Commit Graph

675 Commits

Author SHA1 Message Date
Hansung Kim
b3c328b1be tensor: Assert minimum response queue depth with doc 2024-10-18 23:11:32 -07:00
Hansung Kim
e946403d78 tensor: Fix typo, reduce resp queue depth 2024-10-18 22:55:00 -07:00
Hansung Kim
0aadc6074a tensor: Decouple A and B access states
Get rid of set/stepAccess states and let A and B access progress
independently.
2024-10-18 22:42:41 -07:00
Hansung Kim
c0292dd0aa tensor: Enlarge operand buffer for A for better SMEM reuse 2024-10-18 21:53:24 -07:00
Hansung Kim
93c9bcc32f tensor: Stage B as well for full throughput 2024-10-18 20:12:15 -07:00
Hansung Kim
c4b5a11fde tensor: Replace staging logic for A with FillBuffer 2024-10-18 19:54:20 -07:00
Hansung Kim
7fab6f89ad tensor: Properly route FillBuffer to DPU 2024-10-18 17:33:55 -07:00
Hansung Kim
91d9897c27 tensor: Write FillBuffer for tile buffering 2024-10-18 17:17:41 -07:00
Hansung Kim
c2f39f7474 tensor: Rename substepExecute 2024-10-18 16:21:43 -07:00
Hansung Kim
64ea48ace3 tensor: Consider data reuse for B memory request
B is reused every 4 steps because of the k->i->j iteration order.
2024-10-18 13:46:04 -07:00
Hansung Kim
a2519da58f tensor: SMEM address generation 2024-10-17 16:36:18 -07:00
Hansung Kim
2741af0b2b tensor: Keep set/step in the tag writeback queue 2024-10-17 15:44:05 -07:00
Hansung Kim
7de8e86d4f tensor: Sync rd with DPU using a queue 2024-10-17 15:18:47 -07:00
Hansung Kim
8847278ad1 tensor: Instantiate actual DPU 2024-10-17 14:44:34 -07:00
Hansung Kim
e1e3ac8274 tensor: Fix busy state 2024-10-16 22:22:27 -07:00
Hansung Kim
23edc34c7e tensor: Add two TLRAM config for full throughput test 2024-10-16 22:15:35 -07:00
Hansung Kim
6cad8edd18 tensor: Fix operand alignment in pipelining 2024-10-16 22:01:02 -07:00
Hansung Kim
77dae3e1f9 tensor: Write staging pipeline for A tile 2024-10-16 21:21:48 -07:00
Hansung Kim
444dd5d7e1 tensor: Add destination reg to IO 2024-10-16 14:25:38 -07:00
Hansung Kim
e2abe1cffd tensor: Sequence set/steps in the execute-side 2024-10-15 19:12:45 -07:00
Hansung Kim
efaf599fbe tensor: Assert alignment of A and B response queues 2024-10-15 17:08:32 -07:00
Hansung Kim
de393115cd tensor: Translate TL response source to set/step tag 2024-10-15 16:50:40 -07:00
Hansung Kim
2ca2ee37b0 tensor: Fix writeback datawidth 2024-10-15 15:45:59 -07:00
Hansung Kim
ab8d3554bb Bump vortex to tensor-decoupled 2024-10-15 15:45:52 -07:00
Hansung Kim
90949f488b tensor: Add memory response queue 2024-10-14 22:36:18 -07:00
Hansung Kim
8d2e13b4ee tensor: Hold step until req fired for both A and B 2024-10-14 22:07:20 -07:00
Hansung Kim
14a640bf2d tensor: Do proper source generation
SourceGenerator keeps on givin'
2024-10-14 21:38:54 -07:00
Hansung Kim
bf6f7210b7 tensor: Generate TL traffic, separate edges for A and B 2024-10-14 21:15:07 -07:00
Hansung Kim
9ac8f2492c tensor: Minimal diplomacy config for unittest 2024-10-14 20:54:33 -07:00
Hansung Kim
01f53a8be1 tensor: Sequence through set/steps 2024-10-14 20:20:30 -07:00
Hansung Kim
3165108c8b Add synthesizable unit test for tensor 2024-10-14 19:47:00 -07:00
Hansung Kim
327615e330 Add state regs and init/writeback transition 2024-10-14 17:28:51 -07:00
Hansung Kim
6a3aa549d3 Add skeleton for Hopper Tensor Core 2024-10-14 15:02:25 -07:00
Hansung Kim
447977bd89 addResource hopper tensor core 2024-10-14 15:02:08 -07:00
Hansung Kim
b4bd9ecbc9 Dummy comment 2024-10-02 15:18:56 -07:00
Richard Yan
2929a84ecc get smem params from key 2024-09-26 16:49:06 -07:00
Richard Yan
f11385218f move virgo components into shared mem module, more cleanup 2024-09-26 14:41:46 -07:00
Richard Yan
998f73b54a general cleanup 2024-09-24 18:17:00 -07:00
Richard Yan
3b8c9812b4 refactor smem counter 2024-09-24 17:24:52 -07:00
Richard Yan
85336399c2 refactor radiance cluster shared memory into components 2024-09-24 03:14:32 -07:00
Richard Yan
20cf4609b7 camelCase 2024-09-22 01:21:37 -07:00
Richard Yan
daacae9edc fallback for hint select 2024-09-11 15:09:52 -07:00
Richard Yan
f1a1b77828 actually support large smem subbanks 2024-09-10 23:24:02 -07:00
Richard Yan
13142ab0b9 Merge branch 'main' of https://github.com/ucb-bar/radiance into main 2024-09-10 18:30:54 -07:00
Richard Yan
810db6a1ea new crossbar w/ individual select and group hint, subbanks > num lanes support 2024-09-10 18:30:48 -07:00
Hansung Kim
b335132c34 Parameterize tensor core FP16 2024-09-10 15:38:12 -07:00
Hansung Kim
4b031d1ade radiance.mk: Remove SMEM_LOG_SIZE override 2024-09-10 15:38:12 -07:00
Richard Yan
3fd0fd296b queued cisc commands 2024-09-09 22:38:16 -07:00
Richard Yan
06edba2a78 fix comb loop & revert xbar temporarily 2024-09-09 02:27:08 -07:00
Richard Yan
afc6ba7eca fix ext policy xbar, add rectangular tile support 2024-09-08 13:21:31 -07:00