Anmelden oder registrieren
Alpha Board
Startseite
Foren
>
Board Intern
>
Foren Info
>
Forum-News
>
Wie findet ihr die Regeln?
>
Wir beantworten jede Frage in ca. 10 Minuten.
Klicke hier um deine Frage zu stellen
Auf dieses Thema antworten
Benutzername:
Verifizierung:
Wie heißtdie beannteste Suchmaschiene Weltweit
Beitrag:
<p>[quote="morten, post: 635944, member: 127548"]<b>AW: Wie findet ihr die Regeln?</b></p><p><br /></p><p>It exists Amdahls law und Gustavsons remarks to Amdahls law. </p><p><a href="http://en.wikipedia.org/wiki/Amdahl's_law" target="_blank" class="externalLink" rel="nofollow">http://en.wikipedia.org/wiki/Amdahl's_law</a></p><p><br /></p><p><br /></p><p>More technically, the law is concerned with the speedup achievable from an improvement to a computation that affects a proportion P of that computation where the improvement has a speedup of S. (For example, if an improvement can speed up 30% of the computation, P will be 0.3; if the improvement makes the portion affected twice as fast, S will be 2). </p><p><br /></p><p>Amdahl's law states that the overall speedup of applying the improvement will be</p><p><br /></p><p><br /></p><p><br /></p><p><dl></p><p><dd><img class="tex" alt="\frac{1}{(1 - P) + \frac{P}{S}}" src="http://upload.wikimedia.org/math/8/9/6/89638a25639297d701f9298b68ee9cec.png">.</dd></p><p></dl></p><p><br /></p><p>Parallelization</p><p><br /></p><p>In the case of parallelization, Amdahl's law states that if P is the proportion of a program that can be made parallel (i.e. benefit from parallelization), and (1 − P) is the proportion that cannot be parallelized (remains serial), then the maximum speedup that can be achieved by using N processors is</p><p><br /></p><p><dl></p><p><dd><img class="tex" alt="\frac{1}{(1-P) + \frac{P}{N}}" src="http://upload.wikimedia.org/math/8/9/0/89073e92c6f758701ce24b9ed480ecd2.png"></dd></p><p></dl> </p><p><br /></p><p><br /></p><p><a href="http://en.wikipedia.org/wiki/Gustafson%27s_Law" target="_blank" class="externalLink" rel="nofollow">http://en.wikipedia.org/wiki/Gustafson's_Law</a> </p><p><br /></p><p>is incorrectl stated as a Law, </p><p><br /></p><p>Gustavsson does not prove or state anything new to me.</p><p><br /></p><p><br /></p><p>Gustafson's Law (also known as Gustafson-Barsis' law) is a law in computer engineering which states that any sufficiently large problem can be efficiently parallelized. Gustafson's Law is closely related to Amdahl's law, which gives a limit to the degree to which a program can be sped up due to parallelization. It was first described by John L. Gustafson in 1988.</p><p><br /></p><p> S(P) = P − α * (P − 1).</p><p><br /></p><p>where P is the number of processors, S is the speedup, and α the non-parallelizable part of the process. </p><p><br /></p><p><br /></p><p><br /></p><p>Amdahw was right and still is, if certan loads are fixed and if they will remained fixed. But now the Gigabit/second inflation of bandwith and komplexity of application requires every month more and more, Amdald put a rule for the given assumed constraint, that you don't need real more software. And after some days during using the software that you don't need or will not know anything about new features. </p><p><br /></p><p>The industry is driving us with even more and faster games, adobe flash, and software updates which requires more CPU, GPU and bandtwidth. </p><p><br /></p><p>If you start to measure and then sun gives you a 64 bit java browser plugin, the youtube videos will release a new HDTV standard, all during the Amdahl benchmarking, and you are expected to run that on the same Vista as yesterday. </p><p><br /></p><p>then you need quadcore and multicore. and it will help you more than amdahl predicted. </p><p>By constantly increasing load. </p><p><br /></p><p>The hardware requirementys and esp the monthly increasing requirements of hardware specs and the weekly updates, or even upgrades was not foreseen by Amdahl. </p><p><br /></p><p>Gustavson appendet it</p><p><br /></p><p>which cannot scale to match availability of computing power </p><p>as the machine size increases.</p><p><br /></p><p>this is incorrect. as the software complexity and Bandth with CPU GPU and IO Performance constantly increases exponentially amdahls law . </p><p><br /></p><p>amdahl said, you gan get a ******* incrase in measurable Computing speed at constant load. </p><p><br /></p><p>Gustavson said, A program that does not change or upgrade for a longer period of time is useless. That is not the case today. whith this taken into consideration we can get more speed , higher benchmarks, petter performance by using paralellism. than previously assumed. </p><p><br /></p><p>amdahl has still right. but the input of data streams in a datacenter which needs to be spam filtered, re rendered, on the fly converted to another format, and then this codec is no more used, the new codec must be used instead, and all which every programmer is diong today will be likely to need a more powerful update. </p><p><br /></p><p><br /></p><p>for Instance the word processing program MS Word, more features offered can only be remedied with more hardware paralellizm. and that works better as predicted by amdahl, if the load is constantly increasing. as is the case now. We also run 10 20 different applications simultaneously , when one is calculating the maximum number of subatomic particles in the entire universe, from the latest hubble achievements. the other application is still running to adjust the formulas for how to continue if the outcome would be that the universe is closed. and how to include google earth in next generation scalable mongrel clustering. </p><p><br /></p><p>newer application like simultaneously reviewing youtube, for adult content also gives another result than Amdahls law. </p><p><br /></p><p>So I don't accept the laws as something more than old fashioned facts based on old fashioned traditions. </p><p><br /></p><p><br /></p><p>Even if they are proven to be correct. </p><p><br /></p><p>So My rule is that all rules are limited, based upon a ******* snapshot in time.[/quote]</p><p><br /></p>
[quote="morten, post: 635944, member: 127548"][b]AW: Wie findet ihr die Regeln?[/b] It exists Amdahls law und Gustavsons remarks to Amdahls law. [url]http://en.wikipedia.org/wiki/Amdahl's_law[/url] More technically, the law is concerned with the speedup achievable from an improvement to a computation that affects a proportion P of that computation where the improvement has a speedup of S. (For example, if an improvement can speed up 30% of the computation, P will be 0.3; if the improvement makes the portion affected twice as fast, S will be 2). Amdahl's law states that the overall speedup of applying the improvement will be <dl> <dd><img class="tex" alt="\frac{1}{(1 - P) + \frac{P}{S}}" src="http://upload.wikimedia.org/math/8/9/6/89638a25639297d701f9298b68ee9cec.png">.</dd> </dl> Parallelization In the case of parallelization, Amdahl's law states that if P is the proportion of a program that can be made parallel (i.e. benefit from parallelization), and (1 − P) is the proportion that cannot be parallelized (remains serial), then the maximum speedup that can be achieved by using N processors is <dl> <dd><img class="tex" alt="\frac{1}{(1-P) + \frac{P}{N}}" src="http://upload.wikimedia.org/math/8/9/0/89073e92c6f758701ce24b9ed480ecd2.png"></dd> </dl> [url]http://en.wikipedia.org/wiki/Gustafson%27s_Law[/url] is incorrectl stated as a Law, Gustavsson does not prove or state anything new to me. Gustafson's Law (also known as Gustafson-Barsis' law) is a law in computer engineering which states that any sufficiently large problem can be efficiently parallelized. Gustafson's Law is closely related to Amdahl's law, which gives a limit to the degree to which a program can be sped up due to parallelization. It was first described by John L. Gustafson in 1988. S(P) = P − α * (P − 1). where P is the number of processors, S is the speedup, and α the non-parallelizable part of the process. Amdahw was right and still is, if certan loads are fixed and if they will remained fixed. But now the Gigabit/second inflation of bandwith and komplexity of application requires every month more and more, Amdald put a rule for the given assumed constraint, that you don't need real more software. And after some days during using the software that you don't need or will not know anything about new features. The industry is driving us with even more and faster games, adobe flash, and software updates which requires more CPU, GPU and bandtwidth. If you start to measure and then sun gives you a 64 bit java browser plugin, the youtube videos will release a new HDTV standard, all during the Amdahl benchmarking, and you are expected to run that on the same Vista as yesterday. then you need quadcore and multicore. and it will help you more than amdahl predicted. By constantly increasing load. The hardware requirementys and esp the monthly increasing requirements of hardware specs and the weekly updates, or even upgrades was not foreseen by Amdahl. Gustavson appendet it which cannot scale to match availability of computing power as the machine size increases. this is incorrect. as the software complexity and Bandth with CPU GPU and IO Performance constantly increases exponentially amdahls law . amdahl said, you gan get a ******* incrase in measurable Computing speed at constant load. Gustavson said, A program that does not change or upgrade for a longer period of time is useless. That is not the case today. whith this taken into consideration we can get more speed , higher benchmarks, petter performance by using paralellism. than previously assumed. amdahl has still right. but the input of data streams in a datacenter which needs to be spam filtered, re rendered, on the fly converted to another format, and then this codec is no more used, the new codec must be used instead, and all which every programmer is diong today will be likely to need a more powerful update. for Instance the word processing program MS Word, more features offered can only be remedied with more hardware paralellizm. and that works better as predicted by amdahl, if the load is constantly increasing. as is the case now. We also run 10 20 different applications simultaneously , when one is calculating the maximum number of subatomic particles in the entire universe, from the latest hubble achievements. the other application is still running to adjust the formulas for how to continue if the outcome would be that the universe is closed. and how to include google earth in next generation scalable mongrel clustering. newer application like simultaneously reviewing youtube, for adult content also gives another result than Amdahls law. So I don't accept the laws as something more than old fashioned facts based on old fashioned traditions. Even if they are proven to be correct. So My rule is that all rules are limited, based upon a ******* snapshot in time.[/quote]
Mit Facebook anmelden
Benutzername oder E-Mail-Adresse:
Besitzt du schon ein Benutzerkonto?
Nein, erstelle jetzt ein Benutzerkonto.
Ja, mein Passwort ist:
Hast du dein Passwort vergessen?
Angemeldet bleiben
Alpha Board
Startseite
Foren
>
Board Intern
>
Foren Info
>
Forum-News
>
Wie findet ihr die Regeln?
>
Startseite
Foren
Foren
Direktauswahl
Foren durchsuchen
Recent Posts
Mitglieder
Mitglieder
Direktauswahl
Notable Members
Derzeitige Besucher
Letzte Aktivitäten
Menu
Suche
Nur die Titel durchsuchen
Erstellt von:
Trenne Benutzernamen durch Kommata.
Neuer als:
Nur dieses Thema durchsuchen
Nur dieses Forum durchsuchen
Die Ergebnisse als Themen anzeigen
Nützliche Suchen
Recent Posts
Mehr...