Hugging Face's logo

lrh12580
/
UvmPinAsync

Model card Files
xet
 Community
UvmPinAsync
File size: 8,696 Bytes
5cb6c4b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
//----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------200
//	plasmaKernel_gpu_2
//----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------200

#include <cooperative_groups.h>
#include <cooperative_groups/memcpy_async.h>

using namespace nvcuda::experimental;

#define PREFETCH_COUNT 2

__global__ void kernel_gpu_cuda(par_str d_par_gpu,
								dim_str d_dim_gpu,
								box_str *d_box_gpu,
								FOUR_VECTOR *d_rv_gpu,
								fp *d_qv_gpu,
								FOUR_VECTOR *d_fv_gpu,
								int boxes_per_block)
{
	cooperative_groups::thread_block block = cooperative_groups::this_thread_block();
	pipeline pipe;
	//--------------------------------------------------------------------------------------------------------------------------------------------------------------------------180
	//	THREAD PARAMETERS
	//--------------------------------------------------------------------------------------------------------------------------------------------------------------------------180
	int bx = blockIdx.x;  // get current horizontal block index (0-n)
	int tx = threadIdx.x; // get current horizontal thread index (0-n)
	int wtx = tx;

	//------------------------------------------------------------------------------------------------------------------------------------------------------160
	//	Extract input parameters
	//------------------------------------------------------------------------------------------------------------------------------------------------------160

	// parameters
	fp a2 = 2.0 * d_par_gpu.alpha * d_par_gpu.alpha;

	// home box
	int first_i;
	FOUR_VECTOR *rA;
	FOUR_VECTOR *fA;
	__shared__ FOUR_VECTOR rA_shared[100];

	// nei box
	int pointer;
	int k = 0;
	int first_j;
	FOUR_VECTOR *rB;
	fp *qB;
	int j = 0;
	__shared__ FOUR_VECTOR rB_shared[NUMBER_PAR_PER_BOX * PREFETCH_COUNT];
	__shared__ double qB_shared[NUMBER_PAR_PER_BOX * PREFETCH_COUNT];

	// common
	fp r2;
	fp u2;
	fp vij;
	fp fs;
	fp fxij;
	fp fyij;
	fp fzij;
	THREE_VECTOR d;

	int box = bx * boxes_per_block;
	int end_box = box + boxes_per_block;

	//--------------------------------------------------------------------------------------------------------------------------------------------------------------------------180
	//	DO FOR THE NUMBER OF BOXES
	//--------------------------------------------------------------------------------------------------------------------------------------------------------------------------180
	for (; box < end_box; box++)
	{
		//------------------------------------------------------------------------------------------------------------------------------------------------------160
		//	Home box
		//------------------------------------------------------------------------------------------------------------------------------------------------------160

		//----------------------------------------------------------------------------------------------------------------------------------140
		//	Setup parameters
		//----------------------------------------------------------------------------------------------------------------------------------140

		// home box - box parameters
		first_i = d_box_gpu[box].offset;

		// home box - distance, force, charge and type parameters
		rA = &d_rv_gpu[first_i];
		fA = &d_fv_gpu[first_i];

		//----------------------------------------------------------------------------------------------------------------------------------140
		//	Copy to shared memory
		//----------------------------------------------------------------------------------------------------------------------------------140

		// home box - shared memory
		while (wtx < NUMBER_PAR_PER_BOX)
		{
			rA_shared[wtx] = rA[wtx];
			wtx = wtx + NUMBER_THREADS;
		}
		wtx = tx;

		// synchronize threads  - not needed, but just to be safe
		block.sync();

		//------------------------------------------------------------------------------------------------------------------------------------------------------160
		//	nei box loop
		//------------------------------------------------------------------------------------------------------------------------------------------------------160

		// if (wtx == 0)
		// 	printf("d_box_gpu[%d].nn is %d\n", bx, d_box_gpu[bx].nn);

		int fetch = 0;
		int end_tile = 1 + d_box_gpu[box].nn;

		// loop over neiing boxes of home box
		for (int compute = fetch; compute < end_tile; compute++)
		{
			for (; fetch < end_tile && fetch < compute + PREFETCH_COUNT; fetch++)
			{
				//----------------------------------------50
				//	nei box - get pointer to the right box
				//----------------------------------------50

				if (fetch == 0)
				{
					pointer = box; // set first box to be processed to home box
				}
				else
				{
					pointer = d_box_gpu[box].nei[fetch - 1].number; // remaining boxes are nei boxes
				}

				//----------------------------------------------------------------------------------------------------------------------------------140
				//	Setup parameters
				//----------------------------------------------------------------------------------------------------------------------------------140

				// nei box - box parameters
				first_j = d_box_gpu[pointer].offset;

				// nei box - distance, (force), charge and (type) parameters
				rB = &d_rv_gpu[first_j];
				qB = &d_qv_gpu[first_j];

				//----------------------------------------------------------------------------------------------------------------------------------140
				//	Setup parameters
				//----------------------------------------------------------------------------------------------------------------------------------140

				// nei box - shared memory
				while (wtx < NUMBER_PAR_PER_BOX)
				{
					memcpy_async(rB_shared[(fetch % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + wtx], rB[wtx], pipe);
					memcpy_async(qB_shared[(fetch % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + wtx], qB[wtx], pipe);
					wtx = wtx + NUMBER_THREADS;
				}
				wtx = tx;

				// synchronize threads because in next section each thread accesses data brought in by different threads here
				pipe.commit();
			}
			if (fetch == end_tile)
			{
				for (int i = 0; i < PREFETCH_COUNT - 1; ++i)
				{
					pipe.commit();
				}
				++fetch;
			}
			pipe.wait_prior<PREFETCH_COUNT - 1>();
			block.sync();

			//----------------------------------------------------------------------------------------------------------------------------------140
			//	Calculation
			//----------------------------------------------------------------------------------------------------------------------------------140

			// loop for the number of particles in the home box
			// for (int i=0; i<nTotal_i; i++){
			while (wtx < NUMBER_PAR_PER_BOX)
			{
				// loop for the number of particles in the current nei box
				for (j = 0; j < NUMBER_PAR_PER_BOX; j++)
				{
					r2 = (fp)rA_shared[wtx].v + (fp)rB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j].v - DOT((fp)rA_shared[wtx], (fp)rB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j]);
					u2 = a2 * r2;
					vij = exp(-u2);
					fs = 2 * vij;

					d.x = (fp)rA_shared[wtx].x - (fp)rB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j].x;
					fxij = fs * d.x;
					d.y = (fp)rA_shared[wtx].y - (fp)rB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j].y;
					fyij = fs * d.y;
					d.z = (fp)rA_shared[wtx].z - (fp)rB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j].z;
					fzij = fs * d.z;

					fA[wtx].v += (double)((fp)qB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j] * vij);
					fA[wtx].x += (double)((fp)qB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j] * fxij);
					fA[wtx].y += (double)((fp)qB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j] * fyij);
					fA[wtx].z += (double)((fp)qB_shared[(compute % PREFETCH_COUNT) * NUMBER_PAR_PER_BOX + j] * fzij);
				}

				// increment work thread index
				wtx = wtx + NUMBER_THREADS;
			}

			// reset work index
			wtx = tx;

			// synchronize after finishing force contributions from current nei box not to cause conflicts when starting next box
			block.sync();

			//----------------------------------------------------------------------------------------------------------------------------------140
			//	Calculation END
			//----------------------------------------------------------------------------------------------------------------------------------140
		}
	}
}