[SSE2]Make all bytes positive

[Farabi]

Im substracting 16 8bit integer and want to have the result all positive. How to do that?

[jj2007]

Something like this?

.code
whatever   db 123, -127, 99, 255, 128, 127, 0, 100, 123, -127, 99, 255, 128, 127, 0, 100
start:
   mov eax, 7f7f7f7fh
   movd xmm1, eax
   pshufd xmm1, xmm1, 0
   movups xmm0, oword ptr whatever
   andps xmm0, xmm1

[Farabi]

Something like this?

.code
whatever   db 123, -127, 99, 255, 128, 127, 0, 100, 123, -127, 99, 255, 128, 127, 0, 100
start:
   mov eax, 7f7f7f7fh
   movd xmm1, eax
   pshufd xmm1, xmm1, 0
   movups xmm0, oword ptr whatever
   pandn xmm0, xmm1

:t Well I dont get it how it work yet, but yes that is what I want. I though there are a single instruction out there but that should sufficient, thanks.

[jj2007]

It seems andps is the one to choose, not pandn - see corrected code above.

[Farabi]

Can I know what this code does?

pshufd xmm1, xmm1, 0

It doesnot seems did nothing.

[qWord]

Farabi, do you want to have the absolute value of the difference? (abs(b-a))

[jj2007]

Can I know what this code does?

pshufd xmm1, xmm1, 0

It doesnot seems did nothing.

It propagates the lowest dword to the other three dwords of the xmm reg, so that xmm1 contains
7f7f7f7f7f7f7f7f7f7f7f7f7f7f7fh

[Farabi]

Farabi, do you want to have the absolute value of the difference? (abs(b-a))

I think integer will always be absolute.

[jj2007]

Farabi, do you want to have the absolute value of the difference? (abs(b-a))

I think integer will always be absolute.

That doesn't answer the question, Onan. Can you post a real example what you want to see?

[Farabi]

Farabi, do you want to have the absolute value of the difference? (abs(b-a))

I think integer will always be absolute.

That doesn't answer the question, Onan. Can you post a real example what you want to see?

Sorry maybe Im misunderstood with the term "absolute".

I need to substract a pixel and check if the result was less than ten, and to check all of that I need all to be positive, if is it negative, I need to add another cycle to check wheter it negative or not.

Code Select Expand


mov edx,bm.bmBits
movd xmm1,[edx]
pxor xmm0,xmm0
psubb xmm0,xmm1
lea edx,buff
movd [edx],xmm0


You want me to post the whole project?

[Farabi]

I cant use "pshufd" why?

[dedndave]

perhaps it's the addressing mode ?
it's SSE2 - i figure your CPU does that
you must be using .686/.MMX/.XMM or the other instructions wouldn't work  :P

[Farabi]

perhaps it's the addressing mode ?
it's SSE2 - i figure your CPU does that
you must be using .686/.MMX/.XMM or the other instructions wouldn't work  :P

I can use "psubb" which is SSE2 instruction but not  "pshufd" why?

[qWord]

perhaps it's the addressing mode ?
it's SSE2 - i figure your CPU does that
you must be using .686/.MMX/.XMM or the other instructions wouldn't work  :P

I can use "psubb" which is SSE2 instruction but not  "pshufd" why?

error message?

Also remarks the instruction pcmpGTb (and pcmpEQb)

[Farabi]

perhaps it's the addressing mode ?
it's SSE2 - i figure your CPU does that
you must be using .686/.MMX/.XMM or the other instructions wouldn't work  :P

I can use "psubb" which is SSE2 instruction but not  "pshufd" why?

error message?

Also remarks the instruction pcmpGTb (and pcmpEQb)

Here is the error message "error A2008: syntax error : xmm"