[SciPy-user] Fw: Changes for ufuncobject.h

mfmorss@aep.com mfmorss at aep.com
Fri Jan 13 08:31:28 CST 2006


(* Forgot to sent to everyone *)

Mark F. Morss
Principal Analyst, Market Risk
American Electric Power
----- Forwarded by Mark F Morss/OR3/AEPIN on 01/13/2006 09:30 AM -----
                                                                           
             Mark F                                                        
             Morss/OR3/AEPIN                                               
                                                                        To 
             01/13/2006 09:18          Travis Oliphant                     
             AM                        <oliphant.travis at ieee.org>          
                                                                        cc 
                                                                           
                                                                   Subject 
                                       Re: Changes for ufuncobject.h       
                                       (Document link: Mark F. Morss)      
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           



Thank you, Travis!  With that, I now have numpy working -- albeit without
LAPACK or BLAS, which are absent from my system.

I would appreciate if anyone could comment on the significance of a couple
of messages that I received during the compilation of umathmodule.c:

"build/src/numpy/core/src/umathmodule.c", line 8155.32: 1506-280 (W)
Function argument assignment between types "long double*" and "double*" is
not allowed.

    1500-030: (I) INFORMATION: InitOperators: Additional optimization may
be attained by recompiling and specifying MAXMEM option with a value
greater than 2048.

I wonder if I will encounter any difficulties due to this
double/long-double assignment issue.  Also, I apologize for my ignorance,
but I wonder if it would be advisable to reset MAXMEM, and how to do so.  I
searched the download for a possible answer to the latter question, and
couldn't find any.

Mark F. Morss
Principal Analyst, Market Risk
American Electric Power



                                                                           
             Travis Oliphant                                               
             <oliphant.travis@                                             
             ieee.org>                                                  To 
                                       mfmorss at aep.com                     
             01/12/2006 05:47                                           cc 
             PM                                                            
                                                                   Subject 
                                       Changes for ufuncobject.h           
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           





There are some syntax errors in that section of code.

Here is a patch showing the changes...  You'll notice that a '\'
character is needed at the end of two lines...




Index: ufuncobject.h
===================================================================
--- ufuncobject.h       (revision 1887)
+++ ufuncobject.h       (working copy)
@@ -280,7 +280,7 @@
 #define generate_divbyzero_error() feraiseexcept(FE_DIVBYZERO)
 #define generate_overflow_error() feraiseexcept(FE_OVERFLOW)

-#elif defined(AIX)
+#elif defined(_AIX)

 #include <float.h>
 #include <fpxcp.h>
@@ -288,11 +288,11 @@
 #define UFUNC_CHECK_STATUS(ret) { \
        fpflag_t fpstatus; \
                                                 \
-       fpstatus = fp_read_flag();
+       fpstatus = fp_read_flag(); \
        ret = ((FP_DIV_BY_ZERO & fpstatus) ? UFUNC_FPE_DIVIDEBYZERO : 0) \
                | ((FP_OVERFLOW & fpstatus) ? UFUNC_FPE_OVERFLOW : 0)   \
                | ((FP_UNDERFLOW & fpstatus) ? UFUNC_FPE_UNDERFLOW : 0) \
-               | ((FP_INVALID & fpstatus) ? UFUNC_FPE_INVALID : 0);
+               | ((FP_INVALID & fpstatus) ? UFUNC_FPE_INVALID : 0); \
        fp_clr_flag( FP_DIV_BY_ZERO | FP_OVERFLOW | FP_UNDERFLOW |
FP_INVALID); \
 }





More information about the SciPy-user mailing list