Neko  0.8.99
A portable framework for high-order spectral element flow simulations
fusedcg_cpld_kernel.h
Go to the documentation of this file.
1 #ifndef __KRYLOV_FUSEDCG_CPLD_KERNEL_H__
2 #define __KRYLOV_FUSEDCG_CPLD_KERNEL_H__
3 /*
4  Copyright (c) 2021-2024, The Neko Authors
5  All rights reserved.
6 
7  Redistribution and use in source and binary forms, with or without
8  modification, are permitted provided that the following conditions
9  are met:
10 
11  * Redistributions of source code must retain the above copyright
12  notice, this list of conditions and the following disclaimer.
13 
14  * Redistributions in binary form must reproduce the above
15  copyright notice, this list of conditions and the following
16  disclaimer in the documentation and/or other materials provided
17  with the distribution.
18 
19  * Neither the name of the authors nor the names of its
20  contributors may be used to endorse or promote products derived
21  from this software without specific prior written permission.
22 
23  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
26  FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
27  COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
28  INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
29  BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
30  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
31  CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
33  ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  POSSIBILITY OF SUCH DAMAGE.
35 */
36 
38 
42 template< typename T >
43 __global__ void fusedcg_cpld_part1_kernel(T * __restrict__ a1,
44  T * __restrict__ a2,
45  T * __restrict__ a3,
46  T * __restrict__ b1,
47  T * __restrict__ b2,
48  T * __restrict__ b3,
49  T * __restrict__ tmp,
50  const int n) {
51 
52  const int idx = blockIdx.x * blockDim.x + threadIdx.x;
53  const int str = blockDim.x * gridDim.x;
54 
55  for (int i = idx; i < n; i+= str) {
56  tmp[i] = a1[i]*b1[i] + a2[i]*b2[i] + a3[i]*b3[i];
57  }
58 
59 }
60 
64 template< typename T >
65 __global__ void fusedcg_cpld_update_p_kernel(T * __restrict__ p1,
66  T * __restrict__ p2,
67  T * __restrict__ p3,
68  const T * __restrict__ z1,
69  const T * __restrict__ z2,
70  const T * __restrict__ z3,
71  const T * __restrict__ po1,
72  const T * __restrict__ po2,
73  const T * __restrict__ po3,
74  const T beta,
75  const int n) {
76 
77  const int idx = blockIdx.x * blockDim.x + threadIdx.x;
78  const int str = blockDim.x * gridDim.x;
79 
80  for (int i = idx; i < n; i+= str) {
81  p1[i] = beta*po1[i] + z1[i];
82  p2[i] = beta*po2[i] + z2[i];
83  p3[i] = beta*po3[i] + z3[i];
84  }
85 
86 }
87 
91 template< typename T >
92 __global__ void fusedcg_cpld_update_x_kernel(T * __restrict__ x1,
93  T * __restrict__ x2,
94  T * __restrict__ x3,
95  const T ** p1,
96  const T ** p2,
97  const T ** p3,
98  const T * __restrict__ alpha,
99  const int p_cur,
100  const int n) {
101 
102  const int idx = blockIdx.x * blockDim.x + threadIdx.x;
103  const int str = blockDim.x * gridDim.x;
104 
105  for (int i = idx; i < n; i+= str) {
106  T tmp1 = 0.0;
107  T tmp2 = 0.0;
108  T tmp3 = 0.0;
109  for (int j = 0; j < p_cur; j ++) {
110  tmp1 += p1[j][i] * alpha[j];
111  tmp2 += p2[j][i] * alpha[j];
112  tmp3 += p3[j][i] * alpha[j];
113  }
114  x1[i] += tmp1;
115  x2[i] += tmp2;
116  x3[i] += tmp3;
117  }
118 
119 }
120 
124 template< typename T>
125 __global__ void fusedcg_cpld_part2_kernel(T * __restrict__ a1,
126  T * __restrict__ a2,
127  T * __restrict__ a3,
128  const T * __restrict__ b,
129  const T * __restrict__ c1,
130  const T * __restrict__ c2,
131  const T * __restrict__ c3,
132  const T alpha,
133  T * buf_h,
134  const int n) {
135 
136  const int idx = blockIdx.x * blockDim.x + threadIdx.x;
137  const int str = blockDim.x * gridDim.x;
138 
139  const unsigned int lane = threadIdx.x % warpSize;
140  const unsigned int wid = threadIdx.x / warpSize;
141 
142  __shared__ T buf[32];
143  T tmp = 0.0;
144 
145  for (int i = idx; i < n; i+= str) {
146  T rt1 = a1[i] - alpha * c1[i];
147  T rt2 = a2[i] - alpha * c2[i];
148  T rt3 = a3[i] - alpha * c3[i];
149  tmp = tmp + ((rt1*rt1 + rt2*rt2 + rt3*rt3) * b[i]);
150  a1[i] = rt1;
151  a2[i] = rt2;
152  a3[i] = rt3;
153  }
154 
155  tmp = reduce_warp<T>(tmp);
156  if (lane == 0) {
157  buf[wid] = tmp;
158  }
159  __syncthreads();
160 
161  tmp = (threadIdx.x < blockDim.x / warpSize) ? buf[lane] : 0;
162  if (wid == 0) {
163  tmp = reduce_warp<T>(tmp);
164  }
165 
166  if (threadIdx.x == 0) {
167  buf_h[blockIdx.x] = tmp;
168  }
169 }
170 
171 #endif // __KRYLOV_FUSEDCG_CPLD_KERNEL_H__
const int i
Definition: cdtp_kernel.h:128
const int j
Definition: cdtp_kernel.h:127
__syncthreads()
__global__ void fusedcg_cpld_update_p_kernel(T *__restrict__ p1, T *__restrict__ p2, T *__restrict__ p3, const T *__restrict__ z1, const T *__restrict__ z2, const T *__restrict__ z3, const T *__restrict__ po1, const T *__restrict__ po2, const T *__restrict__ po3, const T beta, const int n)
__global__ void fusedcg_cpld_part2_kernel(T *__restrict__ a1, T *__restrict__ a2, T *__restrict__ a3, const T *__restrict__ b, const T *__restrict__ c1, const T *__restrict__ c2, const T *__restrict__ c3, const T alpha, T *buf_h, const int n)
__global__ void fusedcg_cpld_update_x_kernel(T *__restrict__ x1, T *__restrict__ x2, T *__restrict__ x3, const T **p1, const T **p2, const T **p3, const T *__restrict__ alpha, const int p_cur, const int n)
__global__ void fusedcg_cpld_part1_kernel(T *__restrict__ a1, T *__restrict__ a2, T *__restrict__ a3, T *__restrict__ b1, T *__restrict__ b2, T *__restrict__ b3, T *__restrict__ tmp, const int n)
real * buf
Definition: pipecg_aux.cu:42