freenode/#clasp - IRC Chatlog

16:35:51 beach If so, why do we need multiple values? I guess that depends on what kind of code you want to run on the GPU.

16:40:31 beach I should be quiet. I had a very long day and a very long week, so I am too tired to think clearly. Plus, dinner is imminent.

16:46:51 Bike i cannot understand how precalc value reference is supposed to work. it gets an immediate input, but the "immediate" is one of these "literal" objects somehow even though that's not what translate-datum returns...

16:54:05 Bike guess i'll figure it out

16:57:15 Bike i guess i'll undo all my changes, build again, put in more print statements

17:00:43 drmeister I’m at nvidia a bit early. I’ll be able to help you in a few hours.

17:00:57 Bike sure

17:08:58 drmeister I'm sitting idle here - looking at the precalc value reference stuff.

17:10:59 drmeister I think the input to precalc-value-instruction is outside of the scope of what Cleavir does. Think of a precalc-value-instruction and it's input as one thing - it's a literal value that either comes from the literal vector or it's an immediate.

17:11:49 drmeister So translate-datum doesn't have anything to do with precalc-value-instruction and the input to precalc-value-instruction.

17:12:03 drmeister Does that make any sense?

17:13:39 Bike but the translate-simple-instruction method uses the input.

17:14:16 drmeister Uh huh - right. What problems does that cause?

17:15:25 Bike the input is the one passed into translate-simple-instruction, which is the result of translate-datum

17:15:44 Bike that is, layout-basic-block calls translate-datum to get it

17:15:52 Bike translate-datum returns a constant-int-64 or whatever, the llvm thing

17:16:12 Bike so the input in the translate-simple-instruction method should result in it hitting the path with the warning

17:16:32 drmeister Which warning?

17:17:39 Bike https://github.com/clasp-developers/clasp/blob/dev/src/lisp/kernel/cleavir/translate-instruction.lisp#L74-L79

17:20:11 drmeister That translate-simple-instruction method USED to get a i64, which was a simple index into the literal vector - but with immediates I needed to get a little fancier - I didn't want to put immediates into the literal vector - that would be silly.

17:21:07 Bike but i don't see how the method can get anything but an i64.

17:22:31 drmeister And I don't see how the clasp-cleavir:immediate-literal (an immediate) and clasp-cleavir:arrayed-literal (an index) pass through translate-datum to be inputs to the translate-simple-instruction.

17:22:59 Bike the "inputs" argument to the method is passed from layout-basic-block

17:23:13 Bike https://github.com/clasp-developers/clasp/blob/dev/src/lisp/kernel/cleavir/translate.lisp#L155-L169

17:23:25 drmeister Can you do some tracing and printing of inputs to see how immediate-literal and arrayed-literal find their way in there?

17:23:42 Bike that's what i'm about to do now that it finished rebuilding

17:23:46 drmeister This stuff works - it's a question of how.

17:24:22 Bike ok well, put in a print statement and it's an arrayed-literal most of the time

17:24:29 Bike so we're starting off with the continuation of nonsense

17:24:47 drmeister "continuation of nonsense"?

17:25:01 Bike nevermind

17:25:34 Bike oh

17:25:36 Bike i see it

17:25:42 Bike ensure-jit-constant-i64 is hacked

17:25:50 Bike if it doesn't receive a fixnum as its argument, it just returns what it's passed

17:25:52 Bike that really sucks

17:26:00 drmeister There you go.

17:26:21 Bike it's hacked all the way back in cmp to support this.

17:26:50 drmeister And I'm sure you can do better.

17:27:30 Bike so first i'll fix up this thing so that it doesn't force data to be in memory, and then i can look at that

17:27:35 Bike now that i have some handle on it, i guess

17:29:44 drmeister What will you do with it?

17:30:03 Bike with literals?

17:30:09 drmeister Yeah.

17:31:12 Bike probably: eliminate the immediate-literal class entirely; move any information in the arrayed-literal class into the precalc-value-instruction itself; rewrite hoisting to only generate a precalc-value-instruction for non-immediate constants

17:32:46 Bike apparently ensure-jit-constant-i64 is called with non-fixnums elsewhere as well....

17:33:08 drmeister Ok - that sounds right. I've had a nagging feeling that the input for precalc-value-instruction should be pulled into the instruction itself. Then it will just output a value with no inputs.

17:33:19 Bike it can get the index as an immediate fixnum

17:33:30 Bike no problem with that

17:33:36 drmeister A tagged immediate fixnum?

17:33:47 Bike should be untagged

17:33:53 Bike we could also pull it into the instruction if we want

17:34:54 Bike immediate inputs are mostly nice if we could use a lexical-location or whatnot in the same place

17:34:58 Bike but that's not how precalc works

17:35:23 Bike beach made a similar change a few days ago where the offset argument of memref/set2 was moved from an immediate input into the instruction

17:36:20 Bike either way i'd like to keep the intent of immediates in cleavir, which is that they only have small integers in them

17:36:34 Bike which may be interpreted as tagged fixnums, or characters, or whatever.

17:42:26 drmeister Sounds good.

17:43:40 rudkx drmeister: enjoyed your talk at the LLVM dev meeting, thanks so much for that

19:00:15 Bike my translate-instruction changes results in obviously wrong results but i'm not sure why

19:00:27 Bike looks like return-instruction or F->M or something should usually generate an insertvalue, but it's not

19:00:45 Bike but nothing in translate-instruction actually seems to generate insertvalue, i figure that's cmp level...

19:04:45 Bike or maybe i'm just looking at different optimization results.

19:06:49 Bike it... looks kind of like it's storing things into e.g. G110027, then reading them out of G1100277... what the heck...

19:07:10 Bike oh. wait. there it is. stupid problem