June 10, 2018 
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | >
June 10, 2018 [00:05:14] *** psychicist__ <psychicist__!~psychicis@5356A22B.cm-6-7c.dynamic.ziggo.nl> has quit IRC (Ping timeout: 260 seconds)[00:05:49] *** soulz <soulz!~soulz@unaffiliated/soulz> has quit IRC (Ping timeout: 256 seconds)[00:13:03] *** soulz <soulz!~soulz@unaffiliated/soulz> has joined ##vulkan[00:36:21] *** bpmedley <bpmedley!~bpm@c-24-72-144-115.ni.gigamonster.net> has quit IRC (Ping timeout: 264 seconds)[00:41:17] *** nsf <nsf!~nsf@jiss.convex.ru> has quit IRC (Quit: WeeChat 2.1)[00:49:26] *** bpmedley <bpmedley!~bpm@c-24-72-144-115.ni.gigamonster.net> has joined ##vulkan[00:57:00] *** cramalho <cramalho!~cramalho@2804:54:16f5:8200:bd67:9a34:7fbd:ea22> has joined ##vulkan[01:01:33] *** bpmedley <bpmedley!~bpm@c-24-72-144-115.ni.gigamonster.net> has quit IRC (Ping timeout: 264 seconds)[01:12:27] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[01:24:58] *** ector <ector!~asdf@ua-85-224-236-175.bbcust.telenor.se> has quit IRC ()[01:33:58] *** Thorn__ is now known as Thorn[03:09:25] *** cramalho_ <cramalho_!~cramalho@179.190.254.211> has joined ##vulkan[03:10:37] *** cramalho <cramalho!~cramalho@2804:54:16f5:8200:bd67:9a34:7fbd:ea22> has quit IRC (Ping timeout: 245 seconds)[03:10:37] *** cramalho_ is now known as cramalho[04:18:06] *** ratchet_freak <ratchet_freak!~ratchetfr@ptr-82s3g7nu9uak27g24h0.18120a2.ip6.access.telenet.be> has quit IRC (Ping timeout: 256 seconds)[04:45:36] *** cramalho <cramalho!~cramalho@179.190.254.211> has quit IRC (Quit: cramalho)[04:47:06] *** cramalho <cramalho!~cramalho@179.190.254.211> has joined ##vulkan[05:12:07] *** MrFlibble <MrFlibble!MrFlibble@2.122.47.217> has quit IRC (Ping timeout: 260 seconds)[05:33:44] *** cramalho <cramalho!~cramalho@179.190.254.211> has quit IRC (Quit: cramalho)[05:34:11] *** cramalho <cramalho!~cramalho@2804:54:16f5:8200:bd67:9a34:7fbd:ea22> has joined ##vulkan[05:55:01] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has quit IRC (Quit: This computer has gone to sleep)[05:57:21] *** salamanderrake <salamanderrake!~quassel@cpe-24-165-202-54.neo.res.rr.com> has joined ##vulkan[06:48:19] *** slime <slime!~slime73@24.215.81.93> has joined ##vulkan[06:51:47] *** cramalho <cramalho!~cramalho@2804:54:16f5:8200:bd67:9a34:7fbd:ea22> has quit IRC (Quit: cramalho)[06:52:54] *** slime <slime!~slime73@24.215.81.93> has quit IRC (Client Quit)[07:40:54] *** daxar2 <daxar2!~daxar@193-81-39-231.adsl.highway.telekom.at> has joined ##vulkan[07:41:23] *** xissburg <xissburg!~xissburg@unaffiliated/xissburg> has quit IRC (Quit: ZZZzzz…)[07:43:55] *** daxar1 <daxar1!~daxar@212-197-177-51.adsl.highway.telekom.at> has quit IRC (Ping timeout: 260 seconds)[07:49:10] *** Fats <Fats!fats@gateway/vpn/privateinternetaccess/fats> has quit IRC (Remote host closed the connection)[07:50:24] *** Fats <Fats!fats@gateway/vpn/privateinternetaccess/fats> has joined ##vulkan[07:52:15] *** sla_ro|master <sla_ro|master!~sla.ro@78.96.209.89> has joined ##vulkan[07:57:37] *** ImQ009 <ImQ009!~ImQ009@unaffiliated/imq009> has joined ##vulkan[07:58:24] *** bpmedley <bpmedley!~bpm@c-24-72-144-115.ni.gigamonster.net> has joined ##vulkan[09:03:00] *** lambskin <lambskin!~lambskin@ppp118-211-201-52.bras1.syd2.internode.on.net> has joined ##vulkan[09:05:50] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[09:06:01] <bridger> hi vulkan folks[09:06:52] <bridger> Does vulkan require less or more passthrough power from cpu to gfx card than directx or opengl?[09:07:17] *** nsf <nsf!~nsf@jiss.convex.ru> has joined ##vulkan[09:10:51] *** neurre <neurre!~Timo@mobile-access-5d6aec-56.dhcp.inet.fi> has joined ##vulkan[09:14:04] *** cheakoirccloud <cheakoirccloud!uid293319@gateway/web/irccloud.com/x-belmyslkqqepeuzu> has quit IRC (Quit: Connection closed for inactivity)[09:23:12] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[09:36:45] <neurre> hi[09:38:13] <Lymia> bridger, I don't see how the graphics API in use would affect power usage[09:38:46] <Lymia> Assuming equally efficient uses of the API, at least.