Re: [Scilab-users] spec can crash scilab
Hello all, Well, my workaround works up to a certain point. When I increase the size of the matrix under study, I face another blocking issue: free(): corrupted unsorted chunks Antoine On 05/05/2021 09:23, Antoine Monmayrant wrote: Hello all, I might have found a workaround, but I still need to ensure that I get exactly the same end results in my code. It seems that I can replace: [W2, gammas] = spec(A); //crashes by [al,be,W2]=spec(A,eye(A)); //fine gammas=diag(al./be); If anyone has a solution with eigs, I'll take it. Could my workaround help investigate where the bug is in the gateway between scilab and lapack? Cheers, Antoine On 05/05/2021 08:22, Antoine Monmayrant wrote: On 05/05/2021 00:09, Samuel Gougeon wrote: Le 04/05/2021 à 16:38, Antoine Monmayrant a écrit : Yes, it does the same So this bug is still a problem, at least on Ubuntu. Can you guys try to run spec.tst on his machine and comment on https://bugzilla.scilab.org/show_bug.cgi?id=15330 ? Hello Samuel, Besides, how confusing and messy the notations on the spec () documentation page are! ... * R defined twice but differently among arguments * L defined among arguments but not used in syntaxes * Q and Z used in syntaxes but not described among arguments * al, be used in descriptions, instead of alpha and beta listed in arguments * E used in the description, but not in arguments nor in syntaxes... and so on.. In addition, notations differ from a language version to another one... Some clarification is urgently required. Indeed. It is also a big problem for portability to get such a basic functionality (solving eigenvalue problems in a linear algebra software!) only working on some platforms! In next release, eigs() -- mainly based on the former Arnoldi module, that becomes internal -- is listed in the same help section. Have you tried it? Well, I might be a bit far from my field of expertise, but it does not seem that eigs can compute all the eigenvalues/eigenvectors, does it? It seems that by default it computes the 6 biggest ones and can go up to rank(A)-2. In my use case, I need all the eigenvalues... Antoine Samuel ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] spec can crash scilab
Hello all, I might have found a workaround, but I still need to ensure that I get exactly the same end results in my code. It seems that I can replace: [W2, gammas] = spec(A); //crashes by [al,be,W2]=spec(A,eye(A)); //fine gammas=diag(al./be); If anyone has a solution with eigs, I'll take it. Could my workaround help investigate where the bug is in the gateway between scilab and lapack? Cheers, Antoine On 05/05/2021 08:22, Antoine Monmayrant wrote: On 05/05/2021 00:09, Samuel Gougeon wrote: Le 04/05/2021 à 16:38, Antoine Monmayrant a écrit : Yes, it does the same So this bug is still a problem, at least on Ubuntu. Can you guys try to run spec.tst on his machine and comment on https://bugzilla.scilab.org/show_bug.cgi?id=15330 ? Hello Samuel, Besides, how confusing and messy the notations on the spec () documentation page are! ... * R defined twice but differently among arguments * L defined among arguments but not used in syntaxes * Q and Z used in syntaxes but not described among arguments * al, be used in descriptions, instead of alpha and beta listed in arguments * E used in the description, but not in arguments nor in syntaxes... and so on.. In addition, notations differ from a language version to another one... Some clarification is urgently required. Indeed. It is also a big problem for portability to get such a basic functionality (solving eigenvalue problems in a linear algebra software!) only working on some platforms! In next release, eigs() -- mainly based on the former Arnoldi module, that becomes internal -- is listed in the same help section. Have you tried it? Well, I might be a bit far from my field of expertise, but it does not seem that eigs can compute all the eigenvalues/eigenvectors, does it? It seems that by default it computes the 6 biggest ones and can go up to rank(A)-2. In my use case, I need all the eigenvalues... Antoine Samuel ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] spec can crash scilab
