According to entries in the Apple support area (and some personal experience): Old (iMovie2) projects are trouble. "Reverse clip" is trouble. 800x600 resolution is simply not possible - real trouble for 800x600 LCD displays.
On the bright side there is probably a non-crashing code path in there somewhere.
Sadly, OpenBoot is no longer an IEEE standard. IEEE 1275 lapsed when no one took the initiative to update it on the required basis. I think you can still buy the manuals; IBM, Apple and Sun still use it; and the other adjectives (flexible, etc.) still apply. But not IEEE standard.
It should be mentioned that, even leaving the effect on IP aside, the three instructions
MOV AX,#0
XOR AX,AX
SUB AX,AX are not equivalent. MOV leaves the AF bit alone, SUB sets it to a defined value (0, I think in this case), and XOR leaves it undefined. Or don't assembly hackers notice condition bits these days?
People will use the tools which are convenient. If the debugger is most convenient, code gets fixed via the debugger. Unfortunately, those fixes are expensive. (The cost to fix a bug goes up an order of magnitude with each development step.) So spending money developing good debuggers turns out to be counter-productive!
Better to put the money into front-end tools (HLLs, lint, etc.) so you will have fewer items to debug.
Re:What about "chat" and "talk"?
on
AOL Patents IM
·
· Score: 2
IBMers could query and talk with users on networked VM machines by the early 80's. I don't know whether the code (the "SMSG node CMD othernode CPQ NAMES" and "TELL" commands) was by IBM or by users. Back on the Open Source days many users wrote their own mods to VM. Users did not have to be on the same machine. I don't mean to imply that the VM commands were first, but they were the first that I came across.
San Francisco is well-known for tolerance of illegal parking. About 25 years ago the SF police wanted higher wages but were forbidden to strike. One day the police protested by ticketing all illegally parked cars, especially those around City Hall and the local newspaper offices. That night the Chief of Police appeared on local TV to apologize for the unprofessional conduct of his men!
By comparison, there were - and are - Los Angeles suburbs in which police will ticket a car parked facing the wrong direction. The police infer that something illegal must have happened to place the car in that position.
This difference broke the Patty Hearst case. While the Symbionese Liberation Army held Patty Hearst in San Francisco, police and FBI had no luck finding her. Then the SLA moved to Los Angeles - unfamiliar territory for them. They parked a van in a red zone, which would have been no big deal in SF. Police arrived to ticket the van, determined that it was stolen, and caught or killed a large part of the SLA.
Every so often a memo comes out reminding us that we must have the latest Norton Anti-Virus. NAV is not supported on Linux, so I have to power on the Windows box to update my virus protection. Except for that it stays off.
Pluses: the database takes care of synchronization, and nearly takes care of backup/recover. Very nice. Some DB rigor may rub off on your designers.
Minuses: the DBM is large (in MB, in install/config requirements, and in CPU usage) and your customer may not be running the DBM brand/version which you have tested your app with. Supporting multiple DB vendors is a pain. SQL is sort of standard, but the table definitions tend to vary. Ick.
Actually it is 64K and still 64K in 64bit because you can load from unaligned locations
But the TOC entries are addresses, which are 4 or 8 bytes long. Unless you manage to find addresses which can overlap due to common bit patterns, you have to divide the 64K maximum TOC size (the offset of an lwz instriction is 16 bits long) by 4 or 8.
As for alignment, section 6.4.6.1 here indicates that one response to unaligned integers is an alignment exception. So you can load unaligned, and the PowerPC may fix things up in hardware, but if it doesn't your O/S is expected to simulate the instruction.
Note that Article VI of the Constitution specifies these two components of "the supreme law of the land":
1. the Consitution
1a. laws made pursuant to the Constitution
2. treaties made under the authority of the United States
A treaty, in turn, requires only President approval plus consent of 2/3 of the Senate.
If the President and 2/3 of the Senate were to agree to an international treaty outlawing marijuana, the Feds might have a good case for Federal drug laws regardless of Article I, Section 8.
(The parent needs to be modded up. He may be an AC, but his information is accurate.)
