Fix Index-64 conversion errors by ACSimon33 · Pull Request #1278 · Reference-LAPACK/lapack

ACSimon33 · 2026-05-12T15:48:20Z

Description

Compiling with -Werror=conversion and -DBUILD_INDEX64 results in a lot of conversion warnings/errors. This MR fixes those by adding explicit conversions to the appropriate types or fixing wrong declarations. The conversion errors are mostly in the testing routines, but there are a few in the XBLAS-related routines as well.

The three types of conversion warnings are:

Error: Possible change of value in conversion from COMPLEX(8) to REAL(8) at (1) [-Werror=conversion]
Error: Possible change of value in conversion from INTEGER(8) to REAL(4) at (1) [-Werror=conversion]
Error: Possible change of value in conversion from REAL(4) to INTEGER(8) at (1) [-Werror=conversion]

…Index-64 builds

… builds

…Index-64 builds

… builds

Copilot

Pull request overview

This pull request addresses build failures when compiling LAPACK with -DBUILD_INDEX64 and -Werror=conversion by removing implicit/narrowing conversions across testing code and a few computational/XBLAS-related routines.

Changes:

Add explicit REAL(...), INT(...), and CMPLX(...) conversions (and a few literal-kind adjustments) to eliminate conversion warnings under Index-64 builds.
Fix a few incorrect/misleading declarations that caused problematic implicit conversions (e.g., statement-function result types).
Remove unused intermediate arrays/loops in the *EBCHVXX testing routines.

