TEST

Tests to perform on the supported libraries. [Edit on GitHub]

Subsections

• CP_DBCSR
• CP_FM_GEMM
• DBM
• EIGENSOLVER
• ERI_MME_TEST
• GRID_INFORMATION
• PROGRAM_RUN_INFO
• PW_TRANSFER
• RS_PW_TRANSFER
• SHG_INTEGRALS_TEST

Keywords

• CLEBSCH_GORDON

• COPY

• DGEMM

• ERI

• FFT

• LEAST_SQ_FT

• MATMUL

• MEMORY

• MINIMAX

• MPI

Keyword descriptions

CLEBSCH_GORDON: integer = 0

Aliases: CLEBSCH

Usage: clebsch_gordon 10

Tests the Clebsch-Gordon Coefficients. Tests are repeated the given number of times. [Edit on GitHub]

COPY: integer = 0

Usage: copy 10

Tests the performance to copy two vectors. The results of these tests allow to determine the size of the cache of the CPU. This can be used to optimize the performance of the FFTSG library. Tests are repeated the given number of times. [Edit on GitHub]

DGEMM: integer = 0

Usage: DGEMM 10

Tests the performance of different kinds of matrix matrix multiply kernels for the BLAS INTRINSIC DGEMM. Matrices up to 2**N+1 will be tested. [Edit on GitHub]

ERI: integer = 0

Usage: eri 1

Tests the performance and correctness of ERI libraries [Edit on GitHub]

FFT: integer = 0

Usage: fft 10

Tests the performance of all available FFT libraries for 3D FFTs Tests are repeated the given number of times. [Edit on GitHub]

LEAST_SQ_FT: integer = 0

Usage: MINIMAX 1000

Tests accuracy of the integration weights gamma_ik from Kaltak, Klimes, Kresse, JCTC 10, 2498 (2014), Eq. 30. Printed is the L1-error (=minimax error for a given range and a given number of grid points. The input parameter is the given number of different Rc values. [Edit on GitHub]

MATMUL: integer = 0

Usage: matmul 10

Tests the performance of different kinds of matrix matrix multiply kernels for the F95 INTRINSIC matmul. Matrices up to 2**N+1 will be tested. [Edit on GitHub]

MEMORY: real = 2.56000000E+008

Usage: MEMORY

Set the maximum amount of memory allocated for a given test (in bytes) [Edit on GitHub]

MINIMAX: integer = 0

Usage: MINIMAX 1000

Tests validity of minimax coefficients for approximating 1/x as a sum of exponentials. Checks numerical error against tabulated error, testing the given number of different Rc values. [Edit on GitHub]

MPI: integer = 0

Usage: mpi 6

Tests mpi, quickly adapted benchmark code, will ONLY work on an even number of CPUs. comm is the relevant, initialized communicator. This test will produce messages of the size 8*10**requested_size, where requested_size is the value given to this keyword [Edit on GitHub]