Commit Graph

  • 8847278ad1 tensor: Instantiate actual DPU Hansung Kim 2024-10-17 14:37:33 -07:00
  • e1e3ac8274 tensor: Fix busy state Hansung Kim 2024-10-16 22:22:27 -07:00
  • 23edc34c7e tensor: Add two TLRAM config for full throughput test Hansung Kim 2024-10-16 22:15:35 -07:00
  • 6cad8edd18 tensor: Fix operand alignment in pipelining Hansung Kim 2024-10-16 22:01:02 -07:00
  • 77dae3e1f9 tensor: Write staging pipeline for A tile Hansung Kim 2024-10-16 21:21:48 -07:00
  • 9e86007e90 add imp component to rad smem, add core serialized access, update 2p smem Richard Yan 2024-10-16 16:20:58 -07:00
  • 444dd5d7e1 tensor: Add destination reg to IO Hansung Kim 2024-10-16 14:25:38 -07:00
  • e2abe1cffd tensor: Sequence set/steps in the execute-side Hansung Kim 2024-10-15 19:12:15 -07:00
  • efaf599fbe tensor: Assert alignment of A and B response queues Hansung Kim 2024-10-15 17:08:14 -07:00
  • de393115cd tensor: Translate TL response source to set/step tag Hansung Kim 2024-10-15 16:48:39 -07:00
  • 2ca2ee37b0 tensor: Fix writeback datawidth Hansung Kim 2024-10-15 15:45:59 -07:00
  • ab8d3554bb Bump vortex to tensor-decoupled Hansung Kim 2024-10-15 15:45:52 -07:00
  • 90949f488b tensor: Add memory response queue Hansung Kim 2024-10-14 22:34:11 -07:00
  • 8d2e13b4ee tensor: Hold step until req fired for both A and B Hansung Kim 2024-10-14 22:06:58 -07:00
  • 14a640bf2d tensor: Do proper source generation Hansung Kim 2024-10-14 21:38:54 -07:00
  • bf6f7210b7 tensor: Generate TL traffic, separate edges for A and B Hansung Kim 2024-10-14 21:14:51 -07:00
  • 9ac8f2492c tensor: Minimal diplomacy config for unittest Hansung Kim 2024-10-14 20:54:24 -07:00
  • 01f53a8be1 tensor: Sequence through set/steps Hansung Kim 2024-10-14 20:20:30 -07:00
  • 3165108c8b Add synthesizable unit test for tensor Hansung Kim 2024-10-14 19:47:00 -07:00
  • 327615e330 Add state regs and init/writeback transition Hansung Kim 2024-10-14 17:28:51 -07:00
  • 6a3aa549d3 Add skeleton for Hopper Tensor Core Hansung Kim 2024-10-14 15:02:25 -07:00
  • 447977bd89 addResource hopper tensor core Hansung Kim 2024-10-14 15:02:08 -07:00
  • 0989d90dd2 connect tc nodes and maybe fix distributor node Richard Yan 2024-10-07 02:59:06 -07:00
  • 4f057c6994 Merge branch 'main' of https://github.com/ucb-bar/radiance into main Richard Yan 2024-10-05 02:48:48 -07:00
  • c6df484c00 add tensor core read client Richard Yan 2024-10-05 02:48:47 -07:00
  • b4bd9ecbc9 Dummy comment Hansung Kim 2024-09-12 13:44:04 -07:00
  • 2929a84ecc get smem params from key Richard Yan 2024-09-26 16:49:06 -07:00
  • f11385218f move virgo components into shared mem module, more cleanup Richard Yan 2024-09-26 14:41:46 -07:00
  • 998f73b54a general cleanup Richard Yan 2024-09-24 18:17:00 -07:00
  • 3b8c9812b4 refactor smem counter Richard Yan 2024-09-24 17:24:52 -07:00
  • 85336399c2 refactor radiance cluster shared memory into components Richard Yan 2024-09-24 03:14:32 -07:00
  • 20cf4609b7 camelCase Richard Yan 2024-09-22 01:21:37 -07:00
