Here's my notes for those who are want; they added some 68 instructions for linalg to risc v. Their arc has a specialized reduce / aggregate logic and associate

Here's my notes for those who are want; they added some 68 instructions for linalg to risc v. Their arc has a specialized reduce / aggregate logic and associated memory for vectors and matrices. They also made a c lib for their instruction set. Saw some impressive speedups compared to normal risc v (x12?), x86 (x5?) and fermi gpus (x10?) (In the hpc challenge where they did have to update some of the code, obviously, to use their instructions).