Dear PGI User Forum members,
I am compiling and linking an external program with the pgf90 fortran compiler and succeeded to get a working executable without optimization. As the code will be run extensively in the following years, I have tried to optimize the code as recommended with the -fastsse fortran optimization flag. I have AMD Dual Core Opteron processors with revision E6. These processors should have SSE and SSE2. Compiling and linking with the -fastsse flag resulted in an error:
lcs.a(xyi.o): In function
xyi_':xyi.F:(.text+0xab7): undefined reference to .LB1_563
The name .LB1_563 is most probably a compiler internal reference, and is not mentioned in this function nor in the included files. Looking at the symbol tables in the archive files lcs.a for object code xyi.o with and without optimization (command nm -s -a lcs.a) shows as only difference the undefined reference .LB1_563
My question is what I should change to let the program run at it’s highest speed?
Thanking in advance for your help or suggestions.
With friendly greetings