Skip to content
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

Jump to bottom

TTG device tasks with coroutines #265

Merged
merged 154 commits into from
Feb 24, 2024
Merged

TTG device tasks with coroutines #265

merged 154 commits into from
Feb 24, 2024

Commits on Aug 28, 2023

  1. Add support for HIP 

    Most functionality is an extension/copy of the CUDA support. Also ported
the bspmm example to HIP.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Aug 28, 2023
    Configuration menu
    Copy the full SHA
    31af820 View commit details
    Browse the repository at this point in the history

  2. Bump PaRSEC version to current master 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Aug 28, 2023
    Configuration menu
    Copy the full SHA
    9b817f3 View commit details
    Browse the repository at this point in the history

  3. SPMM: move tiles in the non-cuda version 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Aug 28, 2023
    Configuration menu
    Copy the full SHA
    30d7ee0 View commit details
    Browse the repository at this point in the history

Commits on Aug 30, 2023

  1. Fixes to make Clang happy 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Aug 30, 2023
    Configuration menu
    Copy the full SHA
    81cca94 View commit details
    Browse the repository at this point in the history

Commits on Sep 6, 2023

  1. Try to send from accelerator device directly 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 6, 2023
    Configuration menu
    Copy the full SHA
    975e878 View commit details
    Browse the repository at this point in the history

  2. Fix cpu-only spmm compilation 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 6, 2023
    Configuration menu
    Copy the full SHA
    2b7e67c View commit details
    Browse the repository at this point in the history

Commits on Sep 22, 2023

  1. Add have_hip_op to TT produced by make_tt 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 22, 2023
    Configuration menu
    Copy the full SHA
    7faf783 View commit details
    Browse the repository at this point in the history

Commits on Sep 25, 2023

  1. Add TTG_FORCE_DEVICE_COMM to force use of device-side comm buffers 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 25, 2023
    Configuration menu
    Copy the full SHA
    a423e87 View commit details
    Browse the repository at this point in the history

  2. Bump TA to support ROCM allocators 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 25, 2023
    Configuration menu
    Copy the full SHA
    a7fe69a View commit details
    Browse the repository at this point in the history

  3. Fix device comm code 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 25, 2023
    Configuration menu
    Copy the full SHA
    6d73621 View commit details
    Browse the repository at this point in the history

  4. SPMM: restore write-back to Eigen matrix for verification 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 25, 2023
    Configuration menu
    Copy the full SHA
    88f8da9 View commit details
    Browse the repository at this point in the history

Commits on Sep 28, 2023

  1. We need one cublas/hipblas context per device 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 28, 2023
    Configuration menu
    Copy the full SHA
    3ba4a4f View commit details
    Browse the repository at this point in the history

  2. Add missing header to install 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 28, 2023
    Configuration menu
    Copy the full SHA
    b570108 View commit details
    Browse the repository at this point in the history

Commits on Sep 29, 2023

  1. Bump TA sha to pull in newer Umpire 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 29, 2023
    Configuration menu
    Copy the full SHA
    5afb32b View commit details
    Browse the repository at this point in the history

  2. SPMM: Flush out PaRSEC data between measurements 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 29, 2023
    Configuration menu
    Copy the full SHA
    3aff11c View commit details
    Browse the repository at this point in the history

  3. Minor fixes to device cide 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Sep 29, 2023
    Configuration menu
    Copy the full SHA
    3fa8001 View commit details
    Browse the repository at this point in the history

  4. [cmake] bump MADNESS tag to pull in m-a-d-n-e-s-s/madness#501 

    avoids CMake errors related to lapack including OpenMP_C dependence that MADNESS does not know anything about
    @evaleev
    evaleev committed Sep 29, 2023
    Configuration menu
    Copy the full SHA
    4d60c3f View commit details
    Browse the repository at this point in the history

Commits on Oct 1, 2023

  1. Add missing device::sendk and sendv routines 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 1, 2023
    Configuration menu
    Copy the full SHA
    4f21ccd View commit details
    Browse the repository at this point in the history

  2. Add device-based chain task benchmark 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 1, 2023
    Configuration menu
    Copy the full SHA
    f91fa92 View commit details
    Browse the repository at this point in the history

  3. User needs to include any std:: serialization they use

    @therault @devreal
    therault authored and devreal committed Oct 1, 2023
    Configuration menu
    Copy the full SHA
    d1a6ef3 View commit details
    Browse the repository at this point in the history

  4. Work in progress: adding CUDA version of POTRF task in POTRF example … 

    …-- not compiling
    @therault @devreal
    therault authored and devreal committed Oct 1, 2023
    Configuration menu
    Copy the full SHA
    7576321 View commit details
    Browse the repository at this point in the history

  5. Further steps towards device-enabled POTRF 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 1, 2023
    Configuration menu
    Copy the full SHA
    4a23f42 View commit details
    Browse the repository at this point in the history

Commits on Oct 2, 2023

  1. Add allocator to ttg::buffer 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 2, 2023
    Configuration menu
    Copy the full SHA
    57faac4 View commit details
    Browse the repository at this point in the history

  2. Add first full draft of device-enabled POTRF 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 2, 2023
    Configuration menu
    Copy the full SHA
    194e93d View commit details
    Browse the repository at this point in the history

  3. POTRF: Use the right precompiler guard for CUDA and HIP 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 2, 2023
    Configuration menu
    Copy the full SHA
    e41a4b1 View commit details
    Browse the repository at this point in the history

  4. POTRF: make it run, still fails factorization 

    Notable changes:
- add ttg::device::current_device and ttg::device::current_stream to
  query current streams
- Consistently start counting devices at 0
- The host has ID -2
- Fixes in the broadcast

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 2, 2023
    Configuration menu
    Copy the full SHA
    18577f9 View commit details
    Browse the repository at this point in the history

Commits on Oct 3, 2023

  1. Fix release of data copies in buffer 

    This fixes a segfault in multi-gpu operations. We should not destroy the
full data_t but only remove the host-side copy and release our reference
on the data. The data will be destroyed once all gpu copies have been
removed.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 3, 2023
    Configuration menu
    Copy the full SHA
    b60577c View commit details
    Browse the repository at this point in the history

Commits on Oct 4, 2023

  1. Enable POTRF on hip 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 4, 2023
    Configuration menu
    Copy the full SHA
    a2f4bbe View commit details
    Browse the repository at this point in the history

  2. POTRF: add -n parameters to control number of iterations 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 4, 2023
    Configuration menu
    Copy the full SHA
    9654335 View commit details
    Browse the repository at this point in the history

  3. Minor fixes to compile on Frontier

    @devreal
    devreal committed Oct 4, 2023
    Configuration menu
    Copy the full SHA
    410dc23 View commit details
    Browse the repository at this point in the history

Commits on Oct 5, 2023

  1. buffer: cannot capture this pointer for deleter and move, use raw poi… 

    …nter instead

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 5, 2023
    Configuration menu
    Copy the full SHA
    5b81ac4 View commit details
    Browse the repository at this point in the history

  2. POTRF: construct whole TTG each iteration 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 5, 2023
    Configuration menu
    Copy the full SHA
    403b06b View commit details
    Browse the repository at this point in the history

  3. Add TTG_ENABLE_ASAN CMake option 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 5, 2023
    Configuration menu
    Copy the full SHA
    1998655 View commit details
    Browse the repository at this point in the history

Commits on Oct 6, 2023

  1. Chain benchmark: Disable wait_kernel co_yield 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 6, 2023
    Configuration menu
    Copy the full SHA
    548b65b View commit details
    Browse the repository at this point in the history

