This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Re: Question about -ftree-* optimizations in -O1
- From: Jeff Law <law at redhat dot com>
- To: Jiaqi Tan <jiaqi dot tan at gmail dot com>
- Cc: gcc-help at gcc dot gnu dot org, Libin Dang <libin dot dang at windriver dot com>
- Date: Mon, 20 Jun 2016 11:21:49 -0600
- Subject: Re: Question about -ftree-* optimizations in -O1
- Authentication-results: sourceware.org; auth=none
- References: <CAPfiRNCBDtLakxJs1WWoSD3QCLR7q-xbqQ+WA+vPWpSB+KSu5Q at mail dot gmail dot com> <57635674 dot 20006 at windriver dot com> <79ddc313-e1bc-49e8-fae1-5846a183df64 at redhat dot com> <CAPfiRNCGcHnm3NuwiY_0qmv-cYH5AXcunPGsx-fjSbcKD2-zAg at mail dot gmail dot com>
On 06/18/2016 06:10 AM, Jiaqi Tan wrote:
Thank you for the replies!
In that case, can I assume that the switches that don't begin with
"-ftree" work on the lower level RTL IR?
Generally, yes.
Also, out of curiousity, are there any optimizations that require the
lower level IR in order to work?
Register allocation, low level jump optimization, addressing mode
optimizations, etc. Essentially anything closely related to generating
code for the target processor is handled at the lower level (RTL) IL.
Things which are largely independent of the target are handled at the
higher level IL (trees/gimple).
Jeff