Neko 1.99.7
A portable framework for high-order spectral element flow simulations
Loading...
Searching...
No Matches
Macros | Functions
ax_helm.cu File Reference
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include "ax_helm_kernel.h"
#include "elem_block_tune.h"
#include <device/device_config.h>
#include <device/cuda/check.h>
#include <common/neko_log.h>

Go to the source code of this file.

Macros

#define CASE_1D(LX, C)
 
#define CASE_1D_SEL(LX, SEL)
 
#define CASE_KSTEP(LX, C)
 
#define CASE_KSTEP_PADDED(LX, C)
 
#define CASE_KSTEP_SEL(LX, SEL)
 
#define CASE_KSTEP_PADDED_SEL(LX, SEL)
 
#define CASE(LX)
 
#define CASE_PADDED(LX)
 
#define CASE_LARGE(LX)
 
#define CASE_LARGE_PADDED(LX)
 
#define AX_VEC_EB(LX)   (elem_block<LX, NEKO_AX_HELM_VECTOR_EB_C>::value)
 
#define AX_VEC_NTHRDS(LX)   dim3((LX), (LX), AX_VEC_EB(LX))
 
#define AX_VEC_NBLCKS(LX)    dim3(((*nelv) + AX_VEC_EB(LX) - 1)/AX_VEC_EB(LX), 1, 1)
 
#define CASE_VECTOR_KSTEP(LX)
 
#define CASE_VECTOR_KSTEP_PADDED(LX)
 
#define CASE_VECTOR(LX)
 
#define CASE_VECTOR_PADDED(LX)
 

Functions

template<const int >
int tune (void *w, void *u, void *dx, void *dy, void *dz, void *dxt, void *dyt, void *dzt, void *h1, void *g11, void *g22, void *g33, void *g12, void *g13, void *g23, int *nelv, int *lx, int *eb_sel, int *ch_sel)
 
template<const int >
int tune_padded (void *w, void *u, void *dx, void *dy, void *dz, void *dxt, void *dyt, void *dzt, void *h1, void *g11, void *g22, void *g33, void *g12, void *g13, void *g23, int *nelv, int *lx, int *eb_sel, int *ch_sel)
 
void cuda_ax_helm (void *w, void *u, void *dx, void *dy, void *dz, void *dxt, void *dyt, void *dzt, void *h1, void *g11, void *g22, void *g33, void *g12, void *g13, void *g23, int *nelv, int *lx)
 
void cuda_ax_helm_vector (void *au, void *av, void *aw, void *u, void *v, void *w, void *dx, void *dy, void *dz, void *dxt, void *dyt, void *dzt, void *h1, void *g11, void *g22, void *g33, void *g12, void *g13, void *g23, int *nelv, int *lx)
 
void cuda_ax_helm_vector_part2 (void *au, void *av, void *aw, void *u, void *v, void *w, void *h2, void *B, int *n)
 

Macro Definition Documentation

◆ AX_VEC_EB

#define AX_VEC_EB (   LX)    (elem_block<LX, NEKO_AX_HELM_VECTOR_EB_C>::value)

◆ AX_VEC_NBLCKS

#define AX_VEC_NBLCKS (   LX)     dim3(((*nelv) + AX_VEC_EB(LX) - 1)/AX_VEC_EB(LX), 1, 1)

◆ AX_VEC_NTHRDS

#define AX_VEC_NTHRDS (   LX)    dim3((LX), (LX), AX_VEC_EB(LX))

◆ CASE

#define CASE (   LX)
Value:
case LX: \
if(autotune[LX] == 0 ) { \
autotune[LX]=tune<LX>( w, u, \
dx, dy, dz, \
dxt, dyt, dzt,h1, \
g11, g22, g33, \
g12, g13, g23, nelv, lx, \
&autotune_eb[LX], &autotune_ch[LX]); \
} else if (autotune[LX] == 1 ) { \
CASE_1D_SEL(LX, autotune_ch[LX]); \
} else if (autotune[LX] == 2 ) { \
CASE_KSTEP_SEL(LX, autotune_eb[LX]); \
} \
break
ale_add_kinematics_kernel
__global__ void ale_add_kinematics_kernel(const int n, T *__restrict__ wx, T *__restrict__ wy, T *__restrict__ wz, const T *__restrict__ x_ref, const T *__restrict__ y_ref, const T *__restrict__ z_ref, const T *__restrict__ phi, const T *__restrict__ x, const T *__restrict__ y, const T *__restrict__ z, const kinematics_params_t kin_params)
Definition ale_kinematics_kernel.h:52
w
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ w
Definition ax_helm_full_kernel.h:48
u
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ u
Definition ax_helm_full_kernel.h:46
dx
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dx
Definition ax_helm_full_kernel.h:49
dz
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dz
Definition ax_helm_full_kernel.h:51
h1
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ h1
Definition ax_helm_full_kernel.h:52
dy
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dy
Definition ax_helm_full_kernel.h:50
nelv
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const int nelv
Definition ax_helm_kernel.h:177
g23
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ g23
Definition ax_helm_kernel.h:176
g22
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ g22
Definition ax_helm_kernel.h:172
g13
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ g13
Definition ax_helm_kernel.h:175
g12
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ g12
Definition ax_helm_kernel.h:174
g33
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ g33
Definition ax_helm_kernel.h:173
g11
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ g11
Definition ax_helm_kernel.h:171
dzt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dzt
Definition cdtp_kernel.h:114
dyt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dyt
Definition cdtp_kernel.h:113
dxt
__global__ void const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ const T *__restrict__ dxt
Definition cdtp_kernel.h:112

◆ CASE_1D

#define CASE_1D (   LX,
  C 
)
Value:
ax_helm_kernel_1d<real, LX, NEKO_CHUNKS(LX, C)> \
<<<nblcks_1d, NEKO_CHUNKS_NTHRDS(LX, C), 0, stream>>> \
((real *) w, (real *) u, \
(real *) dx, (real *) dy, (real *) dz, \
(real *) dxt, (real *) dyt, (real *) dzt, (real *) h1,\
(real *) g11, (real *) g22, (real *) g33, \
(real *) g12, (real *) g13, (real *) g23); \
CUDA_CHECK(cudaGetLastError());
ax_helm_kernel_1d
__global__ void ax_helm_kernel_1d(T *__restrict__ w, const T *__restrict__ u, const T *__restrict__ dx, const T *__restrict__ dy, const T *__restrict__ dz, const T *__restrict__ dxt, const T *__restrict__ dyt, const T *__restrict__ dzt, const T *__restrict__ h1, const T *__restrict__ g11, const T *__restrict__ g22, const T *__restrict__ g33, const T *__restrict__ g12, const T *__restrict__ g13, const T *__restrict__ g23)
Definition ax_helm_kernel.h:57
NEKO_CHUNKS
#define NEKO_CHUNKS(LX, C)
Definition elem_block.h:145
NEKO_CHUNKS_NTHRDS
#define NEKO_CHUNKS_NTHRDS(LX, C)
Definition elem_block.h:146
real
double real
Definition device_config.h:12

◆ CASE_1D_SEL

#define CASE_1D_SEL (   LX,
  SEL 
)
Value:
switch (SEL) { \
case 0: CASE_1D(LX, 0); break; \
case 1: CASE_1D(LX, 1); break; \
case 2: CASE_1D(LX, 2); break; \
default: CASE_1D(LX, 3); break; \
}
CASE_1D
#define CASE_1D(LX, C)

