Hello, Sorry for the delayed response and thank you for your interest in Chapel.
Unoptimized Chapel generates many more runtime checks such as array bounds checking, nil checking, and stack overflow checking. For this reason, we recommend using --fast when performance is critical. Did the serial performance also improve when compiled with --fast?
Lydia On 08/11/2014 04:35 PM, Tomsy Paul wrote:
Hello all, Attached herewith are two chapel programs finding the value of Pi using summation. One is a parallel version using reduce and the other, the serial version. The time for execution on a 3.3 GHz Quad core Intel i3 processor is found to be 118 seconds for serial version and 186 seconds for parallel version!! That means, Parallel version takes more time!!!. If the --fast flag is used, the speed for parallel version reduces considerably to just 16 seconds. 1. Why is the unoptimized version so slow? 2.The OpenMP versions take only 36 seconds and 13 seconds respectively. I use Chapel 1.9.0 Please help. ------------------------------------------------------------------------------ _______________________________________________ Chapel-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/chapel-users
------------------------------------------------------------------------------
_______________________________________________ Chapel-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/chapel-users
