.0im -6.666666666666667e-5 + 0.0im   0.0 + 0.0im
3.333333333333333 - 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.0006666666666666666 + 0.0im   -1.0 + 0.0im

L*U
0.001 + 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im -1.0 - 0.30000000000000004im
0.0 + 0.0im 0.00030303030303030303 + 0.0im  0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.30000000000000004im
0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
-0.0002 + 0.0im 0.0 + 0.0im 0.0002 + 0.0im  1.0 + 0.0im 0.0 + 0.0im
0.00030000000000000003 + 0.0im  0.0 + 0.0im -0.0002 + 0.0im 0.0 + 0.0im 0.0 + 0.0im

LoopLU
0.001 + 0.0im   0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im -1.0 + 0.0im
0.0 + 0.0im 0.00030303030303030303 + 0.0im  0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 1.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
-0.0002 + 0.0im 0.0 + 0.0im 0.0002 + 0.0im  1.0 + 0.0im 0.0 + 0.0im
0.00030000000000000003 + 0.0im  0.0 + 0.0im -0.0002 + 0.0im 0.0 + 0.0im 0.0 + 0.0im

Difference LoopLU-L*U

0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.30000000000000004im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 - 0.30000000000000004im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im

XTemp=inv(LoopLU)*b[p,1]
6.666666666666666 + 0.0im
-21.999999999999996 + 0.0im
10.0 + 0.0im
-0.0006666666666666668 + 0.0im
0.006666666666666666 + 0.0im

Xlu=invLU*b[p,1]
6.666666666666666 + 0.0im
-21.999999999999996 + 0.0im
10.0 + 0.0im
-0.0006666666666666668 + 0.0im
0.006116207951070336 - 0.0018348623853211012im

X=inv(M)*b
6.666666666666666 + 0.0im
-21.999999999999996 + 0.0im
10.0 + 0.0im
-0.0006666666666666668 + 0.0im
0.006666666666666666 + 0.0im

MI  Array{Float64,2}
MI  Array{Float64,2}
M  Array{Complex{Float64},2}
L  Array{Complex{Float64},2}
U  Array{Complex{Float64},2}
invM  Array{Complex{Float64},2}
LU=L*U  Array{Complex{Float64},2}
invLU  Array{Complex{Float64},2}
LoopLU  Array{Complex{Float64},2}
b  Array{Complex{Float64},2}
X  Array{Complex{Float64},2}
Xlu  Array{Complex{Float64},1}
Xtemp  Array{Complex{Float64},1}

[andreasnoack]

I cannot reproduce this on my mac or a Windows server. However, it could architecture specific. What is your versioninfo().

[jasax]

Hi Andreas,

Thanks for the quick answer.

I just tried the random matrices program in ubuntu12.10 running in
Virtualbox and the result is ok. That is, LU-LoopLU ~ 0, both with
real and with complex matrices. So this is an architecture dependent
issue by sure.

But below I show the result of running the code in julia 3.0 dev (with
julia 0.2.1 the result is similar) in windows7 64 bits: The complex
subtraction is quite different from 0.

Can it be related with the compiler used in windows (or with cross compiling?)

Best regards

Jose

[3,4,2,1]
[0.0 0.0 0.0 0.0
 0.0 0.0 0.0 0.0
 0.0 0.0 0.0 0.0
 0.0 0.0 0.0 0.0]
[4,2,3,1]
Complex{Float64}[0.0 + 0.0im 0.0 + 0.0im 0.010724319095534562 -
0.00436327616604229im 0.05626670997775374 - 0.04487609382080637im
                 0.0 + 0.0im 0.0 + 1.1102230246251565e-16im
0.00031766699802464327 - 0.0007146125364691867im 0.028237510396673327
- 0.02084902211094586im
                 0.0 + 0.0im 0.0 + 0.0im 0.0719049982965248 -
0.036089242268516175im 0.06891390625263594 - 0.05992628099413577im
                 0.0 + 0.0im -5.551115123125783e-17 +
2.7755575615628914e-17im 0.042979910736967 - 0.016978401082962447im
0.04316866011620124 - 0.020865798633942556im]

#############################################

On 5/29/14, Andreas Noack Jensen [email protected] wrote:

I cannot reproduce this on my mac or a Windows server. However, it could
architecture specific. What is your systeminfo().

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Hello again,

Sorry, I forgot the systeminfo. It goes attached, with a few personal
fields deleted :-)

Yhanks again and best regards.

Jose

On 5/29/14, Andreas Noack Jensen [email protected] wrote:

I cannot reproduce this on my mac or a Windows server. However, it could
architecture specific. What is your systeminfo().

---

Reply to this email directly or view it on GitHub:
#117

[andreasnoack]

Thank you for testing it on Ubuntu. I should have written versioninfo(). It is a function in Julia which prints the version, commit and some info about the platform, but no personal stuff.

On your Windows machine, could you please try to run blas_set_num_threads(1) before you run your test code and see if there is still a problem? I think the problem is in OpenBLAS. There has been Windows specific problems there before.

[jasax]

Hello Andreas,

Before I do what you suggest (I can only do that tomorrow at my
workplace), this is just to inform on another run of the test code.

At home I also have a windows7 machine, running on a AMD A8-3850 APU
with 4 cores.

Just now I've tried the code with the random matrices, and it happens
that the result is OK! This means that both the float64 and the
complex residuals of the matrix subtraction is near 0! Below I show
one result from julia 0.2.1 in this home PC.

The result of versioninfo() here at my home PC is also shown below.
Tomorrow at work I'll run versioninfo() in my work PC and also will
try to run blas_set_num_threads(1) before running the test code, and
see what happens. Then I'll report the findings.

Just to summarize, my home processor has 4 cores, my work processor
has 8 cores. If the problem has to do with multiple threads running in
multiple processors, perhaps that number of processors provides an
answer (matrix multiplication is parallelizable, is open_blas using
this feature in julia?)

Finnally, thank you very much again for your interest and effort.

Jose

---

julia> versioninfo()
Julia Version 0.2.1
Commit e44b593* (2014-02-11 06:30 UTC)
Platform Info:
System: Windows (x86_64-w64-mingw32)
WORD_SIZE: 64
BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY)
LAPACK: libopenblas
LIBM: libopenlibm

---

julia 0.2.1 at home PC ###### runs ok

4
2
1
3

0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0

4
3
2
1

0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 - 5.551115123125783e-17im 0.0 + 0.0im 0.0 + 0.0im
0.0 + 0.0im 0.0 - 1.1102230246251565e-16im 5.551115123125783e-17 +
0.0im 0.0 + 2.7755575615628914e-17im
0.0 + 0.0im -1.1102230246251565e-16 + 0.0im 2.220446049250313e-16 +
0.0im -1.1102230246251565e-16 + 1.1102230246251565e-16im

On 5/29/14, Andreas Noack Jensen [email protected] wrote:

Thank you for testing it on Ubuntu. I should have written versioninfo().
It is a function in Julia which prints the version, commit and some info
about the platform, but no personal stuff.

On your Windows machine, could you please try to run
blas_set_num_threads(1) before you run your test code and see if there is
still a problem? I think the problem is in OpenBLAS. There has been Windows
specific problems there before.

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Hello Andreas,

More info about my workplace Windows 7 machine.

I ran the random-matrix-test with julia021 and julia03-dev after
inserting blas_set_num_threads(1) in the beginning of the code and,
that didn't change the result: that is, the errors are there.

Finally, below are the results of versioninfo() in julia 0.2.1 and
julia 0.3-dev in my work machine.

Kind Regards

Jose

###############

| | || | | | (| | | Version 0.2.1 (2014-02-11 06:30 UTC)
/ |__'|||'_| | Official http://julialang.org/ release
|/ | x86_64-w64-mingw32

julia> versioninfo()
Julia Version 0.2.1
Commit e44b593* (2014-02-11 06:30 UTC)
Platform Info:
System: Windows (x86_64-w64-mingw32)
WORD_SIZE: 64
BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY)
LAPACK: libopenblas
LIBM: libopenlibm

| | |_| | | | (_| |  |  Version 0.3.0-prerelease+3225 (2014-05-27 03:45 UTC)

/ |__'|||'_| | Commit b57777f* (3 days old master)
|/ | x86_64-w64-mingw32

julia> versioninfo()
Julia Version 0.3.0-prerelease+3225
Commit b57777f* (2014-05-27 03:45 UTC)
Platform Info:
System: Windows (x86_64-w64-mingw32)
CPU: AMD FX(tm)-8320 Eight-Core Processor
WORD_SIZE: 64
BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY)
LAPACK: libopenblas
LIBM: libopenlibm

On 5/29/14, Andreas Noack Jensen [email protected] wrote:

Thank you for testing it on Ubuntu. I should have written versioninfo().
It is a function in Julia which prints the version, commit and some info
about the platform, but no personal stuff.

On your Windows machine, could you please try to run
blas_set_num_threads(1) before you run your test code and see if there is
still a problem? I think the problem is in OpenBLAS. There has been Windows
specific problems there before.

---

Reply to this email directly or view it on GitHub:
#117

[andreasnoack]

@jasax Let's keep the discussion of this issue. (I have updated the name). Are you able to compile fortran programs on your work Windows machine?

[jasax]

Hi Andreas,

I never tried compiling fortran. But I have mingw installed in my win7
machines (e.g. I usually successfully compile recent Lua and Ruby from
sources) and I can install the gcc-fortran pack (if its not there).
Now I am at home (weekend...) and I can test the fortran gcc front-end
here. Tomorrow I'll test in the AMD 8-core. So, please tell me what to
compile. Eventually I can try other fortran compiler (G95, Watcom?).

In fact I learned programming using Basic and Fortran in a IBM 360
around 1979 in my college years :-) but since then I don't remember
ever using it seriously... no more spaghetti GOTOs and start writing
code at column 7 since then ;-)

I still am a bit literate in Fortran basics; I think I'll be able to
do a few modifications on sources if needed. So plese tell me what to
compile.

Best regards

Jose

On 6/1/14, Andreas Noack Jensen [email protected] wrote:

@jasax Let's keep the discussion of this issue. (I have updated the name).
Are you able to compile fortran programs on your work Windows machine?

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Keeping up...

Did use gemm() from BLAS and gave exactly the same errors (I think the matrix product in julia recurs to this gemm() function, isn't it?)

using Base.LinAlg.BLAS
...
GemmLU=gemm('N','N',L,U)
println("LU-GemmLU complex\n",LU-GemmLU) # => gives ~[0] matrix
println("GemmLU-LoopLU complex\n",GemmLU-LoopLU) # => gives matrix with errors

Meanwhile tried to use a vectorized LoopLU using dotu() in the test program with random matrices;

DotLU=zeros(Complex{Float64},N,N)
for i in 1:N
for j in 1:N
DotLU[i,j]=dotu(N,L[i,:],1,U[:,j],1)
end
end

and it gives the following error (doesn't find dotu()... but is in julia's manual...):

ERROR: dotu not defined
in anonymous at no file:50

The most bizarre thing happened when I replaced in the above loop dotu() (dot product for complex numbers) by dot() (used for float vectors). With dot() the program ends gracefully in the loop without finishing the execution of the remaining code and without throwing any error or exception B-(

[tkelman]

I think where @andreasnoackjensen was going is, can you reproduce the incorrect results just using gemm from a small Fortran example? Then you can report this bug upstream to OpenBLAS's issue tracker.

[andreasnoack]

Please try to compile the program below. You can do so by saving the program in the file test.f90 in a folder together with a copy of libopenblas.dll from julia\bin and the run the commands

gfortran test.f90 -L . -lopenblas
a.exe

program test

    complex*16 :: a(4,4), b(4,4), c(4,4)

    a(:,:) = (1.0d0, 1.0d0)
    b(:,:) = 1.0d0

    call zgemm('N', 'N', 4_8, 4_8, 4_8, (1.0d0, 0.0d0), a, 4_8, b, 4_8, (0.0d0, 0.0d0), c, 4_8)

    write(*,*) c

end program

[jasax]

Hi,

I used this gfortran version (most recent TDM-mingw)

$ gfortran --version
GNU Fortran (tdm64-2) 4.8.1
Copyright (C) 2013 Free Software Foundation, Inc.

