> This compiled on my 3.4.6 compiler on Linux just fine before the patch. > What was the error you were trying to fix?
"Allignment Value not constant" was the error. This was gcc 3.4.4 > Are you sure this is right? In section 5.31 of the GCC Manual it > indicates that if an alignment size is not given with the aligned > attribute keyword then it defaults to: > > "the maximum useful alignment for the target machine you are compiling > for. Whenever you leave out the alignment factor in an `aligned' > attribute specification, the compiler automatically sets the alignment > for the declared variable or field to the largest alignment which is > ever used for any data type on the target machine you are compiling > for". > > Is this what we want here? I believe so. The alignment of that field needs to be sufficient for any data type put there. The non-gcc version does that with a union. My reading of the gcc manual is that without a value, this attribute does exactly that (and since doubles are not always the most constrained types, vectors tend to be, this is more general, though I doubt altivec vectors are going to be used directly by LLVM in small vectors anytime soon). Andrew _______________________________________________ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits