July 3, 2019 
< | 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 | 31 | >
July 3, 2019 [00:04:48] *** takahe <takahe!~raul@149.36.252.145> has quit IRC (Quit: Saliendo)[00:11:48] <Illya> cjay-: oh right. I thought it was a child of a framebuffer[00:12:18] <Illya> VK_EXT_debug_utils looks interesting I'll check it out[00:23:40] *** gaulishcoin <gaulishcoin!~gaulishco@anice-652-1-21-234.w83-201.abo.wanadoo.fr> has quit IRC (Quit: Leaving)[00:58:22] *** felipealmeida <felipealmeida!~user@2804:14d:badc:2b50::1000> has quit IRC (Remote host closed the connection)[01:23:42] *** ratchetfreak <ratchetfreak!~ratchetfr@d51A4BEAF.access.telenet.be> has quit IRC (Ping timeout: 268 seconds)[01:51:16] *** LunarJetman <LunarJetman!LunarJetma@94.14.216.120> has quit IRC (Ping timeout: 244 seconds)[01:53:00] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has joined ##vulkan[02:38:17] *** wildlander <wildlander!~wildlande@unaffiliated/wildlander> has quit IRC (Ping timeout: 244 seconds)[02:50:36] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has quit IRC (Ping timeout: 272 seconds)[03:23:03] *** saint_go <saint_go!~textual@135-23-210-183.cpe.pppoe.ca> has joined ##vulkan[03:29:10] *** Andrej1 <Andrej1!~Andrej@BSN-143-210-205.dynamic.siol.net> has quit IRC (Read error: Connection reset by peer)[04:04:04] *** mirv_ <mirv_!~mirv@xoreos/mirv> has quit IRC (Ping timeout: 272 seconds)[04:36:52] *** penguin42 <penguin42!~dg@cpc109017-salf6-2-0-cust428.10-2.cable.virginm.net> has quit IRC (Remote host closed the connection)[04:37:01] *** mirv_ <mirv_!~mirv@xoreos/mirv> has joined ##vulkan[04:48:00] *** Snoras <Snoras!~Snoras@gateway/tor-sasl/snoras> has quit IRC (Ping timeout: 260 seconds)[04:49:22] *** Snoras <Snoras!~Snoras@gateway/tor-sasl/snoras> has joined ##vulkan[05:08:29] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has quit IRC (Quit: This computer has gone to sleep)[05:20:35] *** Illya <Illya!~Illya@ffmpeg/developer/january> has quit IRC (Quit: ZNC - https://znc.in)[05:20:48] *** Illya <Illya!~Illya@ffmpeg/developer/january> has joined ##vulkan[05:45:22] *** saint_go <saint_go!~textual@135-23-210-183.cpe.pppoe.ca> has quit IRC (Quit: Textual IRC Client: www.textualapp.com)[06:41:42] *** halbeno <halbeno!~halbeno@d50-93-37-115.abhsia.telus.net> has quit IRC (Ping timeout: 245 seconds)[06:42:01] *** halbeno <halbeno!~halbeno@d50-93-37-115.abhsia.telus.net> has joined ##vulkan[08:27:45] *** early` <early`!~early@static.38.6.217.95.clients.your-server.de> has quit IRC (Ping timeout: 248 seconds)[08:29:13] *** early <early!~early@static.38.6.217.95.clients.your-server.de> has joined ##vulkan[09:59:26] *** spossiba <spossiba!~spossiba@unaffiliated/stryx/x-3871776> has quit IRC (Read error: Connection reset by peer)[10:00:27] *** spossiba <spossiba!~spossiba@unaffiliated/stryx/x-3871776> has joined ##vulkan[10:01:56] *** ratchetfreak <ratchetfreak!c351a8d8@195.81.168.216> has joined ##vulkan[11:21:53] *** pa <pa!~pa@unaffiliated/pa> has quit IRC (Ping timeout: 258 seconds)[11:32:15] *** pa <pa!~pa@unaffiliated/pa> has joined ##vulkan[11:52:57] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has joined ##vulkan[12:18:29] *** [exa] <[exa]!exa@unaffiliated/exa/x-5381537> has joined ##vulkan[12:19:36] <[exa]> Hi everyone! Are there some guarantees on how subgroups (aka warps) are formed in local workgroups?[12:22:32] <[exa]> especially in content of parallel reductions; I'd like to use subgroupAdd for summing up columns of a matrix, but with slightly fatter workgroups spanning multiple columns (to reduce cache load). The question is whether there is a simple way to guarantee that things from more columns will not add together.[12:50:09] <bnieuwenhuizen> [exa]: I don't think there is[12:51:21] <bnieuwenhuizen> also IIRC AMD does it as in a row-major array (i.e. horizontal first) which sounds like it is exactly what you don't want?[12:54:20] *** Andrej1 <Andrej1!~Andrej@BSN-143-210-205.dynamic.siol.net> has joined ##vulkan[12:59:54] <[exa]> bnieuwenhuizen: I wouldn't like to assume that kind of stuff :][13:00:54] <bnieuwenhuizen> yeah, my second comments was more like "you cannot assume this because some HW does the wrong thing for you"[13:01:07] <[exa]> but yes, all machines I have here do it this way[13:01:36] <[exa]> I was kindof wondering about whether there would be some simple kind of workaround[13:01:39] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has joined ##vulkan[13:02:22] <bnieuwenhuizen> you probably can get the subgroup index and the invocation index in the subgroup and do your own numbering?