berkeley-softfloat-3
berkeley-softfloat-3 copied to clipboard
ucb-bar
f32_div and f32_sqrt
I could hardly figure out how does f32_div and f32_sqrt work, especially the two magic functions "softfloat_approxRecip32_1" and "softfloat_apprroxxRecipSqrt32_1" If there is any paper or website help me learn more details?
Do you find it easier to follow something like QEMU's rewrite of float divide or the sqrt function? We worked quite hard to ensure the eventual generated code is as efficient at least for the add/sub/div/mul code but I suspect the sqrt code is non-optimal.
Do you find it easier to follow something like QEMU's rewrite of float divide or the sqrt function? We worked quite hard to ensure the eventual generated code is as efficient at least for the add/sub/div/mul code but I suspect the sqrt code is non-optimal.
my apology that reply so so so late. I have got the div function thanks for you reply !
Do you find it easier to follow something like QEMU's rewrite of float divide or the sqrt function? We worked quite hard to ensure the eventual generated code is as efficient at least for the add/sub/div/mul code but I suspect the sqrt code is non-optimal.
my apology that reply so so so late. I have got the div function thanks for you reply !
It would be great if you can share the paper or the link or key words :)
Do you find it easier to follow something like QEMU's rewrite of float divide or the sqrt function? We worked quite hard to ensure the eventual generated code is as efficient at least for the add/sub/div/mul code but I suspect the sqrt code is non-optimal.
my apology that reply so so so late. I have got the div function thanks for you reply !
It would be great if you can share the paper or the link or key words :)
Hi, I barely remember the solution and link. It's been a while, I need to jog my memory about how things were done. If not, I already got some better solution which is same fast and easy to understood with matlab code