wehub-resource-sync
42 lines
1.6 KiB
Common Lisp
42 lines
1.6 KiB
Common Lisp
#ifdef MNN_SUPPORT_FP16
|
|
#pragma OPENCL EXTENSION cl_khr_fp16 : enable
|
|
#endif
|
|
|
|
#define GLOBAL_SIZE_2_DIMS \
|
|
__private const int global_size_dim0, __private const int global_size_dim1,
|
|
|
|
#define DEAL_NON_UNIFORM_DIM2(input1, input2) \
|
|
if (input1 >= global_size_dim0 || input2 >= global_size_dim1) { \
|
|
return; \
|
|
}
|
|
|
|
__kernel void scale_buf(GLOBAL_SIZE_2_DIMS
|
|
__global const FLOAT* input,
|
|
__global const FLOAT* scale,
|
|
#ifdef BIAS
|
|
__global const FLOAT* bias,
|
|
#endif
|
|
__global FLOAT* output,
|
|
__private const int channelBlock,
|
|
__private const int batch,
|
|
__private const int inside) {
|
|
|
|
const int x = get_global_id(0); // inside(width * height)
|
|
const int y = get_global_id(1); // channelBlock * batch
|
|
|
|
DEAL_NON_UNIFORM_DIM2(x, y);
|
|
|
|
const int out_c_idx = y % channelBlock;
|
|
const int out_b_idx = y / channelBlock;
|
|
const int offset = ((out_b_idx + out_c_idx * batch) * inside + x) * 4;
|
|
COMPUTE_FLOAT4 in_value = CONVERT_COMPUTE_FLOAT4(vload4(0, input+offset));
|
|
COMPUTE_FLOAT4 scale_value = CONVERT_COMPUTE_FLOAT4(vload4(out_c_idx, scale));
|
|
#ifdef BIAS
|
|
COMPUTE_FLOAT4 bias_value = CONVERT_COMPUTE_FLOAT4(vload4(out_c_idx, bias));
|
|
COMPUTE_FLOAT4 out_value = in_value * scale_value + bias_value;
|
|
#else
|
|
COMPUTE_FLOAT4 out_value = in_value * scale_value;
|
|
#endif
|
|
vstore4(CONVERT_FLOAT4(out_value), 0, output+offset);
|
|
}
|