[llvm-dev] [RFC] Aggreate load/store, proposed plan

[David Chisnall via llvm-dev]

On 21 Aug 2015, at 23:27, Hal Finkel via llvm-dev <llvm-dev at lists.llvm.org> wrote:
> 
> I agree; I'd like them for float/double as well.

In general, I’d like them for aggregates as well.  We currently lower early in clang, which can miss later lowering opportunities (e.g. lowering to a memcpy in a transaction in systems with STM, if the sizes are small, but bigger than a register, or lowering to atomic vector operations where available).

David