carlushuang c0adab4850 [CK_TILE] moe sorting ex kernel to support expert > 128 (#1840) ...

* moe sorting ex

* fix bug for race condition

* fix bug and optimze large expert

* fix

* optimize with sub_token_oneshot

* support skip empty tokens for expert sorting

* update moe_sorting

* tidy code

2025-02-11 17:49:17 +08:00

..

reference

[CK_TILE] moe sorting ex kernel to support expert > 128 (#1840)

2025-02-11 17:49:17 +08:00

arg_parser.hpp

Apply Ck-tile argument parser for vectors [I/O] (#1758)

2024-12-19 17:55:35 +01:00

check_err.hpp

Support for dtypes (fp8, bf8, bf16 and fp16) for the ck_tile/03_gemm example. (#1845)

2025-02-07 15:07:06 -07:00

concat.hpp

[CK_TILE] Add GetName for GEMM kernels (#1791)

2025-02-11 09:49:48 +01:00

convolution_host_tensor_descriptor_helper.hpp

Simplify static_cast if-lands (#1828)

2025-01-21 23:23:19 +01:00

convolution_parameter.hpp

[CK_TILE] Fix conv param multiple definition (#1550)

2024-10-07 15:21:21 +02:00

device_memory.hpp

[CK_TILE] fused-moe first version (#1634)

2024-11-26 11:14:56 +08:00

fill.hpp

[CK_TILE] fused-moe first version (#1634)

2024-11-26 11:14:56 +08:00

hip_check_error.hpp

introducing ck_tile! (#1216)

2024-04-15 19:27:12 -05:00

host_tensor.hpp

[CK_TILE] not using structures under ck_tile/ops for ck_tile/host (#1834)

2025-01-24 15:35:54 +08:00

joinable_thread.hpp

[CK_TILE] fused-moe first version (#1634)

2024-11-26 11:14:56 +08:00

kernel_launch.hpp

[CK_TILE] Add PagedAttention kernels (#1387)

2024-08-28 20:50:43 +08:00

ranges.hpp

introducing ck_tile! (#1216)

2024-04-15 19:27:12 -05:00

stream_config.hpp

[CK_TILE] support group from cmdline (#1295)

2024-05-28 11:13:21 +08:00

timer.hpp

Replace hipDeviceSynchronize() by hipStreamSynchronize(stream) calls (#1359)

2024-06-26 22:04:52 +08:00