[llvm-commits] [llvm-gcc-4.0] r42134 - /llvm-gcc-4.0/trunk/gcc/llvm-convert.cpp

[Dale Johannesen]

On Sep 23, 2007, at 3:16 PM, Chris Lattner wrote:

>
> On Sep 23, 2007, at 2:36 PM, Dale Johannesen wrote:
>
>>
>> On Sep 23, 2007, at 11:39 AM, Chris Lattner wrote:
>>> One problem Dale:
>>>
>>> @@ -4350,9 +4351,10 @@
>>> +  case BUILT_IN_POWIL:
>>> +    return false;   // treat long double as normal call
>>>     case BUILT_IN_POWI:
>>>     case BUILT_IN_POWIF:
>>> -  case BUILT_IN_POWIL:
>>>       Result = EmitBuiltinPOWI(exp);
>>>       return true;
>>>
>>> I don't think this works.  There is no "powil" function for llvm-gcc
>>> to fall back on.  Not handling this builtin will result in a call to
>>> __builtin_powil, which will fail at link-time.  I had to handle  
>>> these
>>> explicitly here because "fold" produces these operations in some
>>> cases.  Try long double X*X*X with -ffast-math.
>>
>> You're right.  A quick one:  is changing the Intrinsics definitions
>> going to affect the bitcode format?
>
> It depends on how you change it.  One thing that you may be able to
> do is change this:
>
> let Properties = [IntrNoMem] in {
> ..
>    def int_powi_f32 : Intrinsic<[llvm_float_ty, llvm_float_ty,
> llvm_i32_ty]>;
>    def int_powi_f64 : Intrinsic<[llvm_double_ty, llvm_double_ty,
> llvm_i32_ty]>;
> }
>
> to:
>
> let Properties = [IntrNoMem] in {
> ..
>    def int_powi : Intrinsic<[llvm_anyfloat_ty, LLVMMatchType<0>,
> llvm_i32_ty]>;
> }
>
> I don't think this changes the .ll or .bc syntax for f32/f64, but
> allows any fp type to be used with powi.  Calls to
> Intrinsic::getDeclaration(powi) will have to change though.

I was thinking more of adding int_powi_f80.  What I'm mostly  
concerned about is whether
this intrinsic info somehow maps into an enum in the intermediate  
files, is there anything
like that?  If everything is done by name it should be OK.