GNU Tools Cauldron, 9-11 September 2016, Hebden Bridge UK
Hi all, The GNU Tools Cauldron is barely a month away. This is reminder to sign up if you wish to attend (which is free as always). Details of the meeting and how to register here: https://gcc.gnu.org/wiki/cauldron2016 There is still room for more talks and BoF sessions, so please send in your abstracts ASAP. This year, the GNU Tools Cauldron immediately follows the first ever LLVM Cauldron. An invent inspired by the GNU Tools Cauldron, free to attend and open to anyone to speak. Thanks to our generous sponsors, we have a joint gala reception between the two conferences on Thursday evening (which is thus the closing reception for the LLVM Cauldron and the welcome reception for the GNU Tools Cauldron). This is an excellent opportunity for both communities to mix informally. This is in addition to the GNU Tools Cauldron dinner which will take place as usual on the Saturday evening. Both Cauldrons are part of the Wuthering Bytes festival, a 10-day long celebration of free and open source technology at the heart of the UK's "Northern Powerhouse", running from 2 September. Details at http://wutheringbytes.com/. I have sent this announcement to the main GCC, GDB, binutils, CGEN, DejaGnu, newlib and glibc mailing lists. Please feel free to share with any other groups that are appropriate. I look forward to seeing you in Hebden Bridge next month. Jeremy -- Tel: +44 (1590) 610184 Cell:+44 (7970) 676050 SkypeID: jeremybennett Twitter: @jeremypbennett Email: jeremy.benn...@embecosm.com Web: www.embecosm.com PGP key: 1024D/BEF58172FB4754E1 2009-03-20 signature.asc Description: OpenPGP digital signature
[avr] fno-caller-saves and regression tests
Hi Johann, Turning off -fcaller-saves for AVR makes the testcase I had for PR 71873 pass unless I explicitly add -fcaller-saves to force the compiler to generate the triggering insn patterns. Wonder if we should modify the existing test cases in gcc.target/avr to be tested both ways (with and without caller saves)? At least the register allocation related ones probably won't catch regressions. Regards Senthil
Re: Help with lra
On 03-Aug-16 12:10 AM, Vladimir Makarov wrote: On 08/02/2016 04:41 PM, shmeel gutl wrote: I am trying to enable lra for a propriety backend. I ran into one problem that I can't solve. In lra-constraints.c:split_reg lra_create_new_reg can be called with a hard code rclass of NO_REGS. It then queues a move instruction of the type set TYPE:new_reg TYPE:old_reg But the NO_REGS rclass stops new_reg from matching a register constraint and forces a reload. But the reload will have the same problem. This recurses until the recursion limit is hit. What is my backend missing that will allow a register assignment to new_reg? NO_REGS in this case means memory and the generated RTL move insn finally should be a target load or store insn. It is hard to say w/o looking at the code but, probably, your move insn descriptions do not have memory constraints (or these constraints are quite specific). Currently our memory constraints only match memory operands. I assume that you are suggesting that pseudo registers should match memory constraints. Is this true only for lra, or, would reload also benefit from such a change? Would other passes gain by such a change? Is any extra support needed in patterns or hooks? Thanks, Shmeel
