01:50 alyssa: zmike: vk has spec text "If the current pipeline uses custom sample locations the value of any variable decorated with the
01:50 alyssa: SamplePosition built-in decoration is undefined.
01:50 alyssa: "
01:50 alyssa: I don't see similar text in https://registry.khronos.org/OpenGL/extensions/ARB/ARB_sample_locations.txt , implying that the interaction is well defined in gl
01:51 alyssa: is this broken in zink?
01:52 alyssa: (I don't see any code in zink that would fix up the custom sample loc case.)
01:53 alyssa: (If you resolve this by banning the interaction in gl, I wouldn't be opposed ;p)
01:58 alyssa:could've sworn she had this conversation 6 months ago, maybe i'm losing it
01:59 alyssa: right https://oftc.irclog.whitequark.org/dri-devel/2023-09-08#32468213;
02:02 alyssa: guess we didn't come to consensus then
02:06 alyssa: filed https://gitlab.khronos.org/opengl/API/-/issues/218
02:23 DemiMarie: Update: an AMD developer found a solution to the memory mapping problems!
02:23 DemiMarie: So native contexts with Xen should work.
08:02 eric_engestrom: mareko: have a look at how we build deqp in the CI, it might answer future questions as well: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/.gitlab-ci/container/build-deqp.sh
12:26 zmike: alyssa: you may as well just make a PR against the registry repo proposing matching language to vk since that's most likely going to be the reply
13:01 alyssa: zmike: ack
22:38 alyssa: austriancoder: btw, RA_MULTISAMPLE_UNK00E10 contains the sample locations
22:39 alyssa: each byte is a sample, each nibble is x/y, 4-bit fixed points
22:39 alyssa: agx is same
22:43 austriancoder: alyssa: ah.. thanks .. you are going to join the etnaviv team? 😀
23:44 glehmann: alyssa: I thought agx was based on imgtec hw, not vivante?