This a draft for dgecx.f. #1161
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mgates3
requested changes
Oct 15, 2025
There was a problem hiding this comment.
There is both DGECX.f (incorrect) and dgecx.f (correct) file? This file should be removed.
There was a problem hiding this comment.
You can fix this fairly easily by dropping the 4619f61 commit. Normally I would do that with git rebase -i, but here I had difficulties because of the filename conflict between dgecx.f and DGECX.f. This is what works for me:
git checkout master
# Create a temporary branch pointer.
git switch -c tmp
# Copy the 2 commits we want to keep.
git cherry-pick df8c7393a b36b9b76e
# Force move (capital -C) the branch pointer to the tmp branch.
git switch -C scr2016-dgecx-draft tmp
# The log should show the new branch with 2 commits, the old branch with 3 commits.
git log --oneline --graph --all
* 47b924cd7 (HEAD -> scr2016-dgecx-draft, tmp) Fix documentation for ABSTOL and factor flags
* 69a932e32 Added dgecx.f, a DRAFT for DGECX routine
| * b36b9b76e (igor/scr2016-dgecx-draft) Fix documentation for ABSTOL and factor flags
| * df8c7393a Added dgecx.f, a DRAFT for DGECX routine
| * 4619f613c Addded a DRAFT of DGECS routine.
|/
* 66380df1c (github/master, github/HEAD, master) Merge pull request #1157 from amontoison/precompilation
# If everything looks good, force push to your branch.
# I'm assuming that `origin` points to your repo, [email protected]:scr2016/lapack.git
git push -f origin scr2016-dgecx-draft
# Clean up tmp branch pointer.
git branch -D tmp
| *> Specifies how the factors of CX factorization | ||
| *> are returned. | ||
| *> | ||
| *> = 'P' or 'p' : return only the column permutaion matrix P |
There was a problem hiding this comment.
Pedantic: LAPACK doesn't normally specify both Uppercase and lowercase. Cf.:
lapack/SRC> grep "^\*> += '\w'" lapack/SRC/zpotrf.f
*> = 'U': Upper triangle of A is stored;
*> = 'L': Lower triangle of A is stored.
lapack/SRC> grep "^\*> += '\w'" lapack/SRC/*.f`
| *> matrix in the blocked step auxiliary subroutine DLAQP3RK ). | ||
| *> \endverbatim | ||
| *> | ||
| *> \param[out] LIWORK |
| *> The dimension of the array LIWORK. LIWORK >= N | ||
| *> | ||
| *> If LIWORK = -1, then a workspace query is assumed; the routine | ||
| *> only calculates the optimal size of the WORK array, returns |
SRC/dgecx.f
Outdated
| * | ||
| * DGECX | ||
| * | ||
| END No newline at end of file |
There was a problem hiding this comment.
Files should end with a newline. Notice ⊖ on Github.
| NBOPT = ILAENV( INB, 'DGEQRF', ' ', M, N, -1, -1 ) | ||
| LWKOPT = 1000 | ||
| END IF | ||
| WORK( 1 ) = DBLE( LWKOPT ) |
There was a problem hiding this comment.
Add:
LWORK( 1 ) = N
or whatever the formula is. Also elsewhere.
| * | ||
| END IF | ||
| * | ||
| WORK( 1 ) = DBLE( LWKOPT ) |
There was a problem hiding this comment.
Add:
LWORK( 1 ) = N
as above.
| * MSUB. | ||
| * | ||
| INFO = -6 | ||
| WORK( 1 ) = DBLE( LWKOPT ) |
There was a problem hiding this comment.
Add:
LWORK( 1 ) = N
or whatever the formula is. Also elsewhere.
|
by mistake 2 branches (one with dgecx.f and one with DGECX.f ) were joined
in a single pull request, the files are identical. DGECX.f can be removed.
…On Wed, Oct 15, 2025 at 11:15 AM Mark Gates ***@***.***> wrote:
***@***.**** requested changes on this pull request.
Some initial comments.
------------------------------
On SRC/DGECX.f
<#1161 (comment)>
:
There is both DGECX.f (incorrect) and dgecx.f (correct) file? This file
should be removed.
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +*> array as the first K elements.
+*> b) If the flag COMP_X is set, then the routine also explicitly
+*> computes and returns the factor X = pseudoinv(C) * A.
+*>
+*> \endverbatim
+*
+* Arguments:
+* ==========
+*
+*> \param[in] FACT
+*> \verbatim
+*> FACT is CHARACTER*1
+*> Specifies how the factors of CX factorization
+*> are returned.
+*>
+*> = 'P' or 'p' : return only the column permutaion matrix P
Pedantic: LAPACK doesn't normally specify both Uppercase and lowercase.
Cf.:
lapack/SRC> grep "^\*> += '\w'" lapack/SRC/zpotrf.f
*> = 'U': Upper triangle of A is stored;
*> = 'L': Lower triangle of A is stored.
lapack/SRC> grep "^\*> += '\w'" lapack/SRC/*.f`
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +*>
+*> If LWORK = -1, then a workspace query is assumed; the routine
+*> only calculates the optimal size of the WORK array, returns
+*> this value as the first entry of the WORK array, and no error
+*> message related to LWORK is issued by XERBLA.
+*> \endverbatim
+*>
+*> \param[out] IWORK
+*> \verbatim
+*> IWORK is INTEGER array, dimension (N).