  2. PaRSEC: Increase max AM size to 1MB

    @devreal
    devreal committed Oct 6, 2023
    Configuration menu
    Copy the full SHA
    fdf6e2e View commit details
    Browse the repository at this point in the history

  3. Remove TTG cublas_helper, superseeded by devblas helper in examples 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 6, 2023
    Configuration menu
    Copy the full SHA
    9d21225 View commit details
    Browse the repository at this point in the history

Commits on Oct 11, 2023

  1. [cmake] FindTBB can handle OneTBB

    @evaleev
    evaleev committed Oct 11, 2023
    Configuration menu
    Copy the full SHA
    24155bc View commit details
    Browse the repository at this point in the history

  2. [cmake] introduce TTG_HAVE_DEVICE

    @evaleev
    evaleev committed Oct 11, 2023
    Configuration menu
    Copy the full SHA
    ee82565 View commit details
    Browse the repository at this point in the history

  3. [cmake] introduce TTG_HAVE_DEVICE used to conditionally compile *port… 

    …able* device-specific code
    @evaleev
    evaleev committed Oct 11, 2023
    Configuration menu
    Copy the full SHA
    0993f79 View commit details
    Browse the repository at this point in the history

  4. bspmm: support other reals ... depends on ValeevGroup/BTAS#164 so bum… 

    …p BTAS tag
    @evaleev
    evaleev committed Oct 11, 2023
    Configuration menu
    Copy the full SHA
    366e60e View commit details
    Browse the repository at this point in the history

  5. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' into … 

    …potrf-cuda-wip
    @devreal
    devreal authored Oct 11, 2023
    Configuration menu
    Copy the full SHA
    94f3a81 View commit details
    Browse the repository at this point in the history

  6. Merge pull request #7 from therault/potrf-cuda-wip 

    Add accelerator support for POTRF
    @devreal
    devreal authored Oct 11, 2023
    Configuration menu
    Copy the full SHA
    fa4aedf View commit details
    Browse the repository at this point in the history

Commits on Oct 30, 2023

  1. Add debugging capabilities for POTRF benchmark 

    Set DEBUG_TILES_VALUES=1 to enable norm2 check of tiles.

This also fixes issues with the parsec data wrapper. This needs some
more thought and might come back to haunt us if we try to use data with
extrinsic tracking information.

Also adds ttg::device::Device to query the ID and type of an instance of
the current execution space (e.g., the currently assigned GPU).

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 30, 2023
    Configuration menu
    Copy the full SHA
    148a3ad View commit details
    Browse the repository at this point in the history

  2. Cleanup devblas support in rocm support 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 30, 2023
    Configuration menu
    Copy the full SHA
    57f914e View commit details
    Browse the repository at this point in the history

  3. Lazily compute norm on MatrixTile 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 30, 2023
    Configuration menu
    Copy the full SHA
    162a0a1 View commit details
    Browse the repository at this point in the history

  4. Fix previous commits 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 30, 2023
    Configuration menu
    Copy the full SHA
    9dbe648 View commit details
    Browse the repository at this point in the history

  5. Fix previous commit 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 30, 2023
    Configuration menu
    Copy the full SHA
    fdc3c95 View commit details
    Browse the repository at this point in the history

  6. Fix bspmm-cuda after device and devblas helper changes 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 30, 2023
    Configuration menu
    Copy the full SHA
    7200e1c View commit details
    Browse the repository at this point in the history

Commits on Oct 31, 2023

  1. POTRF: make tile argument to dispatcher const 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Oct 31, 2023
    Configuration menu
    Copy the full SHA
    9d15ba1 View commit details
    Browse the repository at this point in the history

Commits on Nov 3, 2023

  1. Multi-GPU: Play nice with the PaRSEC accelerator 

    We need to a) push tiles back if they move from RW to RO tasks, and b)
create a fake task-class that contains all the flows we actually need
(i.e., that are part of to_device()). This way PaRSEC can select the
right device and we get all the transfers in multi-GPU settings.