Reviewed changes

Copilot reviewed 171 out of 171 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
TESTING/LIN/zebchvxx.f	Remove unused `DIFF` array/loop to avoid unnecessary conversions.
TESTING/LIN/stsqr01.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/strt05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/stpt05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/stbt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/sqrt16.f	Cast `MAX(M,N)` to `REAL` in residual normalization.
TESTING/LIN/sqrt12.f	Replace `ZERO` with `0` in integer comparison.
TESTING/LIN/sqrt05.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/sqrt04.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/sptt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/sppt05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/spot05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/spbt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/sorhr_col02.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/sorhr_col01.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/slqt05.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/slqt04.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/slattr.f	Cast diagonal indices and `(N-2)` to `REAL` to avoid implicit conversions.
TESTING/LIN/slattp.f	Cast diagonal indices to `REAL` to avoid implicit conversions.
TESTING/LIN/slattb.f	Cast diagonal indices to `REAL` to avoid implicit conversions.
TESTING/LIN/sgtt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/sget07.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/sgbt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/sdrvrf4.f	Cast `MAX(N,1)` to `REAL` in normalization.
TESTING/LIN/sdrvrf3.f	Cast `MAX(M,N,1)` to `REAL` in normalization.
TESTING/LIN/sdrvrf1.f	Cast `LDA` to `REAL` when scaling `SMALL/LARGE`.
TESTING/LIN/schksy_rook.f	Use integer literal `0` in integer comparisons.
TESTING/LIN/schksy_rk.f	Use integer literal `0` in integer comparisons.
TESTING/LIN/schkps.f	Cast `N` to `REAL` before `CEILING` rank calculation.
TESTING/LIN/schkeq.f	Make `(-1)**(I+J)` multiplication explicit in `REAL` arithmetic.
TESTING/LIN/cunhr_col02.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/cunhr_col01.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/ctsqr01.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/ctrt05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/ctpt05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/ctbt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/cqrt16.f	Cast `MAX(M,N)` to `REAL` in residual normalization.
TESTING/LIN/cqrt12.f	Replace `ZERO` with `0` in integer comparison.
TESTING/LIN/cqrt05.f	Add explicit `REAL(MAX(...))` scaling factors; minor cleanup at EOF.
TESTING/LIN/cqrt04.f	Add explicit `REAL(MAX(...))` scaling factors; minor cleanup at EOF.
TESTING/LIN/cptt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/cppt05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/cpot05.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/cpbt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/clqt05.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/clqt04.f	Add explicit `REAL(MAX(...))` scaling factors in residual computations.
TESTING/LIN/clattr.f	Cast diagonal indices to `CMPLX(REAL(...))`; cast `(N-2)` to `REAL`.
TESTING/LIN/clattp.f	Cast diagonal indices to `CMPLX(REAL(...))`.
TESTING/LIN/clattb.f	Cast diagonal indices to `CMPLX(REAL(...))`.
TESTING/LIN/cgtt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/cget07.f	Cast `(N+1)` factors to `REAL` in error-bound scaling.
TESTING/LIN/cgbt05.f	Cast `NZ` factors to `REAL` in error-bound scaling.
TESTING/LIN/cebchvxx.f	Remove unused `DIFF` array/loop to avoid unnecessary conversions.
TESTING/LIN/cdrvrf4.f	Cast `MAX(N,1)` to `REAL` in normalization.
TESTING/LIN/cdrvrf3.f	Cast `MAX(M,N,1)` to `REAL` in normalization.
TESTING/LIN/cdrvrf1.f	Cast `LDA` to `REAL` when scaling `SMALL/LARGE`.
TESTING/LIN/cchksy_rook.f	Use integer literal `0` in integer comparisons.
TESTING/LIN/cchksy_rk.f	Use integer literal `0` in integer comparisons.
TESTING/LIN/cchkps.f	Cast `N` to `REAL` before `CEILING` rank calculation.
TESTING/LIN/cchkhe_rook.f	Use integer literal `0` in integer comparisons.
TESTING/LIN/cchkhe_rk.f	Use integer literal `0` in integer comparisons.
TESTING/LIN/cchkeq.f	Make `(-1)**(I+J)` multiplication explicit in `REAL` arithmetic.
TESTING/EIG/ssyt22.f	Cast `M` to `REAL` in ULP scaling to avoid implicit conversions.
TESTING/EIG/ssyt21.f	Cast `N` to `REAL` in ULP scaling; adjust result normalization.
TESTING/EIG/ssvdch.f	Cast `MAX(N/10,1)` to `REAL` in tolerance computation.
TESTING/EIG/sstt22.f	Cast `M` to `REAL` in ULP scaling.
TESTING/EIG/sstt21.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/sspt21.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/ssgt01.f	Cast `N` to `REAL` in ULP scaling for residuals.
TESTING/EIG/ssbt21.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/shst01.f	Cast `N` to `REAL` in `SMLNUM` and residual normalization.
TESTING/EIG/sget54.f	Cast `2*N` to `REAL` in ULP scaling.
TESTING/EIG/sget51.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/sget35.f	Cast integer-valued generators to `REAL` for matrix fills.
TESTING/EIG/sget32.f	Cast sign/ITVAL-generated integers to `REAL` for arithmetic.
TESTING/EIG/sget24.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/sget10.f	Cast `M` to `REAL` in EPS scaling.
TESTING/EIG/sdrvst2stg.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/sdrvst.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/sdrvsg2stg.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/sdrvsg.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/sdrvbd.f	Cast `MNMIN` to `REAL` and cast `(MNMIN-1)` before `SLARND` scaling.
TESTING/EIG/sdrgvx.f	Return workspace sizes as `REAL(MAXWRK)` under index64.
TESTING/EIG/sdrgsx.f	Return workspace sizes as `REAL(MAXWRK)` under index64.
TESTING/EIG/sdrgev3.f	Return workspace sizes as `REAL(MAXWRK)`; cast `N1` in magnitude setup.
TESTING/EIG/sdrgev.f	Return workspace sizes as `REAL(MAXWRK)`; cast `N1` in magnitude setup.
