Zhongdi LUO
9251ba0a24
feat: add scalar fexp support
2026-07-02 07:24:59 +00:00
Zhongdi LUO
97a1eff701
Add scalar TMEM load-store path
2026-06-24 06:24:46 +00:00
Zhongdi LUO
abee301b6e
Support 4-lane Blackwell tensor wrapper
2026-05-27 05:54:24 +00:00
c87fea5c48
Keep masked barriers local under cluster gbar
2026-05-27 09:07:47 +08:00
0ad87bde81
Implement WU architecture support
2026-05-25 19:25:05 +08:00
323ed7d7e9
Update Vortex core for Blackwell tensor instructions
...
- Add Blackwell tensor core support in VX_tensor_blackwell_core.sv
- Update decode, execute, and dispatch logic for new instructions
- Extend VX_define.vh and VX_types.vh with Blackwell ISA definitions
2026-05-06 14:50:54 +08:00
cb912d3b8b
Add Blackwell tensor RTL scaffolding
2026-04-25 10:15:31 +08:00
Hansung Kim
f1d0fac518
Change to 8-core Volta/Ampere config
2025-01-28 22:36:58 -08:00
Hansung Kim
c8529c4339
Disable EXT_T_HOPPER atm for flash runs
2024-11-08 21:52:52 -08:00
Hansung Kim
cf000afc8f
tensor: Remove unused a[2] and a[3] ports for FP32 DPU
2024-11-08 14:34:47 -08:00
Richard Yan
8dc2a25e32
oopsie 3
2024-11-02 14:52:43 -07:00
Richard Yan
d794e055b6
oopsie 2
2024-11-02 14:51:52 -07:00
Richard Yan
ed61418ebf
oopsie
2024-11-01 02:55:38 -07:00
Richard Yan
2e3ea060a5
gate operand read
2024-11-01 02:44:35 -07:00
Hansung Kim
ef902614ff
tensor: Fix race in inflight_tensor counter
2024-10-29 14:14:31 -07:00
Hansung Kim
1013a74abd
tensor: Switch back to hopper + 4 cores
2024-10-28 23:39:39 -07:00
Hansung Kim
19876ab9fd
tensor: Fix wrong writeback bit
2024-10-28 21:47:25 -07:00
Hansung Kim
8a66b5ed89
tensor: Connect SMEM addr/rf IO
2024-10-28 19:42:02 -07:00
Hansung Kim
4376bd33a2
tensor: Decode rs1/rs2 of HGMMA for smem addresses
2024-10-28 19:41:37 -07:00
Hansung Kim
72db04cec0
tensor: Switch to 8cores, non-hopper config
2024-10-27 19:47:22 -07:00
Hansung Kim
3e67ddd6c6
tensor: Properly guard tc_rf_if for non-hopper
2024-10-27 17:55:09 -07:00
Hansung Kim
1bc4afe2bb
tensor: Bore tensor regfile IO to execute units
2024-10-24 20:32:18 -07:00
Hansung Kim
c88fd89f1f
tensor: Don't make initiate_valid depend on ready
2024-10-24 19:29:21 -07:00
Richard Yan
b64e53ff02
Merge branch 'rtl' of github.com:hansungk/vortex-private into rtl
2024-10-24 16:51:22 -07:00
Richard Yan
155cbb0abc
tc rf read port
2024-10-24 16:51:15 -07:00
Hansung Kim
40565de8cd
tensor: Fix initiate sync with meta queue when !commit.ready
2024-10-24 16:41:54 -07:00
Hansung Kim
3ebeb43568
tensor: Fix inflight_tensor decrement, add under/overflow checks
2024-10-24 14:36:29 -07:00
Hansung Kim
8337488ed3
tensor: Don't check invalid writeback reg for ghost writes
2024-10-24 14:36:18 -07:00
Hansung Kim
e855a47295
Add missing commit_if.tensor bit inits
2024-10-24 13:28:30 -07:00
Hansung Kim
c77a25c968
tensor: Add missing HOPPER guard
2024-10-23 20:33:45 -07:00
Hansung Kim
78df981366
tensor: Simply metadata queue
...
