#asahi-dev · 2026-11-15 12:37
jannauwtf, this is depedent on the dst rectangle size or layer? I have two identical surfaces (premultiplied, same format, color space, xfer func), one for top 74 pixels (notch) and one for everything below. the same pixel values in the buffer are rendered differently#asahi-dev · 2026-11-15 12:15
jannaumostly annoyed that it almost works and I fear it might be firmware "bug" which could cause trouble with premult alpha in general#asahi-dev · 2026-11-15 11:15
jannauI'd love to use it in the display driver to support XRGB (not really) #asahi-dev · 2026-11-15 11:14
jannaumain use case is video / display related so VA VPP is a good fit#asahi-dev · 2026-11-15 11:13
jannauyeah, probably nothing in the direction of desktop software which is written with x86 in mind anyway#asahi-dev · 2026-11-15 10:08
jannauisn't apple's kernel driver also named *Scaler*#asahi-dev · 2026-11-15 10:04
jannausofus: could useful for vaapi video processing. vlc currently breaks with vaapi because it expects to be able to do nv12 to planar yuv#asahi-dev · 2026-11-14 21:59
jannauit's one of the early IOMobileFramebufferAP::setProperty_dict() callbacks#asahi-dev · 2026-11-14 21:51
jannauDCP firmware reports now key='IOMFBNumLayers': {'MaxNumLayers': 3, 'FullFrameRequired': 0}#asahi-dev · 2026-11-14 21:25
jannauthe checks for XRGB8888 and XBGR8888 need to stay at least for Pro/Max (it looked like DCP used coverage for ARGB/BGRA on M1 or M2)#asahi-dev · 2026-11-14 21:18
jannauso using DRM_MODE_BLEND_PREMULTI without setting .is_premultiplied was broken#asahi-dev · 2026-11-14 21:17
jannauI don't think clients can at the moment. we would need to add the blend property with `BIT(DRM_MODE_BLEND_PREMULTI) | BIT(DRM_MODE_BLEND_PIXEL_NONE)`#asahi-dev · 2026-11-14 21:06
jannauisn't the blending mode always pre-multiplied?#asahi-dev · 2026-11-14 21:04
jannauis fixing premult alpha blending more than `.is_premultiplied |= format->has_alpha`? but the strange pre-multiplied alpha compositing results might throw a wrench through that too#asahi-dev · 2026-11-14 21:01
jannauI have a few fixes for DCP 14.8.3. the padding you added for swap_submit_dcp is an additional surfaces2. still no idea what those do#asahi-dev · 2026-11-14 20:57
jannautrying to support XRGB8888 on M3 (at least on the primary plane)#asahi-dev · 2026-11-14 18:47
jannaualpha values smaller than 0x28 result in a less bright green and it vanished with 0x00 completly#asahi-dev · 2026-11-14 18:45
jannau0x2800FF00 over 0xFF000000 has the same result as 0xFF00FF00 over 0xFF000000#asahi-dev · 2026-11-14 18:44
jannauthis is strange. it is still somewhat pre-multiplied alpha compositing but somewhat weird#asahi-dev · 2026-11-14 15:29
jannausigh, XRGB works when the framebuffer covers the whole display, i.e. without notch avoidance#asahi-dev · 2026-11-14 13:40
jannaufinally working experiments/dcp.py on t6030, that doesn't have issues displaying XRGB (with X/alpha set to 0) though#asahi-dev · 2026-11-13 23:25
jannauI would have expected that the unused DARTs runtime-pm and power-down ps_ane_sys#asahi-dev · 2026-11-13 23:23
jannauor maybe just https://paste.centos.org/view/170d2353#asahi-dev · 2026-11-13 23:21
jannauhttps://paste.centos.org/view/54fcc0b3 should work as well#asahi-dev · 2026-11-13 23:21
jannauis ane_sys powered on during boot?#asahi-dev · 2026-11-13 23:20
jannauin addition to the devicetree snippet you have an ane driver?#asahi-dev · 2026-11-13 12:07
jannauor possibly it reenables interrupt's too early#asahi-dev · 2026-11-13 12:05
