Skip to content
Permalink
master

Commits on Dec 2, 2020

  1. Stabilize threshold for daint-mc

    alazzaro committed Dec 2, 2020
  2. Update DBCSR to v2.1.0-rc18

    alazzaro committed Dec 2, 2020

Commits on Dec 1, 2020

  1. grid: Move zeroing of results into task_list

    oschuett committed Dec 1, 2020
  2. grid: Streamline task list frontend

    oschuett committed Dec 1, 2020

Commits on Nov 25, 2020

  1. pretty

    hforbert committed Nov 25, 2020
  2. Bug fix: restart PIMD constraints

    duranl21 authored and hforbert committed Nov 25, 2020

Commits on Nov 24, 2020

  1. RI HFX: change defaults (cholesky instead of diagonalization)

    pseewald committed Nov 24, 2020

Commits on Nov 23, 2020

  1. Adjust tolerances

    mkrack committed Nov 23, 2020
  2. Low-scaling RPA/GW: print memory info

    pseewald committed Nov 23, 2020
  3. Adapt regtest tolerances

    Minor loss in accuracy due to compression in cubic scaling code.
    pseewald committed Nov 23, 2020
  4. Use tag instead of sha

    oschuett committed Nov 23, 2020
  5. Fix unused parameter warning when libxsmm is not present

    mtaillefumier authored and oschuett committed Nov 23, 2020

Commits on Nov 22, 2020

  1. grid: Fix grid_create_buffer for length zero

    oschuett committed Nov 22, 2020
  2. grid: Introduce generic buffer type

    oschuett committed Nov 22, 2020

Commits on Nov 20, 2020

  1. se: fix another occurrence of the implicitly defined eval func

    dev-zero committed Nov 20, 2020
  2. optimized some grid reorder function (#1182)

    Improves code-gen with GCC (code still does not fully vectorize due to memory layout/permutation and indirect accesses/reorder). Moved comment to call-side as the comment-related work is now mostly completed upfront of calling general_ci_to_grid.
    hfp committed Nov 20, 2020
  3. Changed "inline" to "static inline" to allow compilation without C99 …

    …standard (#1180)
    hfp committed Nov 20, 2020
  4. account for libxsmm's limitations (#1181)

    Fixed dgemm_simplified according to LIBXSMM's limitations (kernels are limited to Alpha := 1, Beta := { 1, 0 }, and TransA := 'N'; the latter of which applies here).
    hfp committed Nov 20, 2020

Commits on Nov 19, 2020

  1. Added the hybrid cpu/gpu and dgemm backends

    - when running on cpu, the collocate and integrate rely on dgemm for the computation.
    - the GPU backend uses a hybrid solution. Coefficients are computed on the CPU side while all the rest is computed on the GPU side.
    - GPU backend supports multiple GPUs / rank.
    - GPU backend make no difference between orthorhombic and non orthorhombic lattices.
    - use enum for the grid backend and function type
    mtaillefumier committed Nov 19, 2020
    1

Commits on Nov 18, 2020

Commits on Nov 17, 2020

  1. hfx&xas: use new rng_stream_type%shuffle

    dev-zero committed Nov 17, 2020
Older
You can’t perform that action at this time.