forwardcom forum

Mhmm... Question: is there any reason for C to be this highly abstract machine, rather than the real world "sum-of-x86-ARM-MIPS-PPC-68k-SPARC-ALPHA-ITANIUM-PARISC-SUPERH-32/64bit-compiled-on-GCC/LLVM/MSVC/ICC" de-facto architecture (for lack of a better term), OTHER than for compiler opti...

Hi Agner, I really think that NaN propagation is much better than fault trapping. However, I fail to see how avoiding fault trapping would remove the need for speculative execution... I see why it reduces the complexity, but I have the feeling speculative execution is still needed: traps are not the...

Actually, the multiple instruction scheme is also applicable to the division (just call div and rem) The only thing: in that case it will be slow. That's were I started to think about caching to not pay for the second divsion (required for rem). I wonder: is there any performance critical codes that...

I was looking in the intel documentation and I found the function _mm512_fixupimm_ps: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#expand=2380,4470,2357,2360,2380&text=fixup Basically, this function allows to map some value type to other value type. The primary use is to handle ...

What I like with the "caching" approach: it is completely transparent. If it is not implemented by the hardware, the software still behaves exactly the same (only slower). And it does not require any cross-lane communication. After some thinking, such a caching would be beneficial only for...

Hi Agner, Some operations have naturally multiple outputs: like multiplication (low part and high part), division (quotient and remainder), addition (result and carry)... As far as I understand, you want to avoid multiple outputs to simplify the hardware. What if those multiple outputs were single o...