Talk:Hex - Revision history

Ppaatt Lynagh: archive the talk that preceded our dividing the article in two

2007-10-29T20:36:44Z

archive the talk that preceded our dividing the article in two

Ppaatt Lynagh: /* That Red Mist, Dispelled Maybe */ Hope springs eternal ...

2007-10-29T14:19:54Z

That Red Mist, Dispelled Maybe: Hope springs eternal ...

← Older revision		Revision as of 07:19, 29 October 2007
Line 1:		Line 1:
			= That Red Mist, Dispelled Maybe =

			It's next week now.

			Stay tuned and in just a moment you will see, I have followed thru on the idea of disentangling [[hex]] from [[Efficient Hex]].

			I like this change very much. It now makes sense to me to lead the [[Efficient Hex]] article with the Different, Clever and Fast exemplar, presented just past a less astonishing specification, and then trail with a dismissive mention of the brief & clear & conventional code of the [[hex]] article. Also the [[hex]] article now explains another key piece of our history here: in the [[hex]] article we do actually choose requirements that astonish people who usually write clever, small, or fast code by taking as precedent a specification from the far off world of people who usually write brief, clear, and conventional code. And so on and so on. I like it.

			So here now I am hoping we have figured out how to duck all our disputes, and put an end to me deleting your work and you deleting my work. I did much fear to split the article in two -- ordinarily forking the article is no sign of consensus -- but this fork feels right. We now have two distinct collections of exemplars designed for two communities, only related by the coincidence of similar specifications, not related because either community has much interest in the exemplars of the other community, no need to entangle.

			Like it? Love it? Hate it? Most curiously yours, and thanks again for speaking kindly of me volunteering here. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 07:19, 29 October 2007 (PDT)

	= Solution =		= Solution =
	I just realized how to solve all our problems: Rename the page to Python Hex. You being the python expert could write the article anyway you want, I have no love for Python after all. Since it wouldn't be occupying the Hex page we could put in an article about Hexadecimal and provide a link to your page. Your page could be part of a Python to LSL migration Tutorial. I'll wait for your agreement before carrying this out. -- [[User:Strife Onizuka\|Strife Onizuka]] 18:33, 24 October 2007 (PDT)		I just realized how to solve all our problems: Rename the page to Python Hex. You being the python expert could write the article anyway you want, I have no love for Python after all. Since it wouldn't be occupying the Hex page we could put in an article about Hexadecimal and provide a link to your page. Your page could be part of a Python to LSL migration Tutorial. I'll wait for your agreement before carrying this out. -- [[User:Strife Onizuka\|Strife Onizuka]] 18:33, 24 October 2007 (PDT)

Ppaatt Lynagh: for more reasonable run time measures, how about seven trials and toss the anomalies, rather than taking the min max from three

2007-10-28T14:55:10Z

for more reasonable run time measures, how about seven trials and toss the anomalies, rather than taking the min max from three

← Older revision		Revision as of 07:55, 28 October 2007
Line 34:		Line 34:
	Ah. Carrying that exact same Code Racer script over to another sim reports between 15 and 30 ms rather than between 3 and 10 ms. Looks like llGetTime doesn't reliably cancel out all lag. Might be fun to build a lag meter that compared llGetTime with llGetTimestamp. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:59, 25 October 2007 (PDT)		Ah. Carrying that exact same Code Racer script over to another sim reports between 15 and 30 ms rather than between 3 and 10 ms. Looks like llGetTime doesn't reliably cancel out all lag. Might be fun to build a lag meter that compared llGetTime with llGetTimestamp. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:59, 25 October 2007 (PDT)

	= llGetFreeMemory and llGetTime results - ~~Second~~ Round =		= llGetFreeMemory and llGetTime results =

			== Round Three ==

			So my procedure for using the [[Efficiency Tester]] instrument was to measure three times and report the min to max range. Those results felt unreasonable. I suppose the way the way the results vary wildly, like 5X, between sims is inescapable, but I'm hoping we can find some repeatably measurable relative differences within any one sim -- actual results that match our theories of what code feels fast.

