By: philt (ptay1685.delete@this.bigpond.net.au), June 29, 2007 1:58 am
Room: Moderated Discussions
Linus Torvalds (torvalds@osdl.org) on 6/28/07 wrote:
---------------------------
>philt (ptay1685@bigpond.net.au) on 6/28/07 wrote:
>>
>>FYI just had a look at the Intel pdf for Core 2 errata.
>>Amazed at the number of errata listed.
>
>Read that as:
>"Amazed that Intel actually tells the world, rather than
>hiding it under the rug".
>
Is it really so amazing? The bugs exist in reality, whether they admit them or not. Sooner or later the bugs will cause problems for someone, either software developers or end users, that may result in difficulties debugging (hence developing) software or unreliability of systems for end users. Either way the "truth will out".
Publishing the errata helps Intel by facilitating the design of reobust software. Intel would surely not sell many chips if OS developers like Microsft and yourself could not get an OS to work reliably. (Of course the term reliably needs to be defined - in the IT world the word tends to have little value. What is reliable in IT?)
I suspect this has less to do with noble and altruistic motives than simply the bottom line. Note how the attempt to cover up bugs in the past backfired on them.
>They did learn something from the F00F and FDIV bugs, I
>think. They're very good about making them public these
>days.
>
As I said - they were kind of forced to admit the bugs, or completely lose their reputation. Not really something to praise them for. Like naughty children they only came clean after they were "found out".
>>Also, most seem to have the "nofix" code next to them,
>>which if ive read the key correctly, means they are not
>>going to be fixed. The fixed or going-to-be-fixed bugs are
>>in the minority.
>
>That's related to the above thing.
>
>A lot of the errata tend to be for things like "if you
>have a page-crossing task state segment, and take a
>page fault on the second access, the CPU can corrupt
>the ESP register".
>
>And yes, we've occasionally hit real hardware bugs. Not
>just in CPU's either. You'd think something like a "simple"
>ethernet chip wouldn't be buggy. Think again.
>
>Bugs happen.
>
Nothing is perfect, but the IT industry as a whole seems to scrape the bottom of the barrel. I guess this is inevtiable with the pace of change. I for one would be happier if it all got slowed down. It would be nice for once to buy something that didnt become outmoded junk three weeks after it was purchased. One never gets to feel satisfied with anything one buys in this industry. Nothing last long enough to make it worth the effort to get attached to it.
Its fun but very fleeting fun. Hopefully eventually CPU's will get to the stage when they are fast enough to do more or less anything one wants from them, and then the industry can put efforts into bug eradication rather than continual peformance enhancement, and the things will become truly reliable.
One can only dream.
>Linus
---------------------------
>philt (ptay1685@bigpond.net.au) on 6/28/07 wrote:
>>
>>FYI just had a look at the Intel pdf for Core 2 errata.
>>Amazed at the number of errata listed.
>
>Read that as:
>"Amazed that Intel actually tells the world, rather than
>hiding it under the rug".
>
Is it really so amazing? The bugs exist in reality, whether they admit them or not. Sooner or later the bugs will cause problems for someone, either software developers or end users, that may result in difficulties debugging (hence developing) software or unreliability of systems for end users. Either way the "truth will out".
Publishing the errata helps Intel by facilitating the design of reobust software. Intel would surely not sell many chips if OS developers like Microsft and yourself could not get an OS to work reliably. (Of course the term reliably needs to be defined - in the IT world the word tends to have little value. What is reliable in IT?)
I suspect this has less to do with noble and altruistic motives than simply the bottom line. Note how the attempt to cover up bugs in the past backfired on them.
>They did learn something from the F00F and FDIV bugs, I
>think. They're very good about making them public these
>days.
>
As I said - they were kind of forced to admit the bugs, or completely lose their reputation. Not really something to praise them for. Like naughty children they only came clean after they were "found out".
>>Also, most seem to have the "nofix" code next to them,
>>which if ive read the key correctly, means they are not
>>going to be fixed. The fixed or going-to-be-fixed bugs are
>>in the minority.
>
>That's related to the above thing.
>
>A lot of the errata tend to be for things like "if you
>have a page-crossing task state segment, and take a
>page fault on the second access, the CPU can corrupt
>the ESP register".
>
>And yes, we've occasionally hit real hardware bugs. Not
>just in CPU's either. You'd think something like a "simple"
>ethernet chip wouldn't be buggy. Think again.
>
>Bugs happen.
>
Nothing is perfect, but the IT industry as a whole seems to scrape the bottom of the barrel. I guess this is inevtiable with the pace of change. I for one would be happier if it all got slowed down. It would be nice for once to buy something that didnt become outmoded junk three weeks after it was purchased. One never gets to feel satisfied with anything one buys in this industry. Nothing last long enough to make it worth the effort to get attached to it.
Its fun but very fleeting fun. Hopefully eventually CPU's will get to the stage when they are fast enough to do more or less anything one wants from them, and then the industry can put efforts into bug eradication rather than continual peformance enhancement, and the things will become truly reliable.
One can only dream.
>Linus
