I did both Michelson-Morley and Stern-Gerlach as a college junior. They took a little time and skill, but even our old upper-division lab had the equipment necessary. Kids (ugghh..read 20 yr olds..damn, I'm getting old) nowdays can do oodles of stuff: holography, Meissner effect, etc. at any good sized university.
Vector registers are an important part of Cray architecture, but where the machines really shine for scientific computing is memory bandwidth.
The SV-1 (and all it's immediate predecessors - the C90, T90, YMP, etc...) have memory arranged in banks. If you program carefully, you can yank a whole vector from memory at a time, WITHOUT having to wait for a memory refresh.
That said, commodity processors are catching up by just being damn fast!
Slashdot Mirror
I did both Michelson-Morley and Stern-Gerlach
as a college junior. They took a little time and
skill, but even our old upper-division lab
had the equipment necessary. Kids (ugghh..read
20 yr olds..damn, I'm getting old) nowdays
can do oodles of stuff: holography, Meissner
effect, etc. at any good sized university.
Vector registers are an important part of Cray architecture, but where the machines really shine for scientific computing is memory bandwidth. The SV-1 (and all it's immediate predecessors - the C90, T90, YMP, etc...) have memory arranged in banks. If you program carefully, you can yank a whole vector from memory at a time, WITHOUT having to wait for a memory refresh. That said, commodity processors are catching up by just being damn fast!