One problem with TOC is that is you are limited to 16K external addresses. Offset "foo" in the TOC example is 16 bits, and the low two are zero. With 64-bit addressing I suppose that drops to 8K externals.
Another characteristic calling a separately compiled function requires that you load a different TOC. The PC-relative scheme requires that you load one value: the PC; the TOC scheme requires that you load two new values: the new PC and the new TOC.
On the plus side, TOC makes shared libraries easier to manage because external addresses are bound to a non-shared data area.
> The inline keyword is not supported by ANSI C - it is a C++ feature.
Inline is in C99, see section 6.7.4 of the standard.
That said, inlining can reduce code size if the function body object code is small relative to the parameter setup object code. Absent an inline feature in the compiler (what compiler could that be?), I think macros would be less dangerous than any specialized "inline preprocessor", even if there should be one.
Disraeli once said "Nations have no permanent friends and no permanent enemies. Only permanent interests." HP and Dell are nation size wrt revenue. I wonder what will happen when the MS Works bid comes in 5 cents less per copy, rather than 5 cents more.
the BIOS code on such things as PCI cards, needs to be in PPC opcode format
Apple adapter card code is usually in Open Firmware format (FORTH), not PPC native. Whether FORTH is easier to read than PPC opcodes is a separate matter.:-)
"Listen4ever has clearly located itself in China to avoid the ambit of United States copyright law," the suit said.
Interesting. The same device that the U.S. is using to hold prisoners indefinitely at Guantanamo Bay. Too bad we (America) didn't think to patent that practice. Though if we did, RMS would probably object...
Correct in concept, wrong on some details. RS/6000 and Mac were both big-endian.
POWER chips were big-endian only. POWER had a few instructions with a big endian bias, i.e. the used computed values as big-endian indices into registers.
PowerPC removed the biased instructions (and made other changes too) resulting in an "unbiased" architecture.
I don't know what caused PowerPC to become endian-agnostic. Maybe it was a desire for elegance. But it was not a desire to accomodate both MacOS and AIX 3 heritage; they were both big-endian.
When I was a kid (about 1960) candy stores sold waxy figurines filled with juice. You would bite off the head and drink the juice. The sidewalks had black blobs and there wasn't any mystery.
>it most certainly was not a trivial task to:- > [...] Make sure that EVERY y2k bug was identified
Yup. I spent 1998 writing C++ code to plow through COBOL programs looking for dates. A favorite COBOL trick was to reuse the same field for months, years and non-dates (memory was scarce). We called those "modal dates."
> Why were software houses *still* producing non-compliant code in 1995?
It didn't help that there were "Introduction to COBOL" books for sale in 1998 that used 2-digit dates in examples.
As to why things didn't collapse, there certainly was a lot of work involved. A big question was whether errors would propagate - like falling dominoes - or be caught and ignored.
We processed code from a large bank. One program was supposed to produce quarterly reports. Due to a logic error, the reports for the first 3 quarters were all zeroes, and the fourth report actually contained entire year totals. A major bug, but it affected nothing. Whoever was supposed to read the report probably just shrugged and moved on.
That's how it was, mostly. Things went wrong. People looked, shrugged and moved on. Sometime after 1/1/00 I brought up the web site for a "partner" of our parent company. There, on the home page, was today's date: September 2, 101. Shrug and move on.
Other Wired article errors
on
AMD's 64-Bit Chip
·
· Score: 5, Informative
The Wired article has other errors as well. A 32-bit CPU isn't limited to 4GB; that confuses address space with physical memory. The definition of exabyte is wrong (1000 petabytes, not 1000 terabytes). The 8080 in 1981? Closer to 1975. And many have mentioned the bogus "no compatibility" claim.
Slashdot Mirror
Just app crashes, AFAIK. Mine are the result of passing 0x0000004a as the argument to strlen. Sure wish Apple hadn't stripped the executable.
According to entries in the Apple support area (and some personal experience): Old (iMovie2) projects are trouble. "Reverse clip" is trouble. 800x600 resolution is simply not possible - real trouble for 800x600 LCD displays.