			The results that were unreasonable were the min-max range across three trials. I'm now going to try seven trials rather than three, hoping to throw out the few trials that stand apart from the majority as invalid. Maybe that approach will make the average of valid trials feel reasonable ... we'll see. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 07:55, 28 October 2007 (PDT)

			== Round Two ==

	So I measured another round of results with the llGetTime version of the [[Efficiency Tester]]. Surprising results, so maybe not convincing. In these results, clever & fast doesn't pay off: clever & supposedly fast always actually runs more slowly than clever & small. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]]		So I measured another round of results with the llGetTime version of the [[Efficiency Tester]]. Surprising results, so maybe not convincing. In these results, clever & fast doesn't pay off: clever & supposedly fast always actually runs more slowly than clever & small. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]]
Line 127:		Line 135:
	</pre>		</pre>

	= ~~llGetFreeMemory and llGetTime results~~ =		== Round One ==

	So here's the raw data for a new round of summary results, measured with [[llGetTime]] and [[llGetFreeMemory]] rather than with [[llGetTimestamp]] and llGetFreeMemory -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 11:49, 20 October 2007 (PDT)		So here's the raw data for a new round of summary results, measured with [[llGetTime]] and [[llGetFreeMemory]] rather than with [[llGetTimestamp]] and llGetFreeMemory -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 11:49, 20 October 2007 (PDT)

Ppaatt Lynagh: another round of results with the llGetTime version of the Efficiency Tester -- clever & supposedly fast always actually runs more slowly than clever & small

2007-10-28T05:06:45Z

another round of results with the llGetTime version of the Efficiency Tester -- clever & supposedly fast always actually runs more slowly than clever & small

← Older revision		Revision as of 22:06, 27 October 2007
Line 33:		Line 33:

	Ah. Carrying that exact same Code Racer script over to another sim reports between 15 and 30 ms rather than between 3 and 10 ms. Looks like llGetTime doesn't reliably cancel out all lag. Might be fun to build a lag meter that compared llGetTime with llGetTimestamp. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:59, 25 October 2007 (PDT)		Ah. Carrying that exact same Code Racer script over to another sim reports between 15 and 30 ms rather than between 3 and 10 ms. Looks like llGetTime doesn't reliably cancel out all lag. Might be fun to build a lag meter that compared llGetTime with llGetTimestamp. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:59, 25 October 2007 (PDT)

			= llGetFreeMemory and llGetTime results - Second Round =

			So I measured another round of results with the llGetTime version of the [[Efficiency Tester]]. Surprising results, so maybe not convincing. In these results, clever & fast doesn't pay off: clever & supposedly fast always actually runs more slowly than clever & small. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]]

			In summary I saw:

			31.4+-0.7 ms = concise & conventional

			29.2+-1.2 ms = clever & small

			30.5+-0.5 ms = clever & fast

			28.4+-1.0 ms = different & clever & small

			30.7+-0.4 ms = different & clever & fast

			Some of those summary results overlap, again making them less persuasive. Maybe the difference we're measuring is too small to measure in thirty minutes. In those thirty minutes I get just 3 x 10,000 runs, and then the actual results still overlap.

			On the other hand, maybe small matters more than we know. Could be some characteristic of the sims is rewarding small, so that only small matters, not apparent speed.

			Details as follows:

			<pre>
			// Concise & Conventional

			[15:38] 10/25 Script Ball: 14193 free bytes of code at default.state_entry
			[15:43] 10/25 Script Ball: 32.063110+-??% ms may have elapsed on average in each of
			[15:43] 10/25 Script Ball: 10000 trials of running the code in the loop
			[15:49] 10/25 Script Ball: 31.294397+-??% ms may have elapsed on average in each of
			[15:49] 10/25 Script Ball: 10000 trials of running the code in the loop
			[15:54] 10/25 Script Ball: 30.830774+-??% ms may have elapsed on average in each of
			[15:54] 10/25 Script Ball: 10000 trials of running the code in the loop

			// Clever & Small

			[16:28] 10/25 Script Ball: 14193 free bytes of code at default.state_entry
			[16:33] 10/25 Script Ball: 30.362169+-??% ms may have elapsed on average in each of
			[16:33] 10/25 Script Ball: 10000 trials of running the code in the loop
			[16:38] 10/25 Script Ball: 27.982462+-??% ms may have elapsed on average in each of
			[16:38] 10/25 Script Ball: 10000 trials of running the code in the loop
			[16:43] 10/25 Script Ball: 28.738605+-??% ms may have elapsed on average in each of
			[16:43] 10/25 Script Ball: 10000 trials of running the code in the loop

			// Clever & Fast

			[16:45] 10/25 Script Ball: 14193 free bytes of code at default.state_entry
			[16:51] 10/25 Script Ball: 30.944481+-??% ms may have elapsed on average in each of
			[16:51] 10/25 Script Ball: 10000 trials of running the code in the loop
			[16:56] 10/25 Script Ball: 30.126232+-??% ms may have elapsed on average in each of
			[16:56] 10/25 Script Ball: 10000 trials of running the code in the loop
			[17:01] 10/25 Script Ball: 30.744410+-??% ms may have elapsed on average in each of
			[17:01] 10/25 Script Ball: 10000 trials of running the code in the loop

			// Different & Concise & Small

			[17:06] 10/25 Script Ball: 14193 free bytes of code at default.state_entry
			[17:11] 10/25 Script Ball: 29.401611+-??% ms may have elapsed on average in each of
			[17:11] 10/25 Script Ball: 10000 trials of running the code in the loop
			[17:16] 10/25 Script Ball: 28.408644+-??% ms may have elapsed on average in each of
			[17:16] 10/25 Script Ball: 10000 trials of running the code in the loop
			[17:20] 10/25 Script Ball: 27.419189+-??% ms may have elapsed on average in each of
			[17:20] 10/25 Script Ball: 10000 trials of running the code in the loop

			// Different & Clever & Fast

			[21:33] 10/25 Script Ball: 14193 free bytes of code at default.state_entry
			[21:38] 10/25 Script Ball: 31.089897+-??% ms may have elapsed on average in each of
			[21:38] 10/25 Script Ball: 10000 trials of running the code in the loop
			[21:44] 10/25 Script Ball: 30.330477+-??% ms may have elapsed on average in each of
			[21:44] 10/25 Script Ball: 10000 trials of running the code in the loop
			[21:49] 10/25 Script Ball: 31.118830+-??% ms may have elapsed on average in each of
			[21:49] 10/25 Script Ball: 10000 trials of running the code in the loop

			// Summary results

			$ dc -e '9 k 32.063110 d 30.830774 + 2 / p - p'
			31.446942000
			.616168000
			$ dc -e '9 k 30.362169 d 27.982462 + 2 / p - p'
			29.172315500
			1.189853500
			$ dc -e '9 k 30.944481 d 30.126232 + 2 / p - p'
			30.535356500
			.409124500
			$ dc -e '9 k 27.419189 d 29.401611 + 2 / p - p'
			28.410400000
			-.991211000
			$ dc -e '9 k 31.118830 d 30.330477 + 2 / p - p'
			30.724653500
			.394176500
			$
			</pre>

	= llGetFreeMemory and llGetTime results =		= llGetFreeMemory and llGetTime results =

Ppaatt Lynagh: /* Efficiency Tester vs. Code Racer */ run times measured by llGetTime vary 10X across sims, ouch

2007-10-25T12:59:49Z

Efficiency Tester vs. Code Racer: run times measured by llGetTime vary 10X across sims, ouch

← Older revision		Revision as of 05:59, 25 October 2007
Line 31:		Line 31:

	I am now running all five versions of code vs. an empty block in the new N-way Code Racer, I am again choosing the test case of 0x07fffFFFF. The self-proclaimed inaccurate but conveniently early claim there now is between 3 and 10 ms per run. Looking out across llGetTime vs. llGetTimestamp and across versions of Efficiencey Tester and Code Racer, I'm not sure I should believe this claim. In other places I saw claims more like 30ms per run. Likely I'll return when I have more results to share. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:39, 25 October 2007 (PDT)		I am now running all five versions of code vs. an empty block in the new N-way Code Racer, I am again choosing the test case of 0x07fffFFFF. The self-proclaimed inaccurate but conveniently early claim there now is between 3 and 10 ms per run. Looking out across llGetTime vs. llGetTimestamp and across versions of Efficiencey Tester and Code Racer, I'm not sure I should believe this claim. In other places I saw claims more like 30ms per run. Likely I'll return when I have more results to share. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:39, 25 October 2007 (PDT)

			Ah. Carrying that exact same Code Racer script over to another sim reports between 15 and 30 ms rather than between 3 and 10 ms. Looks like llGetTime doesn't reliably cancel out all lag. Might be fun to build a lag meter that compared llGetTime with llGetTimestamp. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:59, 25 October 2007 (PDT)

	= llGetFreeMemory and llGetTime results =		= llGetFreeMemory and llGetTime results =

Ppaatt Lynagh: /* Efficiency Tester vs. Code Racer */ 3 or 10 or 30 ms per run, that is the question