[09:42:47] <bridger> Lymia: Lemme put it this way, is vulkann more efficient because it uses more bandwidth from cpu to gpu interface compared to directx or opengl?[09:44:55] <Lymia> Vulkan is potentially faster because it lets you explicitly ask for what you want, instead of making the driver guess. I doubt it'd be a big bandwidth difference because the majority of CPU->GPU bandwidth is going to be resources and data, not commands.[09:45:30] <bridger> hmm[09:45:33] <bridger> thanks[09:48:48] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[09:48:58] <Lymia> Granted, I'm assuming 3D gaming or something instead of... I don't know, word processing or something like that.[10:14:44] *** ratchet_freak <ratchet_freak!~ratchetfr@ptr-82s3g7nu9uak27g24h0.18120a2.ip6.access.telenet.be> has joined ##vulkan[10:19:04] *** ratchet_freak <ratchet_freak!~ratchetfr@ptr-82s3g7nu9uak27g24h0.18120a2.ip6.access.telenet.be> has quit IRC (Ping timeout: 256 seconds)[10:35:42] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[10:49:18] *** Quetzal2 <Quetzal2!~Quetzal2@unaffiliated/quetzal2> has joined ##vulkan[10:52:38] *** neurre <neurre!~Timo@mobile-access-5d6aec-56.dhcp.inet.fi> has quit IRC (Quit: Leaving)[10:55:20] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has joined ##vulkan[11:23:56] *** Guest4630 <Guest4630!4d3b9548@gateway/web/cgi-irc/kiwiirc.com/ip.77.59.149.72> has joined ##vulkan[11:26:51] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[11:28:05] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[11:42:40] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[11:44:42] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[11:44:59] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[11:47:04] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[11:49:19] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[11:56:41] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[11:58:25] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:00:34] *** psychicist__ <psychicist__!~psychicis@5356A22B.cm-6-7c.dynamic.ziggo.nl> has joined ##vulkan[12:01:53] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:03:51] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[12:11:05] *** borkr <borkr!~borkr@static130-244.mimer.net> has joined ##vulkan[12:13:43] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[12:15:19] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:17:17] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[12:18:56] *** TSS_ <TSS_!~TSS@2a02:2f0a:4050:386:d250:99ff:fe83:4a0a> has quit IRC (Quit: ZNC 1.6.2 - http://znc.in)[12:19:13] *** TSS_ <TSS_!~TSS@2a02:2f0a:4050:be2:d250:99ff:fe83:4a0a> has joined ##vulkan[12:20:05] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:23:51] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[12:25:43] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:30:19] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[12:30:55] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:34:43] *** TSS3000RO <TSS3000RO!~TSS@5-12-100-195.residential.rdsnet.ro> has joined ##vulkan[12:35:52] *** TSS_ <TSS_!~TSS@2a02:2f0a:4050:be2:d250:99ff:fe83:4a0a> has quit IRC (Ping timeout: 265 seconds)[12:39:56] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[12:41:45] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[12:43:07] *** TSS3000RO <TSS3000RO!~TSS@5-12-100-195.residential.rdsnet.ro> has quit IRC (Ping timeout: 245 seconds)[12:51:50] *** TSS_ <TSS_!~TSS@2a02:2f0a:4030:1447:d250:99ff:fe83:4a0a> has joined ##vulkan[13:10:59] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[13:12:57] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[13:15:33] *** bridger <bridger!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[13:20:32] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[13:31:13] *** cramalho <cramalho!~cramalho@2804:54:16f5:8200:bd67:9a34:7fbd:ea22> has joined ##vulkan[13:38:05] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[13:43:55] *** HZun <HZun!