TESTING/EIG/sdrges3.f	Return workspace sizes as `REAL(MAXWRK)` under index64.
TESTING/EIG/sdrges.f	Return workspace sizes as `REAL(MAXWRK)` under index64.
TESTING/EIG/sckcsd.f	Cast `M` to `REAL` when scaling random index selection.
TESTING/EIG/schkst2stg.f	Cast `N` to `REAL` in norm setup and test-constant formulas.
TESTING/EIG/schkst.f	Cast `N` to `REAL` in norm setup and test-constant formulas.
TESTING/EIG/schksb2stg.f	Cast `N` to `REAL` in norm setup.
TESTING/EIG/schksb.f	Cast `N` to `REAL` in norm setup.
TESTING/EIG/schkhs.f	Cast `N` to `REAL` in norm setup.
TESTING/EIG/schkgg.f	Cast `N1` to `REAL` in magnitude setup.
TESTING/EIG/schkdmd.f90	Use `REAL(...,KIND=WP)` in tolerances to avoid kind conversions.
TESTING/EIG/schkbd.f	Cast `MAX(M,N,1)`/`MAX(M,N)` to `REAL`; cast before `SLARND` scaling.
TESTING/EIG/schkbb.f	Cast `MAX(M,N)` to `REAL` in norm setup.
TESTING/EIG/cstt22.f	Cast `M` to `REAL` in ULP scaling.
TESTING/EIG/cstt21.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/csgt01.f	Cast `N` to `REAL` in ULP scaling for residuals.
TESTING/EIG/chst01.f	Cast `N` to `REAL` in `SMLNUM` and residual normalization.
TESTING/EIG/chpt21.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/chet22.f	Cast `M` to `REAL` in ULP scaling.
TESTING/EIG/chet21.f	Cast `N` to `REAL` in ULP scaling and result normalization.
TESTING/EIG/chbt21.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/cget54.f	Cast `2*N` to `REAL` in ULP scaling.
TESTING/EIG/cget51.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/cget24.f	Cast `N` to `REAL` in ULP scaling.
TESTING/EIG/cget10.f	Cast `M` to `REAL` in EPS scaling.
TESTING/EIG/cdrvst2stg.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/cdrvst.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/cdrvsg2stg.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/cdrvsg.f	Cast `N` to `REAL` in norm setup and random index selection.
TESTING/EIG/cdrvbd.f	Cast `(MNMIN-1)` before `SLARND` scaling.
TESTING/EIG/cdrgvx.f	Return complex WORK query sizes with explicit integer→real conversion.
TESTING/EIG/cdrgsx.f	Return complex WORK query sizes with explicit integer→real conversion.
TESTING/EIG/cdrgev3.f	Return complex WORK query sizes as `CMPLX(REAL(MAXWRK))`; cast `N1`.
TESTING/EIG/cdrgev.f	Return complex WORK query sizes as `CMPLX(REAL(MAXWRK))`; cast `N1`.
TESTING/EIG/cdrges3.f	Return complex WORK query sizes with explicit integer→real conversion.
TESTING/EIG/cdrges.f	Return complex WORK query sizes with explicit integer→real conversion.
TESTING/EIG/cckcsd.f	Cast `M` to `REAL` when scaling random index selection.
TESTING/EIG/cchkst2stg.f	Cast `N` to `REAL` in norm setup and test-constant formulas.
TESTING/EIG/cchkst.f	Cast `N` to `REAL` in norm setup and test-constant formulas.
TESTING/EIG/cchkhs.f	Cast `N` to `REAL` in norm setup.
TESTING/EIG/cchkhb2stg.f	Cast `N` to `REAL` in norm setup.
TESTING/EIG/cchkhb.f	Cast `N` to `REAL` in norm setup.
TESTING/EIG/cchkgg.f	Cast `N1` to `REAL` in magnitude setup.
TESTING/EIG/cchkdmd.f90	Use `REAL(...,KIND=WP)` in tolerances to avoid kind conversions.
TESTING/EIG/cchkbd.f	Cast `MAX(M,N,1)`/`MAX(M,N)` to `REAL` in norm setup.
TESTING/EIG/cchkbb.f	Cast `MAX(M,N)` to `REAL` in norm setup.
SRC/zsyrfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/zporfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/zla_lin_berr.f	Fix statement-function result type to avoid complex→real conversion issues.
SRC/zherfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/zgerfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/zgbrfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/ssyrfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/sporfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/sla_syamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/sla_lin_berr.f	Cast `(NZ+1)` to `REAL` for `SAFE1` scaling.
SRC/sla_geamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/sla_gbamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/sgerfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/sgbrfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/dsyrfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/dporfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/dgerfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/dgbrfsx.f	Make `REF_TYPE` assignment explicit with `INT(...)`.
SRC/csyrfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/cporfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/cla_syamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/cla_lin_berr.f	Fix statement-function result type and cast `(NZ+1)` to `REAL` for scaling.
SRC/cla_heamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/cla_geamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/cla_gbamv.f	Cast `(N+1)` to `REAL` for `SAFE1` scaling.
SRC/cherfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/cgerfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
SRC/cgbrfsx.f	Make `REF_TYPE/ITHRESH` handling explicit with `INT(...)`/`REAL(...)`.
CBLAS/testing/c_zblat3.f	Fix `NULL` integer comparison (`0` vs `0.`).
CBLAS/testing/c_zblat2.f	Use double-precision literals in `DCMPLX` arguments.
CBLAS/testing/c_sblat3.f	Cast integer-valued expressions to `REAL` in test matrix generation.
CBLAS/testing/c_sblat2.f	Cast integer-valued expressions to `REAL` in test vector/matrix generation.
CBLAS/testing/c_sblat1.f	Cast integer literals to `REAL` in test-vector initialization.
CBLAS/testing/c_cblat3.f	Cast integer-valued expressions to `REAL` and fix `NULL` comparison.
CBLAS/testing/c_cblat2.f	Cast integer-valued expressions to `REAL` in test matrix/vector generation.
BLAS/TESTING/sblat3.f	Cast integer-valued expressions to `REAL` in test matrix generation.
BLAS/TESTING/sblat2.f	Cast integer-valued expressions to `REAL` in test vector/matrix generation.
BLAS/TESTING/sblat1.f	Cast integer literals to `REAL` in test-vector initialization.
BLAS/TESTING/cblat3.f	Cast integer-valued expressions to `REAL` in test matrix generation.
BLAS/TESTING/cblat2.f	Cast integer-valued expressions to `REAL` in test vector/matrix generation.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