This needs to be revisited in the PaRSEC backend. We should not need to
do the push-back and create a fake task-class to select the correct
device. That requires an overhaul of the device backend though...

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 3, 2023
    Configuration menu
    Copy the full SHA
    166bdec View commit details
    Browse the repository at this point in the history

  2. Bump PaRSEC and MADNESS tags for proper tear-down 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 3, 2023
    Configuration menu
    Copy the full SHA
    33a5a8d View commit details
    Browse the repository at this point in the history

Commits on Nov 6, 2023

  1. Bump PaRSEC reference for TTG-specific fixes 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 6, 2023
    Configuration menu
    Copy the full SHA
    f9375c8 View commit details
    Browse the repository at this point in the history

Commits on Nov 15, 2023

  1. Bump PaRSEC ref for reworked GPU backend 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 15, 2023
    Configuration menu
    Copy the full SHA
    9903f25 View commit details
    Browse the repository at this point in the history

Commits on Nov 16, 2023

  1. Add a LEVEL_ZERO implementation for the GEMM kernel in spmm_cuda.cc

    @therault
    therault committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    e9271ff View commit details
    Browse the repository at this point in the history

  2. Add level zero support and refactor device naming 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    e20758a View commit details
    Browse the repository at this point in the history

  3. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' of gi… 

    …thub.com:devreal/ttg into ttg-device-support-master-coro-with-stream-tasks
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    af6f684 View commit details
    Browse the repository at this point in the history

  4. Attempt at cmakery for level-zero

    @therault
    therault committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    97d0bf6 View commit details
    Browse the repository at this point in the history

  5. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' of gi… 

    …thub.com:devreal/ttg into ttg-device-support-master-coro-with-stream-tasks
    @therault
    therault committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    c2da1f5 View commit details
    Browse the repository at this point in the history

  6. Missing declaration for level zero cmakery

    @therault
    therault committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    6af1828 View commit details
    Browse the repository at this point in the history

  7. Agnostic async copy and non-device support 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    0be1212 View commit details
    Browse the repository at this point in the history

  8. Add missing parsec backend header file to CMake 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    a119828 View commit details
    Browse the repository at this point in the history

  9. Add MKL dependency to spmm-lz

    @therault
    therault committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    4f125ad View commit details
    Browse the repository at this point in the history

  10. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' of gi… 

    …thub.com:devreal/ttg into ttg-device-support-master-coro-with-stream-tasks
    @therault
    therault committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    67434d1 View commit details
    Browse the repository at this point in the history

  11. Work on device_gpu_module instead of device_module 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    03b54bb View commit details
    Browse the repository at this point in the history

  12. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' of gi… 

    …thub.com:devreal/ttg into ttg-device-support-master-coro-with-stream-tasks
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    a9e5959 View commit details
    Browse the repository at this point in the history

  13. Get sycl queue in spmm 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    dde416c View commit details
    Browse the repository at this point in the history

  14. Add parsec level-zero include 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    5569b57 View commit details
    Browse the repository at this point in the history

  15. Return sycl queue as non-const 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    86b92db View commit details
    Browse the repository at this point in the history

  16. Use the right parsec device module types 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    5aac5c2 View commit details
    Browse the repository at this point in the history

  17. Add inline in device.h 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    afeda46 View commit details
    Browse the repository at this point in the history

  18. PaRSEC: Add have_level_zero_op() 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    d1fb80f View commit details
    Browse the repository at this point in the history

  19. Add missing have_level_zero_op_non_type_t 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    bfa3fde View commit details
    Browse the repository at this point in the history

  20. Fix bug in level zero detection of derived TT 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    68e4825 View commit details
    Browse the repository at this point in the history

  21. Don't ever assume that 2 is the first device 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 16, 2023
    Configuration menu
    Copy the full SHA
    9afb6f8 View commit details
    Browse the repository at this point in the history