~HZun@0x3ec72d49.osd.customer.dk.telia.net> has joined ##vulkan[13:45:12] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[13:48:29] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[13:55:48] <HZun> Do GPUs flush their registers and shared memory upon context switches between different applications/processes? (i know that they dont between thread context switches in the same process, so this question is only about between different applications/processes).[13:57:15] *** herbmillerjr <herbmillerjr!~herb@c-76-120-173-141.hsd1.pa.comcast.net> has joined ##vulkan[14:21:35] <HZun> What are the advantages/purpose of the "specialized" shader stages instead of just using compute shaders for everything? (since they use the same SMs anyway)?[14:22:35] *** ratchet_freak <ratchet_freak!~ratchetfr@ptr-82s3g7nu9uak27g24h0.18120a2.ip6.access.telenet.be> has joined ##vulkan[14:55:18] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has joined ##vulkan[15:10:36] <fazias> HZun: https://fgiesen.wordpress.com/2011/07/03/a-trip-through-the-graphics-pipeline-2011-part-3/[15:16:52] <HZun> fazias: thanks, though after reading it it does not seem to answer my question. or else i am just too stupid to understand it haha :D[15:17:12] <HZun> i get the different between the normal different shader stages.[15:17:15] <fazias> There is fixed function hardware on gpu's, specialized shader stages use them to their advantage.[15:18:08] *** lambskin <lambskin!~lambskin@ppp118-211-201-52.bras1.syd2.internode.on.net> has quit IRC (Remote host closed the connection)[15:18:09] <fazias> Anyhow, there are multiple parts to that, if you are interested in how gpu's work or why, most of your questions are answered in that.[15:18:34] <HZun> i will read all of the parts then :)[15:19:59] <HZun> however it is still written prior to vulkan. I thought perhabs that vulkan was low-level enough to use interopate with the fixed function stages from compute shaders in a performant way.[15:20:56] <HZun> interoperate*[15:21:16] <ratchet_freak> createGraphicsPipeline at least has input parameters that set the fixed pipeline bits[15:21:29] <fazias> Nothing has changed that dramatically yet. The blog post holds for the majority of everything.[15:21:47] <ratchet_freak> also the driver is allowed to do a lot of optimization so it's very likely to find the fixed function bits it can use[15:33:23] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[15:38:44] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[15:45:21] *** Ralith__ <Ralith__!~ralith@c-24-143-116-108.customer.broadstripe.net> has quit IRC (Ping timeout: 240 seconds)[15:47:01] *** Ralith_ <Ralith_!~ralith@c-24-143-116-108.customer.broadstripe.net> has joined ##vulkan[15:57:36] <HZun> ok thanks for the help :)[16:13:43] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[16:21:59] *** cramalho <cramalho!~cramalho@2804:54:16f5:8200:bd67:9a34:7fbd:ea22> has quit IRC (Remote host closed the connection)[16:24:21] *** Yamakaja <Yamakaja!~yamakaja@vps.pub.yamakaja.me> has quit IRC (Quit: Bye)[16:26:18] *** Yamakaja <Yamakaja!~yamakaja@vps.pub.yamakaja.me> has joined ##vulkan[16:28:53] *** cramalho <cramalho!~cramalho@179.190.254.211> has joined ##vulkan[16:32:44] *** HZun <HZun!~HZun@0x3ec72d49.osd.customer.dk.telia.net> has quit IRC (Quit: Leaving)[16:35:36] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has joined ##vulkan[16:42:22] *** Guest4630 <Guest4630!4d3b9548@gateway/web/cgi-irc/kiwiirc.com/ip.77.59.149.72> has quit IRC (Ping timeout: 264 seconds)[16:42:36] *** penguin42 <penguin42!~dg@cpc109017-salf6-2-0-cust428.10-2.cable.virginm.net> has joined ##vulkan[16:44:55] *** MrFlibble <MrFlibble!MrFlibble@2.122.47.217> has joined ##vulkan[16:50:27] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has quit IRC (Ping timeout: 240 seconds)[16:53:04] *** nsf <nsf!~nsf@jiss.convex.ru> has quit IRC (Quit: WeeChat 2.1)[17:11:11] *** daxar3 <daxar3!~daxar@62-46-171-112.adsl.highway.telekom.at> has joined ##vulkan[17:14:05] *** daxar2 <daxar2!