[LLVMdev] Loopinfo Analysis

[Hisham Chowdhury]

Hi Tom,
Thanks for your response. The sample code that I was compiling just have 1 loop in it and interestingly this issue only occurs if I run some combination of passes on the .ll code(when running both loop strength reduce and SROA, it doesnt happen if I run one or the other). The following is the machine function dump of the code:
BB#10: derived from LLVM BB %for.cond
    Predecessors according to CFG: BB#9 BB#12
        %reg1039<def> = MOV %reg1099
        %reg1040<def> = MOV  %reg1100
        %reg1041<def> = MOV  %reg1101
        %reg1042<def> = MOV %reg1102
        %reg1043<def> = MOV  %reg1103
        %reg1044<def> = MOV  %reg1104
        %reg1080<def> = FADD_RR %reg1042, %reg1043
        %reg1081<def> = CMP_F32_UGT %reg1080, %reg1038
        IF %reg1081, <BB#13>
    Successors according to CFG: BB#14 BB#13

BB#14: derived from LLVM BB %for.cond
    Predecessors according to CFG: BB#10
        %reg1082<def> = CMP_I32_UGE %reg1044, %reg1026
        IF %reg1082, <BB#13>
    Successors according to CFG: BB#11 BB#13

BB#11: derived from LLVM BB %for.body
    Predecessors according to CFG: BB#14
        %reg1083<def> = FADD %reg1040, %reg1040
        %reg1048<def> = MOV %reg1083
    Successors according to CFG: BB#12

BB#12: derived from LLVM BB %for.inc
    Predecessors according to CFG: BB#11
        %reg1090<def> = ADD %reg1044, 1
        %reg1091<def> = ADD %reg1039, 255
        %reg1049<def> = MOV %reg1090
        %reg1050<def> = MOV %reg1091
        UncondBr <BB#10>
        %reg1099<def> = MOV %reg1050
        %reg1100<def> = MOV %reg1046
        %reg1101<def> = MOV %reg1045
        %reg1102<def> = MOV %reg1047
        %reg1103<def> = MOV %reg1048
        %reg1104<def> = MOV %reg1049
    Successors according to CFG: BB#10

BB#13: derived from LLVM BB %for.end
    Predecessors according to CFG: BB#10 BB#14
        %reg1095<def> = ADD %reg1024, %reg1092
        STORE %reg1048, %reg1095; mem:ST4[%arrayidx]
        return

for the above code its returning true for both BB10 and BB14. but, instead if i use getLoopHdr for the same set of blocks, i get the right result(both blocks returns the same hdr = #BB10).

Thanks for help,
Hisham


--- On Thu, 6/17/10, Thomas B. Jablin <tjablin at cs.princeton.edu> wrote:

> From: Thomas B. Jablin <tjablin at cs.princeton.edu>
> Subject: Re: [LLVMdev] Loopinfo Analysis
> To: 
> Cc: llvmdev at cs.uiuc.edu
> Date: Thursday, June 17, 2010, 6:22 AM
> Hi Hisham,
> Most likely the basic blocks are the headers of two
> different loops. Try running viewCFG() on the function in
> question to see if this is the case.
> Tom
> 
> ----- Original Message -----
> From: "Hisham Chowdhury" <hisham_chow at yahoo.com>
> To: llvmdev at cs.uiuc.edu
> Sent: Wednesday, June 16, 2010 7:22:00 PM GMT -05:00
> US/Canada Eastern
> Subject: [LLVMdev] Loopinfo Analysis
> 
> Hello,
> I have a question regrading the analysis pass that
> generates loop info from an .ll code. My previous
> understanding was there will be just one loop header(in the
> loop info) for a particular loop. But, when i use
> isLoopHeader() member function from the loop info class I
> get 'true' return value for two different basic blocks. Note
> both basic blocks are loop conditional block(break condition
> of the loop), but only one starts the loop. Is this the
> right behavior for the isLoopHeader() member function?
> 
> Thanks for help,
> Hisham
> 
> 
>       
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu 
>        http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu 
>        http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>