mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-26 16:04:58 +00:00
277be4385b
* Add custom target to bundle examples together * Add int4 example conditionally (just copy from int8 example) * Extract common code into common.hpp * Move ref gemm type alias into data-type-specific sources * Add #error directive to prevent compile with wrong setting * Let AddAddFastGelu support int4 parameter type * Let check_err() support int4 parameter type * Add wrapper function to hide value conversion while copying memory * Finish int4 example for GEMM + AddAddFastGelu * Add new DeviceMem API to copy memory * Use new DeviceMem API to implement examples * Fix wrongly use of macro 'CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4' * Revert "Add new DeviceMem API to copy memory" This reverts commite26e7af71e. * Add conversion ctor for Tensor<> * Add 'const' specifier to Tensor<>::CopyAsType() * Convert Tensor<> values before/after transfer between host & device [ROCm/composable_kernel commit:2327f1a640]
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>