Skip to content

DRAFT: MFEM kernel updates #609

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
artv3 wants to merge 10 commits into
base: develop
Choose a base branch
from
Draft

DRAFT: MFEM kernel updates #609

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
b8ed6ee
initial commit for all fem updates
artv3 Dec 26, 2025
e366158
add mfem-v4.9 draft kernel
artv3 Dec 26, 2025
3cb54d8
update diffusion ref impl
artv3 Dec 26, 2025
63562b1
update base impl for diffusion
artv3 Dec 26, 2025
e659836
update seq variant
artv3 Dec 26, 2025
1c046eb
update pass
artv3 Dec 26, 2025
b3c283a
update diffusion kernel to mfem v4.9
artv3 Dec 26, 2025
d027da1
use real_type and index_type
artv3 Dec 26, 2025
27b2090
MASS3DPA.hpp
artv3 Dec 26, 2025
bece84d
validate mfem assemble MASS3DEA kernel
artv3 Dec 26, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions src/apps/CONVECTION3DPA.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,8 +10,8 @@
/// Action of 3D diffusion matrix via partial assembly
///
/// Based on MFEM's/CEED algorithms.
/// Reference implementation
/// https://github.com/mfem/mfem/blob/master/fem/bilininteg_convection_pa.cpp
/// Reference implementation - MFEM-v4.9
/// https://github.com/mfem/mfem/blob/v4.9/fem/integ/bilininteg_convection_kernels.hpp
///
///
/// for(int e = 0; e < NE; ++e) {
Expand Down
50 changes: 25 additions & 25 deletions src/apps/DIFFUSION3DPA-Cuda.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,73 +32,73 @@ __global__ void Diffusion3DPA(const Real_ptr Basis,

DIFFUSION3DPA_0_GPU;

GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_1;
}
}
}

if (threadIdx.z == 0) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_2;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_3;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_4;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_5;
}
}
}
__syncthreads();
if (threadIdx.z == 0) {
GPU_FOREACH_THREAD(d, y, DPA_D1D) {
GPU_FOREACH_THREAD(q, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(d, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(q, x, DPA_Q1D) {
DIFFUSION3DPA_6;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_7;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_8;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_9;
}
}
Expand Down
61 changes: 31 additions & 30 deletions src/apps/DIFFUSION3DPA-Hip.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,81 +28,82 @@ __global__ void Diffusion3DPA(const Real_ptr Basis,
const Real_ptr dBasis, const Real_ptr D,
const Real_ptr X, Real_ptr Y, bool symmetric) {

const Index_type e = hipBlockIdx_x;
const Index_type e = blockIdx.x;

DIFFUSION3DPA_0_GPU;

GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_1;
}
}
}

if (threadIdx.z == 0) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_2;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_3;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_4;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD(qx, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_5;
}
}
}
__syncthreads();
if (threadIdx.z == 0) {
GPU_FOREACH_THREAD(d, y, DPA_D1D) {
GPU_FOREACH_THREAD(q, x, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qx, x, DPA_Q1D) {
DIFFUSION3DPA_6;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(qy, y, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_7;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(qz, z, DPA_Q1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_8;
}
}
}
__syncthreads();
GPU_FOREACH_THREAD(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD(dx, x, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dz, z, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dy, y, DPA_D1D) {
GPU_FOREACH_THREAD_DIRECT(dx, x, DPA_D1D) {
DIFFUSION3DPA_9;
}
}
}

}

template < size_t block_size >
Expand Down Expand Up @@ -248,7 +249,7 @@ void DIFFUSION3DPA::runHipVariantImpl(VariantID vid) {
RAJA::loop<inner_x>(ctx, RAJA::RangeSegment(0, DPA_Q1D),
[&](Index_type qx) {

DIFFUSION3DPA_5;
DIFFUSION3DPA_5;

} // lambda (qx)
); // RAJA::loop<inner_x>
Expand All @@ -262,9 +263,9 @@ void DIFFUSION3DPA::runHipVariantImpl(VariantID vid) {
RAJA::loop<inner_z>(ctx, RAJA::RangeSegment(0, 1),
[&](Index_type RAJA_UNUSED_ARG(dz)) {
RAJA::loop<inner_y>(ctx, RAJA::RangeSegment(0, DPA_D1D),
[&](Index_type d) {
[&](Index_type dy) {
RAJA::loop<inner_x>(ctx, RAJA::RangeSegment(0, DPA_Q1D),
[&](Index_type q) {
[&](Index_type qx) {

DIFFUSION3DPA_6;

Expand Down Expand Up @@ -320,7 +321,7 @@ void DIFFUSION3DPA::runHipVariantImpl(VariantID vid) {
RAJA::loop<inner_x>(ctx, RAJA::RangeSegment(0, DPA_D1D),
[&](Index_type dx) {

DIFFUSION3DPA_9;
DIFFUSION3DPA_9;

} // lambda (dx)
); // RAJA::loop<inner_x>
Expand Down
79 changes: 40 additions & 39 deletions src/apps/DIFFUSION3DPA-OMP.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,70 +38,71 @@ void DIFFUSION3DPA::runOpenMPVariant(VariantID vid) {

DIFFUSION3DPA_0_CPU;

