Assembler problems

[wilbert]

I tried a very simple instruction

Code: Select all

!movq rax, xmm0

which is not working on PB 5.40 Linux (x64).

When I started comparing assembler versions that come with PureBasic, I wasn't surprised.

PB 5.40 (x64) OSX comes with YASM 1.2.0 from 2012
PB 5.40 (x64) Windows comes with FASM 1.69.35 from 2011
PB 5.40 (x64) Linux comes with FASM 1.67.18 from 2006

I just don't understand why those aren't updated.
It makes it very difficult to write cross platform code this way that works on all systems.
I just don't know how to make my sha3 module code work now on all three systems.
I need to move 64 bit values between a xmm register and a 64 bit register and it just doesn't seem possible on Linux out of the box.
Anyone has a suggestion that works with the fasm version that comes with PB ?

[Keya]

+10!!! how nice would it be (but not just nice, important) to have the latest assemblers as part of PB and its nearly Christmas, Fred!

surely all of those assemblers have had a lot of bugs fixed in the 3-9 years since, not to mention new optimizations and features for us to take advantage of and empower our Purebasic abilities

wilbert i know that for forcing undocumented or unsupported stuff you can use raw !db bytes in the code instead, for all assemblers i think? as i know you know!
so i went into Hopper seeing as we're both playing with that today heehee...
First go to Main menu -> Preferences, and click "Show Hex Column" to see the raw bytes alongside the disassembly.
And now with any random binary open go to an instruction, and click Modify menu -> Assemble Instruction, and enter your "movq rax, xmm0", and it will assemble it for you to: 66 48 0F 7E C0
not fun though when you just want to program

[wilbert]

@Keya, I know using hex bytes is possible but in this case rax and xmm0 is just an example.
I need multiple general purpose registers (r8, r9, r10, r11 and r12) and multiple xmm registers.
It would be very inconvenient to have to use a different byte sequence for each combination of registers.

[Keya]

yes and it's slow-going and ugly code too but i think the only workaround besides having fasm updated? fingers crossed!

[davido]

+1 for latest assembler versions.

[Fred]

I will see what can be done.

[Keya]

Thankyou soooo much for looking into it Fred. I think we can all appreciate that if anything major has changed in the assembler itself then it's not going to be just a trivial matter of switching one binary for another! but hopefully it'll be a smooth transition! fingers crossed and best luck

[wilbert]

I will see what can be done.

Thank you very much

[Fred]

All updated to lastest versions

[wilbert]

That's great !!!

[marroh]

All updated to lastest versions

Thank you very much

[Keya]

but Christmas isn't until December heehee exciting times, thankyou sooooo much Fred!!!!!!!!

[ker2x]

It makes it very difficult to write cross platform code this way that works on all systems.

Cross platform assembly ???

[wilbert]

It makes it very difficult to write cross platform code this way that works on all systems.

Cross platform assembly ???

Yes, cross platform assembly.
Windows, OSX and Linux all use a x86 / x64 processor.
There are some differnces between fasm and yasm/nasm but if you pay attention to those you can write cross platform assembly code.

[ker2x]

only very partially then. Since they have different function call conventions.