Here is the result. It seems OK. I used the libopenblas.dll both from
julia 021 and 03 dev and got the same result.

If wnat to test more code, please send it.

Best regards

Jose

###########

$ a.exe
( 0.0000000000000000 , 8.0000000000000000 ) ( 0.0000000000000000
, 8.0000000000000000 ) ( 0.0000000000000000 , 8.0000000000000000
) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.0000000000000000
, 8.0000000000000000 ) ( 0.0000000000000000 , 8.0000000000000000
) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.0000000000000000
, 8.0000000000000000 ) ( 0.0000000000000000 , 8.0000000000000000
) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.00000000000000
00 , 8.0000000000000000 ) ( 0.0000000000000000 , 8.00000000000000
00 ) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.000000000000
0000 , 8.0000000000000000 ) ( 0.0000000000000000 , 8.000000000000
0000 ) ( 0.0000000000000000 , 8.0000000000000000 )

On 6/3/14, Andreas Noack Jensen [email protected] wrote:

Please try to compile the program below. You can do so by saving the program
in the file test.f90 in a folder together with a copy of libopenblas.dll
from julia\bin and the run the commands

gfortran test.f90 -L . -lopenblas
a.exe

program test

    complex*16 :: a(4,4), b(4,4), c(4,4)

    a(:,:) = (1.0d0, 1.0d0)
    b(:,:) = 1.0d0

    call zgemm('N', 'N', 4_8, 4_8, 4_8, (1.0d0, 0.0d0), a, 4_8, b, 4_8,
(0.0d0, 0.0d0), c, 4_8)

  write(*,*) c

end program

---

Reply to this email directly or view it on GitHub:
#117

[andreasnoack]

Thanks. The right result is (4.0d0, 4.0d0) for all entries so it appears that I have an example for the bug report. @tkelman Is it possible for you to provide a build of the latest OpenBLAS. Just in case that they have solved the issue already.

[ViralBShah]

What output did you get for test.f90?

[andreasnoack]

Me? I got

(  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     ) (  4.0000000000000000     ,  4.0000000000000000     )

[tkelman]

I've got one from 4-24 here http://sourceforge.net/projects/juliadeps-win/files/openblas-47b22763f8ab0219-x86_64-w64-mingw32.7z/download

I can build a newer one, sure, give me half an hour or so

[andreasnoack]

There has been some changes to OpenBLAS lately, so it would be great if you could make a new one.

[jasax]

Hello,

It seems I didn't guess the correct result. In fact I didn't look with
care for the matrix values :-)

If you think its useful, I can try the fortran code in my home machine
(around 00h WEST time, I go home late...) where julia didn't show the
errors in complex matrix product.

Meanwhile, in my quest to get complex algebra working in julia in
Win7-64, I turned to sparse matrices and lufact(). Apparently I
uncovered another glitch, which is the nonexistence of a method:
"no method umf_extract(UmfpackLU{Complex{Float64},Int64},)"

Below is the test code and result. Shall I post a separate entry in
the julia-devs list?

Best Regards

Jose

################

N=4
J=0.0+1.0im

MR = rand(Float64,N,N)
MI = rand(Float64,N,N)

SpMR=sparse(MR)
SpMI=sparse(MI)
SpM=SpMR+J*SpMI
println("Sparse M \n",typeof(SpM))
println(SpM)

F=lufact(SpM)
println("F \n",typeof(F),"\n")

println(F[:L])
println(F[:U])
println(F[:p])
println(F[:q])
println(F[:Rs])

#############################

C:\Users...>julia bug4.jl
Sparse M
SparseMatrixCSC{Complex{Float64},Int64}
0.17324610206636093 + 0.556739526277584im 0.3636155270578807 + 0.779898130
0944341im 0.9462204048595892 + 0.0816916240409733im 0.32368934936414
506 + 0.7481843025943526im
0.9912417559106097 + 0.0051733113483452im 0.23271082818867894 + 0.50533292
47279121im 0.16763856188783266 + 0.07775290297860527im 0.73970414046621
24 + 0.243096819167115im
0.783090390333812 + 0.5863584102559694im 0.2521944895936421 + 0.201648277
29072243im 0.8912116317502279 + 0.13636989345048156im 0.27938051396603
88 + 0.4982856834741676im
0.4573921542170194 + 0.4777090518686091im 0.6704835515875494 + 0.388381970
5387357im 0.004182163343978029 + 0.7094696885852512im 0.76423637986060
87 + 0.12410491260558443im

F
UmfpackLU{Complex{Float64},Int64}

ERROR: no method umf_extract(UmfpackLU{Complex{Float64},Int64},)
in getindex at linalg/umfpack.jl:285
WARNING: backtraces on your platform are often misleading or partially incorrect

################################

On 6/3/14, Andreas Noack Jensen [email protected] wrote:

There has been some changes to OpenBLAS lately, so it would be great if you
could make a new one.

---

Reply to this email directly or view it on GitHub:
#117

[tkelman]

Alright, took more like an hour to build openblas. @jasax try with this dll http://sourceforge.net/projects/juliadeps-win/files/openblas-0ac073fa9415571d-x86_64-w64-mingw32.7z/download

The complex umfpack issue is worth opening a separate issue on. I checked that the same issue occurs as of 0c3de45. Looks like umfpack is capable of handling complex matrices but linalg/umfpack.jl is missing a few methods for them?

[jasax]

Hi all.

The error persists, so it seems... Results below.

I'll fill a new issue about umfpack in the dev mailing list.

Thank you Andreas and Tony.

Jose

JAugusto@THOR /c/Users/.../OPenBLAS_Tests
$ date
Tue Jun 3 18:39:57 GMTDT 2014

$ gfortran test.f90 -L . -lopenblas

$ a.exe
( 0.0000000000000000 , 8.0000000000000000 ) ( 0.0000000000000000
, 8.0000000000000000 ) ( 0.0000000000000000 , 8.0000000000000000
) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.0000000000000000
, 8.0000000000000000 ) ( 0.0000000000000000 , 8.0000000000000000
) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.0000000000000000
, 8.0000000000000000 ) ( 0.0000000000000000 , 8.0000000000000000
) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.00000000000000
00 , 8.0000000000000000 ) ( 0.0000000000000000 , 8.00000000000000
00 ) ( 0.0000000000000000 , 8.0000000000000000 ) ( 0.000000000000
0000 , 8.0000000000000000 ) ( 0.0000000000000000 , 8.000000000000
0000 ) ( 0.0000000000000000 , 8.0000000000000000 )

On 6/3/14, Tony Kelman [email protected] wrote:

Alright, took more like an hour to build openblas. @jasax try with this dll
http://sourceforge.net/projects/juliadeps-win/files/openblas-0ac073fa9415571d-x86_64-w64-mingw32.7z/download

The complex umfpack issue is worth opening a separate issue on. I checked
that the same issue occurs as of 0c3de45. Looks like umfpack is capable of
handling complex matrices but linalg/umfpack.jl is missing a few methods for
them?

---

Reply to this email directly or view it on GitHub:
#117

[tkelman]

Don't send the umfpack issue to the mailing list, open a separate issue here on Github.

[andreasnoack]

@jasax Thank you for the feedback so far. Could you also try this line in Julia?

complex64(complex(ones(4,4),ones(4,4)))*complex64(complex(ones(4,4),zeros(4,4)))

[jasax]

Hi Tony,

I already did. Issue JuliaLang/julia#7098.

Thanks

Jose

On 6/3/14, Tony Kelman [email protected] wrote:

Don't send the umfpack issue to the mailing list, open a separate issue here
on Github.

---

Reply to this email directly or view it on GitHub:
#117

[andreasnoack]

Issue filed: OpenMathLib/OpenBLAS#380

[jasax]

Hi,

Image attached. I think the error persists. (0+8im is the result...)

Best

Jose

On 6/3/14, Andreas Noack Jensen [email protected] wrote:

@jasax Thank you for the feedback so far. Could you also try this line in
Julia?

complex64(complex(ones(4,4),ones(4,4)))*complex64(complex(ones(4,4),zeros(4,4)))

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Hi again,

Run the line in the same machine which has Win7-64bits, but this time
in an updated distro of Ubuntu 14.04 in Virtualbox. Runs OK, it seems.
Photo attached.

Now I have to leave for a few hours... ;-)

Best

Jose

On 6/3/14, Andreas Noack Jensen [email protected] wrote:

@jasax Thank you for the feedback so far. Could you also try this line in
Julia?

complex64(complex(ones(4,4),ones(4,4)))*complex64(complex(ones(4,4),zeros(4,4)))

---

Reply to this email directly or view it on GitHub:
#117

[tkelman]

Just FYI, image attachments to email replies don't come through to Github. You can drag-and-drop images if you use the Github comment interface, or upload the file to somewhere like imgur and include a link in your message.

[jasax]

Hi,

you cannot run all coretypes on one machine.
HASWELL only runs on a machine with haswell processor,
PILEDRIVER only on a machine with piledriver processor,
BULLDOZER only on a machine with piledriver or bulldozer processor,
SANDYBRIDGE on all machines with support for AVX.
OPTERON and OPTERON_SSE3 should only be used with an opteron processor.

I know that Werner ;-)

So my surprise when Intel coretypes were successful while others didn't.
Are these invalid coretypes "falling" for some default type, or
Piledriver emulates them (i.e. their assember instructions) well?

Best regards

Jose

Best regards

Werner

On 7/1/14, wernsaar [email protected] wrote:

On 01.07.2014 18:50, jasax wrote:

Hello,

Running linalg4 tests with all the processors in the list. Here is a
summary:

"Gross" errors: Piledriver, Bulldozer;
Had to stop test with CTRL-C: Dunnington;
"Silent" end of test: Haswell, Opteron, Opteron(SSE), Atom;
SUCCESS: All the remaining cores (even Intel cores ??? such as
Sandybridge, Northwood...)

Complete output below.

Regards

Jose

#################################################

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Unknown

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Piledriver

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 169
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 169
while loading
E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Bulldozer

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 169
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 169
while loading
E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Haswell

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Bobcat

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Sandybridge

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Nano

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Barcelona

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Opteron(SSE3)

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Opteron

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Athlon

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Nehalem

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Dunnington

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4

XXXXXXXXXXXXXXXXXXXXXX CTRL-C XXXXXXXXXXXXXXXXXXXXXXX

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Penryn

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Core2

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Atom

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Banias

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Prescott

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Northwood

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Coppermine

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Katmai

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
relty = Float32, elty = Float32
relty = Float32, elty = Complex{Float32}
relty = Float64, elty = Float64
relty = Float64, elty = Complex{Float64}
relty = BigFloat, elty = BigFloat
relty = BigFloat, elty = Complex{BigFloat}
SUCCESS

#################################################

On 7/1/14, Tony Kelman [email protected] wrote:

Another thing you can try now is setting OPENBLAS_CORETYPE to various
other AMD processor families in this list
https://github.com/xianyi/OpenBLAS/blob/53bfa51ee0a47a8b904fa06a03fb57084e97208f/driver/others/dynamic.c#L292-L314,
see if the tests pass with some alternate setting.

(In the Windows command prompt, you set environment variables by set OPENBLAS_CORETYPE=value

---

Reply to this email directly or view it on GitHub:

#117

Reply to this email directly or view it on GitHub:
#117
Hi,

you cannot run all coretypes on one machine.
HASWELL only runs on a machine with haswell processor,
PILEDRIVER only on a machine with piledriver processor,
BULLDOZER only on a machine with piledriver or bulldozer processor,
SANDYBRIDGE on all machines with support for AVX.
OPTERON and OPTERON_SSE3 should only be used with an opteron processor.