~daxar@193-81-39-231.adsl.highway.telekom.at> has quit IRC (Ping timeout: 240 seconds)[17:29:03] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[17:36:17] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has quit IRC (Quit: This computer has gone to sleep)[17:37:25] *** cheakoirccloud <cheakoirccloud!uid293319@gateway/web/irccloud.com/x-picfmupkxngcrchq> has joined ##vulkan[17:45:26] *** Mazon <Mazon!~Mazon@37.205.127.168> has quit IRC (Ping timeout: 268 seconds)[17:45:34] *** sla_ro|master <sla_ro|master!~sla.ro@78.96.209.89> has quit IRC ()[17:45:52] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has joined ##vulkan[17:47:17] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[17:53:45] *** Mazon <Mazon!~Mazon@37.205.127.168> has joined ##vulkan[18:17:52] *** Deluxe_ <Deluxe_!~Deluxe@212.4.150.151> has joined ##vulkan[18:24:54] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has quit IRC (Quit: WeeChat 2.1)[18:38:13] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan[18:38:51] *** Deluxe_ <Deluxe_!~Deluxe@212.4.150.151> has quit IRC (Remote host closed the connection)[18:41:46] *** Deluxe_ <Deluxe_!~Deluxe@212.4.150.151> has joined ##vulkan[18:43:25] *** Deluxe_ <Deluxe_!~Deluxe@212.4.150.151> has quit IRC (Remote host closed the connection)[18:46:34] *** xissburg <xissburg!~xissburg@unaffiliated/xissburg> has joined ##vulkan[19:30:25] *** cramalho <cramalho!~cramalho@179.190.254.211> has quit IRC (Quit: cramalho)[19:30:52] *** cramalho <cramalho!~cramalho@179.190.254.211> has joined ##vulkan[19:30:59] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has quit IRC (Read error: Connection reset by peer)[19:34:54] *** cramalho <cramalho!~cramalho@179.190.254.211> has quit IRC (Client Quit)[19:35:20] *** cramalho <cramalho!~cramalho@179.190.254.211> has joined ##vulkan[19:44:18] *** Fats <Fats!fats@gateway/vpn/privateinternetaccess/fats> has quit IRC (Remote host closed the connection)[19:45:33] *** Fats <Fats!fats@gateway/vpn/privateinternetaccess/fats> has joined ##vulkan[19:48:56] *** sla_ro|master <sla_ro|master!~sla.ro@78.96.209.89> has joined ##vulkan[19:54:48] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has joined ##vulkan[20:02:59] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has joined ##vulkan[20:13:09] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has quit IRC (Remote host closed the connection)[20:17:30] *** nsf <nsf!~nsf@jiss.convex.ru> has joined ##vulkan[20:29:27] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[20:30:54] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[20:47:21] *** herbmillerjr <herbmillerjr!~herb@c-76-120-173-141.hsd1.pa.comcast.net> has quit IRC (Quit: Konversation terminated!)[20:54:52] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has joined ##vulkan[20:56:49] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[20:58:55] *** psychicist__ <psychicist__!~psychicis@5356A22B.cm-6-7c.dynamic.ziggo.nl> has quit IRC (Quit: leaving)[21:05:04] *** ector <ector!~asdf@ua-85-224-236-175.bbcust.telenor.se> has joined ##vulkan[21:11:14] *** glYoda <glYoda!~MTLYoda@c-73-25-27-206.hsd1.or.comcast.net> has quit IRC (Quit: glYoda)[21:12:29] *** glYoda <glYoda!~MTLYoda@c-73-25-27-206.hsd1.or.comcast.net> has joined ##vulkan[21:14:33] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[21:22:19] *** herbmillerjr <herbmillerjr!~herb@c-76-120-173-141.hsd1.pa.comcast.net> has joined ##vulkan[21:24:22] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has quit IRC (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)[21:28:11] *** penguin42 <penguin42!~dg@cpc109017-salf6-2-0-cust428.10-2.cable.virginm.net> has quit IRC (Remote host closed the connection)[21:50:09] *** borkr <borkr!~borkr@static130-244.mimer.net> has quit IRC (Ping timeout: 264 seconds)[22:13:37] *** salamanderrake <salamanderrake!~quassel@cpe-24-165-202-54.neo.res.rr.com> has quit IRC (Remote host closed the connection)[22:23:50] *** salamanderrake <salamanderrake!~quassel@cpe-24-165-202-54.neo.res.rr.com> has joined ##vulkan[22:24:47] *** salamanderrake <salamanderrake!~quassel@cpe-24-165-202-54.neo.res.rr.com> has quit IRC (Remote host closed the connection)[22:27:30] *** davr0s <davr0s!~textual@host86-157-70-142.range86-157.btcentralplus.com> has joined ##vulkan[22:38:43] *** ImQ009 <ImQ009!