Re: ml64, PROC and parameters

I don't know of many MASM books for x64 as well.
Chapter 12 of this book
http://www.amazon.com/gp/product/0321374460/ref=s9_asin_title_1?pf_rd_m=ATVPDKIKX0DER&pf_rd_s=center-1&pf_rd_r=038N71NS0D9BFAST29X7&pf_rd_t=101&pf_rd_p=320448601&pf_rd_i=507846
explains the concepts below, even if it is not an assembler book.

From my experience, the hard part of x64 is not the assembler, but,
to make the generated code to satisfy the requirements of the platform
where it will run.

--

--
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included script samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm


"Peter" <Peter@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:2F7DA4A9-951B-4776-95BC-8B0D060C0FBF@xxxxxxxxxxxxxxxx
Thanks Ivan, your help is great,
I have now all basic inputs to begin with my own functions.

last question:
I found many of MASM books, tutorials and manuals, but all for 32-bit,
does exist some books or manuals where are explained such basics of x64
MASM
- like you described in your posts ?

Peter


"Ivan Brugiolo [MSFT]" wrote:

#1
_test_nested creates 98h bytes of stack space for the sake of the
exercise.
It does not use them, but, the stack space could be used for temporary
storage,
temporary structures, etc etc

#2
exception handling support in amd64/ia64 is not optional.
Every non-leaf function needs to have a function descriptor
(or function-entry, as sometimes referred in older documentation,
and in the `.fnentry` command in cdb/ntsd/windbg, or personality-routine,
if you have ever read the ABI for IA64 from HP/Intel).
The function descriptor is used to perfrom stack unwinding even if your
function does not throw/raise/handle an exception.
Immagine the following two scenarios:
-stack probing and commitment
Your function might take an exception while pushing the registers on the
stack
while in the middle of the prologue (because it hits the guard-page)
-nested calls
Your function may call one other function that raises an exception,
and, your function is being called from a function that handles the
exception.
In this case the compiler needs to generate a function-entry header in
order
for the Exception Unwind code to locate the filter.

#3
I simply wanted to make your function to assemble as it were.
Indeed _MyProc is wrong by the rules and principles set so far.
You should push on the stack only the non-parameter register.

--

--
This posting is provided "AS IS" with no warranties, and confers no
rights.
Use of any included script samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm


"Peter" <Peter@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:55EDEBA5-4AFB-425A-83E2-7BD8F69B7B5F@xxxxxxxxxxxxxxxx
Thanks for more lite on this problem.
I think I understand what is register homing now, and that first four
parameters are stored in mentioned four registers (which is new for
me).
Q1:
In _test_nested function, Why is there created 98h bytes space on the
stack ?
What purpose can be used this stack space for ?
Q2:
keyword "frame": I red paragraph about it in ml64. If I good understand
it
has sense to use it only when I want support for structured exception
handling ? Or it has also other sense ?
Q3:
I am a bit confusing with your sentence: "You do NOT save the parameter
registers by pushing them on the stack..." and with prologue in
_MyProc,because there are pushed two registers: rcx,rdx and these two
can
be
also used for parameters. If _MyProc is changed to have parameters, can
stay
prolog in _MyProc sample the same ?

Peter


"Ivan Brugiolo [MSFT]" wrote:

First of all, I would advise you using the macros
already defined in macamd64.inc.
Then, I wrote the code you posted in a way that it
assembles correctly, and I reported it below.
Notheless, what you are trying to do just tells me
you have not quite understood the calling convention
of the architecture.
The first 4 parameters are in registers rcx, rdx, r8, r9.
You do NOT save the parameter registers by pushing
them on the stack. The caller has prepared space on the stack for you,
and, all you have to do is register homing.
The _test_leaf function shows how to do register homing properly.
In any case, register homing is hardly ever necessary,
and, it usually indicates poor optimization and register handling.
You can push on the stack the other registers besides the first 4,
and, that is up to your function to do that via an explicit `push rbx`
call, or something more complicated.
The fifth param of a function is usually at `rsp+28`, unless you have
altered the stack pointer in the prologue.

