This also includes optimizations to remove unnecessary checks when using GSL::Span and similar bounds-checked containers. It's looking like a great release for folks who like high-speed and safe code generation 
-
-
Prikaži ovu nitHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Pretty cool! Any plans for an additional blog describing further your implementation? Also, is there optimised code specifically for the case when zeroing out memory using memset?
-
I'm not planning on blogging about the memset implementation. You can see how it compiles by using the compiler output though :-). This is optimized for all memset unrolling (0 and non-zero).
- Još 1 odgovor
Novi razgovor -
-
-
You can beat repstos with SSE2?
-
repstos is trivial to beat when the size is known at compile time (which it is in the case of this unrolling code). It's also not particularly hard to beat even if the size isn't known (depending on the size of the memset and if the cpu supports enhanced rep stosb).
- Još 3 druga odgovora
Novi razgovor -
-
-
I think that the ontroduction of SVML is game changer I can finally not need to link with the CRT :)
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Go, Joe!
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
I note that every time you call an SVML function we generate a call to __vdecl_exp4f for example if we have used _mm_exp_ps, where this function got implemented in the CRT? Seem to miss it if I compile with /NODEFAULTLIB
- Još 1 odgovor
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.