Several issues in LAPACK 3.4.2 and ScaLAPACK 2.0.2

[ivan_latkin]

Hi,

Please find my comments regarding issues in LAPACK 3.4.2 and ScaLAPACK 2.0.2 below:

LAPACK:
1. LAPACKE interfaces:
• lapacke_?tprfb.c – ‘myldwork’ parameter is redundant;
• lapacke_?tprfb_work.c – ‘mywork’ and ‘myldwork’ should be renamed to ‘work’ and ‘ldwork’;
• lapacke_ctpqrt.c, lapacke_ctpqrt_work.c – wrong order of parameters (‘t,b,ldb,ldt‘ instead of ‘b,ldb,t,ldt’);
• lapacke_?tpqrt2.c – parameter ‘l’ between ‘n’ and ‘a’ is absent;
2. Tests:
• ?drvhex.f(617), ?drvsyx.f(623), c and z precisions: IWORK(N+1) was changed to RWORK, but RWORK( 2*NRHS+1 ) should be used here, otherwise BERR and FERR ( which are stored from RWORK(1) to RWORK( 2*NRHS )) are destroyed before check, resulting 10^308 residual errors.
• ?drvge.f(d,s –line 555, c,z – line 556), ?drvgb.f (651) – add check ‘.AND. INFO.LE.N’ to compute RPVGRW correctly
3. Comments and source files :
• ?sbevd.f(159): In description comment ‘The dimension of array LIWORK...’ – rename LIWORK with IWORK
• chesvxx.f, zhesvxx.f: Size of WORK array should be 5*N (not 2*N), see line with call of ?hetrf

ScaLAPACK:
• p?laqr5
- STOP statement is used in library function - it may crash program without any message. It should be changed to RETURN, i believe.

Best regards,
Ivan Latkin

[admin]

Dear Ivan,
Thank you very much for reporting those issues.
We corrected them all in our repositories.
Julie