26 commits

Author	SHA1	Message	Date
Jason Volk	d377674748	ircd::simt: Split vector reduce_add to hadd.	2023-01-01 19:16:06 -08:00
Jason Volk	561be9973a	ircd::simt::norm: Barrier for overlapping input and output buffers. (gfx1011)	2022-11-03 18:28:53 +00:00
Jason Volk	e89703aa97	ircd::gpt::gpu: Limit dispatch to the number of control frame buffers.	2022-10-30 18:44:50 +00:00
Jason Volk	2609c21913	ircd::gpt::pipe: Enable mutable model; fixes for backpropagation; range stub.	2022-10-18 22:01:35 +00:00
Jason Volk	6a05fcefeb	ircd::simt: Consolidate timestamp counter sampling into inline.	2022-10-12 18:53:53 -07:00
Jason Volk	54e3b8f5b4	ircd::simt: Consolidate portables and macros from units into headers.	2022-10-11 02:13:46 +00:00
Jason Volk	831141727b	ircd::gpt::gpu: Add assert macro when trapping supported.	2022-10-09 03:29:29 +00:00
Jason Volk	c1168fcc30	ircd::gpt: Resolve behavior of opts.limit: 0=analysis, -n=unlimited, n=limited.	2022-10-09 03:13:09 +00:00
Jason Volk	9682f406b3	ircd::gpt::gpu: Mute printf() on unsupporting platforms.	2022-10-06 22:26:19 +00:00
Jason Volk	442dad869d	ircd::gpt: Resolve cycle count sampling; add debug log; fix count.	2022-10-06 22:01:41 +00:00
Jason Volk	0917a1f041	ircd::gpt::pipe: Resolve control page sync at ends of sample.	2022-10-06 18:54:29 +00:00
Jason Volk	33afa8a4fc	ircd::gpt::gpu: Add global fence between attn and ffnn accumulations (gfx1011).	2022-10-05 20:10:31 +00:00
Jason Volk	c4cceb425c	ircd::gpt::gpu: Use explicit broadcast for local access.	2022-10-05 20:08:40 +00:00
Jason Volk	331a417656	ircd::gpt::gpu: Fix keywording for OpenCL 2.0+.	2022-10-02 01:30:10 +00:00
Jason Volk	e85ed0e0dd	ircd::gpt: Remove various cruft.	2022-09-24 16:40:39 -07:00
Jason Volk	6d2da3b4f1	ircd::gpt::task: Refactor generator interface to member functions.	2022-07-01 20:17:56 -07:00
Jason Volk	78848925ee	ircd::gpt: Various refactoring.	2022-06-19 20:14:22 -07:00
Jason Volk	b7b1328352	ircd::gpt::pipe: Reuse logsm buffer for logexp intermediate values.	2022-06-17 21:11:53 -07:00
Jason Volk	8f90e7c0cd	ircd::gpt: Optimizations for matrix multiply.	2021-10-06 13:13:47 -07:00
Jason Volk	aea6c79fc2	ircd::gpt: Add top N and target label result register control block.	2021-10-06 13:13:47 -07:00
Jason Volk	8bd78af128	ircd::gpt: Additional task header/interface simplification.	2021-10-06 13:13:47 -07:00
Jason Volk	c1f3e580c3	ircd::gpt: Add top_p lmhead selector, quantized for now.	2021-10-06 13:13:47 -07:00
Jason Volk	8a3eeb46f9	ircd::gpt::pipe: Optimize pipeline to cache attention state for generations.	2021-10-06 13:13:47 -07:00
Jason Volk	c5f159ad58	ircd::gpt: Cleanup/improve work item related prologues.	2021-10-06 13:13:47 -07:00
Jason Volk	ce9abfb321	ircd::gpt::model: Optimize left-attention mask.	2021-10-06 13:13:47 -07:00
Jason Volk	20162fd7d5	ircd::gpt: Splits and renames; various reorg.	2021-09-15 01:44:36 -07:00

Renamed from ircd/gpt_cl.cl (Browse further)