◆ CASE_KSTEP

#define CASE_KSTEP (   LX,
  C 
)
Value:
ax_helm_kernel_kstep<real, LX, NEKO_EB(LX, C)> \
<<<NEKO_EB_NBLCKS(*nelv, LX, C), NEKO_EB_NTHRDS(LX, C), 0, stream>>> \
((real *) w, (real *) u, \
(real *) dx, (real *) dy, (real *) dz, (real *) h1, \
(real *) g11, (real *) g22, (real *) g33, \
(real *) g12, (real *) g13, (real *) g23, *nelv); \
CUDA_CHECK(cudaGetLastError());
NEKO_EB
#define NEKO_EB(LX, C)
Definition elem_block.h:104
NEKO_EB_NBLCKS
#define NEKO_EB_NBLCKS(NELV, LX, C)
Definition elem_block.h:106
NEKO_EB_NTHRDS
#define NEKO_EB_NTHRDS(LX, C)
Definition elem_block.h:105

◆ CASE_KSTEP_PADDED

#define CASE_KSTEP_PADDED (   LX,
  C 
)
Value:
ax_helm_kernel_kstep_padded<real, LX, NEKO_EB(LX, C)> \
<<<NEKO_EB_NBLCKS(*nelv, LX, C), NEKO_EB_NTHRDS(LX, C), 0, stream>>> \
((real *) w, (real *) u, \
(real *) dx, (real *) dy, (real *) dz, (real *) h1, \
(real *) g11, (real *) g22, (real *) g33, \
(real *) g12, (real *) g13, (real *) g23, *nelv); \
CUDA_CHECK(cudaGetLastError());

◆ CASE_KSTEP_PADDED_SEL

#define CASE_KSTEP_PADDED_SEL (   LX,
  SEL 
)
Value:
switch (SEL) { \
case 0: CASE_KSTEP_PADDED(LX, 0); break; \
case 1: CASE_KSTEP_PADDED(LX, 1); break; \
default: CASE_KSTEP_PADDED(LX, 2); break; \
}
CASE_KSTEP_PADDED
#define CASE_KSTEP_PADDED(LX, C)

◆ CASE_KSTEP_SEL

#define CASE_KSTEP_SEL (   LX,
  SEL 
)
Value:
switch (SEL) { \
case 0: CASE_KSTEP(LX, 0); break; \
case 1: CASE_KSTEP(LX, 1); break; \
default: CASE_KSTEP(LX, 2); break; \
}
CASE_KSTEP
#define CASE_KSTEP(LX, C)

◆ CASE_LARGE

#define CASE_LARGE (   LX)
Value:
case LX: \
CASE_KSTEP(LX, 0); \
break

◆ CASE_LARGE_PADDED

#define CASE_LARGE_PADDED (   LX)
Value:
case LX: \
CASE_KSTEP_PADDED(LX, 0); \
break

◆ CASE_PADDED

#define CASE_PADDED (   LX)
Value:
case LX: \
if(autotune[LX] == 0 ) { \
autotune[LX]=tune_padded<LX>(w, u, \
dx, dy, dz, \
dxt, dyt, dzt,h1, \
g11, g22, g33, \
g12, g13, g23,nelv,lx, \
&autotune_eb[LX], &autotune_ch[LX]); \
} else if (autotune[LX] == 1 ) { \
CASE_1D_SEL(LX, autotune_ch[LX]); \
} else if (autotune[LX] == 2 ) { \
CASE_KSTEP_PADDED_SEL(LX, autotune_eb[LX]); \
} \
break

