[AT&T ASM] How to handle LOCAL(Intel syntax) in AT&T

[syn_stanzer]

Refer to the title, does any folks here knw how to handle LOCAL(Intel syntax) in GAS AT&T?!?

Thanks in advanced

[snova]

I think GAS has an assembler directive to switch between different syntaxes, but I don't remember what it is.

[QsM]

You want to use Intel syntax? If so enter the .intel_syntax directive

[syn_stanzer]

# Copyright 1998-2004 - Just For Fun Software, Inc., all rights reserved
# Author: George Woltman
# Email: woltman@alum.mit.edu
#

# ************************************************** ******
# ************************************************** ******
# ******************** FFT MACROS **********************
# ************************************************** ******
# ************************************************** ******

## Perform a 32-element FFT.

.intel_syntax
.macro fft32
LOCAL b1b, c1b

## Do a forward FFT only

.if type EQ 1
call x32_fft
jmp xmiddle_1
.endif

## Do a squaring

.if type EQ 2
call x32_fft
call xmiddle_2
## jmp x32_finish_unfft
.endif

## Do a multiply

.if type EQ 3
call x32_fft
call xmiddle_3
jmp x32_finish_unfft
.endif

## Do a multiply with pre-FFTed inputs

.if type EQ 4
call xmiddle_4
jmp x32_finish_unfft
.endif

## Do FFT levels 1,2,3
## Values 0-31 is real data.
##
## On input the 32-byte cache lines hold these data values:
## 0 16 1 17
## 2 ...
## ... ...
## 14 ...
## On output the 32-byte cache lines hold these data values:
## 0 4 1 5
## 2 ...
## 8 ...
## ...

## Do 4 eight_reals_first_fft macros
## distance between fft data elements is 4

.if type EQ 1
x32_fft:
copy_7_words
subl %eax,%eax
b1b: disp eight_reals_first_fft, 4*dist1, 8*dist1, 8
leal dist1(%esi),%esi ## Next source pointer
addb $256/4, %al ## Test loop counter
jnc b1b ## Iterate if necessary
leal esi-4*dist1,%esi ## Restore source pointer

## End common FFT code

ret
.endif

## Do inverse FFT levels 1,2,3
## On input the 32-byte cache lines hold these data values:
## 0 4 1 5
## 2 ...
## 8 ...
## ...
## On output the 32-byte cache lines hold these data values:
## 0 16 1 17
## 2 ...
## ... ...
## 14 ...

## Do 4 eight_reals_last_unfft macros
## distance between fft data elements is 4

.if type EQ 2
x32_finish_unfft:
c1b: disp eight_reals_last_unfft, 8, 4*dist1, 8*dist1
leal dist1(%esi),%esi ## Next source pointer
addb $256/4, %al ## Test loop counter
jnc c1b ## Iterate if necessary
leal esi-4*dist1,%esi ## Restore source pointer
fft_3_ret
.endif
.endm

fft32

.end



I have this portion of macro file need to be output to .o(object file in AT&T). When I assembled it the error followings show up.




test5.mac:107: Error: no such instruction: `local b1b,c1b'
test5.mac:107: Error: non-constant expression in ".if" statement
test5.mac:107: Error: junk at end of line, first unrecognized character is `E'
test5.mac:107: Error: non-constant expression in ".if" statement
test5.mac:107: Error: junk at end of line, first unrecognized character is `E'
test5.mac:107: Error: non-constant expression in ".if" statement
test5.mac:107: Error: junk at end of line, first unrecognized character is `E'
test5.mac:107: Error: non-constant expression in ".if" statement
test5.mac:107: Error: junk at end of line, first unrecognized character is `E'
test5.mac:107: Error: non-constant expression in ".if" statement
test5.mac:107: Error: junk at end of line, first unrecognized character is `E'
test5.mac:107: Error: non-constant expression in ".if" statement
test5.mac:107: Error: junk at end of line, first unrecognized character is `E'


Could anyone understand how to handle the LOCAL in AT&T?
Thanks

[boz]

Are those local symbols? If so, I don't see them being used anywhere... If they are local symbols, AT&T syntax requires them to be in the form .L1, .L2, etc.

[syn_stanzer]

Are those local symbols? If so, I don't see them being used anywhere... If they are local symbols, AT&T syntax requires them to be in the form .L1, .L2, etc.


Hi Boz, did u mean I need to change the "LOCAL b1b, c1b" to ".L1 b1b, c1b"??

[boz]

Well, to begin with, I'm completely unfamiliar with the LOCAL keyword. Is that supposed to declare a local variable or a local symbol? Like I said, I couldn't see where you were using them in your code. If they're supposed to be local variables, I suggest you just use the stack. But, yes, if they're local symbols, you have to do:

.L1 blb
.L2 clb

Oh, never mind. I see them now. Yeah, try the above. They're local symbols.

[syn_stanzer]

Hi Boz, I just tested

.L1 blb
.L2 clb

but it doesn't work as well, it comes up with the error unknown-pseudo op

[boz]

Try commenting out the LOCAL declaration entirely and in place of blb: and cbl: labels, use .1: and .2:.


http://docs.sun.com/app/docs/doc/817-5477/ennab?a=view (http://docs.sun.com/app/docs/doc/817-5477/ennab?a=view)

Even though that's Solaris' x86 asm, I believe it uses the same syntax. Sorry, I have never used local symbols, so this is a learning experience for me, too. ;)