ByRef

[Peter]

Hi Charles,

BYREF is a problem for your  compiler?

Code: [Select]

sub swap( byref a as sys, byref b as sys )
    iEcx = a
    iEax = b
    a = iEax
    b = iEcx
end sub

v1=10
v2=20

swap v1, v2
print  v1 & " " & v2  'correct

sub swop( byref sys a, byref sys b )
    mov Ecx, a
    mov Eax, b
    mov a, Eax
    mov b, Ecx
end sub

v1=10
v2=20

swop v1, v2
print  v1 & " " & v2  'odd


[Charles Pegge]

Hi Peter,

The solution is to use addr for indirect variables

Code: OxygenBasic

1. sub swop( byref a as sys, byref b as sys )
2.     addr esi,a
3.     addr edi,b
4.     mov  ecx,[esi]
5.     mov  edx,[edi]
6.     mov  [esi],edx
7.     mov  [edi],ecx
8. end sub
9.  

[Peter]

Yes: I see it now, never used ADDR

Thanks Charles. 
                                                                                German:     
This is funny:   Charles Insectivorous Tree Finch.    Kleinschnabel-Darwinfink
                        Large Insectivorous Tree Finch.       Papageischnabel-Darwinfink
                        Small Insectivorous Tree Finch.       Zweigdarwinfink
                                                                            

[edcronos]

Hello
I can use byref in macros and functions in DLL that will be used in other programs or the best is to use byval?
to have a higher speed was thinking of sending the array of data that is already worked by other functions for the function in the DLL generated by the oxygen, and have an array with all results also by reference

[Charles Pegge]

Arrays can only be passed by reference, and the same applies to most UDts. With single primitives, you have a choice

[edcronos]

as test to work in excel, it was like this

includepath "$/inc/"
$filename "And32count.dll"
$dll
include "RTL32.inc"
extern export

function And32(int a,b,byref c)
 

   mov eax, a
   mov ebx, b
   and eax, ebx

    addr edi ,c   '<---- ??
                       ' 'mov edi , eax
   mov c, eax    '<---- ??

end function

and to declare the dll

Private Declare Sub And32 Lib "D:\Programação\OxygenBasic\And32count.dll" _
                          (ByVal n As Long, ByVal n2 As Long, ByRef out As Long)


I have to see how to create multiple functions in the same dll

[Charles Pegge]

Almost!

Code: [Select]

includepath "$/inc/"
$filename "And32count.dll"
$dll
include "RTL32.inc"
extern export

function And32(long a,b,*c)
   mov eax, a
   mov ecx, b 'avoid using ebx
   and eax, ecx

    addr edi ,c
     mov [edi] , eax

     'the compiler will produce similar assembly code:
    'c=a and b
end function
and to declare the dll

Code: [Select]

Private Declare Sub And32 Lib "D:\Programação\OxygenBasic\And32count.dll" Alias "And32" _
                          (ByVal n As Long, ByVal n2 As Long, ByRef out As Long)

You can add as many subs or functions as needed.

[edcronos]

I just did not understand why it worked
    addr edi ,c   '<---- ??
   mov c, eax    '<---- ??
I did not understand the "edi" reference with c

in the case would be that when you pass the address "addr edi, c"
  automatically "c" happens to be the same as "edi"?

[Charles Pegge]

addr is a pseudo instruction for loading the address of a variable into a register, in this case, edi. Then you can use [edi], in square brackets, to access the variable.

Try #show to reveal the compiled assembly code:
#show c=a and b

[edcronos]

thanks for the explanation, had worked out I think because of the pointer, not the referral exchange
even with "addr esi, c"
   "mov c, eax" '<---- c continues being edi

anyway I'll have to wait for popcnt and continue using 32bit integers
I would not like to swap my excel, and found no way to work with more than 32bits with the dll
it would be nice if it were possible for an 8-byte array to work directly as a 64-bit integer, but I can not even dream much

[Charles Pegge]

[edi] is used by all byref parameters. So your code works correctly by accident

[edcronos]

I'm sorry, but I'm going to ask a rather stupid question.
I practically know the answer, but the question kills me

why is the binary system that way?
and not from the least significant to the most significant and sequential way?

of the current way that it is complicates the interaction between systems of different bases

[Charles Pegge]

You can blame Intel. Motorola has it the other way round.

https://pt.wikipedia.org/wiki/Extremidade_(ordena%C3%A7%C3%A3o)

https://en.wikipedia.org/wiki/Endianness

[edcronos]

Wow, I did not even know that, I honestly thought I was talking a big bogey
Thank you for showing me this fact.
then this feeling that something is strange is not for nothing?
in my attempts to bular this communication inconsistency or the values are wrong or gives pane in excel, then I realized that a smaller variable does not fit into a larger one properly

https://www.ibm.com/developerworks/aix/library/au-endianc/index.html?ca=drs-

https://support.industry.siemens.com/cs/document/29155499/how-can-you-convert-a-big-endian-word-or-double-word-into-a-little-endian-word-or-double-word-and-vice-versa-?dti=0&lc=en-WW


but ok, I'm just being curious,
I do not consider myself qualified to work directly with ASM, and it is not because of the logic that I find very simple, but because of the care that has to be taken
I think even passing 4 integers of 32 bit in an array with all the data to be worked must have a good gain of performance, still more using popcnt

[Charles Pegge]

You can do 128 bit logic (but not integer arithmetic), using the SIMD instructions and XMM registers:

Code: [Select]

'128 bit logic
'2017-12-24 T 12:02:25
indexbase 0
int a={1,2,3,4,5,6,7,8}
movups xmm0,a[0]
movups xmm1,a[4]
andps  xmm0,xmm1  'and
'andnps xmm0,xmm1 'and not
'orps   xmm0,xmm1 'or
'xorps  xmm0,xmm1 'xor
movups a[0],xmm0
print a[0]