Commit Graph

385 Commits

Author SHA1 Message Date
PoYen, Chen
2192bbc68a Rename RotaryEmbeddingEnum 2024-07-23 07:50:50 +00:00
PoYen, Chen
48c70720b5 Apply RoPE to q_tile 2024-07-23 03:54:11 +00:00
PoYen, Chen
1dbed18555 Remove constness from q_ptr 2024-07-23 03:11:31 +00:00
PoYen, Chen
631f29d527 Handle RoPE half-rotated logics 2024-07-22 08:50:03 +00:00
PoYen, Chen
fffd6799e6 Instantiate multiple kernels for RoPE approaches 2024-07-20 02:28:21 +00:00
PoYen, Chen
23450526c0 Only apply interleaved RoPE on Knew for now 2024-07-18 19:42:14 +00:00
PoYen, Chen
e83c3c7fa0 Add constraint to the rotary_dim option 2024-07-16 06:54:37 +00:00
PoYen, Chen
879710a495 Fix wrong seqlen_k for kvcache 2024-07-16 03:42:51 +00:00
PoYen, Chen
65dac9fb90 Fix wrong boundaries 2024-07-15 01:42:53 +00:00
PoYen, Chen
4e01307e04 Fix compilation error in debug mode 2024-07-15 01:26:46 +00:00
PoYen, Chen
1a093f94b2 Add minimum seqlen_k to generate compliance kvcache 2024-07-15 01:11:16 +00:00
PoYen, Chen
57c6a4125c Fix seqlen_knew enabling check logic 2024-07-15 00:40:39 +00:00
PoYen, Chen
ad61d9d4b2 Randomly generate seqlen_knew if needed 2024-07-15 00:39:03 +00:00
PoYen, Chen
f6850aef29 Add compute data type alias for RoPE 2024-07-15 00:05:33 +00:00
PoYen, Chen
391210ed9e Pass RoPE kernel args 2024-07-14 23:18:32 +00:00
PoYen, Chen
b5ad1411b0 Merge branch 'feature/cond-add-splitkv' into feature/fmha-fwd-appendkv 2024-07-14 22:13:17 +00:00
PoYen, Chen
8c1647d778 Avoid invoking deprecated method 'find_module' 2024-07-14 22:10:30 +00:00
PoYen, Chen
55f55025ee Fix wrong tensor size 2024-07-14 15:40:56 +00:00
PoYen, Chen
93e5125d7a Rename RoPE utility function 2024-07-14 14:48:06 +00:00
PoYen, Chen
83d6acc111 Apply RoPE on host side 2024-07-14 14:45:17 +00:00
PoYen, Chen
3183b68921 Simplify v_host_ref definition 2024-07-12 06:42:41 +00:00
PoYen, Chen
e5885cab83 Simplify K appending logics 2024-07-12 06:37:23 +00:00
PoYen, Chen
3578c6f836 Append K/V in the host verification code 2024-07-12 06:32:35 +00:00
PoYen, Chen
4107bf03a6 Merge remote-tracking branch 'origin/feature/cond-add-splitkv' into feature/fmha-fwd-appendkv 2024-07-12 04:43:04 +00:00
PoYen, Chen
b34ddf5f71 Merge remote-tracking branch 'origin/feature/cond-add-splitkv' into feature/fmha-fwd-appendkv 2024-07-12 04:42:45 +00:00
Po Yen Chen
b4306af655 Merge branch 'develop' into feature/cond-add-splitkv 2024-07-12 12:34:31 +08:00
zjing14
13c1e64daa add gemm_bias_add example (#1361)
* add gemm_bias_add example

* changed strideD

* clang-format

---------

Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>
2024-07-11 18:08:07 -07:00
Rostyslav Geyyer
7a46a91c84 Add instances for grouped conv fwd 3d with ConvScale for bf8@fp8->fp8 (#1369)
* Add an example

* Add instances

* Add a client example
2024-07-11 13:31:39 -07:00
Illia Silin
98a01bbc72 Add CK_TILE tests to daily CI builds. (#1381)
* add ck_tile tests to CI

* build and run ck_tile tests on gfx90a and gfx942 in parallel

* fix groovy syntax

* turn ck_tile tests OFF by default

* skip creating the build folder

* build ck_tile examples with 64 threads

* build ck_tile examples with cmake-ck-dev.sh script

* add video group to docker on mi300

* do not retry to rebuild the early CI stages

* help prevent jenkins false failure

* restore cron trigger
2024-07-11 13:22:40 -07:00
carlushuang
bbdb0a5dc0 Merge branch 'develop' into feature/cond-add-splitkv 2024-07-11 16:01:19 +08:00
PoYen, Chen
8c733fb3be Fix compilation errors 2024-07-10 10:53:58 +00:00
PoYen, Chen
e939082bdc Add RoPE example utilities 2024-07-09 05:20:47 +00:00
Illia Silin
a328df25a1 Fix the cmake logic when building with INSTANCES_ONLY=ON. (#1376)
* fix the cmake logic when building for various targets

* another minor fix
2024-07-08 21:21:16 -07:00
Po Yen Chen
dc72074ec7 Merge branch 'develop' into feature/cond-add-splitkv 2024-07-09 03:42:25 +08:00
carlushuang
8182976c37 [CK_TILE] wa prec, remove sgpr offset for inline asm (#1356)
* wa prec, remove sgpr offset for inline asm

* macro for set tile

* ignore unused param if no kernel instances in host API

* fix more prec issue

* cache buffer resource

* fix

* support pre-nop

* clear tile by vector type members

* add workaround to reduce scratch memory

* conditionally enable workaround code

* enable workaround start from certain build version

* fallback set_tile() implementation from certain build version

* undo template argument changes

* put dummy asm in load_raw()

* fix comments, refactor s_nop inside buffer_load

---------

Co-authored-by: PoYen, Chen <PoYen.Chen@amd.com>
2024-07-08 11:09:55 -07:00
PoYen, Chen
18a3834fb4 Set num_splits=1 if split-kv is not supported 2024-07-08 10:27:32 +00:00
PoYen, Chen
8ac6bacf26 Unify CMakeLists.txt coding style 2024-07-08 10:19:31 +00:00
PoYen, Chen
5d21b4d736 Merge branch 'feature/cond-add-splitkv' into feature/fmha-fwd-appendkv 2024-07-08 10:18:28 +00:00
PoYen, Chen
6ca3910199 Show message if we are ignoring option 2024-07-08 10:17:55 +00:00
PoYen, Chen
fe4ae5dcd9 Early return if 0 < s_k_new is not supported 2024-07-08 10:09:36 +00:00
PoYen, Chen
be076db91c Merge branch 'feature/cond-add-splitkv' into feature/fmha-fwd-appendkv 2024-07-08 10:03:58 +00:00
PoYen, Chen
aba46cd655 Regsiter API handlers automatically 2024-07-08 09:39:15 +00:00
PoYen, Chen
3aefb560e0 Remove "EXAMPLE_" prefix of cmake variables 2024-07-08 07:17:24 +00:00
PoYen, Chen
1c070380fa Merge branch 'feature/cond-add-splitkv' into feature/fmha-fwd-appendkv 2024-07-08 07:13:34 +00:00
PoYen, Chen
82f3b3d0a0 Conditionally add call to fmha_fwd_splitkv() 2024-07-08 06:40:18 +00:00
PoYen, Chen
efd18fa887 Conditionally add fwd_splitkv API in fmha_fwd example 2024-07-08 06:27:44 +00:00
Harisankar Sadasivan
75e622f02f Universal streamk with atomics (#1360)
* universal streamk with atomics with ckprofiler support. grid_size and streamk strategy are tunable. grid_size of -1 leads to #WGs = maximum occupancy X num_CUs. implementation supports many different streamk policies: 1-tile, 2-tile, 3-tile and 4-tile. streamk strategy of -1 leads to default streamk policy (4-tile). 

* Update README.md

* fixing clang-format issues

* removed conflicts in struct members between streamk and universal streamk

* corrected arg parsing for streamk and universal streamk

* added stream-k policies for 3 tile and 4 tile

* fixed argument type issue with parsing cmd args

* changes suggested in PR review are made- removing comments and correcting copyright

* file permissions updated

* added default value support for grid_size and streamk-policy selection set to -1

* print messages for arguments

* print messages for arguments

* print messages for arguments1
2024-07-05 21:40:30 -07:00
Jun Liu
959073842c Fix issue with multiple targets and remove smfmac tests from unsupported test targets (#1372) 2024-07-03 23:34:38 -07:00
Ruturaj Vaidya
2525864fda Update CMakeLists.txt (#1364)
It is a good practice to check if the file CMakeLists.txt is in fact in the directory.
2024-06-27 12:34:25 -07:00
Illia Silin
941d1f7ce0 Merging the gfx12 code into public repo. (#1362) 2024-06-27 00:33:34 -07:00