Best regards

Werner

---

Reply to this email directly or view it on GitHub:
#117

[wernsaar]

<ed: snip>

Hi,

if you have compiled for dynamic arch and you run a program, if the
correct processor is not
found ( because it is too new ), there is a fallback in the processor
family, but to very old and
slow processors. By setting OPENBLAS_CORETYPE, you can avoid this
fallback by choosing better
processor. I you set a processor, that is not compatible on your
machine, your program may receive
the signal SIGILL, when an unkown instruction should be executed, and
will normally die.

Best regards
Werner

[andreasnoack]

Is it right to say the the short version of this is that @jasax has Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels are used?

Our problem is to figure out which call is actually causing the error because the original problem of this thread is solved, right?

I have pushed some debug options to linalg4.jl. @jasax Can you try to build from latest source and run the linalg tests with debug=true in the top of linalg4.jl?

[wernsaar]

On 02.07.2014 13:12, Andreas Noack Jensen wrote:

Is it right to say the the short version of this is that @jasax has Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels are used?

Our problem is to figure out which call is actually causing the error because the original problem of this thread is solved, right?

I have pushed some debug options to linalg4.jl. @jasax Can you try to build from latest source and run the linalg tests with debug=true in the top of linalg4.jl?

---

Reply to this email directly or view it on GitHub:
#117
Hi,

I think, that it's possible, that the operating system has no full
support for the instruction set and the
application gets a SIGILL signal or a segment violation. Could you try
at first Bulldozer on this
machine and then Barcelona.

Best regards
Werner

[tkelman]

Is it right to say the the short version of this is that @jasax has Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels are used?

Yep that's how I see it.

Our problem is to figure out which call is actually causing the error because the original problem of this thread is solved, right?

Right, the recent fixes to OpenBLAS have solved the vast majority of the problems, there are just a couple stragglers left that it'd be nice to pin down.

I have pushed some debug options to linalg4.jl.

👍 thanks those look good

Could you try at first Bulldozer on this machine and then Barcelona.

From @jasax's results above, Bulldozer appeared to have the same failure as Piledriver, Barcelona passed.

[jasax]

Hi,

Only tomorrow I'll be near the machine. Here are some answers to
previous questions.

First thing, I don't compile from sources (I tried a week ago or so,
but mingw in windows 7 wasn't able to compile). What I'll do is to
re-install the most recent 0.3 dev 64 build for windows. I did this a
couple of days ago: the coretype options were tested with the build
from 29/6 (I think...).

I recall that Piledriver and Bulldozer had test errors; Barcelona
passed the linalg4 tests (as well as several other cores). I will
re-test with these options ASAP.

I suppose the new linalg4.jl with debug options will be already
available in the windows binary next morning (otherwise I'll get it
directly from the sources).

Finally, I started this issue a few weeks ago after seeing errors in
complex matrix product; the last 0.3 dev windows binary didn't have
those issues anymore, the matrix product gave correct results.

Regards

Jose

On 7/2/14, Tony Kelman [email protected] wrote:

Is it right to say the the short version of this is that @jasax has
Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels
are used?

Yep that's how I see it.

Our problem is to figure out which call is actually causing the error
because the original problem of this thread is solved, right?

Right, the recent fixes to OpenBLAS have solved the vast majority of the
problems, there are just a couple stragglers left that it'd be nice to pin
down.

I have pushed some debug options to linalg4.jl.

👍 thanks those look good

Could you try at first Bulldozer on this machine and then Barcelona.

From @jasax's results above, Bulldozer appeared to have the same failure as
Piledriver, Barcelona passed.

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Hello,

Results of linalg4.jl tests with debug=true for Piledriver, Bulldozer
and Barcelona.

Success only with Barcelona.

Piledriver and Bulldozer crash in the same place: Bidiagonal matrices,
Eigensystems.

Results below.

Please indicate further tests/debugs.

Kind Regards

Jose

##############

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

E:\math\Julia-0.3.0-prerelease\share\julia\test>ls
Makefile core.jl linalg1.jl random.jl
sorting.jl
TestHelpers.jl dsp.jl linalg2.jl ranges.jl
sparse.jl
arpack.jl euler.jl linalg3.jl readdlm.jl spawn.jl
arrayops.jl examples.jl linalg4.jl reduce.jl
statistics.jl
arrayperf.jl fft.jl lineedit.jl reducedim.jl
strings.jl
backtrace.jl file.jl math.jl regex.jl
suitesparse.jl
bigint.jl float16.jl mod2pi.jl remote.jl
sysinfo.jl
bitarray.jl floatapprox.jl mpfr.jl repl.jl test.jl
blas.jl functional.jl netload replcompletions.jl
test_sourcepath.jl
broadcast.jl git.jl numbers.jl resolve.jl
testdefs.jl
ccall.jl gitutils.jl parallel.jl rounding.jl
unicode.jl
ccalltest.c goto.jl perf runtests.jl
version.jl
collections.jl hashing.jl pkg.jl show.jl
combinatorics.jl iobuffer.jl pollfd.jl simdloop.jl
complex.jl keywordargs.jl priorityqueue.jl socket.jl

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3
_
_ _ ()_ | A fresh approach to technical computing
() | () () | Documentation: http://docs.julialang.org
_ _ | | __ _ | Type "help()" to list help topics
| | | | | | |/ ` | |
| | || | | | (| | | Version 0.3.0-prerelease+4028 (2014-07-02 23:42 UTC)
/ |_'|||__'| | Commit 2185bd1 (0 days old master)
|__/ | x86_64-w64-mingw32

julia> quit()

E:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Piledriver

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Bulldozer

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Barcelona

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{Float32}, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Float64, relty is Float64
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{Float64}, relty is Float64
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is BigFloat, relty is BigFloat
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
Diagonal matrices
elty is Float32, relty is Float32
Linear solve
Simple unary functions
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solve
Simple unary functions
Binary operations
elty is Float64, relty is Float64
Linear solve
Simple unary functions
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solve
Simple unary functions
Binary operations
elty is BigFloat, relty is BigFloat
Linear solve
Simple unary functions
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solve
Simple unary functions
Binary operations
Test interconversion between special matrix types
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
isupper is true
newtype is Bidiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
isupper is false
newtype is Bidiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>

##############

On 7/2/14, Andreas Noack Jensen [email protected] wrote:

Is it right to say the the short version of this is that @jasax has
Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels are
used?

Our problem is to figure out which call is actually causing the error
because the original problem of this thread is solved, right?

I have pushed some debug options
to linalg4.jl. @jasax Can you try to build from latest source and run the
linalg tests with debug=true in the top of linalg4.jl?

---

Reply to this email directly or view it on GitHub:
#117

[andreasnoack]

Please try with debug=false and println("v1:", v1, "v2:", isupper?v2:v2[:,n:-1:1]) pasted right after if elty <: Real in line 212

[jasax]

Hello,

With debug=false and println("v1:", v1, "v2:",
isupper?v2:v2[:,n:-1:1]) pasted right after if elty <: Real in line
212...

Results below for no coretype (i.e. dynamic core recognition, which is
Piledriver, I think...), Piledriver, Bulldozer and Barcelona.

Only Barcelona passes...

Regards

Jose

############

e:\math\Julia-0.3.0-prerelease\share\julia\test>julia3
_
_ _ ()_ | A fresh approach to technical computing
() | () () | Documentation: http://docs.julialang.org
_ _ | | __ _ | Type "help()" to list help topics
| | | | | | |/ ` | |
| | || | | | (| | | Version 0.3.0-prerelease+4028 (2014-07-02 23:42 UTC)
/ |_'|||__'| | Commit 2185bd1 (1 day old master)
|__/ | x86_64-w64-mingw32

julia> quit()

e:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821
0.0 0.0 0.0 0.0 -0.022935344 0.6306952 0.27791148
-0.0028916348 0.67452693 -0.10671823 0.175
70455 -0.06964667
0.0 0.0 0.0 0.0 0.0 0.23474121 0.07669833 6.274062e-5
0.33820927 0.035774022 0.10542922 -0.0
074075013
0.0 0.0 0.0 0.0 0.0 0.0 0.016797159 5.625236e-5 -0.09946941
0.057604518 -0.05467781 -0.00448
2299
0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.00010766533 -0.14416896
-0.22637479 -0.113924906 0.0048884703

    0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.34955943 0.3289738

-0.32831332 0.002901593
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.75734 0.08910849 0.0031541502
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.90202594 -0.01507505
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.92667]v2:[1.0
-0.6775695521181972 -0.09913981
548494394 0.985840280622123 0.0 0.18547599453282315
-0.5493957034241188 0.0 0.0506859143507885 0.084
16566754648515 -0.006229754552608183 -0.25610760159381857
0.0 0.7354587017925246 -0.17985776488270871 0.05455652373436318 0.0
-0.3037137614030213 0.683613617
9215149 0.0 -0.10976032115842474 0.10549426858904311
0.01600160566437595 0.14114745228429257
0.0 0.0 0.9786840559634183 -0.11679375188881413 0.0
-0.3146388615401998 0.20380270522369198 0.0 -0.
22246760574547017 -0.46302377702037906 0.04453672230656894 -0.14152702608770487
0.0 0.0 0.0 0.10724619499380782 0.0 0.5669912425688104
-0.3254253629791597 0.0 0.4622218239136544 0
.13555762838021237 -0.10190634064584442 0.1747682067611283
0.0 0.0 0.0 0.0 0.0 0.630695242033014 -0.27791148658954884 0.0
0.6745269372615431 -0.10671821187988
446 -0.1757045529682005 0.06964666722004574
0.0 0.0 0.0 0.0 0.0 0.23474121053272648 -0.07669833115947837 0.0
0.33820927900919984 0.035774021349
46724 -0.10542922493155694 0.0074075007154691145
0.0 0.0 0.0 0.0 0.0 0.0 -0.016797158798436167 0.0 -0.0994694112485567
0.057604513272400466 0.054677
81452407743 0.0044822988671663335
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.1441689532418754
-0.22637478312306833 0.11392491510759921 -0.00
4888469785453913
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.3495594420482496
0.3289738118136646 0.32831332756277565 -0.0029
01592701145538
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.7573401233330379
-0.08910849321718631 -0.003154149991984479
6
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.902025937083369 0.01507504988414109
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.0 0.9266700197042705]
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading e:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

e:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Piledriver

e:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821
0.0 0.0 0.0 0.0 -0.022935344 0.6306952 0.27791148
-0.0028916348 0.67452693 -0.10671823 0.175
70455 -0.06964667
0.0 0.0 0.0 0.0 0.0 0.23474121 0.07669833 6.274062e-5
0.33820927 0.035774022 0.10542922 -0.0
074075013
0.0 0.0 0.0 0.0 0.0 0.0 0.016797159 5.625236e-5 -0.09946941
0.057604518 -0.05467781 -0.00448
2299
0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.00010766533 -0.14416896
-0.22637479 -0.113924906 0.0048884703

    0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.34955943 0.3289738

-0.32831332 0.002901593
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.75734 0.08910849 0.0031541502
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.90202594 -0.01507505
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.92667]v2:[1.0
-0.6775695521181972 -0.09913981
548494394 0.985840280622123 0.0 0.18547599453282315
-0.5493957034241188 0.0 0.0506859143507885 0.084
16566754648515 -0.006229754552608183 -0.25610760159381857
0.0 0.7354587017925246 -0.17985776488270871 0.05455652373436318 0.0
-0.3037137614030213 0.683613617
9215149 0.0 -0.10976032115842474 0.10549426858904311
0.01600160566437595 0.14114745228429257
0.0 0.0 0.9786840559634183 -0.11679375188881413 0.0
-0.3146388615401998 0.20380270522369198 0.0 -0.
22246760574547017 -0.46302377702037906 0.04453672230656894 -0.14152702608770487
0.0 0.0 0.0 0.10724619499380782 0.0 0.5669912425688104
-0.3254253629791597 0.0 0.4622218239136544 0
.13555762838021237 -0.10190634064584442 0.1747682067611283
0.0 0.0 0.0 0.0 0.0 0.630695242033014 -0.27791148658954884 0.0
0.6745269372615431 -0.10671821187988
446 -0.1757045529682005 0.06964666722004574
0.0 0.0 0.0 0.0 0.0 0.23474121053272648 -0.07669833115947837 0.0
0.33820927900919984 0.035774021349
46724 -0.10542922493155694 0.0074075007154691145
0.0 0.0 0.0 0.0 0.0 0.0 -0.016797158798436167 0.0 -0.0994694112485567
0.057604513272400466 0.054677
81452407743 0.0044822988671663335
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.1441689532418754
-0.22637478312306833 0.11392491510759921 -0.00
4888469785453913
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.3495594420482496
0.3289738118136646 0.32831332756277565 -0.0029
01592701145538
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.7573401233330379
-0.08910849321718631 -0.003154149991984479
6
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.902025937083369 0.01507504988414109
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.0 0.9266700197042705]
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading e:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

e:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Bulldozer

e:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821
0.0 0.0 0.0 0.0 -0.022935344 0.6306952 0.27791148
-0.0028916348 0.67452693 -0.10671823 0.175
70455 -0.06964667
0.0 0.0 0.0 0.0 0.0 0.23474121 0.07669833 6.274062e-5
0.33820927 0.035774022 0.10542922 -0.0
074075013
0.0 0.0 0.0 0.0 0.0 0.0 0.016797159 5.625236e-5 -0.09946941
0.057604518 -0.05467781 -0.00448
2299
0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.00010766533 -0.14416896
-0.22637479 -0.113924906 0.0048884703

