Commits · 60ca6e4e23076a667c9b730c35416c33ff833b87 · KIT / GPI / AG / Software / IFOS3D

Feb 06, 2025
- Merged from Branch from Begatim Bytyqi <begatim.bytyqi@kit.edu>: Added OpenMP... · 60ca6e4e
  Holger Obermaier authored 2 months ago
  
  Merged from Branch from Begatim Bytyqi <begatim.bytyqi@kit.edu>: Added OpenMP SIMD pragma and avoid type conversion
  60ca6e4e
- Remove NR_END padding from f3tensor_aligned · a97151b5
  Holger Obermaier authored 2 months ago and Holger Obermaier committed 2 months ago
  
  a97151b5
- Use SIMD to fill the MPI buffers and add asynchronous MPI communication to exchange_s_acoustic · 9bd44b22
  Holger Obermaier authored 2 months ago
  
  9bd44b22
- Add missing SIMD pragmas · d25a530c
  Holger Obermaier authored 2 months ago
  
  d25a530c
- Correct syntax error in loop_blocking.h · b8084978
  Holger Obermaier authored 2 months ago
  
  b8084978
- Allow to select loop blocking implementation at compile time · 58c7d7e4
  Holger Obermaier authored 2 months ago
  
  58c7d7e4
- Add nontemporal attribute for vector destinations in update_v_acoustic · 28fa183f
  Holger Obermaier authored 2 months ago
  
  28fa183f
- Loop fusion in update_v_acoustic · ed0e0061
  Holger Obermaier authored 2 months ago
  
  ed0e0061
- Add asynchronous MPI communication with MPI_Isend and MPI_Irecv · b868141c
  Holger Obermaier authored 2 months ago
  
  b868141c
- Add asynchronous MPI communication with MPI_Isendrecv_replace · b7f3ddd7
  Holger Obermaier authored 2 months ago
  
  b7f3ddd7
- Check if malloc was successful · 1f4c0959
  Holger Obermaier authored 2 months ago
  
  1f4c0959
- Allow to select buffer push and pop method at build time · 93c02fb7
  Holger Obermaier authored 2 months ago
  
  93c02fb7
- Add asynchronous MPI communication with MPI_Isendrecv_replace · 78d63905
  Holger Obermaier authored 2 months ago
  
  78d63905
- Allow to select asynchronous MPI communication at build time · 823eea87
  Holger Obermaier authored 2 months ago
  
  823eea87
- Add asynchronous MPI communication · 05c6becf
  Holger Obermaier authored 3 months ago
  
  05c6becf
- Improve readability · 88c99c58
  Holger Obermaier authored 3 months ago
  
  88c99c58
- Initialize MPI tags · 4feae75d
  Holger Obermaier authored 3 months ago
  
  4feae75d
- Add restrict hints for the compiler · 6eb7d363
  Holger Obermaier authored 3 months ago
  
  6eb7d363
- Add more OpenMP SIMD pragmas · a30a36e7
  Holger Obermaier authored 3 months ago
  
  a30a36e7
- Updated Makefile to allow compilation on login node which has a different... · d3da122e
  Thomas Hertweck authored 2 months ago
  
  Updated Makefile to allow compilation on login node which has a different architecture than genoa nodes; make compiler flags for AOCC and OpenMPI 4 the default.
  d3da122e
Jan 13, 2025
- Corrected error in buffer_pop for the memcpy implementation · e8bc79a4
  Holger Obermaier authored 3 months ago
  
  e8bc79a4
- Use SIMD to fill the MPI buffers · 50aa387c
  Holger Obermaier authored 3 months ago
  
  50aa387c
Jan 10, 2025
- Create header file for buffer_push and buffer_pop · 10a3f0cc
  Holger Obermaier authored 3 months ago
  
  10a3f0cc
- Use SIMD to fill the MPI buffers · 722fa04e
  Holger Obermaier authored 3 months ago
  
  722fa04e
Jan 08, 2025
- Use flat MPI buffer · 6d1f1f60
  Holger Obermaier authored 3 months ago
  
  6d1f1f60
Dec 20, 2024
- Get L2 cache size from sysconf · 0868a745
  Holger Obermaier authored 4 months ago
  
  0868a745
- Finetune i,k blocking factor · 521e4620
  Holger Obermaier authored 4 months ago
  
  521e4620
Dec 19, 2024
- Add option to block only the i and k loop · cbae866c
  Holger Obermaier authored 4 months ago
  
  cbae866c
Dec 11, 2024
- Add nontemporal attribute for vector destinations. · 091fb73d
  Holger Obermaier authored 4 months ago
  
  091fb73d
- Add comments about processed boundaries · 5d756cf2
  Holger Obermaier authored 4 months ago
  
  5d756cf2
- Make loop blocking configurable · 6ea97949
  Holger Obermaier authored 4 months ago
  
  6ea97949
Dec 10, 2024
- Changed blocking factor · e8e79bfb
  Holger Obermaier authored 4 months ago
  
  e8e79bfb
- Revert some of the loop blockings as they did not show any advantage. · f059030d
  Holger Obermaier authored 4 months ago
  
  f059030d
Dec 09, 2024
- Merge from develop_begatim and cleanup · 3b74fa85
  Holger Obermaier authored 4 months ago
  
  3b74fa85
Dec 06, 2024
- Add cache blocking · ded7b380
  Holger Obermaier authored 4 months ago
  
  ded7b380
- Add cache blocking · 210a3b8d
  Holger Obermaier authored 4 months ago
  
  210a3b8d
- Revert "Avoid divisions as they are expensive" · 892b51f6
  Holger Obermaier authored 4 months ago
  
  This reverts commit c70931e1.
  892b51f6
Dec 05, 2024
- Avoid divisions as they are expensive · c70931e1
  Holger Obermaier authored 4 months ago
  
  c70931e1
Dec 04, 2024

Avoid remark: floating point conversion changes vector width. Mixed floating... · de0d197e

Holger Obermaier authored 4 months ago

Avoid remark: floating point conversion changes vector width. Mixed floating point precision requires an up/down cast that will negatively impact performance. [-Rpass-analysis=loop-vectorize]
Add some prama omp simd

de0d197e

Nov 26, 2024
- Replace all compiler specific SIMD pragmas by OpenMP SIMD pragma · ba0e498b
  Holger Obermaier authored 4 months ago
  
  ba0e498b