Re: PATCH: ra-rewrite.c comment fixes

[Michael Matz <matz at suse dot de>]

Hi,

On Thu, 14 Aug 2003 tm_gccmail@mail.kloo.net wrote:

> The comments in ra-rewrite.c were difficult to read, so
> I removed a bunch of extraneous commas

I'm german.  I like commas and long sentences ;-)  (I agree it's a good
idea, though)

> and reformatted some of the more prepositionally difficult ones.

I prefer if you would do this on the new-regalloc-branch, otherwise a
merge would have to fix possibly many conflicts.  Anyway some comments:

> !    to check for coalescing.  This tries to coalesce two webs, which were
> !    spilled, are connected by a move, and don't conflict.  Greatly
> --- 79,86 ----
> !    to check for coalescing.  This tries to coalesce two web which were
> !    spilled, are connected by a move, and don't conflict.  This greatly

You removed the plural 's' in "webs".

> *************** static unsigned HOST_WIDE_INT spill_load
> *** 383,389 ****
>   static unsigned HOST_WIDE_INT spill_store_cost;
>   static unsigned HOST_WIDE_INT spill_remat_cost;
>
> ! /* In rewrite_program2() we detect if some def us useless, in the sense,
>      that the pseudo set is not live anymore at that point.  The REF_IDs
>      of such defs are noted here.  */
>   static bitmap useless_defs;
> --- 383,389 ----
>   static unsigned HOST_WIDE_INT spill_store_cost;
>   static unsigned HOST_WIDE_INT spill_remat_cost;
>
> ! /* In rewrite_program2() we detect if some def us useless in the sense

This should anyway read "... is useless ..."

> ! 		   and insn.  Happens, when the an insn uses the same web
> --- 436,442 ----
> ! 		   and insn.  This happens when the an insn uses the same web

"... when an insn ..."

> ! /* Given two rtx' S1 and S2, either being REGs or MEMs (or SUBREGs
> !    thereof), return nonzero, if they overlap.  REGs and MEMs don't
> !    overlap, and if they are MEMs they must have an easy address
> --- 550,558 ----
> ! /* Given two RTXs S1 and S2, either being REGs or MEMs (or SUBREGs
> !    thereof), return nonzero if they overlap.  REGs and MEMs don't
> !    overlap, and if they are MEMs they must have an simple address

"... a simple address ..."

> ! /* This deletes from *LIST all rtx's which overlap with X in the sense
> !    of slots_overlap_p().  */
> --- 603,610 ----
> ! /* This deletes all RTXs (which overlap with X in the sense
> !    of slots_overlap_p() from *LIST.  */

Closing ")" missing.

> ! /* A more sophisticated (and slower) method of adding the stores, than
> !    rewrite_program().  This goes backward the insn stream, adding
> --- 633,640 ----
> ! /* A more sophisticated (and slower) method of adding the stores than
> !    rewrite_program().  This scans backward throughthe insn stream, adding

"... through the ..."

> ! 		 the deadness (it anyway is too late, as we already cleared
> --- 1308,1318 ----
> ! 		 the deadness (it ss too late anyway, as we already cleared

"... it is ..."


Ciao,
Michael.