◆ CASE_VECTOR

#define CASE_VECTOR (   LX)
Value:
case LX: \
CASE_VECTOR_KSTEP(LX); \
break

◆ CASE_VECTOR_KSTEP

#define CASE_VECTOR_KSTEP (   LX)
Value:
ax_helm_kernel_vector_kstep<real, LX, AX_VEC_EB(LX)> \
<<<AX_VEC_NBLCKS(LX), AX_VEC_NTHRDS(LX), 0, stream>>> \
((real *) au, (real *) av, (real *) aw, \
(real *) u, (real *) v, (real *) w, \
(real *) dx, (real *) dy, (real *) dz, \
(real *) h1, (real *) g11, (real *) g22, \
(real *) g33, (real *) g12, (real *) g13, \
(real *) g23, *nelv); \
CUDA_CHECK(cudaGetLastError());
AX_VEC_NTHRDS
#define AX_VEC_NTHRDS(LX)
AX_VEC_EB
#define AX_VEC_EB(LX)
AX_VEC_NBLCKS
#define AX_VEC_NBLCKS(LX)
aw
__global__ void T *__restrict__ T *__restrict__ aw
Definition ax_helm_full_kernel.h:45
av
__global__ void T *__restrict__ av
Definition ax_helm_full_kernel.h:44
v
__global__ void T *__restrict__ T *__restrict__ const T *__restrict__ const T *__restrict__ v
Definition ax_helm_full_kernel.h:47

◆ CASE_VECTOR_KSTEP_PADDED

#define CASE_VECTOR_KSTEP_PADDED (   LX)
Value:
ax_helm_kernel_vector_kstep_padded<real, LX, AX_VEC_EB(LX)> \
<<<AX_VEC_NBLCKS(LX), AX_VEC_NTHRDS(LX), 0, stream>>> \
((real *) au, (real *) av, (real *) aw, \
(real *) u, (real *) v, (real *) w, \
(real *) dx, (real *) dy, (real *) dz, \
(real *) h1, (real *) g11, (real *) g22, \
(real *) g33, (real *) g12, (real *) g13, \
(real *) g23, *nelv); \
CUDA_CHECK(cudaGetLastError());

◆ CASE_VECTOR_PADDED

#define CASE_VECTOR_PADDED (   LX)
Value:
case LX: \
CASE_VECTOR_KSTEP_PADDED(LX); \
break

Function Documentation

◆ cuda_ax_helm()

void cuda_ax_helm ( void *  w,
void *  u,
void *  dx,
void *  dy,
void *  dz,
void *  dxt,
void *  dyt,
void *  dzt,
void *  h1,
void *  g11,
void *  g22,
void *  g33,
void *  g12,
void *  g13,
void *  g23,
int *  nelv,
int *  lx 
)

Fortran wrapper for device CUDA Ax

Definition at line 64 of file ax_helm.cu.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cuda_ax_helm_vector()

void cuda_ax_helm_vector ( void *  au,
void *  av,
void *  aw,
void *  u,
void *  v,
void *  w,
void *  dx,
void *  dy,
void *  dz,
void *  dxt,
void *  dyt,
void *  dzt,
void *  h1,
void *  g11,
void *  g22,
void *  g33,
void *  g12,
void *  g13,
void *  g23,
int *  nelv,
int *  lx 
)

Fortran wrapper for device CUDA Ax vector version

Definition at line 217 of file ax_helm.cu.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cuda_ax_helm_vector_part2()

void cuda_ax_helm_vector_part2 ( void *  au,
void *  av,
void *  aw,
void *  u,
void *  v,
void *  w,
void *  h2,
void *  B,
int *  n 
)

Fortran wrapper for device CUDA Ax vector version part2

Definition at line 295 of file ax_helm.cu.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ tune()

template<const int >
int tune ( void *  w,
void *  u,
void *  dx,
void *  dy,
void *  dz,
void *  dxt,
void *  dyt,
void *  dzt,
void *  h1,
void *  g11,
void *  g22,
void *  g33,
void *  g12,
void *  g13,
void *  g23,
int *  nelv,
int *  lx,
int *  eb_sel,
int *  ch_sel 
)

Definition at line 311 of file ax_helm.cu.

Here is the call graph for this function:

◆ tune_padded()

template<const int >
int tune_padded ( void *  w,
void *  u,
void *  dx,
void *  dy,
void *  dz,
void *  dxt,
void *  dyt,
void *  dzt,
void *  h1,
void *  g11,
void *  g22,
void *  g33,
void *  g12,
void *  g13,
void *  g23,
int *  nelv,
int *  lx,
int *  eb_sel,
int *  ch_sel 
)

Definition at line 440 of file ax_helm.cu.

Here is the call graph for this function: