Merge branch 'develop' into ck_tile/fa_bwd_v3

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n1024_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn  tm   tn  vn   pd   2p
+#if 0
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  2,  4,  64, 8,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  4,  4,  64, 4,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  8,  4,  64, 2,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 16,  4,  64, 1,  true, false>>(const S&, A);
+
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  1,  1, 256, 4,  true, false>>(const S&, A);
+#endif
+
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 1, 2,  128, 8,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 2, 2,  128, 4,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 2,  128, 2,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 1,  256, 1,  true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n1536_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn tm  tn  vn  pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 3, 4,  64, 8, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 3, 2, 128, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 3, 1, 256, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 6, 1, 256, 1, true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n2048_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm tn   vn  pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 1, 1, 256, 8, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 2, 1, 256, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 1, 256, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 8, 1, 256, 1, true, false>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n256_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm  rn tm  tn  vn  pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  1,  4, 64, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  2,  4, 64, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  4,  4, 64, 1, true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n3072_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm  tn   vn  pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 3, 1,  128, 8, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 3, 1,  256, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 6, 1,  256, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 3, 1, 1024, 1, true, false>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n4096_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm  tn   vn   pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 2, 1,  256, 8,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 1,  256, 4,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 2, 1, 1024, 2,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 1, 1024, 1,  true, false>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n4096_tp_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm  tn   vn   pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 2, 1,  256, 8,  true, true>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 1,  256, 4,  true, true>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 2, 1, 1024, 2,  true, true>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t,  1, 4, 1, 1024, 1,  true, true>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n512_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn tm tn  vn  pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 1, 4, 64, 8, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 2, 4, 64, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 4, 4, 64, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 8, 4, 64, 1, true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n64_n128_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn  tm  tn  vn   pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  1,  4, 64, 1,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  1,  4, 64, 2,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  2,  4, 64, 1,  true , false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_bf16_n768_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm tn  vn  pd    2p
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  3, 4, 64, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1,  6, 4, 64, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::bf16_t, 1, 12, 4, 64, 1, true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n1024_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn tm    tn  vn   pd    2p
+#if 0
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  2,  4,  64, 8,  true ,false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  4,  4,  64, 4,  true ,false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  8,  4,  64, 2,  true ,false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 16,  4,  64, 1,  true ,false>>(const S&, A);
+
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  1,  1, 256, 4,  true ,false>>(const S&, A);
+#endif
+
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 1, 2,  128, 8,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 2, 2,  128, 4,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 4, 2,  128, 2,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 4, 1,  256, 1,  true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n1536_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn tm   tn  vn   pd    2p
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 3, 4,   64, 8,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 3, 2,  128, 4,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 3, 1,  256, 2,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 6, 1,  256, 1,  true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n2048_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn tm  tn  vn   pd    2p
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 1, 1,  256, 8, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 2, 1,  256, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 4, 1,  256, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 8, 1,  256, 1, true, false>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n256_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn  tm  tn  vn  pd    2p
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  1,  4, 64, 4, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  2,  4, 64, 2, true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  4,  4, 64, 1, true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n3072_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm  tn  vn  pd    2p
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 3, 1,  128, 8,true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 3, 1,  256, 4,true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 6, 1,  256, 2,true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 3, 1, 1024, 1,true, false>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n4096_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm  tn  vn    pd     2p
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 2, 1,  256, 8,  true,  false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 4, 1,  256, 4,  true,  false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 2, 1, 1024, 2,  true,  false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 4, 1, 1024, 1,  true,  false>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n4096_tp_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                   rm rn tm  tn   vn   pd    2p
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 2, 1,  256, 8,  true, true>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 4, 1,  256, 4,  true, true>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 2, 1, 1024, 2,  true, true>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t,  1, 4, 1, 1024, 1,  true, true>>(const S&, A);
+
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n512_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm rn  tm  tn  vn   pd    2p
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  1,  4, 64, 8,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  2,  4, 64, 4,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  4,  4, 64, 2,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  8,  4, 64, 1,  true , false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n64_n128_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm  rn  tm  tn  vn  pd      2p
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  1,  4,  64, 1,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  1,  4,  64, 2,  true, false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  2,  4,  64, 1,  true, false>>(const S&, A);
+// clang-format on

example/ck_tile/12_smoothquant/instances/smoothquant_fp16_n768_instance.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "smoothquant_instance_common.hpp"
+
+// clang-format off
+//                                                  rm  rn  tm  tn  vn  pd       2p
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  3,  4,  64, 4,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1,  6,  4,  64, 2,  true , false>>(const S&, A);
+template float smoothquant_<trait_<ck_tile::fp16_t, 1, 12,  4,  64, 1,  true , false>>(const S&, A);
+// clang-format on