This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Clean-up STORE_FLAG_VALUE usage.
- From: Zack Weinberg <zack at codesourcery dot com>
- To: Roger Sayle <roger at eyesopen dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sat, 14 Jun 2003 14:37:41 -0700
- Subject: Re: [PATCH] Clean-up STORE_FLAG_VALUE usage.
- References: <Pine.LNX.4.44.0306141437320.30777-100000@www.eyesopen.com>
Roger Sayle <roger@eyesopen.com> writes:
> * rtl.h (STORE_FLAG_VALUE): Remove default definition from here.
> * defaults.h (STORE_FLAG_VALUE): Move default definition to here.
> * doc/tm.texi (STORE_FLAG_VALUE): Document the default value.
>
> * config/alpha/alpha.h (STORE_FLAG_VALUE): Remove definition.
> * config/arc/arc.h (STORE_FLAG_VALUE): Likewise.
> * config/arm/arm.h (STORE_FLAG_VALUE): Likewise.
> * config/cris/cris.h (STORE_FLAG_VALUE): Likewise.
> * config/i370/i370.h (STORE_FLAG_VALUE): Likewise.
> * config/i386/i386.h (STORE_FLAG_VALUE): Likewise.
> * config/i960/i960.h (STORE_FLAG_VALUE): Likewise.
> * config/ia64/ia64.h (STORE_FLAG_VALUE): Likewise.
> * config/ip2k/ip2k.h (STORE_FLAG_VALUE): Likewise.
> * config/m32r/m32r.h (STORE_FLAG_VALUE): Likewise.
> * config/mcore/mcore.h (STORE_FLAG_VALUE): Likewise.
> * config/mips/mips.h (STORE_FLAG_VALUE): Likewise.
> * config/mmix/mmix.h (STORE_FLAG_VALUE): Likewise.
> * config/ns32k/ns32k.h (STORE_FLAG_VALUE): Likewise.
> * config/pa/pa.h (STORE_FLAG_VALUE): Likewise.
> * config/pdp11/pdp11.h (STORE_FLAG_VALUE): Likewise.
> * config/sh/sh.h (STORE_FLAG_VALUE): Likewise.
> * config/sparc/sparc.h (STORE_FLAG_VALUE): Likewise.
> * config/v850/v850.h (STORE_FLAG_VALUE): Likewise.
> * config/xtensa/xtensa.h (STORE_FLAG_VALUE): Likewise.
This is OK with minor changes: In the following three files, please
leave a comment behind. Suggested wording:
> ===================================================================
> *** i370/i370.h 7 Jun 2003 17:11:44 -0000 1.62
> --- i370/i370.h 8 Jun 2003 23:23:01 -0000
> *************** enum reg_class
> *** 877,887 ****
>
> #define TRULY_NOOP_TRUNCATION(OUTPREC, INPREC) (OUTPREC != 16)
>
> - /* We assume that the store-condition-codes instructions store 0 for false
> - and some other value for true. This is the value stored for true. */
> -
> - /* #define STORE_FLAG_VALUE (-1) */
> -
/* ??? Investigate defining STORE_FLAG_VALUE to (-1). */
> ===================================================================
> *** mmix/mmix.h 7 Jun 2003 17:11:47 -0000 1.53
> --- mmix/mmix.h 8 Jun 2003 23:23:48 -0000
> *************** typedef struct { int regs; int lib; } CU
> *** 1144,1155 ****
>
> #define TRULY_NOOP_TRUNCATION(OUTPREC, INPREC) 1
>
> - /* We have a choice here too. */
> - #if 0
> - /* FIXME: Revisit, we don't have scc expanders yet. */
> - #define STORE_FLAG_VALUE 1
> - #endif
> -
/* ??? MMIX allows a choice of STORE_FLAG_VALUE. Revisit later,
we don't have scc expanders yet. */
> ===================================================================
> *** xtensa/xtensa.h 4 Jun 2003 15:23:42 -0000 1.36
> --- xtensa/xtensa.h 8 Jun 2003 23:24:44 -0000
> *************** typedef struct xtensa_args {
> *** 1279,1292 ****
> /* Prefer word-sized loads. */
> #define SLOW_BYTE_ACCESS 1
>
> - /* Xtensa doesn't have any instructions that set integer values based on the
> - results of comparisons, but the simplification code in the combiner also
> - uses this macro. The value should be either 1 or -1 to enable some
> - optimizations in the combiner; I'm not sure which is better for us.
> - Since we've been using 1 for a while, it should probably stay that way for
> - compatibility. */
> - #define STORE_FLAG_VALUE 1
> -
/* ??? Xtensa doesn't have any instructions that set integer values
based on the results of comparisons, but the simplification code in
the combiner also uses STORE_FLAG_VALUE. The default value (1) is
fine for us, but (-1) might be better. */
zw