Re: [Qemu-devel] [Qemu-ppc] [V2 PATCH 00/37] target-ppc: Decimal Floating Point

[Alexander Graf <agraf@suse.de>]

On 04/21/2014 10:54 PM, Tom Musta wrote:
> This patch series adds emulation of the PowerPC Decimal Floating Point (DFP) instructions.
> The complete set of DFP instructions defined by the Power ISA is introduced.
>
> The foundation of the emulation code is libdecnumber, a software library that models DFP
> numbers and operations in a manner similar to how softfloat models IEEE Binary Floating
> Point.  A subset of the libdecnumber source (from GCC) is dropped into the QEMU source tree
> and modified slightly; only libdecnumber files containing code that is required to implement
> the PowerPC DFP instructions are copied.
>
> The DFP instructions are implemented via helpers.  A typical helper does the following:
>
>    - The contents of the source registers (PPC FPRs) are converted to libdecnumber's
>      internal representation, the decNumber type.  Inputs are either 64 bit (Long)
>      or 128 bit (Extended or Quad) densely packed decimal (DPD), depending on the specific
>      instruction being emulated.
>    - A libdecnumber operation is invoked, producing a decNumber result as well as status
>      flags.
>    - A chain of post-processors is executed to convert status flags and/or input and output
>      data to PPC status (usually FPSCR bits).
>    - The decNumber result is converted back to DPD format and written into the target
>      FPR(s).
>
> V2:
>    - modified post-processor handling per Richard Henderson's comments; eliminated the use
>      of lists of function pointers, replacing with aggregated, static functions.
>    - cleaned up int64 to decNumber converter.

Thanks, applied to ppc-next.

Since we can't apply any patches from the gcc libdecnumber anyway, maybe 
it makes sense to reformat all libdecformat files to match the QEMU 
coding style. But that can be a follow-up patch.


Alex