//------------------------------------------------
include macamd64.inc

LEAF_ENTRY _test_leaf, _TEXT$00

mov [rsp+ 8h],rcx
mov [rsp+10h],rdx
mov [rsp+18h],r8
mov [rsp+20h],r9

ret
LEAF_END _test_leaf, _TEXT$00

public _test_nested
NESTED_ENTRY _test_nested, _TEXT$00

sub rsp, 98h
.allocstack 98h
END_PROLOGUE

mov rcx, 1
mov rdx, 2
mov r8, 3
mov r9, 4

call _test_leaf

add rsp, 98h
ret
NESTED_END _test_nested, _TEXT$00

_TEXT$00 segment para 'CODE'
public _MyProc
_MyProc proc frame
push rbx
.pushreg rbx
push rcx
.pushreg rcx
push rdx
.pushreg rdx
sub rsp, 98h
.allocstack 98h
.endprolog

mov rcx, 1
mov rdx, 2
mov r8, 3
mov r9, 4

call _test_leaf

add rsp, 98h
pop rdx
pop rcx
pop rbx
ret
_MyProc endp
_TEXT$00 ends

END


--

--
This posting is provided "AS IS" with no warranties, and confers no
rights.
Use of any included script samples are subject to the terms specified
at
http://www.microsoft.com/info/cpyright.htm


"Peter" <Peter@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:C935B3A1-9216-4014-9EF6-3E1096FB0659@xxxxxxxxxxxxxxxx
Thanks Ivan,
I tried to add prologue to my function, but still I have problem to
compile
simple example function, it complains: "A2008 syntax error : FRAME".
I am looking into MSDN MASM Reference -> PROC. There is "FRAME"
after
arguments in PROC definition, I dont understand what is wrong with
my
asm
sample.
I looked for some more detailed x64 MASM doc, but nothing found, in
MSDN
are
only several examples, not helpful for this case.
Can you recommend some links/doc for x64 MASM (I found only
something
for
x86) ?
This is my not-compilable sample, what is wrong with it ?
//////////////////// x64 example
.code

MyFunc PROC dwArg0:DWORD, dwArg1:DWORD
FRAME
push rbx
.pushreg rbx
push rcx
.pushreg rcx
push rdx
.pushreg rdx
.endprolog

; loading first argument:
mov eax, dwArg0
;... another work

functionend:
pop rdx
pop rcx
pop rbx

ret
MyFunc ENDP

END
//////////////////////// end of ml64 example
Peter

"Ivan Brugiolo [MSFT]" wrote:

Can you post a relevant fragment ?

on x64, reading the first 4 parameters from any
on-the-stack location without having done the register-homing first
leads to undefined results.
Starting touching the stack without having completed
the prologue of a function gives also undefined results.

On average, you always read the parameters trough a register
(for the first 4) or trough a direct stack-read (either via ESP or
some
other register) only after you have completed the prologue of the
function.

--

--
This posting is provided "AS IS" with no warranties, and confers no
rights.
Use of any included script samples are subject to the terms
specified
at
http://www.microsoft.com/info/cpyright.htm


"Peter" <Peter@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:A394335C-4DB8-4C62-9921-E5FE02B2A663@xxxxxxxxxxxxxxxx
I have created function in *.asm module included in x64 project.
Problem is that arguments cannot be used when they are defined
after
PROC,
for example:
MyFunc PROC arg1:DWORD, arg2:DWORD
In such function I cannot read argument value by:
1. mov ebx, arg1
But no problem to read argument value by:
2. mov rbx, [rsp+8*1]

I found some information about this topic in forums that in ml64
does
not
support passing parameters by 1. Is it true ? Or I is something
wrong
with
defining my function ?

Peter












.

Relevant Pages

  • Re: ml64, PROC and parameters
    ... _test_nested creates 98h bytes of stack space for the sake of the exercise. ... You should push on the stack only the non-parameter register. ... The first 4 parameters are in registers rcx, rdx, r8, r9. ... mov,rcx ...
    (microsoft.public.development.device.drivers)
  • Re: ml64, PROC and parameters
    ... _test_nested creates 98h bytes of stack space for the sake of the exercise. ... You should push on the stack only the non-parameter register. ... The first 4 parameters are in registers rcx, rdx, r8, r9. ... mov,rcx ...
    (microsoft.public.development.device.drivers)
  • Re: ml64, PROC and parameters
    ... The first 4 parameters are in registers rcx, rdx, r8, r9. ... The _test_leaf function shows how to do register homing properly. ... mov,rcx ...
    (microsoft.public.development.device.drivers)
  • Re: ml64, PROC and parameters
    ... I think I understand what is register homing now, ... The first 4 parameters are in registers rcx, rdx, r8, r9. ... mov,rcx ...
    (microsoft.public.development.device.drivers)
  • Re: Cost of calling a standard library function
    ... >> Thank you for a very thorough and nice explanation Beth! ... and then edi four instructions in a row. ... your code keeps reading the memory through esi register 4 times in a row. ... mov eax D@SourceRect ...
    (alt.lang.asm)
Quantcast