--- Log opened Fri May 22 00:00:33 2009
00:07 -!- ahuillet [n=ahuillet@181.251.80-79.rev.gaoland.net] has quit [Read error: 110 (Connection timed out)]
00:07 -!- ahuillet_ is now known as ahuillet
00:25 -!- ahuillet_ [n=ahuillet@176.169.66-86.rev.gaoland.net] has joined #jato
00:39 -!- ahuillet- [n=ahuillet@103.152.72-86.rev.gaoland.net] has joined #jato
00:40 -!- ahuillet [n=ahuillet@226.169.66-86.rev.gaoland.net] has quit [Read error: 110 (Connection timed out)]
00:40 -!- ahuillet- is now known as ahuillet
00:43 -!- tgrabiec [n=tomekg@rev-189-107.ramtel.pl] has quit ["Leaving"]
00:53 -!- ahuillet_ [n=ahuillet@176.169.66-86.rev.gaoland.net] has quit [Read error: 110 (Connection timed out)]
09:55 < penberg> vegard: so
09:55 < penberg> vegard: I am making the patch now.
09:58 < penberg> vegard: done and pushed.
10:05 < penberg> vegard: here it is in its full glory: http://github.com/penberg/jato/commit/8e77031496c368071135133da3758cd99bd69a6a
10:31 -!- mode/#jato [+o ahuillet] by ChanServ
10:33 -!- mode/#jato [+o penberg] by ChanServ
12:03 -!- tgrabiec [n=tomekg@rev-189-107.ramtel.pl] has joined #jato
12:20 < tgrabiec> ahuillet: do you have any thoughts about the thing I was discussing with penberg last evening (the INSN_CHECKCAST hack vs. bb splitting inside STMT_CHECKCAST) ?
12:21 <@ahuillet> not right now, cooking with a friend
12:21 <@ahuillet> I'll read the backlog and get back to you when I can :)
12:21 < tgrabiec> ahuillet: ok, have a nice cookin' :)
17:19 <@ahuillet> <penberg_home> if we _don't_ maintain basic block properties
17:19 <@ahuillet> <penberg_home> (that is, no jumps in the middle of them)
17:19 <@ahuillet> <penberg_home> we will run into trouble with optimization passes
17:19 <@ahuillet> obviously this will be a problem if we don't take care
17:20 <@ahuillet> the thing is, I feel that there will be more and more cases where we will want to have some "block local jump" support
17:30 <@penberg> ahuillet: I agree
17:30 <@penberg> but
17:30 <@penberg> I am not at all convinced "local jump" is a safe thing to do.
17:31 <@ahuillet> without optimization passes it's not a problem
17:42 <@ahuillet> but it's in fact a bit too hackish
18:01 < tgrabiec> ahuillet, penberg: how can a conditional jump inside checkcast (no bb splitting) be dangerous with optimisation passes ?
18:08 < tgrabiec> hmm, i'm just reading a paper about array check ellimination in HotSpot and it seems to me, that HotSpot cuts basic blocks with regard to HIR branches (goto, if) and not LIR branches - a branch inside arraychect does not split basic block
18:15 <@ahuillet> tgrabiec : it depends on what you do inside the "local" branches
18:15 <@ahuillet> if you do e.g. computations, then CSE might become buggy
19:20 < tgrabiec> a branch inside checkcast is just a branch to exception throwing code - it acts much like any other Potentially Excepting Instruction. The throwing code itself modifies only eax, so it's not a problem in this particullar case.
20:29 -!- tgrabiec [n=tomekg@rev-189-107.ramtel.pl] has quit [Read error: 104 (Connection reset by peer)]
20:30 -!- tgrabiec [n=tomekg@rev-189-107.ramtel.pl] has joined #jato
20:59 < vegard> heh, fun. nr_locals is -4 :)
22:30 < vegard> miranda methods :-D
22:30 < vegard> never heard that term before
--- Log closed Sat May 23 00:00:33 2009