~ImQ009@unaffiliated/imq009> has quit IRC (Quit: Leaving)[22:44:17] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has quit IRC (Remote host closed the connection)[22:47:11] *** borkr <borkr!~borkr@static130-244.mimer.net> has joined ##vulkan[23:04:03] *** HZun <HZun!~HZun@0x3ec72d49.osd.customer.dk.telia.net> has joined ##vulkan[23:05:38] *** sla_ro|master <sla_ro|master!~sla.ro@78.96.209.89> has quit IRC ()[23:06:20] <HZun> Do Vulkan support parallelism similar to NVIDIA MPS (multi-process service) where different applications/processes can execute GPU in parallel as opposed to only being time-sliced?[23:06:35] <HZun> on the GPU*[23:06:42] <ratchet_freak> there are multiple queues you can dispatch to[23:10:01] *** nsf <nsf!~nsf@jiss.convex.ru> has quit IRC (Quit: WeeChat 2.1)[23:13:44] *** ector <ector!~asdf@ua-85-224-236-175.bbcust.telenor.se> has quit IRC ()[23:13:46] <HZun> yes. but are the queues only parallel within the same "context"? for example is it scenario 1: where all the queues "belong" to application A, and then all the queues "belong" to application B such that their multi-tasking is timesliced. or is it scenario 2: where the application A can use queue-1 while application B uses queue-2, etc?[23:14:42] <HZun> similar to how CUDA without MPS only allows concurrency within the same context.[23:15:04] <HZun> even though that you might have multiple streams.[23:15:06] <ratchet_freak> there is only 1 device so at some point stuff has to be timesliced[23:16:42] <HZun> So i guess my question is if shader-A from application-A can execute on SM-A while shader-B from application-B executes on SM-B[23:16:46] <HZun> if that makes sense.[23:16:59] <ratchet_freak> that depends on the driver[23:17:08] <ratchet_freak> though I expect them to be smart about that[23:17:41] <HZun> for example, neither opengl or cuda (without MPS) supports that[23:18:16] <ratchet_freak> so there should be proper dispatching logic to load balance[23:19:56] <HZun> yeah so i guess my question is if any Vulkan drivers you guys know support this parallelism yet.[23:20:03] <sharpneli> For caching purposes it generally doesn't make sense to split the work. It's likely faster if it schedules first the work from A and then from B completely[23:21:09] <HZun> yeah. However i am thinking primarily about a scenario where a user is running multiple windowed applications simultaneously. such as a web browser in a windowed mode side by side with a video game in windowed mode.[23:21:59] <HZun> though if the caches are not flushed upon a context switch, then splitting the work should be fine?[23:22:10] <sharpneli> In WDDM iirc the scheduling goes by drawcall granurality. So there isn't any simultaneous scheduling[23:22:30] <sharpneli> Why run two drawcalls in parallel if it's slower than run then sequentially?[23:22:31] <ratchet_freak> but that isn't that much different from having multiple applications with opengl and/or D3D[23:23:14] <sharpneli> Yup[23:25:02] <HZun> but for example, WDDM has the denial-of-service watchdog that kicks in if a kernel/shaders runs for too long (even if the kernel only uses a single SM).[23:25:29] <sharpneli> Sure. It's few seconds or whatever it was[23:25:53] <sharpneli> Having less than 1 fps is kinda bad for any graphical application[23:25:54] <HZun> so i am trying to determine if that is something that would be solvable on a linux desktop environment by using vulkan (if vulkan supports that kind of parallelism)[23:26:26] <sharpneli> What is the problem actually? Programs running too fast or?[23:27:18] <sharpneli> Some new gpu's are going to have pre-emption support so that long running programs can simply be swapped out if needed.[23:27:43] <HZun> wanting to have support for long-running kernels while simultaneously having support for high-fps low-latency graphics in a different application/context/process.[23:28:30] <HZun> preferable without doing any flushing of registers or caches.