+*> Is a work array. ( IWORK is used by DGEQP3RK to store indices
+*> of "bad" columns for norm downdating in the residual
+*> matrix in the blocked step auxiliary subroutine DLAQP3RK ).
+*> \endverbatim
+*>
+*> \param[out] LIWORK
Input, right? \param[in]
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +*>
+*> \param[out] IWORK
+*> \verbatim
+*> IWORK is INTEGER array, dimension (N).
+*> Is a work array. ( IWORK is used by DGEQP3RK to store indices
+*> of "bad" columns for norm downdating in the residual
+*> matrix in the blocked step auxiliary subroutine DLAQP3RK ).
+*> \endverbatim
+*>
+*> \param[out] LIWORK
+*> \verbatim
+*> LIWORK is INTEGER
+*> The dimension of the array LIWORK. LIWORK >= N
+*>
+*> If LIWORK = -1, then a workspace query is assumed; the routine
+*> only calculates the optimal size of the WORK array, returns
Typo: IWORK array.
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +* Copy matrix C into work array WORK.
+*
+ CALL DLACPY( 'F', M, K, C, LDC, WORK, M )
+*
+ CALL DGELS( 'N', M, K, N, WORK, M, X, LDX,
+ $ WORK( M*K+1 ), LWORK,
+ $ IINFO )
+ INFO = IINFO
+*
+ END IF
+*
+ WORK( 1 ) = DBLE( LWKOPT )
+*
+* DGECX
+*
+ END
Files should end with a newline. Notice ⊖ on Github.
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +* code.
+*
+ IF( INFO.EQ.0 ) THEN
+ MINMN = MIN( M, N )
+ IF( MINMN.EQ.0 ) THEN
+ LWKMIN = 1
+ LWKOPT = 1
+ ELSE
+ LWKMIN = 3*N + 1
+*
+* Assign to NBOPT optimal block size.
+*
+ NBOPT = ILAENV( INB, 'DGEQRF', ' ', M, N, -1, -1 )
+ LWKOPT = 1000
+ END IF
+ WORK( 1 ) = DBLE( LWKOPT )
Add:
LWORK( 1 ) = N
or whatever the formula is. Also elsewhere.
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +* that uses QR factorization. For that purpose, we store C into
+* WORK array WORK(1:M*K), and the matrix A was copied into
+* the array X at the begining of the routine.
+*
+* Copy matrix C into work array WORK.
+*
+ CALL DLACPY( 'F', M, K, C, LDC, WORK, M )
+*
+ CALL DGELS( 'N', M, K, N, WORK, M, X, LDX,
+ $ WORK( M*K+1 ), LWORK,
+ $ IINFO )
+ INFO = IINFO
+*
+ END IF
+*
+ WORK( 1 ) = DBLE( LWKOPT )
Add:
LWORK( 1 ) = N
as above.
------------------------------
In SRC/dgecx.f
<#1161 (comment)>
:
> +*
+ MNSUB = MIN(MSUB, NSUB)
+ MRESID = MSUB-NSEL
+ NRESID = NSUB-NSEL
+ IF( NSEL.GT.0 ) THEN
+ IF( MSUB.LT.NSEL ) THEN
+* TODO: Move this part to the top of the routine.
+* a) Case MSUB < NSEL.
+* When the number of preselected columns
+* is larger than MSUB, hence the factorization of all NSEL
+* columns cannot be completed. Return from the routine with the
+* error of COL_SEL_DESEL parameter. NSEL cannot be larger than
+* MSUB.
+*
+ INFO = -6
+ WORK( 1 ) = DBLE( LWKOPT )
Add:
LWORK( 1 ) = N
or whatever the formula is. Also elsewhere.
—
Reply to this email directly, view it on GitHub
<#1161 (review)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAHYAZG2IG6CWKRVGDULGWT3X2FMPAVCNFSM6AAAAACJHEAYW6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTGNBRGUZDGMZSGU>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
changed the description of DGECX
ilayn
reviewed
Oct 16, 2025
| *> | ||
| *> A * P(K) = C*X + A_resid, where | ||
| *> | ||
| *> C is an M-by-K matrix which is a subset of K columns selected |
There was a problem hiding this comment.
I think with "subset" you mean "spanning the same subspace" or "in the range of K columns of A" here. Is that right?
| *> | ||
| *> DGECX computes a CX factorization of a real M-by-N matrix A: | ||
| *> | ||
| *> A * P(K) = C*X + A_resid, where |
There was a problem hiding this comment.
I think P(K) notation is confusing here. It's not a function of K but depends on K parameter.
| *> from the original matrix A, | ||
| *> | ||
| *> X is a K-by-N matrix that minimizes the Frobenius norm of the | ||
| *> residual matrix A_resid, X = pseudoinv(C) * A, |
There was a problem hiding this comment.
You can just say "minimizes "|A - C*X|" in the Frobenius norm".
| *> Column selection stage 1. | ||
| *> ========================= | ||
| *> | ||
| *> The user can select N_sel columns and deselect N_desel columns |
There was a problem hiding this comment.
Why can't I just supply a list of columns that I want and leave this deselect business out? Then it will be two options, "A" is all columns, "S" selected columns. If "A" then cols array is not referenced otherwise it is read from cols array.
I'm afraid this is getting a bit too complicated with explicit deselection. Same goes for the rows
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
his a draft for dgecx.f. to REVIEW and COMMENT