[LLVMdev] [fwd] Re: LLVM Compiler Infrastructure

[Robert L. Bocchino Jr.]

I would like to know the answer to this too.  One of the optimization  
passes does this, but I don't know which.  If you run gccas with - 
debug-pass=Arguments, it will spit out all the optimizations that it  
runs.  You can do a binary search on these optimizations to figure  
out which one causes this transformation.  Then you can try to  
disable it, but you may have to hack on the optimization pass itself.

If you figure out how to do this, please let me know!

Rob

On Nov 1, 2005, at 3:09 PM, Misha Brukman wrote:

> Hi, Yiping!
>
> I am not sure of the answer to your question, but I am forwarding  
> it to
> the LLVMdev list where I am sure someone will be able to answer you.
>
> Please send development questions directly to LLVMdev and you will  
> get a
> response quicker, as it is read by many LLVM developers.
>
> ----- Forwarded message from Yiping Fan <fanyp at cs.ucla.edu> -----
>
> Date: Mon, 31 Oct 2005 17:20:24 -0800
> From: "Yiping Fan" <fanyp at cs.ucla.edu>
> To: "Misha Brukman" <brukman at uiuc.edu>, "guoling han"  
> <leohgl at cs.ucla.edu>,
>         <cong at cs.ucla.edu>, "'Zhiru Zhang'" <zhiruz at cs.ucla.edu>
> Cc: "Brian Gaeke" <gaeke at uiuc.edu>
> Subject: Re: LLVM Compiler Infrastructure
>
> Hi Misha,
>     How are you doing recently? It has been long time since the  
> last email we exchanged.
> We used your LLVM compiler in our xPilot behavioral synthesis  
> system, and made great
> progress, thanks to your great job.
>    Now we got a small but annoying problem. When we get the  
> following C code to llvm-gcc,
>
> short test(int x) {
>         return (short) (x >> 10);
> }
>
>    It will generate the following LLVM code:
>
> ; ModuleID = 'test.bc'
> target endian = little
> target pointersize = 32
> target triple = "i686-pc-linux-gnu"
> deplibs = [ "c", "crtend" ]
>
> implementation   ; Functions:
>
> short %test(int %x) {
> entry:
>         %x = cast int %x to uint                ; <uint> [#uses=1]
>         %tmp.2 = shr uint %x, ubyte 10          ; <uint> [#uses=1]
>         %tmp.3 = cast uint %tmp.2 to short              ; <short>  
> [#uses=1]
>         ret short %tmp.3
> }
>
> Basically, LLVM frontend will first convert an integer to unsigned  
> integer, and then do shifting.
> It seems that this conversion often occurs for SHIFTING operations.  
> Although it is correct for 32-bit
> general purpose processor, it does not fit well for hardware  
> synthesis, since we need
> to use wider components (up to 32-bit) to implement these  
> operations, otherwise we may lose sign flag.
> Our favorite LLVM code is that without (or with very few) CAST, and  
> doing the SHR directly for the signed integers.
>
> So our question is that do you know there is any option in LLVM to  
> disable this kind of cast-to-unsigned
> code generation?
>
> Thank you very much.
>
> -Yiping Fan
> UCLA Computer Science Department
> 4651 Boelter Hall
> 503 Hilgard Ave, Los Angeles, CA 90095
> Tel: 310-206-5449
> Email: fanyp at cs.ucla.edu
> WWW: http://ballade.cs.ucla.edu/~fanyp
>
> ----- End forwarded message -----
>
> -- 
> Misha Brukman :: http://misha.brukman.net :: http://llvm.cs.uiuc.edu
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

Robert L. Bocchino Jr.
Ph.D. Student, Computer Science
University of Illinois, Urbana-Champaign

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20051101/b8dea8db/attachment.html>