1

Topic: 2

Greetings if it is still interesting to you max (x, y) = 0.5 (x + y + abs (x - y)) min (x, y) = 0.5 (x + y - abs (x - y))

2

Re: 2

Hello, Erop, you wrote: E> max (x, y) = 0.5 (x + y + abs (x - y)) E> min (x, y) = 0.5 (x + y - abs (x - y)) generally, presence abs implies if and then it is not necessary to calculate, it is necessary to do presence at once if

3

Re: 2

Hello, C0s, you wrote: C0s> generally, presence abs implies presence if and then it is not necessary to calculate, it is necessary to do at once if At first, it is a lie. For example, it is possible to take a root from a square. Secondly,  interested as through normal a floor-mat. Operations to express min and max. abs in the mathematician normally registers vertical sticks, but I am not assured that so it would be clear: max (x, y) = 0.5 (x + y + | x - y |) min (x, y) = 0.5 (x + y - | x - y |)

4

Re: 2

Hello, Erop, you wrote: C0s>> generally, presence abs implies presence if and then it is not necessary to calculate, it is necessary to do at once if E> At first, it is a lie. For example, it is possible to take a root from a square. E> Secondly,  interested as through normal a floor-mat. Operations to express min and max. E> <...> All would be clear clearly. My remark was to that it is necessary to look at cost of operations at once. And that these rings, squares, root systems and .....

5

Re: 2

Hello, C0s, you wrote: C0s> all is clear. My remark was to that it is necessary to look at cost of operations at once. And that these rings, squares, root systems and ..... While there is no executor, operations do not have cost... For example, at any SIMD devices can not be min|max, and abs can and to be...

6

Re: 2

Hello, Erop, you wrote: E> While there is no executor, operations do not have cost... E> For example, at any SIMD devices can not be min|max, and abs can and to be... So speech about cost sgn, instead of any delicacy.

7

Re: 2

Hello, C0s, you wrote: E>> While there is no executor, operations do not have cost... C0s> so speech about cost sgn, instead of any delicacy. I there above selected about what speech

8

Re: 2

Hello, Erop, you wrote: E>>> While there is no executor, operations do not have cost... E> I there above selected about what speech not a question, especially at this time days

9

Re: 2

Hello, Erop, you wrote: E> Hello, C0s, you wrote: C0s>> generally, presence abs implies presence if and then it is not necessary to calculate, it is necessary to do at once if E> At first, it is a lie. For example, it is possible to take a root from a square. E> Secondly,  interested as through normal a floor-mat. Operations to express min and max. E> abs in the mathematician normally registers vertical sticks, but I am not assured that so it would be clear: E> max (x, y) = 0.5 (x + y + | x - y |) E> min (x, y) = 0.5 (x + y - | x - y |) it seems To me, something in operation abs is  the wrong. This operation implies a condition (if), and it likely just that for  is inadmissible. It is a bit better to take a root from a square, but also here  a dirty trick.

10

Re: 2

Hello, Erop, you wrote: E> Greetings if it is still interesting to you E> max (x, y) = 0.5 (x + y + abs (x - y)) E> min (x, y) = 0.5 (x + y - abs (x - y)) I already through a LAN sent to it

11

Re: 2

Hello, Khimik, you wrote: K> Hello, Erop, you wrote: E>> Hello, C0s, you wrote: C0s>>> generally, presence abs implies presence if and then it is not necessary to calculate, it is necessary to do at once if E>> At first, it is a lie. For example, it is possible to take a root from a square. E>> Secondly,  interested as through normal a floor-mat. Operations to express min and max. E>> abs in the mathematician normally registers vertical sticks, but I am not assured that so it would be clear: E>> max (x, y) = 0.5 (x + y + | x - y |) E>> min (x, y) = 0.5 (x + y - | x - y |) K> it seems To me, something in operation abs is  the wrong. This operation implies a condition (if), and it likely just that for  is inadmissible. K> it is a bit better to take a root from a square, but also here  a dirty trick. Simply reset of high bit without checks

12

Re: 2

Hello, Erop, you wrote: reminded widely known in narrow circles sqrt (x^2 + y^2) ~ = max (x, y) + 1/2*min (x, y); x, y> = 0 well or even so: sqrt (x^2 + y^2) ~ = alpha * max (x, y) + beta*min (x, y) http://dspguru.com/dsp/tricks/magnitude-estimator

13

Re: 2

Hello, Maniacal, you wrote: K>> it is a bit better to take a root from a square, but also here  a dirty trick. M> simply reset of high bit without checks depends On representation. If speech about whole, for example, that in binary branching code hardly is more difficult. It will be necessary high bit where to shutter, number to invert and then shuttered bit to add. Hardly longer, of course, but too without branchings...

14

Re: 2

Hello, Erop, you wrote: E>>> max (x, y) = 0.5 (x + y + | x - y |) E>>> min (x, y) = 0.5 (x + y - | x - y |) And what type x, y? And what such 0.5 for the whole? And how overflow of a bit grid is processed? C0s>> generally, presence abs implies presence if and then it is not necessary to calculate, it is necessary to do at once if E> At first, it is a lie. For example, it is possible to take a root from a square. And what such generally IF? If it is passage - that it is possible to manage without it here and there CMP r0, r1 MOV?? r0, r1?? = LT-> MAX signed?? = GT-> MIN signed?? = CS-> MAX unsigned?? = CC-> MIN unsigned

15

Re: 2

16

Re: 2

17

Re: 2

Hello, Evgeny. Panasyuk, you wrote: EP> the Field, a ring, a vector space is all concepts, same as well as Forward Iterator, Random Access Iterator As  : unless fields mutually are isomorphic?

18

Re: 2

19

Re: 2

20

Re: 2

Hello, Erop, you wrote: E> If speech about whole, for example, that in binary branching code hardly will be more difficult. It will be necessary high bit where to shutter, number to invert and then shuttered bit to add. Hardly longer, of course, but too without branchings... It is necessary to do This operation only for the negative.

21

Re: 2

Hello, Evgeny. Panasyuk, you wrote: EP> Hello, Erop, you wrote: E>> If speech about whole, for example, that in binary branching code hardly will be more difficult. It will be necessary high bit where to shutter, number to invert and then shuttered bit to add. Hardly longer, of course, but too without branchings... EP> it is necessary to do This operation only for the negative. Yes, it I . Still the shuttered bit should be spread to the register and with it .