Lines Matching refs:float

3668 ;; Note that the `cvtsi2s{s,d}` instruction is not just an int-to-float
3675 ;; This means that the first operand to all of the int-to-float conversions here
3680 ;; between consecutive int-to-float operations but that's not easy
3707 ;; each of those into a float. Afterwards re-pack the two results into the final
3749 ;; As for the bit tricks a float is represented where the low 53 bits are the
3750 ;; decimal of the float, basically:
3757 ;; number. The final step in combining via float arithmetic will chop off the
3758 ;; leading `1.` at the start of the float that we constructed, one for the low
3771 ;; Algorithm uses unpcklps to help create a float that is equivalent
3791 ;; converting packed signed integers to float (cvtdq2ps). In
3806 ;; -> Convert(Al) // Convert int to float
3851             ;; Sets tmp to zero if float is NaN
3855             ;; Sets top bit of tmp if float is positive
3856             ;; Setting up to set top bit on negative float values
3859             ;; Convert the packed float to packed doubleword.
3900 ;; Step 8 - Add the original converted src and the converted tmp1 where float values originally less
3901 ;;          than and equal to INT_MAX will be unchanged, float values originally between INT_MAX+1 …
3902 ;;          UINT_MAX will add together (INT_MAX) + (SRC - INT_MAX), and float values originally
3922             ;; Converting to unsigned int so if float src is negative or NaN
4740 ;; Remove the extractlane instruction, leaving the float where it is. The upper
4797 ;; Case 1: when moving a scalar float the `movss` and `movsd` variants with