ACSimon33 added 7 commits May 12, 2026 11:49

Fix -Wconversion errors in BLAS/TESTING during Index-64 builds

e4dac95

Fix -Wconversion errors in CBLAS/testing during Index-64 builds

acdcf84

Fix -Wconversion errors in single complex TESTING/EIG sources during …

9bbf60c

…Index-64 builds

Fix -Wconversion errors in single TESTING/EIG sources during Index-64…

f9ade3e

… builds

Fix -Wconversion errors in single complex TESTING/LIN sources during …

9549c35

…Index-64 builds

Fix -Wconversion errors in single TESTING/LIN sources during Index-64…

46ec5a4

… builds

Fix -Wconversion errors in XBLAS sources during Index-64 builds

c131da9

Copilot AI review requested due to automatic review settings May 12, 2026 15:48

Copilot started reviewing on behalf of ACSimon33 May 12, 2026 15:49 View session

Copilot AI reviewed May 12, 2026

View reviewed changes

Initialize complex variables with CMPLX, not just REAL

db9aa25

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix Index-64 conversion errors#1278

Fix Index-64 conversion errors#1278
ACSimon33 wants to merge 8 commits into
Reference-LAPACK:masterfrom
ACSimon33:index_64_conversion_errors

ACSimon33 commented May 12, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

ACSimon33 commented May 12, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants