Talk:Intel 8087

Copyright Infringement?

The image on this page seems to have been copied from http://www.cpu-collection.de where the image clearly has a copyright notice attached.

Number of Transistors

I want to meaning intel 8087 in malay.but i canot faind in the internet.so to everwone plese healp me to find the meaning intel 8087 in malay. How many transistors in the 8087?

There were 45.000 transistors in a 8087 (you can find it in the article). Intel 8087 (at least) were indeed manufactured in Malaysia, see here. --Dirk P Broer (talk) 23:22, 12 December 2011 (UTC)

IBM

Should add that the 8087 (and later the 80287) received a significant boost when IBM included coprocessor slots on all its early IBM PC motherboards... AnonMoos (talk) 06:22, 18 March 2008 (UTC)

When this was directly copied into the article, it created an inaccuracy, since the IBM PCjr did not include a numeric coprocessor slot (though the PC, PC-XT, and PC-AT did). AnonMoos (talk) 16:58, 13 December 2011 (UTC)

Neither did the IBM PC Convertible (according to the Mueller book), though that didn't cost Intel many 8087 sales. --Wtshymanski (talk) 19:42, 16 December 2011 (UTC)

Bus access and the 8087

The very first text I ever read about the 8087 architecture, explicitly stated that the 8087 did not directly issue the address onto the bus, but relied on the main CPU to interpret the instruction's addressing requirements and then generate the necessary address and read/write signals on the bus. The main CPU then went idle while the FPU locked the bus, read or wrote the data, and then unlocked the bus.

If someone with low-level 8087 experience could clarify, that would be fantastic. 24.166.235.75 (talk) 07:16, 29 October 2010 (UTC)

The 8087 was directly connected to the 12 bit address bus and the 16 bit data bus (8 bit data bus when used with an 8088). It was normally idle and its bus lines were high impedance allowing the 8086/8 to control the buses. When an instruction that started '11011' (an 'escape' instruction) was placed on the data bus, the main processor immediately placed its address and data buses in a high state and sent a discrete signal to the 8087 effectively waking it up. It read the instruction and then executed that instruction placing any required adress and data on the buses. Once that instruction had finished, control was handed back to the main processor which then fetched the next instruction. If that was an escape instruction, the above repeated.

You may be a bit confused (and I grant that it is quite easy) by the fact that the 8087 never placed the address on the bus to fetch its instructions. These were placed by the main processor which then recognised instructions intended for (any) co-processor. However the 8087 did place addresses that were required as part of the execution of those instructions.

Later co-processors (80287 onwards - there was no 80187) did indeed rely on the main processor to handle all bus traffic. 86.176.155.137 (talk) 19:12, 13 January 2011 (UTC)

Intel 80187 links to x87, where there's a section on it, so apparently somebody thinks that it did exist... AnonMoos (talk) 12:01, 15 March 2011 (UTC)

Indeed there was a 80187. However, it appeared some considerable time after the 80186 did (in fact after the 80287), so some early references would claim that there was no 80187. Incidentally, a 80187 would not work with a 80188 (unlike the situation with 8087 and 8088) because the 80188 lacked the logic to do so. 109.156.49.202 (talk) 17:12, 2 October 2011 (UTC)

Page 17-53 of the Intel "Embedded controller handbook" gives as a bullet point feature of the 80188 "High performance numerical processing capability through 8087 interface", and does not list a lack of coprocessor capability as one of the three differences between the 80188 and 80186. --Wtshymanski (talk) 19:28, 2 October 2011 (UTC)

The references that I have suggest that the 80187 did not work with the 80188. However as discussed elsewhere, it was not clear whether the interface was not present or did not work correctly. Hearsay evidence suggests that it was the latter because the 80188 apparently did work with a standard 8087. Hearsay evidence is not, of course, citeable. Incidentally, the 80187 actually had a 80387 core but the 8087 interface. 109.156.49.202 (talk) 16:08, 22 October 2011 (UTC)

That's the problem. Citations for all this would make a worthwile addition to the article. --Wtshymanski (talk) 21:03, 22 October 2011 (UTC)

Info specifically on the 80187 should go on the x87 artcile... AnonMoos (talk) 08:15, 23 October 2011 (UTC)

Reference to 80487

80487 is not a math co-processor as it appears from the Successors paragraph. 80487 is a complete 80486DX CPU with slightly different pin out. — Preceding unsigned comment added by Skiselev (talk • contribs) 22:12, 2 November 2011 (UTC)

I believe I have this covered in here. --Dirk P Broer (talk) 23:17, 12 December 2011 (UTC)

Table of opcodes and instructions

(To find out how many cells this table should have: 3 bits in the first bite of the opcode and 8 bits in the second bite, means 2^11 = 2048 opcodes reserved for the FPU. The table uses 2 cells per opcode in order to pair the opcodes and mnemonics.)

I wonder whether this is the place for a table which has (or should have) 4096 cells. (Even if many are blank.)

In any case, the cells are not arranged properly. One expects to see all the D8xx opcodes together, followed by the D9xx, and so forth. It seems that, because of the width of the table, it was decided to do it in two halves, with yy00 through yy7F in the first half and yy80 through yyFF in the second half, or something like that. But even so, many things are in the wrong columns, which confused me quite a lot.

If you are going to keep the table, it should be repaired. (Also, the wiki software sometimes interprets the indirection brackets as wikilinks.) Note that there is no reference for the data in the table.

It is clear that there is a full set of operations for 32- and 64-bit reals, and perhaps also for 16- and 32-bit signed integers, but the only obvious operations for 64-bit integers and packed BCD numbers are loading and storing. What are all the operations with a P suffix? Are the 64-bit integers and BCDs handled together by the P operations? If so, what about decimal adjust operations for BCDs? This would be interesting information, even without the very fine detail of the table.

Solo Owl 06:38, 2 September 2015 (UTC)

I agree that such a large table seems like overkill. A better approach would be to list all of the instructions (i.e., instead of all of the opcodes), including a short description of each, perhaps also including the addressing modes they use. This would be a shorter list/table and accomplish the same thing. — Loadmaster (talk) 17:02, 2 September 2015 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified 2 external links on Intel 8087. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive https://web.archive.org/web/20110930172701/http://coprocessor.cpu-info.com/index2.php?mainid=Copro&tabid=1&page=1 to http://coprocessor.cpu-info.com/index2.php?mainid=Copro&tabid=1&page=1
• Added archive https://web.archive.org/web/20110930172701/http://coprocessor.cpu-info.com/index2.php?mainid=Copro&tabid=1&page=1 to http://coprocessor.cpu-info.com/index2.php?mainid=Copro&tabid=1&page=1

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 15:00, 14 November 2017 (UTC)

Summary of significance

Maybe there should be a sentence in the first paragraph explaining how the 8087 was the first "serious" numerics widely-available in hardware form for microcomputer users (somebody else might be able to phrase this better than I could). I remember reading in a mid-1980s computer magazine about how a scientist got a personal computer in his lab, and could then run his calculations overnight, instead of having to reserve time on the university mainframe. Then he plugged an 8087 into his PC-XT (I believe it was), and could run the calculations in an hour or two... AnonMoos (talk) 08:55, 5 December 2019 (UTC)

INTEL Floating Point Developement

Prior to the 8087, in '77, While under John Palmer at Intel, I lead the team that developed the first floating point tests for Intel's floating point. Please add the following text where appropriate:

"Earlier, the SBC-310 Math co-processor board was used during the development of the Intel Floating point to implement test and, eventually, release floating point capabilities for the 8080 and 8085."

Reference [1]https://wikimili.com/en/Intel_8087 [2]http://www.bitsavers.org/pdf/intel/ISIS_II/9800481A_Fortran-80_Programming_Manual_Dec79.pdf Gaither (talk) 19:53, 23 June 2024 (UTC)

I'm sure that your product was very handy for those that needed it, but as I understand it, in the form it was released to programmers outside Intel, it was a software library, and it was not aligned with IEEE Standard 754. It doesn't do much to detract from the status of the 8087 as "the first `serious' numerics widely-available in hardware form for microcomputer users" (see talk-page section immediately above) -- all that tens of thousands (maybe hundreds of thousands) of scientific- or engineering-minded IBM PC owners had to do was a plug a chip into an empty slot, and they instantly had hardware-accelerated 64-bit floating point capability, and to a certain degree the mathematical calculations could sometimes run in parallel to main CPU instruction execution. AnonMoos (talk) 23:10, 23 June 2024 (UTC)