| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| mchehab | b-rad: thanks | [11:12] |
| ........... (idle for 54mn) | ||
| *** | ChanServ sets mode: +o mchehab | [12:06] |
| ............. (idle for 1h0mn) | ||
| ChanServ sets mode: +o mchehab | [13:06] | |
| ......................... (idle for 2h0mn) | ||
| ChanServ sets mode: +o mchehab | [15:06] | |
| ............................ (idle for 2h15mn) | ||
| mchehab | b-rad: please answer Alex too... he works at AMD
(with GPU stuff, but still...) | [17:21] |
| b-rad | Sorry I thought I did
ahh the iommu question on the patch | [17:24] |
| mchehab | yep
I *suspect* that you can't disable IOMMU on non-developer's bioses but I never touched any Ryzen machine | [17:24] |
| b-rad | i have a cheap motherboard, but i'll give it a look | [17:25] |
| mchehab | when I used to work at Red Hat, only a handful systems (the development) ones had a different BIOS that allowed to setup advanced stuff - although I never seek for IOMMU disabling there
(RH had a lab with hundreds of machines - most of them servers) s/had/has/ with respect to P2P, the bugs are present with DVB? if so, I can't see how to do a PCI to PCI transfer with DVB API | [17:26] |
| b-rad | actually the moment i looked this up I remembered, that iommu kernel parameters were required to work at all | [17:30] |
| mchehab | I mean: the Kernel will always request the card to store data into the main board's RAM.
For V4L2, it is possible to do P2P transfers... via either Overlay or DMABUF making the cx23885 to try to write directly at the GPU private memory inside the PCIe board (of course, for that, you need a PCIe GPU card with memory inside it) I guess gstreamer could be used for such purpose via DMABUF | [17:30] |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |