Unless you come up with your own C-compiler, calling them idiots is not valid

??? XMM6+ are required to be saved due to MS x64 ABI. In fact, that's the entire reason for Intel adding the "vzeroupper" crap and complicating themselves with it -- instead of just doing it, by default, zero-extension when they added ymm registers.

(MS ABI, being "old", only requires the XMM regs to be saved, as in the 128-bit lowerpart since that's what existed when it was designed, this is why any ABI reliant on piece of shit SSE is garbage and not future-proof; Linux x64 ABI has no problems at all here)

Page 4, my post:

Furs wrote:

??? XMM6+ are required to be saved due to MS x64 ABI. In fact, that's the entire reason for Intel adding the "vzeroupper" crap and complicating themselves with it -- instead of just doing it, by default, zero-extension when they added ymm registers.

(MS ABI, being "old", only requires the XMM regs to be saved, as in the 128-bit lowerpart since that's what existed when it was designed, this is why any ABI reliant on piece of shit SSE is garbage and not future-proof; Linux x64 ABI has no problems at all here)

Indeed, AVX in your brain only. (just FYI, AVX2 registers are YMM, and AVX is not SSE)

Now please continue with your hysteria because you got so schooled you have nothing other than bullshit to spew.

There's another problem I discovered with GCC. When it saves the XMM registers on the stack (the bullshit x64 MS ABI being the only one in existence requiring so, none other ABI (x64 or 32-bit) need it, I fucking hate it)... it emits shit code like (translated from AT&T):

Code:

push rbp 
mov rbp, rsp 
sub rsp, 64  ; space for XMM regs saved (4 of them in this case) 
and rsp, -32 ; realign stack to 32-bytes (yes, regs are saved unaligned!) 
sub rsp, 32  ; stack frame (random stuff, one AVX2 vector in this case)    

I have no idea what you're talking about. That code was translated from GCC's output. GCC outputs code in AT&T syntax, hence the exact thing I said.

Next time I'll simply copy-paste AT&T code directly from GCC just for you instead of "translating" it. I bet you will whine like how AT&T is "garbage" or other completely off topic stuff in that case, just to avoid the point.

@revo is deleting my posts?

I saw someone got extremely MAD! That's scary!!

Ok okay. You're correct. MS-64 ABI does use AVX2 vectors and 32-byte alignment. Happy?

system error wrote:

@revo is deleting my posts?

Yes.

system error wrote:

I saw someone got extremely MAD! That's scary!!

With all the name calling would you be surprised?

revolution wrote:

system error wrote:

@revo is deleting my posts?

Yes.

system error wrote:

I saw someone got extremely MAD! That's scary!!

With all the name calling would you be surprised?

The atmosphere generally on this board has been toxic these last few weeks. So much hatred. I would delete even more posts than revolution has. I think he's been remarkably restrained actually. It makes you not want to post your opinions because you know most people are more concerned with their next flame attack than having a reasonable balanced discussion.

MS-64 ABI uses (mandates they are saved) XMM regs, so using the same AVX2 registers implicitly clobbers them since they alias them.

MS-64 bit ABI don't use no AVX2 instructions or registers.

system error wrote:

MS-64 bit ABI don't use no AVX2 instructions or registers.

It uses the XMM registers, and the AVX registers alias the XMM registers.

Are you mentally challenged or something?

• I use YMM6 register in my function, and implicitly, XMM6. Deal with it.
  
• Function is a callback using MS ABI (because that's what the caller of the fucking callback expects); no, I'd never use this piece of shit ABI if it wasn't required in stuff out of my control (such as callbacks for a DSP processing function in a plugin, why else would I use AVX?!)
  
• My function thus trashes XMM6 since it's aliased to YMM6
  
• Since the caller of the callback (callback happens to be my function, so it calls my function) expects XMM6 to be unchanged (due to ABI), GCC must save XMM6 since the stupid ABI requires it
  
• If the ABI did not require it, then the caller wouldn't expect it to be unchanged, and thus none of this would be an issue (see: Linux x64 ABI for a REAL EXAMPLE where GCC will not save XMM6, just use attribute sysv_abi)

This is just a simplified example of course, but something tells me you still won't understand it (on purpose), grasping at straws aye.