Enqueue all different-warp reqs into the queue. There is a slight chance
that an HGMMA_WAIT might be blocked from commit when there are multiple
different-warp HGMMAs blocking the dequeue end, but it should be
uncommon.
2024-10-22 22:01:18 -07:00
Hansung Kim
69cbbdd89b
tensor: Consider inflight ops for HGMMA blocking
...
This allows for back-to-back issue of HGMMA past the scoreboard, which
helps to minimize downtime in DPU activity in-between operations.
HGMMA_WAIT now only unblocks when *all* previous HGMMAs have finished
writeback.
2024-10-22 21:32:33 -07:00
Hansung Kim
98eb7cb594
tensor: Block both HGMMA/HGMMA_WAIT at scoreboard
...
If we let back-to-back HGMMAs pass at scoreboard, we can't accurately
keep track of the busy state of the tensor core and block WAITs
accordingly.
TODO: Distinguish "ready-to-fire" from "ready-to-use-writeback".
2024-10-22 21:10:55 -07:00
Hansung Kim
83979c3341
tensor: Fully connect writeback IO
2024-10-22 20:17:00 -07:00
Hansung Kim
47dff74d3a
tensor: Fix commit/metadata logic for HGMMA
...
Block HGMMA commit until previous ones are all done; always commit
HGMMA_WAIT after it passes the scoreboard.
2024-10-22 20:01:37 -07:00
Hansung Kim
3abaaff16f
tensor: Fix tag and data assignment for p0/p1 bus
2024-10-22 17:47:04 -07:00
Hansung Kim
8a8f682194
tensor: Bore smem IO from core to tensor core
2024-10-22 17:42:30 -07:00
Hansung Kim
9131558950
tensor: Connect Chisel-generated TensorCoreDecoupled module
...
Elaborates, but most of the IOs are tied to fake.
2024-10-22 15:16:24 -07:00
Hansung Kim
32ccdeef01
Merge branch 'tensor-decoupled' into rtl
2024-10-21 22:57:07 -07:00
Hansung Kim
0f06afc3ef
Update doc
2024-10-21 22:37:20 -07:00
Richard Yan
cde8da1f3b
add tag to tc smem interface
2024-10-17 14:48:39 -07:00
Hansung Kim
4dcbc31a88
tensor: Separate async commit from tensor commit
...
With this we can prioritize commit of the async hgmma instructions over
the "ghost" commits from the TC.
2024-10-11 21:32:20 -07:00
Hansung Kim
717fe7ff29
tensor: Fix FSM when commit not ready
2024-10-11 20:24:31 -07:00
Hansung Kim
2934b1bd94
tensor: Split execution module from pipeline logic
2024-10-11 20:09:09 -07:00
Hansung Kim
f7f23e0c05
tensor: Doc update
2024-10-11 18:00:36 -07:00
Hansung Kim
42b9d23f83
tensor: Write release logic for hgmma
...
Upon completion of an op, tensor_core_hopper sends a "ghost" commit
signal down the pipeline with the `wb` and `tensor` bit set in
commit_if. The scoreboard receives this signal via writeback_if and
resets the inuse_tensor status bit back to zero, which unblocks the
HGMMA_WAIT instruction.
2024-10-11 17:58:44 -07:00
Hansung Kim
408a9b5d2a
tensor: Write stall logic for hgmma_wait
...
HGMMA_WAIT instruction stalls at issue when inuse_tensor is set, which
is done by the previous HGMMA insn. Currently inuse_tensor is never set
back to zero.
2024-10-11 17:18:01 -07:00
Hansung Kim
72f9dedce3
tensor: Disable micro-ops for hopper
...
Have an uarch FSM handle the stepping mechanism entirely.
2024-10-11 15:59:31 -07:00
Hansung Kim
100d69ef21
Doc update on accumulator regs
2024-10-11 15:47:58 -07:00
Hansung Kim
d9ad4809ec
Add 'tensor' bit to commit_if and writeback_if
...
For use in the asynchronous tensor instruction. When 1'b1, sets/unsets
the inuse_tensor status bit in the scoreboard to signal
kickoff/completion of the asynchronous tensor op.
2024-10-11 15:42:25 -07:00
