I wish they explained or link to somewhere where I could learn what a compiler does when it vectorizes appropriately. I imagine it’s at least putting handling params in a way that takes advantage of cache locality, but are some compilers making use of gpus now? Or do cpus have some support for vectorization?
My dad is in HPC and I remember him making fun of people who would buy expensive Intel chips and then not pay to use the Intel compilers. They are so much better than GNU c++ because they understood how to properly vectorize stuff for the cpu to have the best chance at being fast
I wish they explained or link to somewhere where I could learn what a compiler does when it vectorizes appropriately. I imagine it’s at least putting handling params in a way that takes advantage of cache locality, but are some compilers making use of gpus now? Or do cpus have some support for vectorization?
My dad is in HPC and I remember him making fun of people who would buy expensive Intel chips and then not pay to use the Intel compilers. They are so much better than GNU c++ because they understood how to properly vectorize stuff for the cpu to have the best chance at being fast
Yes, all x86 64 cpus support at least SSE2