On 3/21/2015 11:02 AM, "Ola Fosheim =?UTF-8?B?R3LDuHN0YWQi?=
<[email protected]>" wrote:
On Saturday, 21 March 2015 at 17:55:09 UTC, Walter Bright wrote:
On 3/20/2015 3:50 AM, "Ola Fosheim =?UTF-8?B?R3LDuHN0YWQi?=
<[email protected]>" wrote:
High level constructs may be cleaner if done right, and sometimes saves
programmer time, but it will never be as fast on the standard CPU architectures
we have today. The hardware favours carefully planned iterative, imperative
approaches. That was true before SIMD and caching, and it is even more true now.
It's less true for SIMD. To take advantage of SIMD, compilers have to reverse
engineer low level loops into a higher level construct, then re-compile for
SIMD.
No. You have to design so that you don't get dependencies on the same vector
register.
I know I shouldn't, but I'll bite. Show me the "low level C code" that
effectively uses SIMD vector registers.
