llvm dev - Apr 2010 - [LLVMdev] indirect jumps

Hi,

What kind of C/C++ high level code can generate a computed jump, such as:
  jmpq *%r14 
or
  jmpq    *(%r14,%rbx,8)
?

I imagine that any calls (including virtual) would use something like 'call
*%r14',
and the above jumps are mostly from 'switch' statements.

Is this correct?
Anything else?

Thank you,
Dan

 		 	   		  
_________________________________________________________________
The New Busy is not the too busy. Combine all your e-mail accounts with Hotmail.
http://www.windowslive.com/campaign/thenewbusy?tile=multiaccount&ocid=PID28326::T:WLMTAGL:ON:WL:en-US:WM_HMP:042010_4
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20100414/d03d3f45/attachment.html>

On Wed, Apr 14, 2010 at 2:43 PM, dan mihai <dnmh68 at hotmail.com>
wrote:
> Hi,
>
> What kind of C/C++ high level code can generate a computed jump, such as:
>   jmpq *%r14
> or
>   jmpq    *(%r14,%rbx,8)
> ?
>
> I imagine that any calls (including virtual) would use something like
'call
> *%r14',
> and the above jumps are mostly from 'switch' statements.
>
> Is this correct?
> Anything else?
Ways I can think of to generate computed jumps:
int f1(int (*b)(void)) { return b(); } // Tail call (recently
implemented in LLVM)
void f(void), g(void), h(void), i(void);
void f2(int x) { switch (x) { case 1: f(); case 2: g(); case 3: h();
case 4: i(); } } // switch
void f3(int n) { void* x[] = { &&L1, &&L2, &&L3 }; goto
*x[n]; L1:
f(); L2: g(); L3: h(); } // computed goto

Note that virtual thunks, which show up with C++ class hierarchies,
internally look similar to f1.

-Eli

 
Hello,
I am trying to clone a machine basic block when I ran into some issues, where I
am not able to make some headway. Any of yours help is highly appreciated here:
My question is about Machine Basic Block Duplication:
 
-          Is there a utility to clone a MachineBasicBlock in LLVM? I found
utility to clone machineInstrs, but couldn’t find similar utility for
MachineBasicBlock. So, I created a utility myself for cloning a
MachineBasicBlock, but I am running into an issue when LLVM is trying to
 destroy the use list for registers (RemoveRegOperandsFromUseLists-->
RemoveRegOperandFromRegInfo()-->NextOp->getReg() – where nextOp seems like
invalid but not NULL). Its hitting assert in one of the registers that is
getting reused in the cloned block
What I am trying to do is to use exact set of virtual registers in both the
blocks. Will it cause any issues?
 
Here is the snippet of what the code does:
 MachineBasicBlock* I;
 MachineBasicBlock* copy;
 
copy = createMachineBasicBlock(I->getBasicBlock());
 
Clone all the instrs from I --> copy
Copy all the successor of ‘I’ to ‘copy’ as well
 
If ‘I’ has 2 predecessors, then update the succ list of the second pred to point
to ‘copy’ and insert ‘copy’ after the second pred (‘I’ physically resides after
the first pred):
 
MachineFunction::iterator It = (predMBB+1);
MF.insert(It,copy);
predMBB->addSuccessor(copy);
predMBB->removeSuccessor(I)
 
MF.RenumberBlocks();
 
Am I missing anything here?
 
Please note: I am doing this step as a pass just before the asm printer pass in
the target machine layer after all the instructions has been generated,
registers has been allocated.
 
Appreciate your help a lot,
 
Thanks,
Hisham


 

 


      
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20100414/7615143f/attachment.html>

On Wed, 2010-04-14 at 17:30 -0700, Hisham Chowdhury
wrote:
>  
> 
> 
> -          Is there a utility to clone a MachineBasicBlock in LLVM
> 
> 
> 
>  
>          
> 
There is CloneBasicBlock routine
in ./lib/Transforms/Utils/CloneFunction.cpp

- Sanjiv
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev