This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: BCT optimization
- To: dje at watson dot ibm dot com (David Edelsohn)
- Subject: Re: BCT optimization
- From: Joern Rennecke <amylaar at cygnus dot co dot uk>
- Date: Thu, 1 Oct 1998 14:05:55 +0100 (BST)
- Cc: amylaar at cygnus dot co dot uk, law at cygnus dot com, m dot hayes at elec dot canterbury dot ac dot nz, michaelh at ongaonga dot chch dot cri dot nz, egcs at cygnus dot com
> Joern> Could you elaborate what you mean with an 'empty RTX' ?
>
> I mean an uninitialized RTX. The pattern would have to initialize
How do you allocate it? The gen_rtx* function also initialize.
> it. The BCT optimization would call gen_decrement_and_branch_init() with
> an RTX that was uninitialized. The pattern would know this and initialize
> that operand[] with gen_rtx_REG() or GEN_INT() whatever cookie it wanted
> to hand back.
gen_rtx_REG and GEN_INT allocate, so that is no way to initialize
an uninitialized RTX.