On Apr 11, 2023, at 1:15 PM, robert bristow-johnson wrote: > So the fact that, to be responsible DSPers, that we have to tickle the > samples to keep them away from denormals, that just makes me grumpy. And I > am still skeptical about Sampo's defense of the hardware decisions not to > simply fix this. It really *isn't* that much logic to fix denorms going in > and out of the ALU. It shouldn't be a problem. I just cannot fathom why, in > 2023, that denorms cannot be dealt with accurately and routinely, with no > MIPS penalty, in a modern general-purpose FPU.
In my experience, designers never "decide not to fix something." If a fix is possible, then someone will sell a product that has the fix, and if it's an important fix then everyone else will follow suit or go out of business. At least that's the case in any competitive market where companies are constantly trying to get customers to choose their product over another (SHARC, TMS320, ARM, etc). To put it another way, if you have a floating point ALU design that fixes this, then implement it in FPGA (at the very least) and let us use it. If you don't have a fix, then I suspect you may be underestimating the difficulty (or perhaps impossibility). Brian
