Hi, thanks for you answer!
> Do you have any examples of what you did? The optimization I did transforms numeric loops into equivalent loops using SIMD load/store and arith. both for floats and integers. It is less concerned with string vector operations. So these are the basic optimizations gcc (or other compiler) also would do when turning a scalar loop operating on vectors into a vector loop. > You mean something that goes beyond what can be found in the principles of > operation for the z13? I would expect to get 2x the speedup for a vector loop (128 bits are processed at once instead of 64), but it is far less (most of the time not significant). I wanted to be sure that the vector facility on the marist s390x servers is not "emulated" or configured differently? My assumption is that running the OS virtually, does not change anything right? > or kernel (e.g. for crc32) > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/s390/crypto/crc32be-vx.S Thank you, that one is really a good example. Currently the optimization only considers that it has 128 bit vector registers available, not that it can load/store X times 128 bit at once! Cheers, Richard ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
signature.asc
Description: OpenPGP digital signature