[13:02:47] <bnieuwenhuizen> just be aware that subgroup size = 1 is perfectly valid for a vulkan driver[13:03:24] <[exa]> yeah[13:05:14] <[exa]> btw is there some standardized docs about "how to get shared memory" ?[13:05:28] <[exa]> I didn't find anything that would look official[13:06:00] <bnieuwenhuizen> what do you mean "how to get shared memory?"[13:06:02] <[exa]> workgroup-shared memory a.k.a. __shared a.k.a. LDS[13:06:11] <[exa]> the 48kB of L1-like cache[13:06:36] <[exa]> that's funniest, I don't even know what name to put into google. :][13:06:37] <bnieuwenhuizen> in your shader just declare global variables in the workgroup storageclass (shared in GLSL)[13:07:13] <[exa]> these are automagically shared?[13:07:47] <bnieuwenhuizen> yup, these are atuomatically shared between invocations in a workgroup[13:08:35] <bnieuwenhuizen> and in the physicaldevicelimits there is a maxComputeSharedMemorySize field to say how much you can use per workgroup[13:11:30] <[exa]> ok that's interesting[13:12:17] <[exa]> I guess I could use clustered add[13:20:57] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has quit IRC (Quit: This computer has gone to sleep)[13:33:13] *** ratchetfreak <ratchetfreak!c351a8d8@195.81.168.216> has quit IRC (Remote host closed the connection)[13:40:56] *** ratchetfreak <ratchetfreak!c351a8d8@195.81.168.216> has joined ##vulkan[13:42:43] *** ville <ville!~ville@212-149-214-47.bb.dnainternet.fi> has quit IRC (Ping timeout: 244 seconds)[13:42:49] *** ryp <ryp!ryp@gateway/vpn/privateinternetaccess/ryp> has joined ##vulkan[13:46:24] *** drewbarbs <drewbarbs!~drewbarbs@2601:14a:700:eaa1:b4d1:3650:4556:2d71> has quit IRC (Quit: ZNC 1.6.2 - http://znc.in)[13:47:27] *** drewbarbs <drewbarbs!~drewbarbs@2601:14a:700:eaa1:b4d1:3650:4556:2d71> has joined ##vulkan[13:51:20] *** RicardasSim <RicardasSim!~RicardasS@78-63-172-44.static.zebra.lt> has quit IRC (Quit: Leaving)[13:52:42] *** penguin42 <penguin42!~dg@cpc109017-salf6-2-0-cust428.10-2.cable.virginm.net> has joined ##vulkan[14:47:35] *** ville <ville!~ville@62-113-181-125.bb.dnainternet.fi> has joined ##vulkan[15:03:43] *** ville <ville!~ville@62-113-181-125.bb.dnainternet.fi> has quit IRC (Ping timeout: 246 seconds)[16:09:21] *** takahe <takahe!~raul@149.36.252.145> has joined ##vulkan[16:36:26] *** ville <ville!~ville@188-67-23-147.bb.dnainternet.fi> has joined ##vulkan[16:47:22] *** borkr <borkr!~borkr@static243-215-65.mimer.net> has joined ##vulkan[17:08:26] *** ville <ville!~ville@188-67-23-147.bb.dnainternet.fi> has quit IRC (Ping timeout: 268 seconds)[17:10:12] *** RicardasSim <RicardasSim!~RicardasS@78-63-172-44.static.zebra.lt> has joined ##vulkan[17:24:59] *** ImQ009 <ImQ009!~ImQ009@unaffiliated/imq009> has joined ##vulkan[17:43:41] *** mefesto <mefesto!~user@107.145.49.162> has joined ##vulkan[17:58:41] *** ville <ville!~ville@212-149-214-47.bb.dnainternet.fi> has joined ##vulkan[18:00:51] *** takahe <takahe!~raul@149.36.252.145> has quit IRC (Quit: Saliendo)[18:17:03] *** ratchetfreak <ratchetfreak!c351a8d8@195.81.168.216> has quit IRC (Ping timeout: 260 seconds)[18:44:05] *** LunarJetman <LunarJetman!LunarJetma@94.14.216.120> has joined ##vulkan[18:56:01] *** ratchetfreak <ratchetfreak!~ratchetfr@d51A4BEAF.access.telenet.be> has joined ##vulkan[20:08:43] *** RicardasSim <RicardasSim!~RicardasS@78-63-172-44.static.zebra.lt> has quit IRC (Quit: Leaving)[21:00:16] *** borkr <borkr!~borkr@static243-215-65.mimer.net> has quit IRC (Remote host closed the connection)[21:23:55] *** mefesto <mefesto!~user@107.145.49.162> has quit IRC (Quit: ERC (IRC client for Emacs 26.1))[21:47:36] *** RicardasSim <RicardasSim!~RicardasS@78-63-172-44.static.zebra.lt> has joined ##vulkan[21:56:22] *** ImQ009 <ImQ009!~ImQ009@unaffiliated/imq009> has quit IRC (Quit: Leaving)[22:17:26] *** slime <slime!~slime73@blk-215-81-93.eastlink.ca> has joined ##vulkan[23:38:36] *** Deluxe <Deluxe!~Deluxe@212.4.150.151> has quit IRC (Read error: Connection reset by peer)[23:53:36] *** symb0l <symb0l!~symb0l@dsl-dhcp-katytxxchrc-64-92-54-147.consolidated.net> has joined ##vulkan[23:54:30] <symb0l> This the right place to ask development questions?[23:54:58] <Yaniel> question about developing against the vulkan api, yes[23:56:39] <symb0l> Yes, so I'm for some reason getting a seg fault with the VK_LAYER_KHRONOS_validation layer enabled on a vkCreateDevice call. I've disabled the layer and don't get the problem, and gdb shows the issue happening in the validation layer. Not sure if this is something I'm missing or a known problem?[23:57:57] <Ralith> the most likely explanation is that your code is invoking UB in an unusual way