c - Difference between xxxxx_(), LAPACK_xxxxx() and LAPACKE_xxxxx() functions -

let's want use lapack solve system of linear equations in c (gcc). set problem follows:

/* want solve ax=b */ int n = ...;      // size double *a = ...;  // nxn matrix double *b = ...;  // length-n vector int m = 1;        // number of columns in b (needs in variable) double *pivot;    // records pivoting int info;         // return value 

now seems can use 1 of 3 functions solve problem. first 1 this:

dgesv_( &n, &m, a, &n, pivot, b, &n, &info ); 

i surprised see did not require #includes, seems... weird.

the second function has same signature, except prefix lapack_ think causes less ambiguity , possibly less error-prone:

#include <lapack/lapacke.h> lapack_dgesv( &n, &m, a, &n, pivot, b, &n, &info ); 

note requires me include lapacke.h.

the third function changes signature returning info , not taking arguments pointers:

#include <lapack/lapacke.h> info = lapacke_dgesv( lapack_col_major, n, m, a, n, pivot, b, n); 

again, function requires lapacke.h. requires linking library -llapacke. 3 functions need -llapack.

i'm trying figure out differences between these functions. did snooping around , found following macros in lapacke.h , related header files:

#define lapack_global(name,name) name##_ #define lapack_dgesv lapack_global(dgesv,dgesv) 

so seems lapack_dgesv() , dgesv_() different names exact same function. however, appears lapacke_dgesv() else possibly different implementation, considering fact needs library.

so question is: differences between these 2 functions? documentation says lapacke c interface lapack, function dgesv_()? can use without needing lapacke , without compiling in fortran, how different?

thanks.

---

update

curiously, function dgemm_() (matrix multiplication) not have lapack_dgemm() equivalent. what's going on?

• notice lapacke_dgesv() features additional flag can lapack_col_major (usual fortran order) or lapack_row_major (usual c order). in case of lapack_col_major, calls lapack_dgesv() directly. in case of lapack_row_major, lapacke_dgesv() transpose matrix before calling lapack_dgesv(). not new implementation of dgesv_(). take @ lapack-3.5.0/lapacke/src/dgesv_work.c in file, there minor additional changes error handling.

• lapack_dgesv() defined in header lapacke.h lapack_global(dgesv,dgesv). macro lapack_global defined in lapacke_mangling.h : wraps dgesv_ , cares naming convention if other conventions used.

so, basically, function lapack_dgesv() requires headers of lapacke. compared dgesv_, problems related naming conventions in libraries may avoided. lapack_dgesv() same dgesv_(). function lapacke_dgesv() enlarges scope of lapack_dgesv() handle usual c matrix.but still calls dgesv_ in end.

the function dgemm() part of blas library. wrapped c version cblas_dgemm() can found in cblas . again, additionnal flag cblas_order required, possible values cblasrowmajor , cblascolmajor.