[llvm-commits] [llvm] r58232 - in /llvm/trunk: lib/CodeGen/SelectionDAG/ test/CodeGen/Generic/ test/CodeGen/Mips/ test/CodeGen/X86/

[Bill Wendling]

On Mon, Oct 27, 2008 at 11:00 AM, Duncan Sands <baldrick at free.fr> wrote:
> Hi Bill,
>
>> If you're working on a fix, could you XFAIL this for now?
>
> the cse-libcalls is fixed.  I will xfail the vec_splat one tomorrow.
> The reason for not doing that now is that as nightly tester results
> come in I can see if they have LegalizeTypes turned on or not by
> looking at whether vec_spat.ll fails!  I hope this is acceptable.
>
It's not ideal, but I suppose it's okay for now. I did see this
failure now happening on my PPC box:

$ llvm-dis -o - bug.bc
; ModuleID = 'bugpoint-reduced-simplified.bc'
target datalayout =
"E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f128:64:128"
target triple = "powerpc64-apple-darwin9.5"

define void @__divtc3({ ppc_fp128, ppc_fp128 }* noalias sret
%agg.result, ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c, ppc_fp128 %d)
nounwind {
entry:
	%imag59 = load ppc_fp128* null, align 8		; <ppc_fp128> [#uses=1]
	%0 = mul ppc_fp128 0xM00000000000000000000000000000000, %imag59		;
<ppc_fp128> [#uses=1]
	%1 = mul ppc_fp128 0xM00000000000000000000000000000000,
0xM00000000000000000000000000000000		; <ppc_fp128> [#uses=1]
	%2 = add ppc_fp128 %0, %1		; <ppc_fp128> [#uses=1]
	store ppc_fp128 %2, ppc_fp128* null, align 16
	unreachable
}
$ llc -o - bug.bc
Assertion failed: (NodeID != ReadyToProcess && NodeID != Processed &&
"Invalid node id for user of unprocessed node!"), function run, file
/Volumes/Gir/devel/llvm/llvm.src/lib/CodeGen/SelectionDAG/LegalizeTypes.cpp,
line 145.
0   llc                                 0x0080de25
_ZN4llvm3sys20SetInterruptFunctionEPFvvE + 87
1   llc                                 0x0080df91
_ZN4llvm3sys20SetInterruptFunctionEPFvvE + 451
2   libSystem.B.dylib                   0x92f0109b _sigtramp + 43
3   ???                                 0xffffffff 0x0 + 4294967295
4   libSystem.B.dylib                   0x92f79ec2 raise + 26

Does this look related to your stuff?

-bw