On the bright side there is probably a non-crashing code path in there somewhere.
Sadly, OpenBoot is no longer an IEEE standard. IEEE 1275 lapsed when no one took the initiative to update it on the required basis. I think you can still buy the manuals; IBM, Apple and Sun still use it; and the other adjectives (flexible, etc.) still apply. But not IEEE standard.
It should be mentioned that, even leaving the effect on IP aside, the three instructions
MOV AX,#0
XOR AX,AX
SUB AX,AX
are not equivalent. MOV leaves the AF bit alone, SUB sets it to a defined value (0, I think in this case), and XOR leaves it undefined. Or don't assembly hackers notice condition bits these days?
People will use the tools which are convenient. If the debugger is most convenient, code gets fixed via the debugger. Unfortunately, those fixes are expensive. (The cost to fix a bug goes up an order of magnitude with each development step.) So spending money developing good debuggers turns out to be counter-productive!
Better to put the money into front-end tools (HLLs, lint, etc.) so you will have fewer items to debug.
IBMers could query and talk with users on networked VM machines by the early 80's. I don't know whether the code (the "SMSG node CMD othernode CPQ NAMES" and "TELL" commands) was by IBM or by users. Back on the Open Source days many users wrote their own mods to VM. Users did not have to be on the same machine. I don't mean to imply that the VM commands were first, but they were the first that I came across.
San Francisco is well-known for tolerance of illegal parking. About 25 years ago the SF police wanted higher wages but were forbidden to strike. One day the police protested by ticketing all illegally parked cars, especially those around City Hall and the local newspaper offices. That night the Chief of Police appeared on local TV to apologize for the unprofessional conduct of his men!
By comparison, there were - and are - Los Angeles suburbs in which police will ticket a car parked facing the wrong direction. The police infer that something illegal must have happened to place the car in that position.
This difference broke the Patty Hearst case. While the Symbionese Liberation Army held Patty Hearst in San Francisco, police and FBI had no luck finding her. Then the SLA moved to Los Angeles - unfamiliar territory for them. They parked a van in a red zone, which would have been no big deal in SF. Police arrived to ticket the van, determined that it was stolen, and caught or killed a large part of the SLA.
I keep a bunch of "classic" bookmarks around. Some are undisputed gems, others are, well, to my taste. Bytes being cheap here's a batch.
Eric Weisstein's World of Mathematics
Every so often a memo comes out reminding us that we must have the latest Norton Anti-Virus. NAV is not supported on Linux, so I have to power on the Windows box to update my virus protection. Except for that it stays off.
Pluses: the database takes care of synchronization, and nearly takes care of backup/recover. Very nice. Some DB rigor may rub off on your designers.
Minuses: the DBM is large (in MB, in install/config requirements, and in CPU usage) and your customer may not be running the DBM brand/version which you have tested your app with. Supporting multiple DB vendors is a pain. SQL is sort of standard, but the table definitions tend to vary. Ick.
But the TOC entries are addresses, which are 4 or 8 bytes long. Unless you manage to find addresses which can overlap due to common bit patterns, you have to divide the 64K maximum TOC size (the offset of an lwz instriction is 16 bits long) by 4 or 8.
As for alignment, section 6.4.6.1 here indicates that one response to unaligned integers is an alignment exception. So you can load unaligned, and the PowerPC may fix things up in hardware, but if it doesn't your O/S is expected to simulate the instruction.Note that Article VI of the Constitution specifies these two components of "the supreme law of the land":
1. the Consitution
1a. laws made pursuant to the Constitution
2. treaties made under the authority of the United States
A treaty, in turn, requires only President approval plus consent of 2/3 of the Senate.
If the President and 2/3 of the Senate were to agree to an international treaty outlawing marijuana, the Feds might have a good case for Federal drug laws regardless of Article I, Section 8.
IANAL.
(The parent needs to be modded up. He may be an AC, but his information is accurate.)
