PCI express equipment
On one hand ,hardwarely,as it is always said, PCIE provides the superexcellent performance
comparing with the PCI and PCI -X
On the other ,softwarely,in most the application scenarios i have met , the PCIE equipment will
have to use the pci-compatible driver first in linux. because the Configuration head is pci compatible in the first 256 byte , it is allowed and really works.
i know if we use the conventional Device Driver for pcie equipment, the advanced option such as MSI ,Flow Control ,TC ,VC and hotplug are unavailable, i also know
that the speed is mainly the problem of Bandwidth and Clocking. Still i know that the Trafic
control mechanism is good for the arbitration and thus improve the performance just like the
TCP protocol ways.
So my conclusion is :
the conventional Device driver will not be the bottle neck of the PCIE performance as long as
only speed is concerned ,and the advanced options are only Application-related.
here allow me just compare it with the network protocol , oh ,i just use the IP protocol ,with out the TCP /UDP way. is that right ?or any wise can give some advice and
Signature · 1 line
we who r about to die,salute u!
Quoted 7 lines · kylin
On one hand ,hardwarely,as it is always said, PCIE provides the superexcellent performance
comparing with the PCI and PCI -X
On the other ,softwarely,in most the application scenarios i have met , the PCIE equipment will
Quoted 5 lines
have to use the pci-compatible driver first in linux. because the Configuration head is pci compatible in the first 256 byte , it is allowed and really works.
i know if we use the conventional Device Driver for pcie equipment, the advanced option such as MSI ,Flow Control ,TC ,VC and hotplug are
rubbish. MSI and Hotplug are independent of PCI Express and are used by conventional PCI drivers today. Flow control, TC and VC are all part of the same thing.
Quoted 9 lines
unavailable, i also know
that the speed is mainly the problem of Bandwidth and Clocking. Still i know that the Trafic
control mechanism is good for the arbitration and thus improve the performance just like the
TCP protocol ways.
Um, no, not really. There's very little hardware support for VC/TC at the moment, so having software support for it isn't a priority. In addition, it really depends how deep your PCI bus hierarchy is whether VC/TC will have any effect on performance at all.
Why is this question on linux-scsi anyway?
Signature · 6 lines
"Next the statesmen will invent cheap lies, putting the blame upon the nation that is attacked, and every man will be glad of those conscience-soothing falsities, and will diligently study them, and refuse to examine any refutations of them; and thus he will by and by convince himself that the war is just, and will thank God for the better sleep he enjoys after this process of grotesque self-deception." -- Mark Twain
Quoted 14 lines · Matthew Wilcox
Quoted 7 lines · kylin
On one hand ,hardwarely,as it is always said, PCIE provides the superexcellent performance
comparing with the PCI and PCI -X
On the other ,softwarely,in most the application scenarios i have met , the PCIE equipment willQuoted 5 lines
have to use the pci-compatible driver first in linux. because the Configuration head is pci compatible in the first 256 byte , it is allowed and really works.
i know if we use the conventional Device Driver for pcie equipment, the advanced option such as MSI ,Flow Control ,TC ,VC and hotplug are
Quoted 3 lines
rubbish. MSI and Hotplug are independent of PCI Express and are used by conventional PCI drivers today. Flow control, TC and VC are all part of the same thing.
sorry for my innocence,but all i want to say here is if the unMSI and unHotplug device driver would influence the normal speed in ONE common scenarios
Quoted 5 lines
Quoted 5 lines
unavailable, i also know
that the speed is mainly the problem of Bandwidth and Clocking. Still i know that the Trafic
Quoted 7 lines
Quoted 4 lines
control mechanism is good for the arbitration and thus improve the performance just like the
TCP protocol ways.
Um, no, not really. There's very little hardware support for VC/TC at the moment, so having software support for it isn't a priority.
yeah ,even the bios will rarely support this
Quoted 3 lines
In addition,
it really depends how deep your PCI bus hierarchy is whether VC/TC will have any effect on performance at all.
have you seen the E7520 chipset ,sir?
Quoted 1 line
Why is this question on linux-scsi anyway?
because myu SCSI interface is connecting to the pcie bus:),love it and love the dog :)
Signature · 1 line
we who r about to die,salute u!