[llvm-dev] LLC Version 3.8 : Unsupported library call operation for a mul instruction

Sun Dec 16 13:13:43 PST 2018

Hi Alberto,

Thanks for your reply. So this piece of code is a legacy IIVM IR code which
is not compiled from C project.. I don't even know which clang they use to
produce the IR.

But on the other hand, I tried to collect those related code pieces and put
them into a simple hello world function, it works fine (attached below). So
it seems quite wired to me...

Also, I made a mistake in previous email. It should be llc version 3.8.1,
not 3.8.0.

Irene.

    ; ModuleID = 'hello.c'
    target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
    target triple = "x86_64-unknown-linux-gnu"

    @.str = private unnamed_addr constant [13 x i8] c"Hello
World\0A\00", align 1

    ; Function Attrs: nounwind uwtable
    define i32 @main() #0 {
      %1 = alloca i32, align 4
      store i32 0, i32* %1, align 4
      %2 = add i256 0, 1
      %3 = call i256 @llvm.bswap.i256(i256 %2)
      %4 = mul i256 %3, 190
      %5 = call i256 @test.udiv.i256(i256 %4, i256 100)
      %6 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([13
x i8], [13 x i8]* @.str, i32 0, i32 0))
      ret i32 0
    }

     ; Function Attrs: nounwind readnone
declare i256 @llvm.bswap.i256(i256) #1

declare i256 @llvm.ctlz.i256(i256, i1) #1

define private <2 x i256> @test.udivrem.i256(i256 %x, i256 %y) {
Entry:
  %0 = icmp ule i256 %y, %x
  br i1 %0, label %Main, label %Return

Main:                                             ; preds = %Entry
  %y.lz = call i256 @llvm.ctlz.i256(i256 %y, i1 true)
  %r.lz = call i256 @llvm.ctlz.i256(i256 %x, i1 true)
  %i0 = sub nuw i256 %y.lz, %r.lz
  %1 = shl i256 %y, %i0
  br label %Loop

Loop:                                             ; preds = %Continue, %Main
  %y.phi = phi i256 [ %1, %Main ], [ %8, %Continue ]
  %r.phi = phi i256 [ %x, %Main ], [ %r1, %Continue ]
  %i.phi = phi i256 [ %i0, %Main ], [ %6, %Continue ]
  %q.phi = phi i256 [ 0, %Main ], [ %7, %Continue ]
  %2 = sub nuw i256 %r.phi, %y.phi
  %3 = or i256 %q.phi, 1
  %4 = icmp uge i256 %r.phi, %y.phi
  %r1 = select i1 %4, i256 %2, i256 %r.phi
  %q = select i1 %4, i256 %3, i256 %q.phi
  %5 = icmp eq i256 %i.phi, 0
  br i1 %5, label %Return, label %Continue

Continue:                                         ; preds = %Loop
  %6 = sub nuw i256 %i.phi, 1
  %7 = shl i256 %q, 1
  %8 = lshr i256 %y.phi, 1
  br label %Loop

Return:                                           ; preds = %Loop, %Entry
  %q.ret = phi i256 [ 0, %Entry ], [ %q, %Loop ]
  %r.ret = phi i256 [ %x, %Entry ], [ %r1, %Loop ]
  %ret0 = insertelement <2 x i256> undef, i256 %q.ret, i64 0
  %ret = insertelement <2 x i256> %ret0, i256 %r.ret, i64 1
  ret <2 x i256> %ret
}

define private i256 @test.udiv.i256(i256 %x, i256 %y) {
  %1 = call <2 x i256> @test.udivrem.i256(i256 %x, i256 %y)
  %2 = extractelement <2 x i256> %1, i64 0
  ret i256 %2
}

declare i32 @printf(i8*, ...) #1

attributes #0 = { nounwind uwtable "disable-tail-calls"="false"
"less-precise-fpmad"="false" "no-frame-pointer-elim"="true"
"no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false"
"no-nans-fp-math"="false" "stack-protector-buffer-size"="8"
"target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2"
"unsafe-fp-math"="false" "use-soft-float"="false" }
@
attributes #1 = { "disable-tail-calls"="false"
"less-precise-fpmad"="false" "no-frame-pointer-elim"="true"
"no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false"
"no-nans-fp-math"="false" "stack-protector-buffer-size"="8"
"target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2"
"unsafe-fp-math"="false" "use-soft-float"="false" }

!llvm.ident = !{!0}

!0 = !{!"clang version 3.8.0 (tags/RELEASE_380/final)"}

On Sun, Dec 16, 2018 at 10:02 PM Alberto Barbaro <barbaro.alberto at gmail.com>
wrote:

> Hi Irene,
> I'm absolutely not but have you tried to compile your .c project using -O3
> -march=native and see that call to bswap is not present after? Are we sure
> that bswap was present in version 3.8.1?
>
> Thanks
>
> On Sun, Dec 16, 2018, 17:47 div code via llvm-dev <llvm-dev at lists.llvm.org
> wrote:
>
>> Hello List,
>>
>> I am on the hook to instrument a piece of legacy LLVM IR code, and then
>> we are planning to feed to the SeaHorn framework for some model checking
>> tasks.
>>
>> After the instrumentation, I tried to use llc (version 3.9) to compile
>> the IR code, and it works fine. However, when I try to use llc (version
>> 3.8.1, the default llvm version of SeaHorn) to compile the IR code, it
>> shows the following error:
>>
>> LLVM ERROR: Unsupported library call operation!
>>
>> I tried to do some "delta debugging" and eventually locate the buggy line
>> of the code. So as shown below, the following code cause llc (version
>> 3.8.1) throws the "Unsupported library call operation" issue, but it works
>> fine for llc version 3.9.
>>
>>   %20 = call i256 @llvm.bswap.i256(i256 %msg.value)
>>   %21 = mul i256 %20, 190
>>
>> I tried to tweak the buggy lines into the following code, and it can also
>> pass the compilation:
>>
>>   %20 = call i256 @llvm.bswap.i256(i256 %msg.value)
>>   %21 = add i256 %20, 190
>>
>> This seems really wired to me.. Any idea on that? Thank you!
>>
>>
>> Best,
>>
>> Irene
>> _______________________________________________
>> LLVM Developers mailing list
>> llvm-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181216/5e485ccb/attachment.html>