He glossed _VERY_ quickly over the power supply. But even the power supply was magnificent, it was designed to provide precisely balanced power with an absolute minimum of noise or ripple. This was critical to the operation of the machine.
The highly pipelined nature of the system must make it very difficult to write software that uses the capabilities efficiently. I wonder what that looked like in reality.
The Fortran compiler generated vector instructions when it could. It was fairly easy to write Fortran code for vector algorithms. For problems which were not vectored, the Cray-1 was a very fast scalar machine as well.
Fascinating - thanks!
Great talk 👍 The first generation Crays are just beautiful 😀
Good stuff!
He glossed _VERY_ quickly over the power supply. But even the power supply was magnificent, it was designed to provide precisely balanced power with an absolute minimum of noise or ripple. This was critical to the operation of the machine.
You seem to be conflating Lawrence Livermore Labs with LANL (Los Alamos National Labs). I think Serial 1 of the Cray-1 went to LANL.
The highly pipelined nature of the system must make it very difficult to write software that uses the capabilities efficiently. I wonder what that looked like in reality.
Obviously the compilers were optimized for vector processing and yes, worked very well. Ah, I remember well.
The Fortran compiler generated vector instructions when it could. It was fairly easy to write Fortran code for vector algorithms. For problems which were not vectored, the Cray-1 was a very fast scalar machine as well.