>>>>>>>>>>>>>>>>>>> PRELIMINARY <<<<<<<<<<<<<<<<<<<<<<< MSC Meeting, Draft 00: 18:00-21:00 PDT, Monday, 2005Jan10 Toll House Hotel, Los Gatos With teleconferencing Secretary: David V James [dvj@alum.mit.edu] Attending: Betty Cheng ???? Bob Davis bob@scsi.com Bob Ferguson bobf@consultantsInCognition.com Martin Freeman martin.freeman@philips.com David Gustavson dbg@scizzl.com Dick Karpinski dick@cfcl.com Steve M. Mills steve_mills@hp.com Leonard Tsai tsai@ieee.org Dan Zuras r754@nonabelian.com I--Dinner: Another fine meal was had by all. III--Introductions We were pleased to have Steven Mills, from the IEEE Standards Board, in attendance. IV--Approval of Agenda Approved, without dissent. V--Approval of Minutes Approved, without dissent. VI--Call for Patents Bob Davis presented the IEEE blessed "call for patents" and read their text to the audience. VII--Officers reports VII.1--Chair: Bob Davis, Report for 2004. Report presented. VII.2--Vice-Chair: Dave Gustavson Nothing to report. VII.3--Treasurer - Bob Ferguson Nothing to report. VII.4--Secretary - Dave James Wished to ensure that "IEEE Templates" would be discussed. IX--Liaison Reports IX.1--Standards Board - Davis Bob Davis presented a report. IX.2--IEEE RAC: Dave James/Dave Gustavson DVJ noted that we have new representatives from 802, so the balloting on new web pages is expected to progress. (Note: the web page content has since been approved.) IX.3--Microprocessor TAG Report: Peter Johansson Peter was not present. Bob Davis noted that he has been unsuccessful in contacting Peter for several months. Concern was expressed. X--Old Business X.1--Working group proposal on MSC OUI for MSC Projects - Leonard Tsai A preliminary writeup was discussed. DVJ requested detailed clarifications. XI--New Business XI.1--MyBallot An automated ballot system is being ramped up by the IEEE Staff. Some concern was expressed that volunteers had not been able to "play" before this became real. It was noted that it remains unclear how non-IEEE members can use MyBallot. There are apparently ways, but they are not clear to the casual web surfer. (Ed note: Further communication with the IEEE failed to resolve this concern). TBDs, from Secretary's notes: Additional cryptographic techniques: William White Structure onthology for macro-molecules in the life sciences: Suzanne Luis (UC Berkeley) XI.2--IEEE Template Requirements The MSC voted to have Bob Davis forward the message that forcing use of the IEEE Templates is overly restrictive and not easily enforceable. XII--Working Group Reports Dan Zuras presented a detailed report on the progress of 754 (this is appended). DVJ expressed a concern that simple tables be included in the draft. Dan noted he could present this input one of the following 754 meetings. (Ed note: DVJ will be at the April 27 meeting, for this purpose). Martin Freeman noted that P1285 was ready and the Sponsor ballot group could be formed (Ed note: The P1285 Sponsor ballot has started since then). XIII--Study Group Reports XIV--For the Good Of The Society XV--Next Meetings April 11, 2005 July 11, 2005 October 10, 2005 (election of Officers) January 9, 2006 XVI--Adjournment Adjourned at approximately 21:00 PT. P754 REPORT FROM DAN ZURAS: We introduced the notion of a storage format using ILM's 1+5+10 format. Decimal32 became one as well. In September, we discussed possible other names for the payload of a NaN: Symptom, Load, Annex, Attribute, Aux Info, NaN load, NaNage, message, Cargo, Payload, NaNt, Burden, Baggage, Link, Import, Freight, Code, Slurry, Diagnostic code, Swag, NaN info, Data, Symbol, Entropy, Noise, Significand, Significance, goods. In our vote: Payload 7, Index & Cargo 5, Significance & Data 4, Message & Attribute 3. So we're staying with Payload for now. And that was it. Annex Z was accepted in September & generalized to decimal in October. Correctly rounded binary <-> decimal conversion had to be clarified in October. We started with my embarrassing discovery of yesterday about decimal to binary conversions. Then draft review continued while I worked on a stronger counter example for directed rounding. n | 2^n ULP 2^n + ULP/2 2^n + ULP --|------------------------------------ 24| 16777216 2 16777217 16777218 25| 33554432 4 33554434 33554436 26| 67108864 8 67108868 67108872 27| 134217728 16 134217736 134217744 28| 268435456 32 268435472 268435488 29| 536870912 64 536870944 536870976 30|1073741824 128 1073741888 1073741952 31|2147483648 256 2147483776 2147483904 32|4294967296 512 4294967552 4294967808 n | 5^10*2^n ULP 5^10*2^n + U/2 5^10*2^n + U -- | -------- --- -------------- ------------ 0 | 9765625 1 9765625.5 9765626 1 | 19531250 2 19531251 19531252 2 | 39062500 4 39062502 39062504 3 | 78125000 8 78125004 78125008 4 | 156250000 16 156250008 156250016 5 | 312500000 32 312500016 312500032 6 | 625000000 64 625000032 625000064 7 | 1250000000 128 1250000064 1250000128 8 | 2500000000 256 2500000128 2500000256 9 | 5000000000 512 5000000256 5000000512 10 | 10000000000 1024 10000000512 10000001024 11 | 20000000000 2048 20000001024 20000002048 12 | 40000000000 4096 40000002048 40000004096 13 | 80000000000 8192 80000004096 80000008192 14 | 160000000000 16384 160000008192 160000016384 15 | 320000000000 32768 320000016384 320000032768 16 | 640000000000 65536 640000032768 640000065536 17 | 1280000000000 131072 1280000065536 1280000131072 18 | 2560000000000 262144 2560000131072 2560000262144 19 | 5120000000000 524288 5120000262144 5120000524288 20 | 10240000000000 1048576 10240000524288 10240001048576 21 | 20480000000000 2097152 20480001048576 20480002097152 22 | 40960000000000 4194304 40960002097152 40960004194304 23 | 81920000000000 8388608 81920004194304 81920008388608 24 |163840000000000 16777216 163840008388608 163840016777216 25 |327680000000000 33554432 327680016777216 327680033554432 26 |655360000000000 67108864 655360033554432 655360067108864 This means that there is a counter example to convert from decimal strings to single precision with directed rounding. In particular, for round-to-zero, the string "1073741825" will convert to 2^30 - 64 if the last digit is ignored because 1073741820 < 2^30 but we have 2^30 < 1073741825 = 2^30 + 1 < 2^30 + 128. Therefore, ignoring the tenth digit means that the string is converted to a binary number that is not even one of the two surrounding numbers. Annex D went into the standard in November. Changed basic formats in December. My integer work was shot down. We are now trying to simplify the notions of underflow (clarify IFF & flush to zero or abrupt underflow), contractions (FMA, shorten_add, add3, etc.).