News for hackers. Stuff that matters.
Re:MathSharp, a vector and matrix library written in C# usin... by djmips on Thursday September 12, 2019 @06:12PM (#20957014)
I took a look at the code and I was wondering about the implementation performance where there are a lot of if() statements in the low level functions. Like Vector4F DotProduct2D() has if (Sse41.IsSupported) else if (Sse3.IsSupported) else if (Sse.IsSupported) and finally a default implementation. This seems inefficient but I don't know if somehow the booleans are more like compile time flags and the extraneous code disappears in a puff of logic.
Re:MathSharp, a vector and matrix library written in C# usin... by jdmichal on Thursday September 12, 2019 @06:24PM (#20957118)
Those might fall out during JIT compilation as constant expressions. Would be worth running it through ngen to see what happens. (I'd do it myself but I'm not at a PC at the moment.)

EDIT: Just realized the Intrinsics namespace was added in Core 3.0 preview. I don't have the setup to test this myself right now. I did find an online compiler, but it looked like it optimized away everything in release mode.

Re:MathSharp, a vector and matrix library written in C# usin... by Faark on Thursday September 12, 2019 @07:56PM (#20957739)
There was some discussion here on hn a few days ago [0] that linked to a blog post about those intrinsics [1]. Those should indeed be optimized away. The relevant section:

> The IsSupported checks are treated as runtime constants by the JIT (when optimizations are enabled) and so you do not need to cross-compile to support multiple different ISAs, platforms, or architectures. Instead, you just write your code using if-statements and the unused code paths (any code path which is not executed, due to the condition for the branch being false or an earlier branch being taken instead) are dropped from the generated code (the native assembly code generated by the JIT at runtime).

[0] https://news.ycombinator.com/item?id=20915430 [1] https://devblogs.microsoft.com/dotnet/hardware-intrinsics-in...

Re:MathSharp, a vector and matrix library written in C# usin... by legulere on Friday September 13, 2019 @08:15AM (#20961372)
The JIT optimizes them away:

> The IsSupported checks are treated as runtime constants by the JIT (when optimizations are enabled) and so you do not need to cross-compile to support multiple different ISAs, platforms, or architectures. Instead, you just write your code using if-statements and the unused code paths (any code path which is not executed, due to the condition for the branch being false or an earlier branch being taken instead) are dropped from the generated code (the native assembly code generated by the JIT at runtime).

https://devblogs.microsoft.com/dotnet/hardware-intrinsics-in...

Re:MathSharp, a vector and matrix library written in C# usin... by gameswithgo on Thursday September 12, 2019 @07:07PM (#20957445)
branch predictor gonna work well on those at least, if the jit doesn’t make then go away entirwlyy
Re:MathSharp, a vector and matrix library written in C# usin... by nick_ on Thursday September 12, 2019 @09:00PM (#20958112)
came to say this. +1
Re:MathSharp, a vector and matrix library written in C# usin... by johnkellyoxford on Monday September 16, 2019 @04:41PM (#20988354)
yep, the JIT fully removes those checks when native code gen occurs :)