2007-10-25T12:39:51Z

Efficiency Tester vs. Code Racer: 3 or 10 or 30 ms per run, that is the question

← Older revision		Revision as of 05:39, 25 October 2007
Line 29:		Line 29:

	Now with the change to [[llGetTime]], somehow I don't immediately see consistent results ... That could just be the naturally variability in measuring time on a time-shared system ... or the change introduced a significant bug of some kind ... -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 18:12, 20 October 2007 (PDT)		Now with the change to [[llGetTime]], somehow I don't immediately see consistent results ... That could just be the naturally variability in measuring time on a time-shared system ... or the change introduced a significant bug of some kind ... -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 18:12, 20 October 2007 (PDT)

			I am now running all five versions of code vs. an empty block in the new N-way Code Racer, I am again choosing the test case of 0x07fffFFFF. The self-proclaimed inaccurate but conveniently early claim there now is between 3 and 10 ms per run. Looking out across llGetTime vs. llGetTimestamp and across versions of Efficiencey Tester and Code Racer, I'm not sure I should believe this claim. In other places I saw claims more like 30ms per run. Likely I'll return when I have more results to share. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:39, 25 October 2007 (PDT)

	= llGetFreeMemory and llGetTime results =		= llGetFreeMemory and llGetTime results =

Ppaatt Lynagh: /* Solution */ ack kind words, hope, red mist

2007-10-25T12:10:20Z

Solution: ack kind words, hope, red mist

← Older revision		Revision as of 05:10, 25 October 2007
Line 17:		Line 17:

	:::I do appreciate your contributions, they may drive me a bit bonkers but thats not your problem. Compromise at times seems to be about making everyone unhappy. If I'm a bit tense you can tell me to chill or take a few hours to think things over. I can be tough to get along with and you do extraordinarily well at it. We may not agree but you do a better job of staying calm. -- [[User:Strife Onizuka\|Strife Onizuka]] 20:40, 24 October 2007 (PDT)		:::I do appreciate your contributions, they may drive me a bit bonkers but thats not your problem. Compromise at times seems to be about making everyone unhappy. If I'm a bit tense you can tell me to chill or take a few hours to think things over. I can be tough to get along with and you do extraordinarily well at it. We may not agree but you do a better job of staying calm. -- [[User:Strife Onizuka\|Strife Onizuka]] 20:40, 24 October 2007 (PDT)

			::::Hey lookathat. I think you have found a way to connect us closer on the non-technical issues. I appreciate your kind words, I'm encouraged by your hope that we neither of us have to go elsewhere, I look forward to the red mist lifting away to leave you able to verbalise how we should act next to take care of both newbies & experts together.

			::::If the red mist in fact doesn't lift by next week, then to get us unstuck I might try dividing the content into a page for newbies and a page for experts, maybe with a horizontal separator shown in between. Once we have the content more sharply divided, maybe how to name the pages will become clear. Article names like [[hex]] and [[efficient hex]] would work for me. Article names like [[simple hex]] and [[hex]] would not work so well, for me. I like the simpler name being dedicated to the less expert visitor.

			::::-- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 05:10, 25 October 2007 (PDT)

	= Efficiency Tester vs. Code Racer =		= Efficiency Tester vs. Code Racer =

Strife Onizuka: /* Solution */

2007-10-25T03:40:35Z

Solution

← Older revision		Revision as of 20:40, 24 October 2007
Line 3:		Line 3:

	::Sorry, not a consensus solution. If you just want the page without regard to what a newbie expects, just take it. Throw out my by-newbie for-newbie stuff and say what you want to say. I'll drop out and go work somewhere else. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 18:58, 24 October 2007 (PDT)		::Sorry, not a consensus solution. If you just want the page without regard to what a newbie expects, just take it. Throw out my by-newbie for-newbie stuff and say what you want to say. I'll drop out and go work somewhere else. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 18:58, 24 October 2007 (PDT)

			:::You going elsewhere is not a good solution. -- [[User:Strife Onizuka\|Strife Onizuka]] 20:40, 24 October 2007 (PDT)

	::I imagine my earlier suggestion got lost in the noise, here it is again.		::I imagine my earlier suggestion got lost in the noise, here it is again.

