Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
04/10/12 20:32
Read: times


 
#187101 - Always danger with low-level code for platform
Responding to: ???'s previous message
Pretty smart as Win/x86 programmers still doesn't tell anything about experience of unix development.

Next thing is that the chip manufacturers are normally involved in initial porting of the low-level layer of Linux to their specific processors. If they do a crap job, the result will be crap. It isn't unreasonable to think that they would do a similar crap job with the low-level code for running Windows on that same target processor - our famous BSOD is normally (unless the hw is broken) caused by the huge amounts of badly written Windows drivers.

So yes - I have fixed a number of direct bugs in Linux kernel code. But normally not code from the mainstream Linux distribution but code from the chip manufacturer or from some consults hired by the chip manufacturer. There are also "pretty smart" people who implement SPI drivers strictly simplex - never ever thinking that a device may send data back at the same time that you send data from master to slave.

In the end, anyone who do select a hw + an OS have to make some form of evaluation of the quality of the low-level drivers for that specific combination. Especially if the drivers aren't part of the standard OS distribution.

The serious issue here is that in many cases, the chip documentation doesn't contain enough information to let you fix driver issues yourself even when you have the source code. You don't know what bits in what registers will give what result.

In the end, the weakest spot of both Linux and Windows platforms are the hw-specific driver layer. And a Windows release can get through the hw compatibility tests by just ignoring existing hw functionality. So the user will read documentation about the chip and select it because it supports a critically important function, and then end up with drivers that do not support this feature. That is a good reason for getting a reference hw platform as soon as possible to test all critical functionality on when it's still time to throw out the processor (or maybe an external hw chip) and use something else that is supported by really working code.

List of 92 messages in thread
TopicAuthorDate
has linux had its chips?            01/01/70 00:00      
   some weirdo in sandals a ponytail            01/01/70 00:00      
      Problem is            01/01/70 00:00      
      possibly, but not only            01/01/70 00:00      
      It's all in the history ... and "read the code" doesn't work            01/01/70 00:00      
         Good points!            01/01/70 00:00      
            Why not a firm objective?            01/01/70 00:00      
               Lots of projects have a large percentage research            01/01/70 00:00      
                  I knew you'd have to come in with something irrelevant            01/01/70 00:00      
                     Try document an invention before it's invented...            01/01/70 00:00      
                        Here's some research for you, Per            01/01/70 00:00      
                           Richard to give an example            01/01/70 00:00      
                              I'd like YOU, Erik, to come up with one example ...            01/01/70 00:00      
                                 Always prejudice from Richard            01/01/70 00:00      
                                 'documenting' means many things            01/01/70 00:00      
                                 here we go again            01/01/70 00:00      
                                    when you're wrong, you're wrong            01/01/70 00:00      
                                       I do not have a microscope and probes that small            01/01/70 00:00      
                                          So you've made no observations ... you just guessed ...            01/01/70 00:00      
                                             and that irks you immensely            01/01/70 00:00      
                                                What did you do, aside from guessing?            01/01/70 00:00      
                                                   then please, tell me            01/01/70 00:00      
                                                      Are you willing to explore this in detail?            01/01/70 00:00      
                                                         now you are jumping            01/01/70 00:00      
                           Your references aren't exactly backing your view            01/01/70 00:00      
                              It's not about me ... it's about process            01/01/70 00:00      
                                 But processes contains feedback loops            01/01/70 00:00      
                                    I believe you've gone off-the-rails, Per            01/01/70 00:00      
                                       But getty isn't Linux            01/01/70 00:00      
                                          it's a small piece, but it's an example            01/01/70 00:00      
                                             But not of Linux            01/01/70 00:00      
                                                It was part of the distribution.            01/01/70 00:00      
                                 'Research' can mean many things            01/01/70 00:00      
                                    Yes, but that's in a different context            01/01/70 00:00      
                                       You still haven't told what Linux documentation you miss            01/01/70 00:00      
                                          I don't know what you mean            01/01/70 00:00      
                                             Still claims based on assumptions and not facts            01/01/70 00:00      
                                                Not everyone is completely stupid            01/01/70 00:00      
                                                   But what is the relevance today?            01/01/70 00:00      
                                                      I've no opinion about the current LINUX            01/01/70 00:00      
                                                         and, you Richard, who loves living in the past            01/01/70 00:00      
                                                            just a minute, Erik            01/01/70 00:00      
                                                               Examples?            01/01/70 00:00      
                                                                  Nothing has changed since 15 years ago ...            01/01/70 00:00      
                                                                     at least not Richards opinions :)            01/01/70 00:00      
                                                                     Still lots of assumptions and unbacked claims            01/01/70 00:00      
                                 Oh, Richard, I have a job for you            01/01/70 00:00      
                           I had Yourdons first book as manuscript and ...            01/01/70 00:00      
                              be careful ...            01/01/70 00:00      
                                 were you once a bartender ...            01/01/70 00:00      
                                    You have to accept the difference ...            01/01/70 00:00      
                                 We are careful            01/01/70 00:00      
                     Hog Wash.....            01/01/70 00:00      
                        Odd that you see it that way ...            01/01/70 00:00      
                     I just documented fully            01/01/70 00:00      
               Because an "Objective" is not a final product specification            01/01/70 00:00      
                  I have to disagree ... the objective specification is step 1            01/01/70 00:00      
            documentation            01/01/70 00:00      
               Definitely not the "usual response"            01/01/70 00:00      
   Android            01/01/70 00:00      
      I don't think so            01/01/70 00:00      
         Don't agree            01/01/70 00:00      
      all due respect, no.            01/01/70 00:00      
         Apple may have a price match on the superluxourious            01/01/70 00:00      
            re: Apple may have a price match on the superluxourious            01/01/70 00:00      
               but 95% of the population does not need....            01/01/70 00:00      
                  Web browsing normally the most power-hungry you can do            01/01/70 00:00      
                  re: 95%            01/01/70 00:00      
               So how exactly am I wrong?            01/01/70 00:00      
                  But Android is Linux            01/01/70 00:00      
                     Just the opposite            01/01/70 00:00      
                        {sigh}            01/01/70 00:00      
                           just like the preacher said to the atheist            01/01/70 00:00      
                  re: How exactly?            01/01/70 00:00      
      850000 Android phones activated per day, linux video            01/01/70 00:00      
   anecdotes...            01/01/70 00:00      
      These guys...            01/01/70 00:00      
         digital audio consoles            01/01/70 00:00      
   The Rasperry Pi Foundation clearly doesn't think so!            01/01/70 00:00      
      The world isn't just a few companies            01/01/70 00:00      
   the basic problem with free software is...            01/01/70 00:00      
      Careful with the use of "Linux". Most things "Linux" aren't!            01/01/70 00:00      
         I did refer to linux itself            01/01/70 00:00      
            Wrong hw selected, or just big lack of platform knowledge?            01/01/70 00:00      
               a port            01/01/70 00:00      
                  Always danger with low-level code for platform            01/01/70 00:00      
      True - but "paid-for" is not necessarily any better!            01/01/70 00:00      
         when selecting any tool            01/01/70 00:00      
      As Stallman said.....            01/01/70 00:00      
         free            01/01/70 00:00      
            All about volume or already existing knowledge/experience            01/01/70 00:00      
   Well...say what you like about me, and many people do.            01/01/70 00:00      

Back to Subject List