[llvm-commits] [llvm] r110576 - in /llvm/trunk: lib/Target/CellSPU/SPUISelLowering.cpp lib/Target/CellSPU/SPUISelLowering.h lib/Target/CellSPU/SPUInstrInfo.td lib/Target/CellSPU/SPUNodes.td lib/Target/CellSPU/SPURegisterInfo.cpp test/CodeGen/CellSP...

[Chris Lattner]

On Aug 11, 2010, at 4:19 AM, Kalle.Raiskila at nokia.com wrote:

> On Tue, 2010-08-10 at 19:12 +0200, Bob Wilson wrote:
>> On Aug 10, 2010, at 3:53 AM, Kalle.Raiskila at nokia.com wrote:
>>> b) to the best of my understanding, v2i32 has to be legal if it is to be
>>> used in return values and function parameters.
>> 
>> The types of return values and function parameters are also legalized.  If you pass an argument with a non-legal type, it should be promoted or expanded to make it legal.
> 
> The problem here was that v2i32 gets expanded to two i32s instead of
> promoted to one v4i32. Is it possible to force promotion somehow?

This is something that is on my plate to implement, but again, you can handle this in the frontend.

> Ok, I'll have a look. Just removing the v2i32 code from the SPU backend
> is simple. Getting it to be promoted instead of expanded... lets see...

Vector widening should already work (it does on x86 at least) for everything except arguments and return values.

> I wonder if I should have a go at trying to get i32 promoted to v4i32 at
> the same time :)

While interesting, I'd steer away from that.  There are assumptions throughout the code generator that the pointer type is an integer etc.

>> Beyond Bob's post of "it should work", this is something that can
>> easily be handled by the frontend (by having the frontend lower v2i32
>> arg/ret values into v4i32 values with shufflevector to extract). This
>> is how the X86-64 backend handles its often-crazy ABI constraints.
> 
> Certainly, when strict ABI compliance is required the frontend should do
> this lowering. However, not asserting on code that uses LLVM native
> types (including v2i32) would be nice. Especially as v2i32 is so easily
> mapped to v4i32. 

Yes, I agree completely.  This is something I'd also like on X86-64.  It's on my todo list, but sadly my todo list isn't getting a lot of attention lately.

-Chris