jannauI don't think it's pathological. there is a reason why apple uses sio for most of the SPI HID traffic#asahi-dev · 2026-11-13 12:02
jannaumyria: is you have a M1* Macbook the spi-hid driver would be an option as well. Requires cleanup though#asahi-dev · 2026-11-13 11:59
jannauhttps://github.com/jannau/linux/tree/b4/iommu_reserved_mem_translated is an old partail b4 branch, I think it is not my latest state#asahi-dev · 2026-11-13 11:55
jannauI should have a branch where I started this#asahi-dev · 2026-11-13 11:54
jannau fa103eb1e243 7d89f32a6981 8e45db1af9b8 638ec6fb69ed are the chenges ISP needs but that's probably not upstreamable as is. there are other places that make use reserved memory and it wasn't clear to me how to adapt those. for example vfio#asahi-dev · 2026-11-13 11:52
jannausure but it is not useful to upstream the whole bits/020-dart in one go#asahi-dev · 2026-11-13 11:50
jannaunot upstream dart but upstream the reserved memory handling of dart which is still kinda annoying#asahi-dev · 2026-11-13 10:40
jannauhmm, bcm4388 series was merged with the conflicting DMA_DIRECTION change. resolved correctly in linux next though#asahi-dev · 2026-11-13 10:39
jannauthanks, upstream plan sounds good and I've asked because I was planning to use bcm4388 + 6GHz as base for v7.4 at the latest. I was even wondering whether it would make sense to switch for the next v7.3.x rebase to give it more testing#asahi-dev · 2026-11-13 10:23
jannaumaz: Apple M6 feature regs: https://gist.github.com/jannau/d96eb2fbd778f911ca039d50f81cf5a6#asahi-dev · 2026-11-13 10:05
jannauMichaelReeves[m]: thanks for the 6GHz patches. what is left in our branch once that's accepted and not already upstream like FBT/OKC?#asahi-dev · 2026-11-12 13:52
jannauyuka: `git config b4.prep-perpatch-check-cmd './scripts/checkpatch.pl -q --terse --strict --no-summary --mailback --showfile'`#asahi-dev · 2026-11-11 17:23
jannauI can trace atc on m3 ultra. given that m3 ultra was released after m4 pro/max I think (hope?) it's unlikely that differs significantly#asahi-dev · 2026-11-11 13:25
jannauonly checked bugzilla after writing a similar patch#asahi-dev · 2026-11-11 13:24
jannausomeone already wrote a patch for Firefox sandbox HW decoding: https://bugzilla.mozilla.org/show_bug.cgi?id=2080696#asahi-dev · 2026-11-10 15:54
jannauatcN-dpin[01], see https://f.svenpeter.dev/d09f1fa677f4a843e9c9eb42840f8992861131d2474ec1c200760a5c7fd9c13f_dpin.h how it works#asahi-dev · 2026-11-10 15:30
jannaunot sure how the squencing there is supposed to work but I'd assume the there have to start things on USB4 side before you can connect/HPD dcp#asahi-dev · 2026-11-10 15:27
jannauDPTX_APCALL_DEVICE_NOT_RESPONDING (22) indicates that the signal path between dcp and host router and/or display is not up#asahi-dev · 2026-11-10 15:24
jannauI would expect it to work mostly the same as dp-altmode with selecting dpin0 in the display crossbar, not interacting with atc-phy and then handling everything on usb4 side#asahi-dev · 2026-11-10 12:49
jannauwe even merged the uapi in mesa before v6.16-rc1#asahi-dev · 2026-11-10 12:46
jannauI don't see how the compressed formats help with memory pressure. they actual use more memory. they help with memory bandwidth#asahi-dev · 2026-11-10 12:32
jannauchadmed: I think landing in v7.4-rc1 is enough. I think we merged the driver itself after rc1 with asahi uapi was out#asahi-dev · 2026-11-10 11:03
jannauconfig symbol for 10gbps is AQTION, driver is called atlantic