Is there any advantage to aligning the buffers pasted into HSE_REQ_VECTOR_SEND on cache line boundaries, etc.? Ian
Hi Ian, What's your purpose to align the buffers? Is this in order to make sure the response can be cached in Kernel? David had some detailed instruction on this topic. For your reference: HOWTO: Use the HTTP.SYS Kernel Mode Response Cache with IIS 6 http://blogs.msdn.com/david.wang/archive/2005/07/07/HOWTO_Use_Kernel_Respons e_Cache_with_IIS_6.aspx Best regards, WenJun Zhang Microsoft Online Partner Support This posting is provided "AS IS" with no warranties, and confers no rights.
I suspect it would probably improve the copying which no doubt happens when it gets cached in the kernel, but my reason for asking is that I'm assuming that the buffers passed in will eventually be accessed via DMA to send them to the network card. If this is the case then is there an advantage in aligning them? Ian ""WenJun Zhang[msft]"" wrote in message news:rTgwmpJiGHA.4896@TK2MSFTNGXA01.phx.gbl... > Hi Ian, > > What's your purpose to align the buffers? Is this in order to make sure > the > response can be cached in Kernel? > > David had some detailed instruction on this topic. For your reference: > > HOWTO: Use the HTTP.SYS Kernel Mode Response Cache with IIS 6 > http://blogs.msdn.com/david.wang/archive/2005/07/07/HOWTO_Use_Kernel_Respons > e_Cache_with_IIS_6.aspx > > Best regards, > > WenJun Zhang > Microsoft Online Partner Support > > This posting is provided "AS IS" with no warranties, and confers no > rights. > > >
Hi Ian, I'm currently researching this issue and plan on discussing it in our internal. Please wait for my update. Thanks. Best regards, WenJun Zhang Microsoft Online Partner Support This posting is provided "AS IS" with no warranties, and confers no rights.
Hi Ian, We are still researching this issue. Thanks. Best regards, WenJun Zhang Microsoft Online Partner Support This posting is provided "AS IS" with no warranties, and confers no rights.
I asked the HTTP.SYS developer involved with this (as well as perf). He indicated that the benefit would be very small or none unless you have another app/thread writing heavily into the same cache line that the buffer resides in and to not worry about cache lining the VectorSend buffers. -- //David IIS http://blogs.msdn.com/David.Wang This posting is provided "AS IS" with no warranties, and confers no rights. // "Ian" <Ixpah@newsgroup.nospam> wrote in message news:usjyGitiGHA.4572@TK2MSFTNGP05.phx.gbl... >I suspect it would probably improve the copying which no doubt happens when >it gets cached in the kernel, but my reason for asking is that I'm assuming >that the buffers passed in will eventually be accessed via DMA to send them >to the network card. If this is the case then is there an advantage in >aligning them? > > Ian > > > ""WenJun Zhang[msft]"" wrote in message > news:rTgwmpJiGHA.4896@TK2MSFTNGXA01.phx.gbl... >> Hi Ian, >> >> What's your purpose to align the buffers? Is this in order to make sure >> the >> response can be cached in Kernel? >> >> David had some detailed instruction on this topic. For your reference: >> >> HOWTO: Use the HTTP.SYS Kernel Mode Response Cache with IIS 6 >> http://blogs.msdn.com/david.wang/archive/2005/07/07/HOWTO_Use_Kernel_Respons >> e_Cache_with_IIS_6.aspx >> >> Best regards, >> >> WenJun Zhang >> Microsoft Online Partner Support >> >> This posting is provided "AS IS" with no warranties, and confers no >> rights. >> >> >> > >
