| ... | @@ -99,7 +99,7 @@ Discusses "Best practices" of using OpenMP to offload to GPUs: |
... | @@ -99,7 +99,7 @@ Discusses "Best practices" of using OpenMP to offload to GPUs: |
|
|
- Don't copy scalars, map tham as `firstprivate(scalar)` -- they will become the kernel arguments.
|
|
- Don't copy scalars, map tham as `firstprivate(scalar)` -- they will become the kernel arguments.
|
|
|
- Fortran array operation and sections `(:)` must be avoided in the target regions.
|
|
- Fortran array operation and sections `(:)` must be avoided in the target regions.
|
|
|
|
|
|
|
|
### 2.10 [Thinking OpenMP with NVIDIA compilers by NVIDIA](https://www.nas.nasa.gov/assets/nas/pdf/ams/2021/AMS_20210504_Ozen.pdf)
|
|
### b.10 [Thinking OpenMP with NVIDIA compilers by NVIDIA](https://www.nas.nasa.gov/assets/nas/pdf/ams/2021/AMS_20210504_Ozen.pdf)
|
|
|
|
|
|
|
|
> Advanced Modeling & Simulation (AMS) Seminar Series; NASA Ames Research
|
|
> Advanced Modeling & Simulation (AMS) Seminar Series; NASA Ames Research
|
|
|
> Center, May 2021; Presentation slides, 43 slides; by Dr. Güray Özen
|
|
> Center, May 2021; Presentation slides, 43 slides; by Dr. Güray Özen
|
| ... | | ... | |
