A surprise weekend commit by Intel to the LLVM Clang compiler frontend project Git reveals support for the company's 8th generation Core "Cannonlake" processors. Cannonlake succeeds the 7th generation "Kaby Lake" chips which come out later this year; and is targeted at a 2017 release. Intel's three-letter designation for the architecture, "CNL" tagged a list of processors added to Clang, besides a few processors tagged with "SKL" (Skylake).
The commit confirms that Cannonlake will introduce some new instruction sets, such as AVX-512 (abx512ifma and avx512vbmi), new SHA extensions, and UMIP. With IFMA and VBMI extensions, the AVX-512 instruction set on Cannonlake is more extensive than the one found on current "Skylake" enterprise CPUs (limited to avx512f, avx512cd, avx512dq, avx512bw, and avx512vl).
http://ift.tt/1XHdz40
Комментариев нет:
Отправить комментарий