    0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.34955943 0.3289738

-0.32831332 0.002901593
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.75734 0.08910849 0.0031541502
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.90202594 -0.01507505
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.92667]v2:[1.0
-0.6775695521181972 -0.09913981
548494394 0.985840280622123 0.0 0.18547599453282315
-0.5493957034241188 0.0 0.0506859143507885 0.084
16566754648515 -0.006229754552608183 -0.25610760159381857
0.0 0.7354587017925246 -0.17985776488270871 0.05455652373436318 0.0
-0.3037137614030213 0.683613617
9215149 0.0 -0.10976032115842474 0.10549426858904311
0.01600160566437595 0.14114745228429257
0.0 0.0 0.9786840559634183 -0.11679375188881413 0.0
-0.3146388615401998 0.20380270522369198 0.0 -0.
22246760574547017 -0.46302377702037906 0.04453672230656894 -0.14152702608770487
0.0 0.0 0.0 0.10724619499380782 0.0 0.5669912425688104
-0.3254253629791597 0.0 0.4622218239136544 0
.13555762838021237 -0.10190634064584442 0.1747682067611283
0.0 0.0 0.0 0.0 0.0 0.630695242033014 -0.27791148658954884 0.0
0.6745269372615431 -0.10671821187988
446 -0.1757045529682005 0.06964666722004574
0.0 0.0 0.0 0.0 0.0 0.23474121053272648 -0.07669833115947837 0.0
0.33820927900919984 0.035774021349
46724 -0.10542922493155694 0.0074075007154691145
0.0 0.0 0.0 0.0 0.0 0.0 -0.016797158798436167 0.0 -0.0994694112485567
0.057604513272400466 0.054677
81452407743 0.0044822988671663335
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.1441689532418754
-0.22637478312306833 0.11392491510759921 -0.00
4888469785453913
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.3495594420482496
0.3289738118136646 0.32831332756277565 -0.0029
01592701145538
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.7573401233330379
-0.08910849321718631 -0.003154149991984479
6
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.902025937083369 0.01507504988414109
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.0 0.9266700197042705]
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading e:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

e:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Barcelona

e:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821
0.0 0.0 0.0 0.0 -0.022935344 0.6306952 0.27791148
-0.0028916348 0.67452693 -0.10671823 0.175
70455 -0.06964667
0.0 0.0 0.0 0.0 0.0 0.23474121 0.07669833 6.274062e-5
0.33820927 0.035774022 0.10542922 -0.0
074075013
0.0 0.0 0.0 0.0 0.0 0.0 0.016797159 5.625236e-5 -0.09946941
0.057604518 -0.05467781 -0.00448
2299
0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.00010766533 -0.14416896
-0.22637479 -0.113924906 0.0048884703

    0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.34955943 0.3289738

-0.32831332 0.002901593
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.75734 0.08910849 0.0031541502
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.90202594 -0.01507505
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.92667]v2:[1.0
-0.6775695521181972 -0.09913981
548494394 0.9858402806221229 -0.9514781983503483 0.18547599453282318
0.5493957034241188 -0.996691750
623864 -0.05068591435078851 0.08416566754648515 -0.006229754552608183
-0.25610760159381857
0.0 0.7354587017925246 -0.17985776488270871 0.05455652373436318
0.11083306952378438 -0.303713761403
0213 -0.6836136179215149 0.027725675415920723 0.10976032115842475
0.10549426858904311 0.016001605664
37595 0.14114745228429257
0.0 0.0 0.9786840559634183 -0.11679375188881415 -0.201531827468068
-0.3146388615401999 -0.203802705
223692 -0.05502798384063495 0.2224676057454702 -0.46302377702037906
0.04453672230656894 -0.141527026
08770487
0.0 0.0 0.0 0.10724619499380782 0.2031358202102392 0.5669912425688105
0.3254253629791597 0.05291863
482581453 -0.4622218239136544 0.13555762838021237 -0.10190634064584442
0.1747682067611283
0.0 0.0 0.0 0.0 0.02293534017199715 0.6306952420330141
0.2779114865895489 0.0028916346825857775 -0.
6745269372615431 -0.10671821187988446 -0.1757045529682005 0.06964666722004574
0.0 0.0 0.0 0.0 0.0 0.2347412105327265 0.07669833115947837
-6.274061808247623e-5 -0.338209279009199
84 0.03577402134946724 -0.10542922493155694 0.0074075007154691145
0.0 0.0 0.0 0.0 0.0 0.0 0.016797158798436167 -5.6252362773142516e-5
0.09946941124855671 0.057604513
272400466 0.05467781452407743 0.0044822988671663335
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.00010766533506060247 0.1441689532418754
-0.22637478312306833 0.113924
91510759921 -0.004888469785453913
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.3495594420482497 0.3289738118136646
0.32831332756277565 -0.002901
592701145538
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.7573401233330379
-0.08910849321718631 -0.0031541499919844796
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.902025937083369 0.01507504988414109
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.9266700197042705]
v1:Float32[-0.10435905 -0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.2173675 0.44780803 -0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.06368588 -0.15674347 0.29170203 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.0515887 -0.07684996 0.04723677 -0.8569027 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0
0.037946656 0.10141277 0.02024641 0.375485 -0.04855015 -0.0
0.0 0.0 0.0 0.0 0.0 0.0
0.024085311 0.07714987 -0.7692005 0.19004807 -0.049251325
-0.7978891 0.0 0.0 0.0 0.0 0.0 0.0

    -0.037657373 -0.25559032 -0.53675175 -0.16419467 -0.39879608

-0.570325 -0.51456445 0.0 0.0 0
.0 0.0 0.0
-0.029381434 -0.1577869 -0.15989143 -0.21515498 -0.19493997
-0.171636 -0.2677123 -0.96352 -0
.0 0.0 0.0 0.0
0.018234724 0.15400963 -0.08036459 0.085685425 0.68266326
-0.08891187 0.53529936 0.25123572
0.8477533 0.0 0.0 0.0
0.30235898 -0.22334856 0.01769589 0.07080985 -0.41780844
0.019909456 -0.39368397 0.086565495
-0.42412183 0.26950645 0.0 0.0
0.90318745 0.751786 -0.0049948324 0.05431823 0.38231027
-0.005722551 0.45681432 0.031714622
0.3028671 0.94913125 0.9830645 0.0
0.15277706 0.15410393 0.011621907 0.007246026 0.10687958
0.0175769 0.11550637 0.003226884 0.
09852172 0.16283782 0.18326028 1.0]v2:[0.10435903406197171 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0
.0
-0.21736747087606736 0.44780806147681473 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0
0.06368587220119458 -0.15674347658408933 0.2917020079267694 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.05158869456498528 -0.07684995237001097 0.04723676761238894
0.856902692527982 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0
-0.03794665580426363 0.10141276468967186 0.02024641084282024
-0.3754849778953141 0.0485501460091958
5 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.02408531068133337 0.07714986155312042 -0.7692005171206235
-0.19004805405098008 0.049251321769994
77 0.7978891410380274 0.0 0.0 0.0 0.0 0.0 0.0
0.03765737463545521 -0.2555902931492239 -0.5367516973373543
0.16419465662973412 0.3987960672564004
0.5703249939908089 0.514564474309506 0.0 0.0 0.0 0.0 0.0
0.02938143359655504 -0.1577868880645742 -0.15989140643784225
0.21515496341786292 0.1949399521421234
5 0.17163600444670735 0.26771227645461804 0.9635200073488135 0.0 0.0 0.0 0.0
-0.018234724742740713 0.15400961400479252 -0.0803645802024666
-0.08568541923554857 -0.6826632246470
811 0.08891186868523294 -0.5352993552732612 -0.25123570706608994
0.8477533271940895 0.0 0.0 0.0
-0.30235897587027344 -0.2233485604442196 0.0176958873425955
-0.07080984746812212 0.4178084157251453
-0.019909455956916886 0.39368398750977324 -0.08656549477405531
-0.4241217971339471 0.26950644567423
426 0.0 0.0
-0.9031874693443103 0.7517860023927552 -0.004994831872518743
-0.054318229516047124 -0.3823102274210
1173 0.005722551155727579 -0.4568143321052316 -0.031714622308903075
0.30286708452173644 0.9491312435
377935 0.9830644285489673 0.0
-0.15277707043708272 0.1541039370570079 0.011621905772110886
-0.00724602579346015 -0.10687956726632
494 -0.0175768995907174 -0.1155063765328756 -0.003226883925855374
0.09852170593592807 0.162837828161
77106 0.18326027753360102 1.0]
v1:[1.0 0.8328452297384424 0.10410224878723878 0.30300034406647014
0.1793548975698909 0.249389872832
6044 0.4954626646844483 0.5456045905060148 0.2285581302704708
0.9863452475809965 0.9256125252945819
0.4047808380830982
0.0 -0.5535059379102641 0.1999656050257831 0.24948135413257155
0.23171475000233402 0.24387810480578
96 0.2479326799944397 0.24069379165427393 0.24093114575410884
-0.01571605529717448 -0.03313989717670
367 0.2531176639762361
0.0 0.0 -0.9742568853254835 -0.6995399648508701 -0.8543222278742065
-0.7548459380539035 -0.54430470
56043007 -0.5015358807670509 -0.7803372381706697 0.019210730246539614
0.03926814063904774 -0.6152696
052879725
0.0 0.0 0.0 -0.5971546558574441 -0.41792082496226957
-0.5536375996866562 -0.601371749815208 -0.5772
345890457918 -0.5260621251296269 0.028940183863341712
0.05976273810034115 -0.619997595259935
0.0 0.0 0.0 0.0 -0.09806023848613772 -0.0428415812925448
0.09725815675366715 0.11048275968379193 -0
.060790676691847324 -0.012163257567474836 -0.025712288510048323
0.0614884515221648
0.0 0.0 0.0 0.0 0.0 -0.01363004966785573 0.07799121879619157
0.09522853645222285 -0.014645387238351
596 -0.016116369691630692 -0.034586481629573944 0.041526128523344634
0.0 0.0 0.0 0.0 0.0 0.0 -0.1399598542844887 -0.19210071622398955
-0.00419616604401066 0.06019654821
625756 0.13176737534435043 -0.055026090763614484
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.03829516254383423 -0.007816887945234096
-0.10456625308875277 -0.23769
861600150205 -0.023124032527056845
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.007556059457142271
-0.07537021766958452 -0.17877836586025747 0.00
67157370378890646
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.07087590677519363
-0.17123954597038377 0.0025307582390986672