Commits on Nov 20, 2023

  1. [cmake] cleanup FindOrFetchTiledArray.cmake

    @evaleev
    evaleev committed Nov 20, 2023
    Configuration menu
    Copy the full SHA
    961b0d3 View commit details
    Browse the repository at this point in the history

  2. [cmake] bump TA tag to pull in ValeevGroup/tiledarray#434

    @evaleev
    evaleev committed Nov 20, 2023
    Configuration menu
    Copy the full SHA
    78876e5 View commit details
    Browse the repository at this point in the history

Commits on Nov 27, 2023

  1. Proper use of device ID conversion functions 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    cf757ee View commit details
    Browse the repository at this point in the history

  2. Bump parsec commit to rebased fork 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    d660b07 View commit details
    Browse the repository at this point in the history

  3. There is no allocators for level zero in TiledArray yet 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    5 Configuration menu
    Copy the full SHA
    0ab8f47 View commit details
    Browse the repository at this point in the history

  4. Fix misplaced assert 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    fb5f111 View commit details
    Browse the repository at this point in the history

  5. ze: add the missing link-libraries to spmm 

    Signed-off-by: Aurelien Bouteiller <[email protected]>
    @abouteiller
    abouteiller committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    383d36d View commit details
    Browse the repository at this point in the history

  6. Use std::size_t in parsec matrix wrapper 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    f4b482a View commit details
    Browse the repository at this point in the history

  7. MatrixTile: use std::size_t consistently 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    0b1325e View commit details
    Browse the repository at this point in the history

  8. Fun with signed integers in parsec matrix wrapper 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 27, 2023
    Configuration menu
    Copy the full SHA
    1f3619a View commit details
    Browse the repository at this point in the history

Commits on Nov 29, 2023

  1. Fix is_dummy name collision in parsec_ttg_task_base_t 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Nov 29, 2023
    Configuration menu
    Copy the full SHA
    81c25e3 View commit details
    Browse the repository at this point in the history

Commits on Dec 4, 2023

  1. Bump PaRSEC version to get first set of eviction fixes 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 4, 2023
    Configuration menu
    Copy the full SHA
    bf821a5 View commit details
    Browse the repository at this point in the history

Commits on Dec 5, 2023

  1. task dummy() -> is_dummy() 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 5, 2023
    Configuration menu
    Copy the full SHA
    6727faf View commit details
    Browse the repository at this point in the history

  2. Fix usage of wrong device ID in buffer 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 5, 2023
    Configuration menu
    Copy the full SHA
    049297e View commit details
    Browse the repository at this point in the history

  3. Bump PaRSEC tag for more memory oversubscription fixes 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 5, 2023
    Configuration menu
    Copy the full SHA
    8146481 View commit details
    Browse the repository at this point in the history

Commits on Dec 13, 2023

  1. Map blas/solver handles to streams, not threads and streams 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 13, 2023
    Configuration menu
    Copy the full SHA
    b9727a4 View commit details
    Browse the repository at this point in the history

Commits on Dec 18, 2023

  1. Naming convetions: rename ttg::Buffer 

    Objects in TTG start with a capital letter so rename ttg::buffer to
ttg::Buffer to be consistent.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 18, 2023
    Configuration menu
    Copy the full SHA
    603af4e View commit details
    Browse the repository at this point in the history

  2. Remove legacy device implementation files 

    Both view.h and make_device_tt.h were the initial implementation and are
not used anymore.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 18, 2023
    Configuration menu
    Copy the full SHA
    9afe0c0 View commit details
    Browse the repository at this point in the history

  3. Put all device functions into ttg::device namespace 

    Except for ttg::Buffer, which is separate and may be used by non-device
tasks/data structures.