[23:29:35] <sharpneli> Vulkan doesn't help with that as it requires HW support[23:29:49] <sharpneli> And new gpu's are going to have it[23:30:00] <ratchet_freak> and you are left with drivers doing proper timesharing[23:30:06] <HZun> yeah i read about pascal gpus having support for fine-grained pre-emption. however i am unsure about if that is only pre-emption within the same application (for example an application can pre-empt one of its one kernels only, as opposed to pre-empting the kernel of another application).[23:30:29] <sharpneli> You cannot pre-empt it. Driver can[23:30:35] <sharpneli> It's tool for timesharing for the driver.[23:35:11] <HZun> it was more of an example. like is the driver able to replace the pre-emptied kernel with a kernel from a different application. or is the new kernel always from within the same "context".[23:35:15] <HZun> if that makes sense.[23:36:09] <HZun> a solution that only supports Vulkan capable GPUs would of course be fine.[23:36:44] <HZun> so when you talk about that new GPUs are going to have support for it, which gpus are you talking about?[23:37:34] *** ville <ville!~ville@87-95-249-90.bb.dnainternet.fi> has quit IRC (Ping timeout: 264 seconds)[23:38:38] *** ville <ville!~ville@37-136-13-131.rev.dnainternet.fi> has joined ##vulkan[23:38:38] <HZun> Anyway, my question still stands: Do you guys know of any Vulkan driver + GPU combination that supports running shader-A from application-A on SM-A while running shader-B from application-B on SM-B?[23:39:46] <sharpneli> Depends. If AMD handles their scheduling by queues like they do in OpenCL then they do it like that[23:39:49] <sharpneli> At least for compute[23:39:57] *** borkr <borkr!~borkr@static130-244.mimer.net> has quit IRC (Ping timeout: 240 seconds)[23:40:21] <sharpneli> But not surprisingly the amount of time given to each kernel is undefined.[23:40:31] <sharpneli> What would you gain from that?[23:40:52] <sharpneli> And yes. The driver is able to replace the kernel from whatever it has coming[23:46:29] *** watered <watered!~dagger@gateway/tor-sasl/watered> has quit IRC (Ping timeout: 250 seconds)[23:46:35] <HZun> The gain would be if shader-A is long-running it does not stall shader-B which might be very short in duration. and that if shader-A only uses a single SM and shader-B also only uses a single SM, and the GPU has more than two SMs, then pre-emption would be an unneccesary non-performant way of archieving the concurrency.[23:49:01] <sharpneli> Or likely way more performant[23:49:08] <sharpneli> As all SM's share same memory bus[23:50:30] <HZun> you are arguing that pre-emption would be way more performant?[23:50:35] <sharpneli> But if a shader only uses single SM and another only single it's already done like that[23:50:38] <sharpneli> They run at the same time[23:50:51] *** salamanderrake <salamanderrake!~quassel@cpe-24-165-202-54.neo.res.rr.com> has joined ##vulkan[23:51:21] <sharpneli> In case of both shaders being able to fill all SM's pre-emption is likely more performant due to caching issues[23:52:11] <HZun> hmmm, can you elaborate?[23:52:46] <HZun> the same i see it, then pre-emption would flush the registers and the shared memory. wheras spatial-multitasking would not.[23:52:48] <sharpneli> In general on graphics the memory accesses by instances are coherent. They want same memory or access it in linear fashion[23:53:03] <sharpneli> Split into two and you get essentially random accesses.[23:53:16] <sharpneli> Sure if the granularity is too small[23:53:40] <HZun> the way i see it*[23:54:58] <sharpneli> In the end one must test what works the best. And that's what the driver guys are doing[23:55:07] <sharpneli> For api user it makes no difference.[23:56:54] *** Quetzal2 <Quetzal2!~Quetzal2@unaffiliated/quetzal2> has quit IRC (Quit: ?? Bye!)[23:57:10] <HZun> I agree. However it is worth noting if your are using an API that CAN support spatial multi-tasking if the driver developers perceive it as being the most performant vs. using an API that will probably never support (such as opengl without extensions, etc)[23:57:36] *** watered <watered!~dagger@gateway/tor-sasl/watered> has joined ##vulkan