One problem with TOC is that is you are limited to 16K external addresses. Offset "foo" in the TOC example is 16 bits, and the low two are zero. With 64-bit addressing I suppose that drops to 8K externals.
Another characteristic calling a separately compiled function requires that you load a different TOC. The PC-relative scheme requires that you load one value: the PC; the TOC scheme requires that you load two new values: the new PC and the new TOC.
On the plus side, TOC makes shared libraries easier to manage because external addresses are bound to a non-shared data area.
> The inline keyword is not supported by ANSI C - it is a C++ feature.
Inline is in C99, see section 6.7.4 of the standard.
That said, inlining can reduce code size if the function body object code is small relative to the parameter setup object code. Absent an inline feature in the compiler (what compiler could that be?), I think macros would be less dangerous than any specialized "inline preprocessor", even if there should be one.
Disraeli once said "Nations have no permanent friends and no permanent enemies. Only permanent interests." HP and Dell are nation size wrt revenue. I wonder what will happen when the MS Works bid comes in 5 cents less per copy, rather than 5 cents more.
the BIOS code on such things as PCI cards, needs to be in PPC opcode format
:-)
Apple adapter card code is usually in Open Firmware format (FORTH), not PPC native. Whether FORTH is easier to read than PPC opcodes is a separate matter.
Interesting. The same device that the U.S. is using to hold prisoners indefinitely at Guantanamo Bay. Too bad we (America) didn't think to patent that practice. Though if we did, RMS would probably object
> Windows NT for PowerPC? AFAIK it ran in little-endian mode.
Of course! How could I forget? That and OS/2 for PowerPC.
Correct in concept, wrong on some details. RS/6000 and Mac were both big-endian.
POWER chips were big-endian only. POWER had a few instructions with a big endian bias, i.e. the used computed values as big-endian indices into registers.
PowerPC removed the biased instructions (and made other changes too) resulting in an "unbiased" architecture.
I don't know what caused PowerPC to become endian-agnostic. Maybe it was a desire for elegance. But it was not a desire to accomodate both MacOS and AIX 3 heritage; they were both big-endian.
- Clem (POWER/AIX user since about 1992)
When I was a kid (about 1960) candy stores sold waxy figurines filled with juice. You would bite off the head and drink the juice. The sidewalks had black blobs and there wasn't any mystery.
>it most certainly was not a trivial task to:-
> [...] Make sure that EVERY y2k bug was identified
Yup. I spent 1998 writing C++ code to plow through COBOL programs looking for dates. A favorite COBOL trick was to reuse the same field for months, years and non-dates (memory was scarce). We called those "modal dates."
> Why were software houses *still* producing non-compliant code in 1995?
It didn't help that there were "Introduction to COBOL" books for sale in 1998 that used 2-digit dates in examples.
As to why things didn't collapse, there certainly was a lot of work involved. A big question was whether errors would propagate - like falling dominoes - or be caught and ignored.
We processed code from a large bank. One program was supposed to produce quarterly reports. Due to a logic error, the reports for the first 3 quarters were all zeroes, and the fourth report actually contained entire year totals. A major bug, but it affected nothing. Whoever was supposed to read the report probably just shrugged and moved on.
That's how it was, mostly. Things went wrong. People looked, shrugged and moved on. Sometime after 1/1/00 I brought up the web site for a "partner" of our parent company. There, on the home page, was today's date: September 2, 101. Shrug and move on.
The Wired article has other errors as well. A 32-bit CPU isn't limited to 4GB; that confuses address space with physical memory. The definition of exabyte is wrong (1000 petabytes, not 1000 terabytes). The 8080 in 1981? Closer to 1975. And many have mentioned the bogus "no compatibility" claim.
One wonders if the whole thing wasn't a troll.
Take a black felt-tip marker and carefully ink out the lower 90 pixels of you 17" iMac monitor. This will give you a 16:9 ratio, which matches HDTV!
70-83 were removed in the early 80's. And then there was channel 1, which disappeared (to become the 6 meter ham band) in 1948.
Good question! The party line at IBM is that Linux is a server solution. If there are desktops involved this would be significant.