  • daacae9edc fallback for hint select Richard Yan 2024-09-11 15:09:52 -07:00
  • f1a1b77828 actually support large smem subbanks Richard Yan 2024-09-10 23:24:02 -07:00
  • 13142ab0b9 Merge branch 'main' of https://github.com/ucb-bar/radiance into main Richard Yan 2024-09-10 18:30:54 -07:00
  • 810db6a1ea new crossbar w/ individual select and group hint, subbanks > num lanes support Richard Yan 2024-09-10 18:30:48 -07:00
  • b335132c34 Parameterize tensor core FP16 Hansung Kim 2024-09-10 15:37:23 -07:00
  • 4b031d1ade radiance.mk: Remove SMEM_LOG_SIZE override Hansung Kim 2024-09-07 18:10:09 -07:00
  • 3fd0fd296b queued cisc commands Richard Yan 2024-09-09 22:38:16 -07:00
  • 06edba2a78 fix comb loop & revert xbar temporarily Richard Yan 2024-09-09 02:27:08 -07:00
  • afc6ba7eca fix ext policy xbar, add rectangular tile support Richard Yan 2024-09-08 13:21:31 -07:00
  • 378b3531d4 balanced shared memory across cores Richard Yan 2024-09-07 20:29:27 -07:00
  • 84972181a5 large smem size, fix single gemmini, bump vortex Richard Yan 2024-09-05 16:50:03 -07:00
  • 24df14d7af Bump vortex Hansung Kim 2024-08-28 16:23:49 -07:00
  • e31f25b432 Switch to FP32 tensor core for use in flash Hansung Kim 2024-08-28 16:23:23 -07:00
  • ec0c8750d3 Bump vortex Hansung Kim 2024-08-20 14:47:18 -07:00
  • 2364cd213e Bump vortex Hansung Kim 2024-08-15 13:40:09 -07:00
  • 2e6221661a radiance.mk: Reenable LSU_DUP_DISABLE Hansung Kim 2024-08-15 13:38:02 -07:00
  • d8823a0416 Add back generated verilog for FP32 TensorDPU Hansung Kim 2024-08-12 18:48:21 -07:00
  • 7b06c1778c Bump vortex Hansung Kim 2024-08-07 11:30:09 -07:00
  • c1d95ff205 Revert rename Hansung Kim 2024-08-07 11:29:42 -07:00
  • 477f3955ed Update generated SV for tensordpu Hansung Kim 2024-08-07 11:08:27 -07:00
  • 32c7aed263 Fix fp exception by rounding right after MulRawFN Hansung Kim 2024-08-06 18:17:42 -07:00
  • b7a342fcf6 bump vortex Richard Yan 2024-08-06 02:33:48 -07:00
  • b7239917c0 temporary fix to rename fp units Richard Yan 2024-08-06 02:33:13 -07:00
  • af60ae3332 fp16 gemmini support Richard Yan 2024-08-06 02:32:35 -07:00
  • 0d9c2ca6ad rename TensorDPU.scala Richard Yan 2024-08-06 02:30:37 -07:00
  • d8674b753e Support fp16 input, fp32 output in TensorDPU Hansung Kim 2024-08-01 21:59:41 -07:00
  • ee79fdaa05 Fix typetag for half Hansung Kim 2024-07-31 17:17:03 -07:00
  • 4e3ed96c94 Support fp16 operand/accum in TensorDPU Hansung Kim 2024-07-31 16:37:01 -07:00
  • 33aa4e08db Bump vortex Hansung Kim 2024-07-26 16:27:40 -07:00
  • 42285c7193 Instantiate StallingPipe as a separate module Hansung Kim 2024-07-25 16:27:20 -07:00
  • 79604f51be Fix possible CIRCT bug on SourceGenerator Hansung Kim 2024-07-23 15:09:03 -07:00
  • a375da16a6 radiance.mk: Fix source lookup logic when fd not available Hansung Kim 2024-07-23 15:05:43 -07:00
  • 4566f954ee Update doc; move code out of copypaste block Hansung Kim 2024-07-23 15:04:47 -07:00
  • 37fd0c6200 Bump vortex Hansung Kim 2024-07-19 16:39:32 -07:00
  • 21baeae758 Separate read and write counter for smem Hansung Kim 2024-07-02 14:52:58 -07:00