	::To settle the hex article, I now guess it would make sense to divide all our content in two. In the article tab, we could present only the basic specification and demo together with brief & clear & conventional exemplars. Here in the discussion tab, we could present the clever & small & fast exemplars and the raw data that backs our small & fast claims. -- Ppaatt Lynagh 16:25, 24 October 2007 (PDT)		::To settle the hex article, I now guess it would make sense to divide all our content in two. In the article tab, we could present only the basic specification and demo together with brief & clear & conventional exemplars. Here in the discussion tab, we could present the clever & small & fast exemplars and the raw data that backs our small & fast claims. -- Ppaatt Lynagh 16:25, 24 October 2007 (PDT)

			:::It's workable but I don't think it should be on the discussion page, a subpage would work better (the idea has been nagging at the back of my head but then the red mist descends and it runs away). -- [[User:Strife Onizuka\|Strife Onizuka]] 20:40, 24 October 2007 (PDT)

	::I guess I should hold back from trying that structure until Strife comments. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:21, 24 October 2007 (PDT)		::I guess I should hold back from trying that structure until Strife comments. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:21, 24 October 2007 (PDT)

	::I guess I should mention, I appreciate your every technical contribution, including the technical portions of your last burst of replies to the [[Talk:Hex#Third Talk]]. I don't know how I can helpfully answer the non-technical material such as rants, I do appreciate you again trying to connect with me on non-technical issues also, but I don't know how to begin to give us more connection there. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:31, 24 October 2007 (PDT)		::I guess I should mention, I appreciate your every technical contribution, including the technical portions of your last burst of replies to the [[Talk:Hex#Third Talk]]. I don't know how I can helpfully answer the non-technical material such as rants, I do appreciate you again trying to connect with me on non-technical issues also, but I don't know how to begin to give us more connection there. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:31, 24 October 2007 (PDT)

			:::I do appreciate your contributions, they may drive me a bit bonkers but thats not your problem. Compromise at times seems to be about making everyone unhappy. If I'm a bit tense you can tell me to chill or take a few hours to think things over. I can be tough to get along with and you do extraordinarily well at it. We may not agree but you do a better job of staying calm. -- [[User:Strife Onizuka\|Strife Onizuka]] 20:40, 24 October 2007 (PDT)

	= Efficiency Tester vs. Code Racer =		= Efficiency Tester vs. Code Racer =

Ppaatt Lynagh: /* Solution */ ack the Talk:Hex#Third Talk

2007-10-25T02:31:03Z

Solution: ack the Talk:Hex#Third Talk

← Older revision		Revision as of 19:31, 24 October 2007
Line 9:		Line 9:

	::I guess I should hold back from trying that structure until Strife comments. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:21, 24 October 2007 (PDT)		::I guess I should hold back from trying that structure until Strife comments. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:21, 24 October 2007 (PDT)

			::I guess I should mention, I appreciate your every technical contribution, including the technical portions of your last burst of replies to the [[Talk:Hex#Third Talk]]. I don't know how I can helpfully answer the non-technical material such as rants, I do appreciate you again trying to connect with me on non-technical issues also, but I don't know how to begin to give us more connection there. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:31, 24 October 2007 (PDT)

	= Efficiency Tester vs. Code Racer =		= Efficiency Tester vs. Code Racer =

Ppaatt Lynagh: /* Solution */ copy an older alternative in from the rest of the talk

2007-10-25T02:21:43Z

Solution: copy an older alternative in from the rest of the talk

← Older revision		Revision as of 19:21, 24 October 2007
Line 3:		Line 3:

	::Sorry, not a consensus solution. If you just want the page without regard to what a newbie expects, just take it. Throw out my by-newbie for-newbie stuff and say what you want to say. I'll drop out and go work somewhere else. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 18:58, 24 October 2007 (PDT)		::Sorry, not a consensus solution. If you just want the page without regard to what a newbie expects, just take it. Throw out my by-newbie for-newbie stuff and say what you want to say. I'll drop out and go work somewhere else. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 18:58, 24 October 2007 (PDT)

			::I imagine my earlier suggestion got lost in the noise, here it is again.

			::To settle the hex article, I now guess it would make sense to divide all our content in two. In the article tab, we could present only the basic specification and demo together with brief & clear & conventional exemplars. Here in the discussion tab, we could present the clever & small & fast exemplars and the raw data that backs our small & fast claims. -- Ppaatt Lynagh 16:25, 24 October 2007 (PDT)

			::I guess I should hold back from trying that structure until Strife comments. -- [[User:Ppaatt Lynagh\|Ppaatt Lynagh]] 19:21, 24 October 2007 (PDT)

	= Efficiency Tester vs. Code Racer =		= Efficiency Tester vs. Code Racer =