Re: [PATCH 0/8] MR4007: d3d10/effect: Add support for 'imul' instruction.

[Matteo Bruni (＠Mystral)]

Matteo Bruni (@Mystral) commented about dlls/d3d10/tests/effect.c:
>  {
>      SetBlendState( NULL, rsqrt(g_var), 0 );
>  }
> - pass p31
> - {
> -
>    SetBlendState( NULL, log(g_var), 0 );
> - }

It's not super obvious from a quick look that this is testing logn() while the opcode is effectively log2() (the compiler adds a multiplication by logn(2) to get the intended result, as you'd expect). It turns out that log2() is an intrinsic as well; what do you think about introducing a test for log2() before this one (in this same patch)?

Same would also apply to exp() in the next patch.