OpenEuphoria

1. windows threads work!!

• Posted by jacques deschênes <desja at globetrotter.net> Jul 10, 2006
• 664 views

In the past I tested threads in euphoria without any success. At my first try
and didn't insist because At that time i thought it was a limitation of euphoria
interpreter.
But this morning reading "windows internals" I learned that windows create a new
stack for each thread a process create. if so multithread should in euphoria!
I decided to revisit it and IT WORKED!!

-- testing windows system threads in euphoria

without warning

include misc.e
include machine.e
include dll.e

constant kernel32=open_dll("kernel32.dll")

if kernel32 = -1 then 
  puts(1,"failed to open kernel32.dll\n")
  abort(1)
end if

constant
iCreateThread=define_c_func(kernel32,"CreateThread",{C_POINTER,C_UINT,C_POINTER,C_POINTER,C_UINT,C_POINTER},C_UINT),
  iSleep=define_c_proc(kernel32,"Sleep",{C_UINT}),
  iCloseHandle = define_c_func(kernel32,"CloseHandle",{C_UINT},C_UINT)
  
global function CreateThread(atom lpThreadAttributes,
atom dwStackSize,  -- initial thread stack size, in
                            bytes
atom lpStartAddress,      -- pointer to thread
                            function
                            atom lpParameter, -- argument for new thread 
                            atom dwCreationFlags,      -- creation flags 
atom lpThreadId  -- pointer to returned thread
                            identifier
                            )
return
  c_func(iCreateThread,{lpThreadAttributes,dwStackSize,lpStartAddress,lpParameter,dwCreationFlags,lpThreadId})
end function --CreateThread()

global function CloseHandle(atom handle)
  return c_func(iCloseHandle,{handle})
end function -- CloseHandle()

global procedure Sleep(integer ms)
  c_proc(iSleep,{ms})
end procedure


object fnVal
atom counter, spinlock

counter = 0
spinlock = 0


function IncCounter(atom param)-- the thread function
while 1 do
  if spinlock = 1 then
    counter +=1
    if counter = 10 then spinlock = 2 return 0 end if
    spinlock = 0
  else
    --sleep(500)  this does not work
    Sleep(500)
  end if
end while
end function

atom lpStartAddress   lpStartAddress = call_back(routine_id("IncCounter")) 
atom lpThreadId  lpThreadId = allocate(4)
atom lpParam   lpParam = allocate(4)  poke4(lpParam,0)
atom hThread

hThread = CreateThread(0,0,lpStartAddress,lpParam,0,lpThreadId)
printf(1,"thread handle\n", hThread)
if not hThread then  puts(1, "echec creation thread\n") abort(0) end if

while spinlock < 2 do
 if not spinlock then
   ? counter
   spinlock = 1 
 end if
end while

fnVal = CloseHandle(hThread)

Regards,
Jacques Deschênes

3. Re: windows threads work!!

• Posted by Pete Lomax <petelomax at blueyonder.co.uk> Jul 10, 2006
• 653 views

On Mon, 10 Jul 2006 07:00:41 -0700, jacques desch=EAnes
<guest at RapidEuphoria.com> wrote:

>In the past I tested threads in euphoria without any success. At my
>first try and didn't insist because At that time i thought it was a
>limitation of euphoria interpreter.
It is. Btw, your example does not run on 2.4.
>But this morning reading "windows internals" I learned that windows
>create a new stack for each thread a process create. if so multithread
>should in euphoria!
>I decided to revisit it and IT WORKED!!
If you use Eu to create pure windows/c_func threads then all will be
fine. However Eu vars are not held on the windows stack, so using Eu
code will fail; eg in function IncCounter after while 1 do add:

if not equal(lower("AAA"),"aaa") then
    puts(1,"lower(AAA)!=aaa\n")
    if getc(0) then end if
    abort(0)
 end if

and after while spinlock < 2 do add:

if not equal(lower("BBB"),"bbb") then
    puts(1,"lower(BBB)!=bbb\n")
    if getc(0) then end if
    abort(0)
 end if

Also, replace Sleep(500) with c_proc(iSleep,{500}) to avoid an almost
inevitable type check on ms.
The resulting program will fail or crash at random.

