Intel has details about the next generation of AVX-instructions. This should future generations of processors more effectively to deal with vectorberekeningen. Especially servers would need to benefit.
The new instruction set is going to full the Intel Advanced Vector Extensions 512 are called, abbreviated as AVX-512, as reported Intel on his softwareblog. The new instruction follows the 512bit wide simd-set that Knights Corners was introduced. In the successor of Xeon Phi coprocessor, codenamed Knights Landing, will the AVX-512 instruction set, days. Also future generations of Xeon processors with AVX-512 will be equipped.
The instruction set supports eight double precision or sixteen single precision floating point numbers in the 512bit vectors. Therein also apply the eight 64bit integers, or sixteen 32bit integers, which doubles AVX-512 the amount of processed data compared to the AVX2 instruction set. AVX-512 is compatible with previous AVX-instruction sets. Intel was not known whether the AVX-512 instructions also in future generations consumentenprocessors used. That processors, codenamed Broadwell and later Skylake, would AVX 3.2: or that the new 512bit instruction set is, is to wait and see.