The following were renamed:
- to_device -> select
- wait_kernel -> wait

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 18, 2023
    Configuration menu
    Copy the full SHA
    f424bcd View commit details
    Browse the repository at this point in the history

Commits on Dec 19, 2023

  1. Add device-related fwd-decl to madness backend 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 19, 2023
    Configuration menu
    Copy the full SHA
    3bf1e16 View commit details
    Browse the repository at this point in the history

Commits on Dec 20, 2023

  1. Include fwd.h instead of impl_selector.h 

    fwd.h provides the symbols from the implementation namespace

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    913f792 View commit details
    Browse the repository at this point in the history

  2. Bump catch2 version to latest release 

    This is needed to get the fix for catchorg/Catch2#2421

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    7c6044a View commit details
    Browse the repository at this point in the history

  3. Fix clang compile issues 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    ed055fb View commit details
    Browse the repository at this point in the history

  4. Forward keyT in make_tt overload 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    0c7d291 View commit details
    Browse the repository at this point in the history

  5. Bump github CI Ubuntu to 22.04 to get boost 1.74 needed for TA 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    b486a2f View commit details
    Browse the repository at this point in the history

  6. Add stub implementations of ttvalue and buffer for madness

    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    555ae2f View commit details
    Browse the repository at this point in the history

  7. More fixes to ttvalue and ptr 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    5cf2502 View commit details
    Browse the repository at this point in the history

  8. More catch2 include fixes 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    220c038 View commit details
    Browse the repository at this point in the history

  9. Fix ttvalue and buffer in parsec 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    e9b3233 View commit details
    Browse the repository at this point in the history

  10. Add derived_has_*_op() to madness backend 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    65ffd4c View commit details
    Browse the repository at this point in the history

  11. Allow for non-coro device tasks

    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    e7bba1b View commit details
    Browse the repository at this point in the history

  12. Re-enbale disabled unit tests

    @devreal
    devreal committed Dec 20, 2023
    Configuration menu
    Copy the full SHA
    b653b9b View commit details
    Browse the repository at this point in the history

Commits on Dec 21, 2023

  1. Enable madness serialization for madness backend buffer 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    82a7b2a View commit details
    Browse the repository at this point in the history

  2. Only build device_coro test if device support is detected 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    0d64297 View commit details
    Browse the repository at this point in the history

  3. Fix possible use-after-free in TT::release 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    5f876e3 View commit details
    Browse the repository at this point in the history

  4. Use key directly in task_snprintf instead of locals 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    6f645df View commit details
    Browse the repository at this point in the history

  5. Remove legacy devicemock example 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    ba6f7a1 View commit details
    Browse the repository at this point in the history

  6. Fix device chain benchmark 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    c899f07 View commit details
    Browse the repository at this point in the history

  7. Comment preliminary mra device code pieces 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    bd454dd View commit details
    Browse the repository at this point in the history

  8. Gracefully handle missing TA allocators 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    ec67c6b View commit details
    Browse the repository at this point in the history

  9. Use stock doxygen for relase test 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    8ccab44 View commit details
    Browse the repository at this point in the history

  10. Fix const handling in reducer tasks 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    beed36a View commit details
    Browse the repository at this point in the history

  11. Only enable cuda op if cuda was detected 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 21, 2023
    Configuration menu
    Copy the full SHA
    f30f901 View commit details
    Browse the repository at this point in the history

Commits on Dec 22, 2023

  1. Fix streams unit test 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 22, 2023
    Configuration menu
    Copy the full SHA
    ece8cc3 View commit details
    Browse the repository at this point in the history

  2. Make sure new reducer target copies are marked mutable 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 22, 2023
    Configuration menu
    Copy the full SHA
    db4f736 View commit details
    Browse the repository at this point in the history

  3. Add ttg/ttvalue.h to cmake 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 22, 2023
    Configuration menu
    Copy the full SHA
    4547d8c View commit details
    Browse the repository at this point in the history

  4. Reset readers on reduced copy 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Dec 22, 2023
    Configuration menu
    Copy the full SHA
    f9900fe View commit details
    Browse the repository at this point in the history