  • 7aad800a2d Fix module imports after rocket-chip bump Hansung Kim 2024-06-27 17:17:27 -07:00
  • e1342e431c Bump radiance Hansung Kim 2024-06-26 14:27:59 -07:00
  • 018167480d Add access counter to smem banks Hansung Kim 2024-06-26 14:27:49 -07:00
  • a66a8c1732 Remove NUM_FPU_BLOCKS=2 from radiance.mk Hansung Kim 2024-06-15 17:54:16 -07:00
  • 74cd633246 Merge branch 'main' of https://github.com/ucb-bar/radiance into main Richard Yan 2024-06-20 01:45:18 -07:00
  • 5eb581c7e5 different sized gemminis Richard Yan 2024-06-20 01:45:14 -07:00
  • 5a79d7ff79 Add new opcodes for tensor-core DMA Hansung Kim 2024-06-12 09:52:18 -07:00
  • 5b4d34864f Merge branch 'main' of https://github.com/ucb-bar/radiance into main Richard Yan 2024-06-12 02:24:20 -07:00
  • 19852693b7 dual core gemmini, unpeg gemmini size and smem width Richard Yan 2024-06-12 02:17:40 -07:00
  • 1401c4a090 Separate out core id from tile id in TileParams Hansung Kim 2024-06-11 17:13:51 -07:00
  • 7ced63bd62 Remove clbus definition from RadianceCluster Hansung Kim 2024-06-11 16:23:07 -07:00
  • a0bff40a87 Set correct static tileId for core and Gemmini tiles Hansung Kim 2024-06-11 16:17:08 -07:00
  • a8e35b1f5a oopsie doopsie Richard Yan 2024-06-09 15:35:11 -07:00
  • 17756d5f53 Merge branch 'main' of https://github.com/ucb-bar/radiance into main Richard Yan 2024-06-09 15:26:07 -07:00
  • 3badd75473 framebuffer, shared memory key, etc Richard Yan 2024-06-09 15:05:31 -07:00
  • ca3fd8b515 Bump vortex with doc changes Hansung Kim 2024-06-09 13:41:18 -07:00
  • 98889ebad9 radiance.mk: Change back to NUM_CORES=2 Hansung Kim 2024-06-09 13:39:59 -07:00
  • 3254ac3741 Doc changes Hansung Kim 2024-06-09 13:39:53 -07:00
  • 3e4a04351b Enable GPR_DUPLICATED for simulation Hansung Kim 2024-06-03 13:13:15 -07:00
  • c7e2cd2387 Bump vortex with dispatch_unit fix Hansung Kim 2024-05-30 22:00:10 -07:00
  • 503a3ff246 Bump vortex Hansung Kim 2024-05-30 18:33:56 -07:00
  • 3b08d5ca70 radiance.mk: Remove debug rules Hansung Kim 2024-05-30 18:14:14 -07:00
  • 091115bb90 Bump vortex Hansung Kim 2024-05-29 17:06:27 -07:00
  • 2c196bb9a0 Update generated dpu verilog for stalls Hansung Kim 2024-05-29 16:22:24 -07:00
  • 44bcca9319 radiance.mk: Differentiate verilog copy target by CONFIG Hansung Kim 2024-05-29 16:21:40 -07:00
  • 17886dc050 Implement proper stalls for dpu Hansung Kim 2024-05-29 16:21:12 -07:00
  • ae573aeea1 radiance.mk: Copy over when content differs, not just newer Hansung Kim 2024-05-29 15:03:44 -07:00
  • 0f36e86f63 Change config to 1core/8warp/8thread Hansung Kim 2024-05-29 13:36:03 -07:00
  • 8e79e620cb Bump vortex Hansung Kim 2024-05-29 13:35:43 -07:00
  • d34c5836a8 Add chisel-generated verilog for dpu Hansung Kim 2024-05-29 13:33:23 -07:00
  • 4a43d0126d Make dpu 2-stage Hansung Kim 2024-05-29 13:31:38 -07:00
  • 8dd3994012 radiance.mk: Don't fail when fd not installed Hansung Kim 2024-05-29 00:56:44 -07:00
  • 98618d1483 radiance.mk: Add target that copies Vortex verilog sources Hansung Kim 2024-05-29 00:53:05 -07:00