TDLL in the archives shows how you can multi-OS-thread C/asm code from
Eu, but you cannot multi-OS-thread Eu code, not without doing
something like converting it to a dll first.

Regards,
Pete
PS It is also not as simple as two threads calling the same routine
(lower) simultaneously; the glitch in Sleep() shows it is re-using
internal temporaries or something; if you comment out the lower call
inside IncCounter, it works until the end, but the 'return 0' seems to
spanner the last lower call from the main pgm.

PPS Al: the example did not work for me at all on Eu 2.4 but did on Eu
2.5 (on win98), until as above I replaced Sleep() with c_func and then
2.4 nearly got to the end before crashing )

5. Re: windows threads work!!

• Posted by jacques deschênes <desja at globetrotter.net> Jul 10, 2006
• 662 views

Hi Pete,
I modified my sample code adding the code you suggested but don't get any crash.
But the line puts(1,"lower(AAA)!=aaa\n") doesn't print anything.
Further reading form MSDN learned me that C run time library (CRT) must be a
multi-threaded version for a program using multi-thread with CRT functions to
work properly.
I think that those functions of euphoria that fails are those that call crt
functions and I guess
that exwc.exe version 2.5 is not linked with a multi-threaded version of CRT.

Concerning euphoria variables not being accesible by the thread function, its
wrong.  A thread function is a callback function like any windows functions and
it can access all euphoria object.
In my example spinlock and counter are euphoria global variables and
IncCounter() access them successfully.

In fact the only problem here is the fact that the euphoria interpreter is not
linked with a multi-threaded version of CRT.
By the way it work without any crash on my machine because all CRT dll that
comes with windows xp are multi-threaded versions.
But it fail with crt functions that are statically linked with exwc.exe

Maybe Robert Craig could confirm my hypothesis about static linkage of CRT in
exwc.exe

Regards,
Jacques Deschênes


Pete Lomax wrote:
> 
> On Mon, 10 Jul 2006 07:00:41 -0700, jacques desch=EAnes
> <guest at RapidEuphoria.com> wrote:
> 
> >In the past I tested threads in euphoria without any success. At my
> >first try and didn't insist because At that time i thought it was a
> >limitation of euphoria interpreter.
> It is. Btw, your example does not run on 2.4.
> >But this morning reading "windows internals" I learned that windows
> >create a new stack for each thread a process create. if so multithread
> >should in euphoria!
> >I decided to revisit it and IT WORKED!!
> If you use Eu to create pure windows/c_func threads then all will be
> fine. However Eu vars are not held on the windows stack, so using Eu
> code will fail; eg in function IncCounter after while 1 do add:
> }}}
<eucode>
>  if not equal(lower("AAA"),"aaa") then
>     puts(1,"lower(AAA)!=aaa\n")
>     if getc(0) then end if
>     abort(0)
>  end if
> </eucode>
{{{

> and after while spinlock < 2 do add:
> }}}
<eucode>
>  if not equal(lower("BBB"),"bbb") then
>     puts(1,"lower(BBB)!=bbb\n")
>     if getc(0) then end if
>     abort(0)
>  end if
> </eucode>
{{{

> Also, replace Sleep(500) with c_proc(iSleep,{500}) to avoid an almost
> inevitable type check on ms.
> The resulting program will fail or crash at random.
> 
> TDLL in the archives shows how you can multi-OS-thread C/asm code from
> Eu, but you cannot multi-OS-thread Eu code, not without doing
> something like converting it to a dll first.
> 
> Regards,
> Pete
> PS It is also not as simple as two threads calling the same routine
> (lower) simultaneously; the glitch in Sleep() shows it is re-using
> internal temporaries or something; if you comment out the lower call
> inside IncCounter, it works until the end, but the 'return 0' seems to
> spanner the last lower call from the main pgm.
> 
> PPS Al: the example did not work for me at all on Eu 2.4 but did on Eu
> 2.5 (on win98), until as above I replaced Sleep() with c_func and then
> 2.4 nearly got to the end before crashing )
> 
>