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.004931789583317615
-0.0023522983817159185
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.002310032459877136]v2:[1.0 -0.8328452297384424 -0.10
410224878723878 -0.3030003440664701 -0.17935489756989093
-0.24938987283260433 -0.4954626646844483 0.
5456045905060147 0.22855813027047076 -0.9863452475809965
-0.9256125252945818 -0.4047808380830983
0.0 0.5535059379102641 -0.19996560502578312 -0.24948135413257155
-0.23171475000233405 -0.2438781048
0578956 -0.2479326799944397 0.24069379165427393 0.24093114575410884
0.015716055297174483 0.033139897
17670366 -0.2531176639762361
0.0 0.0 0.9742568853254835 0.6995399648508701 0.8543222278742064
0.7548459380539037 0.5443047056043
006 -0.5015358807670508 -0.7803372381706697 -0.019210730246539614
-0.039268140639047724 0.6152696052
879725
0.0 0.0 0.0 0.5971546558574441 0.4179208249622695 0.5536375996866562
0.6013717498152079 -0.57723458
90457917 -0.5260621251296269 -0.028940183863341712
-0.059762738100341134 0.6199975952599351
0.0 0.0 0.0 0.0 0.09806023848613772 0.042841581292544804
-0.09725815675366714 0.11048275968379191 -
0.060790676691847324 0.012163257567474838 0.025712288510048316
-0.061488451522164804
0.0 0.0 0.0 0.0 0.0 0.013630049667855733 -0.07799121879619157
0.09522853645222283 -0.01464538723835
1596 0.016116369691630685 0.03458648162957394 -0.04152612852334464
0.0 0.0 0.0 0.0 0.0 0.0 0.1399598542844887 -0.19210071622398955
-0.004196166044010659 -0.0601965482
1625754 -0.13176737534435037 0.05502609076361448
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.03829516254383423 -0.007816887945234094
0.10456625308875271 0.2376986
1600150205 0.023124032527056842
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.007556059457142271
0.07537021766958447 0.17877836586025747 -0.006
715737037889063
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0708759067751936
0.17123954597038374 -0.002530758239098667
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0049317895833176145
0.002352298381715918
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.002310032459877136]
v1:[0.003688647956899844 -0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.004740159784379087 -0.44187851109422704 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0
0.009874515948691959 -0.05385192525512327 0.012217385730351213 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.014990221803501617 0.2758681062247914 0.01997826455832316
-0.12517672934788143 -0.0 0.0 0.0 0.0 0
.0 0.0 0.0 0.0
-0.05959144977494442 0.5805419782907039 -0.09451195124613689
-0.40502520880885984 -0.07870210567172
11 -0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.04890281365714445 -0.38439532312353647 -0.08851523208271764
0.45689924683616456 -0.2872100363981
948 0.3299402664598085 -0.0 0.0 0.0 0.0 0.0 0.0
0.9727761152082935 0.48954375577572534 0.9738615940601784
-0.7402709005905348 0.9427250383798739 -0
.8717968356932118 0.9612756604419457 0.0 0.0 0.0 0.0 0.0
0.1872756647774995 0.023132335699570815 0.16025591420717353
-0.04227123766219195 0.0879279422621960
1 -0.07073364568486122 0.23433839926395406 0.790210200780212 0.0 0.0 0.0 0.0
0.10702994641952854 0.007376808016633514 0.08774620827364173
-0.014995296980244065 0.03943812519075
379 -0.029861780901777192 0.14159931548023652 0.6083799489685016
0.9989300250333178 -0.0 0.0 0.0
0.008601307744977978 0.00396025988484366 0.007399335416668204
-0.0035023908013139776 0.004491255373
987864 -0.0038688066063161065 0.010753947021171148 0.03829840076478014
0.03871761556689694 0.3214899
4253191976 -0.0 0.0
-0.021245522464019048 0.007116860708537115 -0.021037477649636222
-0.010401895962540032 -0.062284448
283146006 -0.1584033616025683 -0.02285114875073297
-0.053373216425821386 -0.023923393849791343 0.443
0078052849678 -0.4727710018430864 0.0
0.018782353122311343 0.025346441867563067 0.019976588083727896
0.24778066105443117 0.09695185890355
137 0.31639868934887394 0.01861265032376814 0.03355903072502379
0.00821112406781778 0.83689204877774
95 0.8811853266006445 1.0]v2:[0.003688647956899843 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
-0.004740159784379086 0.441878511094227 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.009874515948691957 0.05385192525512327 0.012217385730351213 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.014990221803501612 -0.27586810622479147 0.01997826455832316
0.12517672934788146 0.0 0.0 0.0 0.0 0
.0 0.0 0.0 0.0
-0.05959144977494442 -0.5805419782907039 -0.09451195124613689
0.4050252088088599 0.0787021056717210
8 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.048902813657144446 0.3843953231235365 -0.08851523208271762
-0.4568992468361646 0.287210036398194
7 0.3299402664598085 0.0 0.0 0.0 0.0 0.0 0.0
0.9727761152082933 -0.4895437557757253 0.9738615940601784
0.7402709005905348 -0.9427250383798739 -0
.8717968356932118 0.9612756604419458 0.0 0.0 0.0 0.0 0.0
0.18727566477749946 -0.02313233569957081 0.1602559142071735
0.04227123766219195 -0.0879279422621960
2 -0.07073364568486122 0.23433839926395408 0.7902102007802121 0.0 0.0 0.0 0.0
0.10702994641952852 -0.007376808016633513 0.0877462082736417
0.01499529698024407 -0.039438125190753
79 -0.029861780901777185 0.14159931548023655 0.6083799489685016
0.9989300250333176 0.0 0.0 0.0
0.008601307744977976 -0.003960259884843659 0.007399335416668203
0.0035023908013139785 -0.0044912553
73987864 -0.0038688066063161056 0.01075394702117115
0.03829840076478013 0.03871761556689692 0.321489
9425319198 0.0 0.0
-0.021245522464019038 -0.007116860708537113 -0.02103747764963622
0.010401895962540035 0.06228444828
314601 -0.15840336160256827 -0.022851148750732974 -0.05337321642582139
-0.023923393849791332 0.44300
780528496786 0.47277100184308646 0.0
0.018782353122311333 -0.02534644186756306 0.019976588083727893
-0.24778066105443125 -0.096951858903
5514 0.31639868934887383 0.018612650323768142 0.03355903072502379
0.008211124067817778 0.83689204877
77495 -0.8811853266006445 1.0]
v1:BigFloat[1e+00
2.124861110142520175900802120368613086311783085646432056190190141403048323994773e-
01 2.461076878787208508340017244254529800553731800910643172931449031889103366993002e-02
7.1930122583
65381695472592236319230349644713900174923793701444724659081151483079e-03
4.2087053524233440344212216
95050977192284803287468845032364966945865354918345e-02
1.3185234688579519714404071954865164079108294
33373127288897178020659870722179756e-02
1.0406386972014702317063136661325889914617741940178374063801
81325852635283224235e-02
3.8616772584959234861009193113974953298598258457215555308010616522774906996
08494e-03 2.141523330352537880480404537633924965281808701482227181084640813184992577336789e-02
3.923
74442237956674464397554065334241314658637085157082764408258663769932076449e-03
1.6302151454364987224
64968140354515416174417060571248634706820281123322404075234e-02
1.9461234435449726251419570540776568
22937597697290131727754077675807423975314132e-04
0e+00 9.771640868482831053908448597217175767558571871424413621032580718105206688233718e-01
7.300027229314478885572990305754945055293384695289021523229232689291064985228619e-01
-1.757196818292
317515737166324008133334405383699511952580604893035782969287454387e-01
4.543941983483218392922500217
650739897622926630081652053946677952592377137936933e-01
-1.51806450075183116035430160341004216736580
724141461472240340868434189713982061e-01
2.011056234068603635733637228856432401757369852594440099729
193997570674258171076e-01
1.048061861279414302847305555720960172981899094854966371715161000225822787
595987e-01 2.950064357205149261739461807901160302234742605334599422163678067869164323150737e-01
9.88
0030737411593639682185340331017011353366726337750084512625838482766688328195e-02
1.64964386546518147
4734342066654966829175242006584281482448407706976370247916617e-01
-4.6019637264489937410655975650430
39853955946564330732996050206079637634964152748e-04
0e+00 0e+00
6.83000976992230684039206367790858123537085608402551193296305196671783095179277
7e-01 1.904131343631428717251244884238286448235647219475343081615947449177941974635355e-01
1.0513115
30135707125329517977701380712300029885428320143488734423163082361966255e-01
9.1306627060988303827022
20575865856868086239611864300266176905254759387345346503e-02
1.1055636958345403429786198352614511982
64562520621643938838294535601227448347208e-01
8.8191573752101916201656603134577831677817930166643882
15886563506126411617316536e-02
1.0106110888380516318235108190583932704931203300207751576360142861761
37683474936e-01
7.590899273900849217771094368488582474061130237300427660779170231208745144768735e-02
3.399583120505385563976888206739768910574875701546397394980848345721932375863524e-02
1.196258237943
435555782807678912257327256356478875240964552146520782431628037728e-04
0e+00 0e+00 0e+00
9.65822805826456070032322373730063954828108930844467183450842857003491839
0525741e-01 1.859823105747782273044246213061331886244315237370372346274384877591352753415259e-01
3.5
25440274077685903038316840731870048578902982507619340621908589489890808721549e-01
1.0716194112812390
35411643832448766770574257701708364716591245447432055421393091e-01
2.0855539844782004682360274195510
37708089376420167840593289547816761202599303002e-02
1.5055092691804752142095481733393378081376217401
3328790870736999609982209860625e-01
3.19022672667991051422845058105790108943339888182116068393010427
4146180685352211e-02
6.22996488914764613904163034419333039337502313502504826822055023584940481576524
2e-02 3.592611373913653562390104161934868117744341798421133128503190680402114019441659e-04
0e+00 0e+00 0e+00 0e+00
8.63778104175104786072105500288300931890162321628156742003277877332
6185562565233e-01
6.003530734972294204966850291196452544874861522935319368624163651315656477094771e-
01 6.182055447616237168587402664474709653087851947890676181881621135983852809082744e-01
1.4180216478
14149016762383217797780261024511290642496286151417943082161199012693e-01
7.5835265534037036517945245
99929653399079509621880262564087326381382753825239937e-01
2.0866767396209074674640347260945443124933
08666193701932550116879880799781953666e-01
2.8378132320966020984286584540595068973514010235258210855
52989033822135895586263e-01
1.0451421441061503022721901051187116229650430089053954929538934831619525
27759085e-03
0e+00 0e+00 0e+00 0e+00 0e+00
-6.9550912793295845602331803681008883800043478548062023986684
52426083440862994261e-01
2.7378722755403862724401688209090198158327264750585888515654040076250272942
51534e-01 1.203448231660805679376735342699991797302773468830559780896981930604661318396814e-01
1.173
117937129555933171070758418181029769564701318180056356294896381202534632063e-01
1.558549324724075537
183509954017851229611912826128588667463137938919578205054869e-01
-9.98152994766151994020334715692436
1119051762181053581326423093540308976454497973e-03
-7.1428215586546995846941957029948881374087586663
36527893212551444815262698440053e-04
0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
6.91812330995019987482907237106154659812075489483320650
4203279052839440732811651e-01
3.81150317479922621174153394899120989289103633450396186638099627976215
201181652e-01 2.430818501555945805634018073120288187122244206353944145531076737247838538350923e-01
4
.685869204532751207167204019449904537537515201129913876142833478030369472016939e-01
-1.7692349937459
57733199763373618531221833742239879848609754441419465176285251686e-02
-5.354380635570488765280291250
564751738539291941704507723564736128644572398411385e-04
0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
8.94942315828433847784329219920284973820148180805
8363262568233104016805175135483e-01
-4.0531933978093229798933803600196389587764860553604378683191672
60175882650700995e-01
8.2430339250784421387140569061111626309184172946880405724603701327189022704298
91e-01 4.89368131543111226973716995184405559513016105782511988850593626012987523969661e-02
3.4893484
12991727590304539114684461006214058615723437682991353856399421488904525e-03
0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
-2.5961568030430339699693823659613945685472
87530235861404129092837602252074105375e-01
7.7430934412968420018341524983279857955456160463903024522
7470643383407747309729e-02
3.99206808146952787091077583501812558127409752607821135286208367344204181
5764697e-02 4.934194728078620337191888746863245418001046462761393345681659607709956735126217e-03
0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
1.02785606971289153030595725573985239
9662743908886901095703797064602386422499541e-01
-1.6989720607862239575215216785244326298650656286280
51204621150629107449865316946e-02
-9.269682771680203988234650053895596212379521057082632252163985525
845661493765602e-03
0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
9.39288249476268574111727229980
2837521761590825033485773078015766175932444178829e-01
9.37271115702405936158435353949883340213950656
0209617928543452532261795956065323e-01
0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
0e+00 -3.4842233650087610852970
55313027579601629374767420569074412359113548658299231012e-01]v2:[1.0
0.21248611101425205 0.024610768
787872085 0.00719301225836538 0.04208705352423344
-0.013185234688579517 0.010406386972014701 0.00386
1677258495924 -0.021415233303525377 0.003923744422379567
0.01630215145436499 -0.00019461234435449721

0.0 0.9771640868482832 0.7300027229314479 -0.17571968182923173
0.4543941983483218 0.151806450075183
1 0.20110562340686036 0.10480618612794146 -0.29500643572051494
0.09880030737411595 0.164964386546518
15 0.0004601963726448992
0.0 0.0 0.6830009769922306 0.19041313436314286 0.10513115301357072
-0.09130662706098829 0.110556369
58345404 0.08819157375210193 -0.10106110888380515 0.0759089927390085
0.03399583120505385 -0.00011962
582379434353
0.0 0.0 0.0 0.965822805826456 0.1859823105747782 -0.3525440274077685
0.1071619411281239 0.020855539
844782008 -0.15055092691804753 0.03190226726679911 0.06229964889147646
-0.0003592611373913653
0.0 0.0 0.0 0.0 0.8637781041751047 -0.6003530734972293
0.6182055447616237 0.14180216478141494 -0.75
83526553403703 0.2086676739620908 0.2837813232096602 -0.0010451421441061502
0.0 0.0 0.0 0.0 0.0 0.6955091279329584 0.27378722755403856
0.1203448231660806 -0.11731179371295558
0.1558549324724076 -0.00998152994766152 0.00071428215586547
0.0 0.0 0.0 0.0 0.0 0.0 0.6918123309950199 0.3811503174799226
-0.24308185015559458 0.46858692045327
52 -0.017692349937459575 0.0005354380635570489
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.8949423158284339 0.40531933978093226
0.8243033925078442 0.04893681315
431112 -0.003489348412991728
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2596156803043034
0.07743093441296843 0.03992068081469528 -0.00493
4194728078619
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.10278560697128918
-0.01698972060786224 0.009269682771680204
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.9392882494762684 -0.937271115702406
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.3484223365008761]
v1:BigFloat[-9.103845827144652919271358346606043735713054394586691600819295899208185288878328e-01
0e
+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
-7.48792724994097615634985013875378348706919135026818897576977014488403723087575e-02
-1.569
941123603518676086087264373019362055324876404495351524255399226799355425058e-03
0e+00 0e+00 0e+00 0e
+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
1.862468261319071838193538909655118792004563052857634458175473088520548926554748e-01
-5.844
775711035577785073467191541196597052708587956090567826655374442176020123182e-02
3.130746372930040772
626626062611281499593740559120851156990969683259411280122217e-03
-0e+00 0e+00 0e+00 0e+00 0e+00 0e+0
0 0e+00 0e+00 0e+00
-1.207939367119769882493917890473087811872668687746778107147354443420401533715423e-01
1.020
547118572660140647872131417206504038998536767493794167612901292278086916312e-01
-4.94273315933749878
500624218259804965795509530848256656676905542727889238445586e-03
-6.45836616223543577434448286005498
6396153310641453584735379231600331502868086677e-01 -0e+00 0e+00 0e+00
0e+00 0e+00 0e+00 0e+00 0e+00
2.074061012573043132116334638108462871516390972734514162257312579118176632796356e-02
-6.954
196305096463618469544104400733199340926695536790373916691747678092545773176e-02
2.840066903432437749
559756018523128626174799293648478766955227532052729936833101e-03
-5.83454490983336707686818059389612
5241957280720850563084896279258825993583170946e-01
-5.7401504163242029909130194108199921652550558268
56491975096583308235696585980221e-01 -0e+00 0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
1.195301253547046452341073976524795753382351535985461550345634202114451581994091e-02
6.3423
25861400650719461772232486874062698233785962804982952364463836587473013443e-01
4.8243811791964480415
39392277952519005924209629701789470588683389753712275132722e-01
4.8627908181854758965626405361523155
64885999814006438961717938204318334662943641e-01
7.8514582947834381607829787728179660212861904011886
91145272063475031242345371883e-01
4.9110394230019393537833843458346038857780206010252395812448526173
05385539522462e-01 -0e+00 0e+00 0e+00 0e+00 0e+00 0e+00
-2.110629287736768707368976350560504643484308013348324872566503091112732987884477e-02
-6.25
1427369699880255492742835824312911509619907634172869186846613263825121301417e-02
-5.0540876319852556
60817266054141641656001976775154910631277330338904720409249675e-02
-3.076528710402504516242985180751
549439275192750107478214804667005655118291496475e-02
-5.87147696137551484680584595270914205954032105
6765385484488148805031843420275357e-02
-5.1284493666811568530023184084619290067586567634971059156431
28033616068978098665e-02
5.6625024074685529973138386285409687711802154518415423817273321660316592974
27189e-02 -0e+00 0e+00 0e+00 0e+00 0e+00
3.699461986835154803314436894284010781799063756021512773094398540836751388657439e-02
-2.876
312315741562531832985761435890027894394222266979138167452931119246201733061e-01
-3.00837116184465771
7645621275518112478615021784432310453791661090631148973957193e-01
-4.5049277617689501939424089543290
62536829603716918333883931653143500940528995615e-02
-1.216278547910630320840538155299385667125057979
008059077463202181316141085033729e-01
-3.00762525974209584495794422085226657137577449590401458492997
071939535382525038e-01
-9.17594877484617183660567200233654029876132413556133944783167804756432476514
1698e-02 -2.086690256644475341329795530045224071995621019970090205192722360987499931147415e-01
0e+00
0e+00 0e+00 0e+00
-1.325523790540737763439170766754658006836836996698387328020720325704074895466257e-01
-7.00
513958275292168446753991860130671388875466833512298608797304745112694618844e-01
-8.18810770057470109
8445661022194001671927435746451007687150859608284137908491819e-01
-5.5060982544008759397026914988040
74853413812841866167468323961049635955524771839e-02
-1.891774157674031061831302298958508775359251974
842694547382885137606002340633841e-01
-8.13733335607355779652210388584999065687729607231151756775074
8969050426761976222e-01
2.86865143397444092461026283863077163546803693180294615445915621235230842999
1648e-01 -9.470320040335584590653344612125914806078449658410351421738345138209640025058366e-01
9.301
08890993905303267124921083273191216990666735273375025582448492547267813735e-01
-0e+00 0e+00 0e+00
-5.453085663829424908954422060502736757439225004098210784517749642021619032866943e-02
-1.09
7335494348642763366760578062741618586192184856104538023977291081761108874737e-02
-1.3648656829694165
90097615370053234713090475794602902847216209659736305891652895e-02
-5.499184019055350451573402434562
298392158060197108545012611564994657697850349935e-04
-2.23817152838087531356706565168573208282280211
2810121003290060399026497071910239e-03
-1.3519861836667861402287607184736803447125479836483198229817
22002639839284188511e-02
-2.391218164303716271110724156334255609958449370940271495510281293048507220
808475e-01 -2.019348420938856369510872453513098339789318410853303659412908998672171722747986e-02
3.4
09498201464932072023184953335121232997416034191689082916343062880032433574708e-02
-2.271571952727995
994314614169284076104564337856399978566957483655452795823646609e-01 -0e+00 0e+00
-1.285351422071569421841670140957709756708814449831770035217978003356311703094273e-01
3.796
500118018002764028177662330115834092894234809974200051835463342861414095202e-02
5.585281706959819815
271236488803932241143029113509489687307015680181929281094737e-02
7.739323739441245627122924520355249
007124125352862521947135261000603863225428797e-04
4.228422736177874221996231514002468574223138714614
150051917697682509860071044391e-03
5.481519783777964164032785879596208190475438622517379523061767936
064349051252062e-02
-5.12676124834950402424490215860631297370928378917375112095530022540530227550591
9e-01 2.188518786374113368804651545548177406767818765190979775396239572364825160959956e-01
2.5142463
50779379337463132328608620675283558779420026925755072038200082351156289e-01
-5.020227036115256769254
179843349264435096488498826598153463746617964205247087284e-01
8.493899270733372324691193132507779042
880869523606111624932221506167471922120276e-01 0e+00
2.772395994425710469119683606969649596132527004786371971777738527860063433127481e-01
1.2563
89469460509298570463000700629953462561487545886975814812186861194291890965e-02
1.9922685332619099237
19302233255707457813035684146080584017768194735566689314102e-02
1.5230646890979178483550203899173683
00631573159670022199085510640995592757937053e-04
1.0077666212015035556027853254525748299141856658254
80114574872245978114085832175e-03
1.9475304347003962381988115951771586762005876844111993821007249904
81117004226468e-02
7.65549567625443199764141124625735230748469073231928999457419441443385990794888e-
01 1.062247592456354071552710144330808510686747065877882799156562925881230492036736e-01
2.6555721788
61536113036954203540546545803906811831644536761613757115827616428654e-01
8.3448955277724683770010173
0006145831228218473438016226998787585488304841537143e-01
5.27765811498197425562079853088806374103590
5898400748117583027798981748090570532e-01 1e+00]v2:[0.9103845827144653
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0
.0 0.0 0.0 0.0
0.07487927249940976 0.0015699411236035187 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0
-0.18624682613190718 0.058447757110355776 0.003130746372930041 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.120793936711977 -0.10205471185726601 -0.004942733159337498
0.6458366162235436 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0
-0.020740610125730436 0.06954196305096463 0.0028400669034324374
0.5834544909833368 0.57401504163242
03 0.0 0.0 0.0 0.0 0.0 0.0 0.0
-0.011953012535470467 -0.6342325861400651 0.48243811791964475
-0.4862790818185477 -0.78514582947834
38 0.49110394230019394 0.0 0.0 0.0 0.0 0.0 0.0
0.02110629287736769 0.0625142736969988 -0.05054087631985255
0.03076528710402505 0.05871476961375514
4 -0.05128449366681157 0.05662502407468554 0.0 0.0 0.0 0.0 0.0
-0.036994619868351564 0.2876312315741562 -0.30083711618446574
0.04504927761768951 0.121627854791063
04 -0.3007625259742096 -0.09175948774846172 0.20866902566444753 0.0 0.0 0.0 0.0
0.1325523790540738 0.7005139582752921 -0.8188107700574702
0.05506098254400878 0.18917741576740313 -
0.8137333356073557 0.2868651433974441 0.9470320040335585
0.9301088909939053 0.0 0.0 0.0
0.05453085663829427 0.010973354943486429 -0.013648656829694168
0.0005499184019055353 0.002238171528
3808752 -0.01351986183666786 -0.23912181643037161 0.020193484209388567
0.03409498201464932 0.2271571
952727996 0.0 0.0
0.128535142207157 -0.03796500118018003 0.0558528170695982
-0.0007739323739441249 -0.004228422736177
874 0.054815197837779626 -0.5126761248349504 -0.21885187863741135
0.25142463507793794 0.502022703611
5257 0.8493899270733373 0.0
-0.27723959944257115 -0.01256389469460509 0.019922685332619098
-0.00015230646890979186 -0.001007766
6212015034 0.019475304347003956 0.7655495676254432 -0.1062247592456354
0.2655572178861536 -0.8344895
527772469 0.5277658114981975 1.0]
SUCCESS

e:\math\Julia-0.3.0-prerelease\share\julia\test>

############

On 7/3/14, Andreas Noack Jensen [email protected] wrote:

Please try with debug=false and println("v1:", v1, "v2:", isupper?v2:v2[:,n:-1:1]) pasted right after if elty <: Real in line 212

---

Reply to this email directly or view it on GitHub:
#117

[wernsaar]

I found and fixed a bug in the daxpy kernel under Windows. Only
bulldozer and piledriver were
affected. You could try my development fork:
https://github.com/wernsaar/OpenBLAS.git

Best regards
Werner

[tkelman]

Binary build from Werner's branch posted here http://sourceforge.net/projects/juliadeps-win/files/openblas-13348b21373848f3-x86_64-w64-mingw32.7z

@jasax you know the drill, hopefully we're all set now

[jasax]

Hi Werner,

I could not build the sources with mingw a couple weeks ago when I tried.
So if you (or anyone) could build the libopenblas.dll for windows 64
bits I would appreciate: otherwise I cannot test this fix.