CPU_FOREACH(dz, z, DPA_D1D) {
CPU_FOREACH(dy, y, DPA_D1D) {
CPU_FOREACH(dx, x, DPA_D1D) {
DIFFUSION3DPA_1;
CPU_FOREACH(dz,z,DPA_D1D) {
CPU_FOREACH(dy,y,DPA_D1D) {
CPU_FOREACH(dx,x,DPA_D1D) {
DIFFUSION3DPA_1
}
}
}

CPU_FOREACH(dy, y, DPA_D1D) {
CPU_FOREACH(qx, x, DPA_Q1D) {
DIFFUSION3DPA_2;
CPU_FOREACH(dy,y,DPA_D1D) {
CPU_FOREACH(qx,x,DPA_Q1D) {
DIFFUSION3DPA_2
}
}

CPU_FOREACH(dz, z, DPA_D1D) {
CPU_FOREACH(dy, y, DPA_D1D) {
CPU_FOREACH(qx, x, DPA_Q1D) {
DIFFUSION3DPA_3;
CPU_FOREACH(dz,z,DPA_D1D) {
CPU_FOREACH(dy,y,DPA_D1D) {
CPU_FOREACH(qx,x,DPA_Q1D) {

DIFFUSION3DPA_3
}
}
}

CPU_FOREACH(dz, z, DPA_D1D) {
CPU_FOREACH(qy, y, DPA_Q1D) {
CPU_FOREACH(qx, x, DPA_Q1D) {
DIFFUSION3DPA_4;
CPU_FOREACH(dz,z,DPA_D1D) {
CPU_FOREACH(qy,y,DPA_Q1D) {
CPU_FOREACH(qx,x,DPA_Q1D) {
DIFFUSION3DPA_4
}
}
}

CPU_FOREACH(qz, z, DPA_Q1D) {
CPU_FOREACH(qy, y, DPA_Q1D) {
CPU_FOREACH(qx, x, DPA_Q1D) {
DIFFUSION3DPA_5;
CPU_FOREACH(qz,z,DPA_Q1D) {
CPU_FOREACH(qy,y,DPA_Q1D) {
CPU_FOREACH(qx,x,DPA_Q1D) {
DIFFUSION3DPA_5
}
}
}

CPU_FOREACH(d, y, DPA_D1D) {
CPU_FOREACH(q, x, DPA_Q1D) {
DIFFUSION3DPA_6;
CPU_FOREACH(dy,y,DPA_D1D) {
CPU_FOREACH(qx,x,DPA_Q1D) {
DIFFUSION3DPA_6
}
}

CPU_FOREACH(qz, z, DPA_Q1D) {
CPU_FOREACH(qy, y, DPA_Q1D) {
CPU_FOREACH(dx, x, DPA_D1D) {
DIFFUSION3DPA_7;
CPU_FOREACH(qz,z,DPA_Q1D) {
CPU_FOREACH(qy,y,DPA_Q1D) {
CPU_FOREACH(dx,x,DPA_D1D) {
DIFFUSION3DPA_7
}
}
}

CPU_FOREACH(qz, z, DPA_Q1D) {
CPU_FOREACH(dy, y, DPA_D1D) {
CPU_FOREACH(dx, x, DPA_D1D) {
DIFFUSION3DPA_8;
CPU_FOREACH(qz,z,DPA_Q1D) {
CPU_FOREACH(dy,y,DPA_D1D) {
CPU_FOREACH(dx,x,DPA_D1D) {
DIFFUSION3DPA_8
}
}
}

CPU_FOREACH(dz, z, DPA_D1D) {
CPU_FOREACH(dy, y, DPA_D1D) {
CPU_FOREACH(dx, x, DPA_D1D) {
DIFFUSION3DPA_9;
CPU_FOREACH(dz,z,DPA_D1D) {
CPU_FOREACH(dy,y,DPA_D1D) {
CPU_FOREACH(dx,x,DPA_D1D) {
DIFFUSION3DPA_9
}
}
}
Expand Down Expand Up @@ -221,7 +222,7 @@ void DIFFUSION3DPA::runOpenMPVariant(VariantID vid) {
RAJA::loop<inner_x>(ctx, RAJA::RangeSegment(0, DPA_Q1D),
[&](Index_type qx) {

DIFFUSION3DPA_5;
DIFFUSION3DPA_5;

} // lambda (qx)
); // RAJA::loop<inner_x>
Expand All @@ -235,11 +236,11 @@ void DIFFUSION3DPA::runOpenMPVariant(VariantID vid) {
RAJA::loop<inner_z>(ctx, RAJA::RangeSegment(0, 1),
[&](Index_type RAJA_UNUSED_ARG(dz)) {
RAJA::loop<inner_y>(ctx, RAJA::RangeSegment(0, DPA_D1D),
[&](Index_type d) {
[&](Index_type dy) {
RAJA::loop<inner_x>(ctx, RAJA::RangeSegment(0, DPA_Q1D),
[&](Index_type q) {
[&](Index_type qx) {

DIFFUSION3DPA_6;
DIFFUSION3DPA_6;

} // lambda (q)
); // RAJA::loop<inner_x>
Expand All @@ -257,7 +258,7 @@ void DIFFUSION3DPA::runOpenMPVariant(VariantID vid) {
RAJA::loop<inner_x>(ctx, RAJA::RangeSegment(0, DPA_D1D),
[&](Index_type dx) {

DIFFUSION3DPA_7;
DIFFUSION3DPA_7;

} // lambda (dx)
); // RAJA::loop<inner_x>
Expand Down
Loading