On 05/05/2021 00:09, Samuel Gougeon wrote: Le 04/05/2021 à 16:38, Antoine Monmayrant a écrit : Yes, it does the same So this bug is still a problem, at least on Ubuntu. Can you guys try to run spec.tst on his machine and comment on https://bugzilla.scilab.org/show_bug.cgi?id=15330 ? Hello Samuel, Besides, how confusing and messy the notations on the spec () documentation page are! ... * R defined twice but differently among arguments * L defined among arguments but not used in syntaxes * Q and Z used in syntaxes but not described among arguments * al, be used in descriptions, instead of alpha and beta listed in arguments * E used in the description, but not in arguments nor in syntaxes... and so on.. In addition, notations differ from a language version to another one... Some clarification is urgently required. Indeed. It is also a big problem for portability to get such a basic functionality (solving eigenvalue problems in a linear algebra software!) only working on some platforms! In next release, eigs() -- mainly based on the former Arnoldi module, that becomes internal -- is listed in the same help section. Have you tried it? Well, I might be a bit far from my field of expertise, but it does not seem that eigs can compute all the eigenvalues/eigenvectors, does it? It seems that by default it computes the 6 biggest ones and can go up to rank(A)-2. In my use case, I need all the eigenvalues... Antoine Samuel ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] spec can crash scilab
Le 04/05/2021 à 16:38, Antoine Monmayrant a écrit : Yes, it does the same So this bug is still a problem, at least on Ubuntu. Can you guys try to run spec.tst on his machine and comment on https://bugzilla.scilab.org/show_bug.cgi?id=15330 ? Besides, how confusing and messy the notations on the spec () documentation page are! ... * R defined twice but differently among arguments * L defined among arguments but not used in syntaxes * Q and Z used in syntaxes but not described among arguments * al, be used in descriptions, instead of alpha and beta listed in arguments * E used in the description, but not in arguments nor in syntaxes... and so on.. In addition, notations differ from a language version to another one... Some clarification is urgently required. In next release, eigs() -- mainly based on the former Arnoldi module, that becomes internal -- is listed in the same help section. Have you tried it? Samuel ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] spec can crash scilab
Yes, it does the same So this bug is still a problem, at least on Ubuntu. Can you guys try to run spec.tst on his machine and comment on https://bugzilla.scilab.org/show_bug.cgi?id=15330 ? Antoine On 04/05/2021 16:32, Stéphane Mottelet wrote: Hi, it seems to be the same bug as https://bugzilla.scilab.org/show_bug.cgi?id=15330 S. Le 04/05/2021 à 16:27, Antoine Monmayrant a écrit : Hello all, I've been fighting during the last few days to nail down a segmentation fault in scilab, that seems to plague linux but not windows. Could you try to reproduce it? Just save the two attached file in a folder, cd to it and run the script. On my machine¹, the call to spec with one output argument is OK, but the second one with two output arguments leads to a seg fault. Can you let me know how it goes on your system? Here ¹scilab 6.1.0 on Linux Ubuntu 18.04 64bits ___ users mailing list users@lists.scilab.org https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users -- Stéphane Mottelet Ingénieur de recherche EA 4297 Transformations Intégrées de la Matière Renouvelable Département Génie des Procédés Industriels Sorbonne Universités - Université de Technologie de Compiègne CS 60319, 60203 Compiègne cedex Tel : +33(0)344234688 http://www.utc.fr/~mottelet ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] spec can crash scilab
Hi, it seems to be the same bug as https://bugzilla.scilab.org/show_bug.cgi?id=15330 S. Le 04/05/2021 à 16:27, Antoine Monmayrant a écrit : Hello all, I've been fighting during the last few days to nail down a segmentation fault in scilab, that seems to plague linux but not windows. Could you try to reproduce it? Just save the two attached file in a folder, cd to it and run the script. On my machine¹, the call to spec with one output argument is OK, but the second one with two output arguments leads to a seg fault. Can you let me know how it goes on your system? Here ¹scilab 6.1.0 on Linux Ubuntu 18.04 64bits ___ users mailing list users@lists.scilab.org https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users -- Stéphane Mottelet Ingénieur de recherche EA 4297 Transformations Intégrées de la Matière Renouvelable Département Génie des Procédés Industriels Sorbonne Universités - Université de Technologie de Compiègne CS 60319, 60203 Compiègne cedex Tel : +33(0)344234688 http://www.utc.fr/~mottelet ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
[Scilab-users] spec can crash scilab
Hello all, I've been fighting during the last few days to nail down a segmentation fault in scilab, that seems to plague linux but not windows. Could you try to reproduce it? Just save the two attached file in a folder, cd to it and run the script. On my machine¹, the call to spec with one output argument is OK, but the second one with two output arguments leads to a seg fault. Can you let me know how it goes on your system? Here ¹scilab 6.1.0 on Linux Ubuntu 18.04 64bits A_kills_spec.sod Description: Binary data bug_spec.sce Description: application/scilab-sce ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users