Which W3SDS4 implementations are timed

Which W3SDS4 implementations are timed An architecture diagram generated by Archify. W3SDS4 fixture · 1 000 points, NK/NTH 36/24, 32/36 · Architecture component W3SDS4 fixture 1 000 points, NK/NTH 36/24, 32/36 Fortran W3SDS4 · gfortran -O3, 1 thread · Architecture component Fortran W3SDS4 gfortran -O3, 1 thread Kokkos W3SDS4 · 1, 32 threads, RTX 4090 (P1.6) · Architecture component Kokkos W3SDS4 1, 32 threads, RTX 4090 (P1.6) Triton C · triton-cpu, 1 and 32 threads · Architecture component Triton C triton-cpu, 1 and 32 threads Triton G, fusion off · RTX 4090 · Architecture component Triton G, fusion off RTX 4090 Triton G, fusion on · RTX 4090, the speed ceiling · Architecture component Triton G, fusion on RTX 4090, the speed ceiling agrees with the Fortran? · within the L1 tolerance · Architecture component agrees with the Fortran? within the L1 tolerance row in the timing table · kernel-only and end-to-end ms · Architecture component row in the timing table kernel-only and end-to-end ms left out of the table · Architecture component left out of the table reference output yes no Legend Frontend Backend Security External