Best Regards

Jose

On 7/6/14, wernsaar [email protected] wrote:

On 03.07.2014 18:51, jasax wrote:

Hello,

Results of linalg4.jl tests with debug=true for Piledriver, Bulldozer
and Barcelona.

Success only with Barcelona.

Piledriver and Bulldozer crash in the same place: Bidiagonal matrices,
Eigensystems.

Results below.

Please indicate further tests/debugs.

Kind Regards

Jose

##############

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

E:\math\Julia-0.3.0-prerelease\share\julia\test>ls
Makefile core.jl linalg1.jl random.jl
sorting.jl
TestHelpers.jl dsp.jl linalg2.jl ranges.jl
sparse.jl
arpack.jl euler.jl linalg3.jl readdlm.jl
spawn.jl
arrayops.jl examples.jl linalg4.jl reduce.jl
statistics.jl
arrayperf.jl fft.jl lineedit.jl reducedim.jl
strings.jl
backtrace.jl file.jl math.jl regex.jl
suitesparse.jl
bigint.jl float16.jl mod2pi.jl remote.jl
sysinfo.jl
bitarray.jl floatapprox.jl mpfr.jl repl.jl
test.jl
blas.jl functional.jl netload replcompletions.jl
test_sourcepath.jl
broadcast.jl git.jl numbers.jl resolve.jl
testdefs.jl
ccall.jl gitutils.jl parallel.jl rounding.jl
unicode.jl
ccalltest.c goto.jl perf runtests.jl
version.jl
collections.jl hashing.jl pkg.jl show.jl
combinatorics.jl iobuffer.jl pollfd.jl simdloop.jl
complex.jl keywordargs.jl priorityqueue.jl socket.jl

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3
_
_ _ ()_ | A fresh approach to technical computing
() | () () | Documentation: http://docs.julialang.org
_ _ | | __ _ | Type "help()" to list help topics
| | | | | | |/ ` | |
| | || | | | (| | | Version 0.3.0-prerelease+4028 (2014-07-02 23:42
UTC)
/ |_'|||__'| | Commit 2185bd1 (0 days old master)
|__/ | x86_64-w64-mingw32

julia> quit()

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Piledriver

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading
E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Bulldozer

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading
E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Barcelona

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{Float32}, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Float64, relty is Float64
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{Float64}, relty is Float64
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is BigFloat, relty is BigFloat
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
Diagonal matrices
elty is Float32, relty is Float32
Linear solve
Simple unary functions
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solve
Simple unary functions
Binary operations
elty is Float64, relty is Float64
Linear solve
Simple unary functions
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solve
Simple unary functions
Binary operations
elty is BigFloat, relty is BigFloat
Linear solve
Simple unary functions
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solve
Simple unary functions
Binary operations
Test interconversion between special matrix types
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
isupper is true
newtype is Bidiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
isupper is false
newtype is Bidiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>

##############

On 7/2/14, Andreas Noack Jensen [email protected] wrote:

Is it right to say the the short version of this is that @jasax has
Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels
are
used?

Our problem is to figure out which call is actually causing the error
because the original problem of this thread is solved, right?

I have pushed some debug
options
to linalg4.jl. @jasax Can you try to build from latest source and run
the
linalg tests with debug=true in the top of linalg4.jl?

---

Reply to this email directly or view it on GitHub:

#117

Reply to this email directly or view it on GitHub:
#117
Hi,

I found and fixed a bug in the daxpy kernel under Windows. Only
bulldozer and piledriver were
affected. You could try my development fork:
https://github.com/wernsaar/OpenBLAS.git

Best regards
Werner

---

Reply to this email directly or view it on GitHub:
#117

[tkelman]

@jasax that's exactly what I did - http://sourceforge.net/projects/juliadeps-win/files/openblas-13348b21373848f3-x86_64-w64-mingw32.7z

[wernsaar]

Hi,

I now have published the dll for windows at sourceforge as version
v0.2.10-pre4.

Best regards
Werner

On 07.07.2014 05:09, jasax wrote:

Hi Werner,

I could not build the sources with mingw a couple weeks ago when I tried.
So if you (or anyone) could build the libopenblas.dll for windows 64
bits I would appreciate: otherwise I cannot test this fix.

Best Regards

Jose

On 7/6/14, wernsaar [email protected] wrote:

On 03.07.2014 18:51, jasax wrote:

Hello,

Results of linalg4.jl tests with debug=true for Piledriver, Bulldozer
and Barcelona.

Success only with Barcelona.

Piledriver and Bulldozer crash in the same place: Bidiagonal matrices,
Eigensystems.

Results below.

Please indicate further tests/debugs.

Kind Regards

Jose

##############

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

E:\math\Julia-0.3.0-prerelease\share\julia\test>ls
Makefile core.jl linalg1.jl random.jl
sorting.jl
TestHelpers.jl dsp.jl linalg2.jl ranges.jl
sparse.jl
arpack.jl euler.jl linalg3.jl readdlm.jl
spawn.jl
arrayops.jl examples.jl linalg4.jl reduce.jl
statistics.jl
arrayperf.jl fft.jl lineedit.jl reducedim.jl
strings.jl
backtrace.jl file.jl math.jl regex.jl
suitesparse.jl
bigint.jl float16.jl mod2pi.jl remote.jl
sysinfo.jl
bitarray.jl floatapprox.jl mpfr.jl repl.jl
test.jl
blas.jl functional.jl netload replcompletions.jl
test_sourcepath.jl
broadcast.jl git.jl numbers.jl resolve.jl
testdefs.jl
ccall.jl gitutils.jl parallel.jl rounding.jl
unicode.jl
ccalltest.c goto.jl perf runtests.jl
version.jl
collections.jl hashing.jl pkg.jl show.jl
combinatorics.jl iobuffer.jl pollfd.jl simdloop.jl
complex.jl keywordargs.jl priorityqueue.jl socket.jl

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3
_
_ _ ()_ | A fresh approach to technical computing
() | () () | Documentation: http://docs.julialang.org
_ _ | | __ _ | Type "help()" to list help topics
| | | | | | |/ ` | |
| | || | | | (| | | Version 0.3.0-prerelease+4028 (2014-07-02 23:42
UTC)
/ |_'|||__'| | Commit 2185bd1 (0 days old master)
|__/ | x86_64-w64-mingw32

julia> quit()

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Piledriver

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading
E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Bulldozer

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading
E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>set
OPENBLAS_CORETYPE=Barcelona

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl
linalg4
* linalg4
Triangular matrices
elty is Float32, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Float64, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
Linear solver
Eigensystems
Condition number tests - can be VERY approximate
Binary operations
elty is BigFloat, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solver
Eigensystems
Binary operations
Linear solver
Eigensystems
Binary operations
Tridiagonal matrices
relty is Float32, elty is Float32
Simple unary functions
Binary operations
relty is Float32, elty is Complex{Float32}
Simple unary functions
Binary operations
relty is Float64, elty is Float64
Simple unary functions
Binary operations
relty is Float64, elty is Complex{Float64}
Simple unary functions
Binary operations
SymTridiagonal (symmetric tridiagonal) matrices
elty is Float32, relty is Float32
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
elty is Float64, relty is Float64
Idempotent tests
Simple unary functions
Eigensystems
This tests eigenvalue and eigenvector computations using stebz! and
stein!
stein! call using iblock and isplit
Binary operations
Bidiagonal matrices
elty is Float32, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{Float32}, relty is Float32
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Float64, relty is Float64
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{Float64}, relty is Float64
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is BigFloat, relty is BigFloat
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Test upper and lower bidiagonal matrices
isupper is: true
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
isupper is: false
Idempotent tests
Linear solver
Eigensystems
Singular systems
Binary operations
Diagonal matrices
elty is Float32, relty is Float32
Linear solve
Simple unary functions
Binary operations
elty is Complex{Float32}, relty is Float32
Linear solve
Simple unary functions
Binary operations
elty is Float64, relty is Float64
Linear solve
Simple unary functions
Binary operations
elty is Complex{Float64}, relty is Float64
Linear solve
Simple unary functions
Binary operations
elty is BigFloat, relty is BigFloat
Linear solve
Simple unary functions
Binary operations
elty is Complex{BigFloat}, relty is BigFloat
Linear solve
Simple unary functions
Binary operations
Test interconversion between special matrix types
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
isupper is true
newtype is Bidiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
isupper is false
newtype is Bidiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
newtype is Diagonal{T}
newtype is Bidiagonal{T}
newtype is SymTridiagonal{T}
newtype is Tridiagonal{T}
newtype is Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}
newtype is Array{T,2}
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>

##############

On 7/2/14, Andreas Noack Jensen [email protected] wrote:

Is it right to say the the short version of this is that @jasax has
Piledriver cores, but the tests fails when OpenBLAS' Piledriver kernels
are
used?

Our problem is to figure out which call is actually causing the error
because the original problem of this thread is solved, right?

I have pushed some debug
options
to linalg4.jl. @jasax Can you try to build from latest source and run
the
linalg tests with debug=true in the top of linalg4.jl?

---

Reply to this email directly or view it on GitHub:

#117

Reply to this email directly or view it on GitHub:
#117
Hi,

I found and fixed a bug in the daxpy kernel under Windows. Only
bulldozer and piledriver were
affected. You could try my development fork:
https://github.com/wernsaar/OpenBLAS.git

Best regards
Werner

---

Reply to this email directly or view it on GitHub:

#117

Reply to this email directly or view it on GitHub:
#117

[timholy]

@wernsaar, I don't run Windows, but just wanted to comment: thanks for the fix!

[tkelman]

Ditto to what @timholy said (except that I use both Windows and Linux), these fixes are much appreciated. For these BLAS/LAPACK tests especially, I think it could make an interesting project to try auto-generating from Julia a C-reproduction of the inputs to tests that fail involving a ccall.

Assuming this updated version passes tests on the problematic AMD processor (I went and verified that it doesn't cause any regressions on Windows Haswell or Sandy Bridge either), @wernsaar would you recommend that we apply the fixes across all architectures, or would these latest changes only effect Windows?

[wernsaar]

Hi,

I think, you can safely apply the fixes to all x86_64 architectures and
processors.
There are still a few lapack issues for 32bit platforms ( x86, ARMV6,
ARMV7), but it's
very difficult to find these errors, because Netlib-Refblas also has a
lot of issues.

Best regards
Werner

On 07.07.2014 13:36, Tony Kelman wrote:

Ditto to what @timholy said, these fixes are much appreciated. For these BLAS/LAPACK tests especially, I think it could make an interesting project to try auto-generating from Julia a C-reproduction of the inputs to tests that fail inside a ccall.

Assuming this updated version passes tests on the problematic AMD family (I went and verified that it doesn't cause any regressions on Windows Haswell or Sandy Bridge either), @wernsaar would you recommend that we apply the fixes across all architectures, or would these latest changes only effect Windows?

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Hi all,

Sorry, I hadn't read the message when I replied about not having the .dll.

OK, to summarize the results below, it seems the issue is solved now.
I re-installed today's 0.3 pre-release 64 bits build for windows.
First I tested linalg4 with the delivered libopenblas.dll and there
was still an error like in the previous tests. Then I replaced that
dll by the one sent by Tony
(openblas-13348b21373848f3-x86_64-w64-mingw32.7z) and the test passed!

Congratulations to all.

Report below, with the output truncated. My messages are enclosed in
XXXXXX .... XXXXXX

Best Regards

jose

#################

| | | | | | |/ ` | |
| | || | | | (| | | Version 0.3.0-prerelease+4079 (2014-07-06 10:37 UTC)
/ |_'|||'_| | Commit 1d94ff9 (1 day old master)
|/ | x86_64-w64-mingw32

julia> quit()

XXXXXXXXX TEST with today's 0.3 dev DLL => Still Gives Error XXXXXXXXX

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
.................................................................

0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.7573401233330379
-0.08910849321718631 -0.003154149991984479
6
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.902025937083369 0.01507504988414109
0.0 0.0 0.0 0.0 0.0 0.0 -0.0 0.0 -0.0 0.0 0.0 0.9266700197042705]
exception on 1: ERROR: assertion failed: |deviation - 0.0| <=
0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
ERROR: assertion failed: |deviation - 0.0| <= 0.00020599365272744308
deviation = 1.0000000521112276
0.0 = 0.0
difference = 1.0000000521112276 > 0.00020599365272744308
in error at error.jl:22
while loading linalg4.jl, in expression starting on line 174
while loading E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

XXXXXXX FROM NOW ON I REPLACED 0.3-dev libopenblas.dll XXXXXXXXXXX
XXXXXXX with openblas-13348b21373848f3-x86_64-w64-mingw32.7z -> Success XXXXXXX
XXXXXXX with implicit core (Piledriver IMHO) and explicit Piledriver,
Bulldozer XXXXXXX
XXXXXXX Results (matrix displays) are chopped for clarity XXXXXXXXXXXXX

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821
0.0 0.0 0.0 0.0 -0.022935344 0.6306952 0.27791148
-0.0028916348 0.67452693 -0.10671823 0.175
70455 -0.06964667
0.0 0.0 0.0 0.0 0.0 0.23474121 0.07669833 6.274062e-5
0.33820927 0.035774022 0.10542922 -0.0
074075013
0.0 0.0 0.0 0.0 0.0 0.0 0.016797159 5.625236e-5 -0.09946941
0.057604518 -0.05467781 -0.00448
2299
0.0 0.0 0.0 0.0 0.0 0.0 0.0 -0.00010766533 -0.14416896
-0.22637479 -0.113924906 0.0048884703

...............................................................

5257 0.8493899270733373 0.0
-0.27723959944257115 -0.01256389469460509 0.019922685332619098
-0.00015230646890979186 -0.001007766
6212015034 0.019475304347003956 0.7655495676254432 -0.1062247592456354
0.2655572178861536 -0.8344895
527772469 0.5277658114981975 1.0]
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Bulldozer

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821

   ..............................................................................

0.128535142207157 -0.03796500118018003 0.0558528170695982
-0.0007739323739441249 -0.004228422736177
874 0.054815197837779626 -0.5126761248349504 -0.21885187863741135
0.25142463507793794 0.502022703611
5257 0.8493899270733373 0.0
-0.27723959944257115 -0.01256389469460509 0.019922685332619098
-0.00015230646890979186 -0.001007766
6212015034 0.019475304347003956 0.7655495676254432 -0.1062247592456354
0.2655572178861536 -0.8344895
527772469 0.5277658114981975 1.0]
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>set OPENBLAS_CORETYPE=Piledriver

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
v1:Float32[1.0 0.67756957 0.09913982 0.9858403 0.95147824 0.185476
0.54939574 0.9966917 0.05068591 0
.08416568 0.006229756 0.25610757
0.0 -0.73545873 0.17985778 0.054556526 -0.11083308 -0.30371377
-0.68361366 -0.027725674 -0.1
0976031 0.10549428 -0.016001608 -0.14114743
0.0 0.0 -0.97868407 -0.116793774 0.20153184 -0.31463885
-0.2038027 0.055027988 -0.22246759 -
0.46302384 -0.04453673 0.14152703
0.0 0.0 0.0 0.10724621 -0.20313585 0.56699127 0.32542536
-0.05291864 0.4622218 0.13555765 0.
10190635 -0.17476821
0.0 0.0 0.0 0.0 -0.022935344 0.6306952 0.27791148
-0.0028916348 0.67452693 -0.10671823 0.175
70455 -0.06964667

...................................................................

5257 0.8493899270733373 0.0
-0.27723959944257115 -0.01256389469460509 0.019922685332619098
-0.00015230646890979186 -0.001007766
6212015034 0.019475304347003956 0.7655495676254432 -0.1062247592456354
0.2655572178861536 -0.8344895
527772469 0.5277658114981975 1.0]
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>

#################

On 7/7/14, Tony Kelman [email protected] wrote:

@jasax that's exactly what I did -
http://sourceforge.net/projects/juliadeps-win/files/openblas-13348b21373848f3-x86_64-w64-mingw32.7z

---

Reply to this email directly or view it on GitHub:
#117

[tkelman]

🍻 fantastic, thanks for being such a champ and testing this repeatedly. As soon as there's a v0.2.10.rc2 to bump to, we can close this one (5th most commented issue yet).

[andreasnoack]

Great. Thank you for all the tests @jasax. Also thanks to @tkelman for the persistence in this issue and to @wernsaar for fixing it.

[jasax]

Hi all,

My pleasure. I'll try to resume working with and learning :-) Julia in
a couple weeks. Whenever you want to test julia issues in Piledriver
architecture feel free to
ask me. And congratulations to @tkelman, @wernsaar and
@andreasnoackjensen (and everybody else who helped) for being so
persistent in hunting this perfidious
bug :-)

Best Regards,

Jose

On 7/7/14, Andreas Noack Jensen [email protected] wrote:

Great. Thank you for all the tests @jasax. Also thanks to @tkelman for the
persistence in this issue and to @wernsaar for fixing it.

---

Reply to this email directly or view it on GitHub:
#117

[ViralBShah]

This is a really amazing bugfix. A big thanks to all who tracked it down and fixed it.

[ViralBShah]

Julia is now bumped to use openblas v0.2.10.rc2 that has this bugfix. @jasax if you can verify in a couple of days with the nightlies that use this, we can finally close this issue.

[jasax]

Hi Viral,

I'll try in a few hours with this night's build of julia 0.3-dev and report.

Best regards

Jose

On 7/9/14, Viral B. Shah [email protected] wrote:

Julia is now bumped to use v0.2.10.rc2 that has this bugfix. @jasax if you
can verify in a couple of days with the nightlies that use this, we can
finally close this issue.

---

Reply to this email directly or view it on GitHub:
#117

[tkelman]

@jasax may as well run the entire test set, if you don't mind. I expect a failure at spawn that looks like could not spawn echo hello | sort: no such file or directory (ENOENT) which there's a fix pending for, but all of the linalg tests should pass now.

[jasax]

Hi,

Just installed the most recent julia 0.3-dev 64 bits for windows.
Still some failures in linalg tests... And I don't think the error
when testing all "linalg" is the one you mention...

Test to "linalg4" only has success.
Test to "linalg" fails.
Test to "all" also fails (this one with the "echo" error you
mentioned, I think..).

Report below.

Best Regards

Jose

           _

_ _ ()_ | A fresh approach to technical computing
() | () () | Documentation: http://docs.julialang.org
_ _ | | __ _ | Type "help()" to list help topics
| | | | | | |/ ` | |
| | || | | | (| | | Version 0.3.0-prerelease+4095 (2014-07-09 02:16 UTC)
/ |_'|||__'| | Commit 2b6d708 (0 days old master)
|__/ | x86_64-w64-mingw32

julia> quit()

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg4
* linalg4
SUCCESS
E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg
From worker 5: * linalg4
From worker 3: * linalg2
From worker 2: * linalg1
From worker 4: * linalg3
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
jl_profile_is_running at ???:1840906582
jl_profile_is_running at ???:1840906198
jl_profile_is_running at ???:1840905764
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840907044
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840907044
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840907044
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840907044
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906594
jl_profile_is_running at ???:1840906818
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840906707
jl_profile_is_running at ???:1840905959
jl_profile_is_running at ???:1840905946
jl_gc_collect at ???:1840908877
allocobj at ???:1840912677
jl_alloc_tuple_uninit at ???:1840820997
jl_subtype at ???:1840532712
jl_reinstantiate_inner_types at ???:1840534132
jl_reinstantiate_inner_types at ???:1840537387
jl_reinstantiate_inner_types at ???:1840538432
jl_reinstantiate_inner_types at ???:1840535660
jl_type_match_ at ???:1840551049
jl_type_intersection_matching at ???:1840554065
jl_init_types at ???:1840574610
jl_init_types at ???:1840575724
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Please submit a bug report with steps to reproduce this fault, and any
error messages that follow (i
n their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0x6dba0156 --
jl_profile_is_running at ???:1840906582
_methods at reflection.jl:81
Worker 2 terminated.
ERROR: ProcessExitedException()
in wait at task.jl:279
in stream_wait at stream.jl:257
while loading E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl all
From worker 2: * linalg1
From worker 3: * linalg2
From worker 5: * linalg4
From worker 6: * core
From worker 8: * numbers
From worker 7: * keywordargs
From worker 9: * strings
From worker 4: * linalg3
From worker 7: * collections
From worker 6: * hashing
From worker 9: * remote
From worker 9: * iobuffer
From worker 9: * arrayops
From worker 7: * reduce
From worker 6: * reducedim
From worker 7: * simdloop
From worker 7: * blas
From worker 7: * fft
From worker 6: * dsp
From worker 7: * sparse
From worker 6: * bitarray
From worker 9: * random
From worker 4: * math
From worker 9: * functional
From worker 8: * bigint
From worker 9: * sorting
From worker 8: * statistics
From worker 7: * spawn
exception on 7: From worker 9: * backtrace
From worker 9: * priorityqueue
ERROR: test error during readall(@cmd "echo hello | sort") == "hello | sort\n"
could not spawn echo hello | sort: no such file or directory (ENOENT)
in _jl_spawn at process.jl:217
while loading spawn.jl, in expression starting on line 16
ERROR: test error during readall(@cmd "echo hello | sort") == "hello | sort\n"
could not spawn echo hello | sort: no such file or directory (ENOENT)
while loading spawn.jl, in expression starting on line 16
while loading E:\math\Julia-0.3.0-prerelease\share\julia\test\runtests.jl,
in expression starting on
line 46

On 7/9/14, Tony Kelman [email protected] wrote:

@jasax may as well run the entire test set, if you don't mind. I expect a
failure at spawn that looks like could not spawn echo hello | sort: no such file or directory (ENOENT) which there's a fix pending for, but all
of the linalg tests should pass now.

---

Reply to this email directly or view it on GitHub:
#117

[jasax]

Hello,

Upgrade results on tests.

I retried twice the test to "linalg" and this time it was successfull
in both trials. Results below.

Since the previous error was an "access violation", can it be the case
that the error is related with dynamic memory allocation (stack size,
garbage collection,...?) and so it is somewhat a sort of a random
error? Which would be bad...

Regards

Jose

##############

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg
From worker 2: * linalg1
From worker 4: * linalg3
From worker 3: * linalg2
From worker 5: * linalg4
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>julia3 runtests.jl linalg
From worker 3: * linalg2
From worker 5: * linalg4
From worker 2: * linalg1
From worker 4: * linalg3
SUCCESS

E:\math\Julia-0.3.0-prerelease\share\julia\test>

##############

On 7/9/14, Tony Kelman [email protected] wrote:

@jasax may as well run the entire test set, if you don't mind. I expect a
failure at spawn that looks like could not spawn echo hello | sort: no such file or directory (ENOENT) which there's a fix pending for, but all
of the linalg tests should pass now.

---

Reply to this email directly or view it on GitHub:
#117

[vtjnash]

Yes, the stack backtrace shows that it failed in the gc mark pass.

[ViralBShah]

Let's have a new issue for the memory violation. I am closing this one.