mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-13 09:45:56 +00:00
627054b941
* Add basic support for direct loads from global to LDS * Clean the code and comments * Add support for fp16 * Add comments * Add check for thread cluster lengths * Align non-direct-load fp16 example * Small fixes * Extend IsSupported to check for supported GPU gens * Build examples only on the supported HW * Do not throw when instance not supported in 04 example * Review: Apply review suggestions * Review: small fix * Review: small fix
Instructions for example_gemm_add_add_fastgelu_xdl_fp16
Run example_gemm_add_add_fastgelu_xdl_fp16
#arg1: verification (0=no, 1=yes)
#arg2: initialization (0=no init, 1=integer value, 2=decimal value)
#arg3: time kernel (0=no, 1=yes)
#arg4 to 11: M (256x), N(128x), K(32x), StrideA, StrideB, StrideD0, StrideD1, StrideE"
./bin/example_gemm_add_add_fastgelu_xdl_fp16 1 1 1
Result (MI100 @ 1087Mhz, 133.5TFlops peak FP16)
a_m_k: dim 2, lengths {3840, 4096}, strides {4096, 1}
b_k_n: dim 2, lengths {4096, 4096}, strides {1, 4096}
d0_m_n: dim 2, lengths {3840, 4096}, strides {0, 1}
d1_m_n: dim 2, lengths {3840, 4096}, strides {4096, 1}
e_m_n: dim 2, lengths {3840, 4096}, strides {4096, 1}
launch_and_time_kernel: grid_dim {480, 1, 1}, block_dim {256, 1, 1}
Warm up 1 time
Start running 10 times...
Perf: 1.26914 ms, 101.525 TFlops, 100.804 GB/s, DeviceGemmMultipleD_Xdl_CShuffle<256, 256, 128, 32, 8, 8>