Re: [dataflow] partial register handling

4 msg	Mainline build problems
1 msg	Profil system call - Adding profiling support G...
1 msg	GCC 4.2.0 Branch Frozen for Release
1 msg	GCC 4.2.0 Status Report (2007-05-13)
2 msg	Likely code generation bug in GCC 4.0.1
1 msg	[patches] fine grained bounds checking patches ...
2 msg	Missed optimisation
3 msg	Re: pr11135: make PIC register a pseudo
2 msg	Status of the pointer_plus branch
4 msg	GCC 4.2.0 boostrap problems on FreeBSD/ia64
7 msg	genautomata.c bug
1 msg	gcc-4.3-20070511 is now available
42 msg	GCC 4.2.0 Status Report (2007-05-11)
1 msg	The Linux binutils 2.17.50.0.16 is released
9 msg	Supporting 'MAC' instruction on gcc v4.1.1
1 msg	RE: gcc 4.3 on Interix

[dataflow] partial register handling
\ Roman Zippel (10 May 2007)
. \ Roman Zippel (11 May 2007)
. . \ Rask Ingemann Lambertsen (11 May 2007)
. . . \ Roman Zippel (11 May 2007)
. . \ Kenneth Zadeck (11 May 2007)
. . . \ Roman Zippel (11 May 2007)
. . . \ Paolo Bonzini (11 May 2007)
. . . . \ Richard Sandiford (12 May 2007)
. . . . . \ Paolo Bonzini (12 May 2007)
. \ Paolo Bonzini (11 May 2007)
. . \ Roman Zippel (11 May 2007)
. . . \ Paolo Bonzini (25 May 2007)
. . \ Paolo Bonzini (11 May 2007)
. . \ Paolo Bonzini (11 May 2007)
. \ Rask Ingemann Lambertsen (11 May 2007)

1 msg	Re: At penrose
5 msg	Clarification request for ipa/cgraph code
1 msg	Web Site Traffic For Free? - Article Submission...

Subject:	Re: [dataflow] partial register handling
Group:	Gcc
From:	Paolo Bonzini
Date:	25 May 2007

> But I don't think that's enough, with the current loop it would strip the
> subreg of a multiword subreg and there is some logic in df_ref_record,
> which wouldn't see it. An alternative might be:
>
> while (GET_CODE (dst) == STRICT_LOW_PART
> || GET_CODE (dst) == ZERO_EXTRACT)
> {
> flags |= DF_REF_READ_WRITE | DF_REF_PARTIAL;
> loc = &XEXP (dst, 0);
> dst = *loc;
> }
>
> if (df_read_modify_subreg_p (dst))
> flags |= DF_REF_READ_WRITE | DF_REF_PARTIAL;
>
> That would only leave singleword subreg and paradoxical subreg, if they
> should require anything.

This seems also correct, if you want to test it.

> One could also restrict STRICT_LOW_PART to subregs, as after reload the
> subreg part is usually gone.

No, a strict_low_part of a reg is also a partial set.

Paolo