Commits on Jan 19, 2024

  1. Fix keystr string name 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Jan 19, 2024
    Configuration menu
    Copy the full SHA
    d42cb92 View commit details
    Browse the repository at this point in the history

Commits on Jan 29, 2024

  1. PaRSEC task: copy task locals to comply with profiling reqs 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Jan 29, 2024
    Configuration menu
    Copy the full SHA
    7b9c248 View commit details
    Browse the repository at this point in the history

  2. SPMM: Make writing C back into Eigen optional 

    The Eigen format is not particularly suitable for thread-parallel
writes so for benchmarks we drop the last part of pushing tiles back
into the Eigen matrix. Tiles are still communicated and reduced properly.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Jan 29, 2024
    Configuration menu
    Copy the full SHA
    641c81f View commit details
    Browse the repository at this point in the history

  3. SPMM: distribute A/B over all procs and throttle broadcasts 

    We distribute A and B over all three dimensions instead of just the first
two. The 2.5D approach does not seem suitable for our approach
since matrices will likely be distributed across all processes.

Also, add the ability to throttle the number of parallel bcasts to
avoid overloading the network. This should be revisited, since it seems
to the an issue esp with UCX as a backend in OMPI.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Jan 29, 2024
    Configuration menu
    Copy the full SHA
    d584e06 View commit details
    Browse the repository at this point in the history

Commits on Jan 30, 2024

  1. btas::Tensor: do not fill with zeroes when constructing (e.g., in zer… 

    …o-copy serialization), unless necessary
    @evaleev
    evaleev committed Jan 30, 2024
    Configuration menu
    Copy the full SHA
    a85c2c4 View commit details
    Browse the repository at this point in the history

  2. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' of gi… 

    …thub.com:devreal/ttg into ttg-device-support-master-coro-with-stream-tasks
    @devreal
    devreal committed Jan 30, 2024
    Configuration menu
    Copy the full SHA
    7a5e2a5 View commit details
    Browse the repository at this point in the history

Commits on Jan 31, 2024

  1. [cmake] bump MADNESS, BTAS, and TA tags to sync with latest; MADNESS … 

    …points to main repo now
    @evaleev
    evaleev committed Jan 31, 2024
    Configuration menu
    Copy the full SHA
    1f2a4fb View commit details
    Browse the repository at this point in the history

  2. fixup

    @evaleev
    evaleev committed Jan 31, 2024
    Configuration menu
    Copy the full SHA
    1125f96 View commit details
    Browse the repository at this point in the history

  3. introduce initialize_fpe to turn on/off floating-point exceptions

    @evaleev
    evaleev committed Jan 31, 2024
    Configuration menu
    Copy the full SHA
    549ba94 View commit details
    Browse the repository at this point in the history

Commits on Feb 1, 2024

  1. can use bundled Boost.CallableTraits only if don't have real Boost an… 

    …d TTG_FETCH_BOOST is not set
    @evaleev
    evaleev committed Feb 1, 2024
    Configuration menu
    Copy the full SHA
    f05530f View commit details
    Browse the repository at this point in the history

  2. bump MADNESS tag to fix up CI issues

    @evaleev
    evaleev committed Feb 1, 2024
    Configuration menu
    Copy the full SHA
    9765f75 View commit details
    Browse the repository at this point in the history

  3. [ci] install libboost-random-dev to help TA build itself

    @evaleev
    evaleev committed Feb 1, 2024
    Configuration menu
    Copy the full SHA
    b70444c View commit details
    Browse the repository at this point in the history

Commits on Feb 2, 2024

  1. implemented Read{A,B} that read A and B more intelligently

    @evaleev
    evaleev committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    1ffaf6a View commit details
    Browse the repository at this point in the history

  2. Bump PaRSEC tag 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    41cc814 View commit details
    Browse the repository at this point in the history

  3. Implement support for inlining of objects with SplitMetadata 

    For small objects it is beneficial of packing the data into the
activation message instead of going through an extra GET.


Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    8993e49 View commit details
    Browse the repository at this point in the history

  4. SPMM: Allow changing the work distribution 

    Work may be performed by the owner of A, B, or C, or the default
honoring the R parameter.

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    becdbf4 View commit details
    Browse the repository at this point in the history

  5. Read{A,B} can use MultipleK schedule

    @evaleev
    evaleev committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    3dad2c1 View commit details
    Browse the repository at this point in the history

  6. Enforce strict ordering of broadcasts in column/row 

    For matrix A, we make sure that tiles are sent out in column-order (left to right).
For matrix B, we want to make sure that tiles are sent out in row-order (top to bottom).

Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    4416f3b View commit details
    Browse the repository at this point in the history

  7. Merge branch 'ttg-device-support-master-coro-with-stream-tasks' of gi… 

    …thub.com:devreal/ttg into ttg-device-support-master-coro-with-stream-tasks
    @devreal
    devreal committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    8fd7be9 View commit details
    Browse the repository at this point in the history

  8. Bump PaRSEC sha 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 2, 2024
    Configuration menu
    Copy the full SHA
    59779c0 View commit details
    Browse the repository at this point in the history

Commits on Feb 14, 2024

  1. Bump PaRSEC version for MPI backend instrumentation 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 14, 2024
    Configuration menu
    Copy the full SHA
    94b2b75 View commit details
    Browse the repository at this point in the history

Commits on Feb 23, 2024

  1. Add missing <numeric> header for CI 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    c06654f View commit details
    Browse the repository at this point in the history

  2. Order broadcast keys so that we send to myrank+1 first 

    Signed-off-by: Joseph Schuchart <[email protected]>
    @devreal
    devreal committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    b8c2cc8 View commit details
    Browse the repository at this point in the history

  3. Reverted ReadA/B additions 

     Partially reverts 1ffaf6a and 3dad2c1
    @evaleev
    evaleev committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    67b0fb3 View commit details
    Browse the repository at this point in the history

  4. [cleanup] Design-Device.md

    @evaleev
    evaleev committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    b741e78 View commit details
    Browse the repository at this point in the history

  5. [dox] new vars config.in.h [skip ci]

    @evaleev
    evaleev committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    ed270da View commit details
    Browse the repository at this point in the history

  6. [dox] stub for device::Device

    @evaleev
    evaleev committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    12d1c1f View commit details
    Browse the repository at this point in the history

  7. [dox] stub for ttg_madness::Buffer

    @evaleev
    evaleev committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    27c03f7 View commit details
    Browse the repository at this point in the history

  8. [cleanup] device/task.h

    @evaleev
    evaleev committed Feb 23, 2024
    Configuration menu
    Copy the full SHA
    fcbd2e8 View commit details
    Browse the repository at this point in the history

Commits on Feb 24, 2024

  1. unit test informational logging interferes with catch_discover_tests,… 

    … only produce if Catch's `verbosity` != `quiet`
    @evaleev
    evaleev committed Feb 24, 2024
    Configuration menu
    Copy the full SHA
    9cd39b8 View commit details
    Browse the repository at this point in the history

  2. pass required Boost version to FindOrFetchBoost

    @evaleev
    evaleev committed Feb 24, 2024
    Configuration menu
    Copy the full SHA
    8104b49 View commit details
    Browse the repository at this point in the history

  3. HIP LANGUAGE needs CMake 3.21 or later

    @evaleev
    evaleev committed Feb 24, 2024
    Configuration menu
    Copy the full SHA
    1826821 View commit details
    Browse the repository at this point in the history

  4. update INSTALL.md

    @evaleev
    evaleev committed Feb 24, 2024
    Configuration menu
    Copy the full SHA
    56ed942 View commit details
    Browse the repository at this point in the history