#KBLAS - Twitter Hashtag

over 4 years ago

We are happy to report that recursive Level-3 #BLAS for triangular matrices from our #KBLAS library was incorporated to @NEC_corp numerical library collection v2.3! We have also released a new branch w/ #HIP support for @AMD GPUs. Download at https://t.co/ZtR0Zvh8Ns. #SC21 #HPC

KAUST_ECRC's tweet photo. We are happy to report that recursive Level-3 #BLAS for triangular matrices from our #KBLAS library was incorporated to @NEC_corp numerical library collection v2.3!

We have also released a new branch w/ #HIP support for @AMD GPUs. Download at https://t.co/ZtR0Zvh8Ns.

#SC21 #HPC https://t.co/6GZzcckKXW

1

4

1

0

Hatem Ltaief @HatemLtaief

over 5 years ago

The nice think in #KBLAS approach with the recursive formulation is portability. It relies on the optimized BLAS libraries available in the system, so no need for low-level implementations. Performance may still be achieved though, as shown in the graph.

0

1

0

Hatem Ltaief @HatemLtaief

over 5 years ago

Performance comparisons of tall-and-skinny DTRMM triangular matrix-matrix multiplication (m x 512). Important Level-3 #BLAS kernel for eigenvalue, inversion, etc. #KBLAS Vs vendors: #BLIS/#OpenBLAS on #AMD, #MKL on #Intel, #NLC on #NEC, #cuBLAS on #NVIDIA. https://t.co/jwnccYvz8w

HatemLtaief's tweet photo. Performance comparisons of tall-and-skinny DTRMM triangular matrix-matrix multiplication (m x 512). Important Level-3 #BLAS kernel for eigenvalue, inversion, etc. #KBLAS Vs vendors: #BLIS/#OpenBLAS on #AMD, #MKL on #Intel, #NLC on #NEC, #cuBLAS on #NVIDIA. https://t.co/jwnccYvz8w https://t.co/cwHYdzx6Gk

3

8

1

0

Hatem Ltaief @HatemLtaief

over 5 years ago

@five9a2 Thanks for pointing out the mixed precision support in #BLIS! Certainly something cool to try out. Looks like #BLIS improved big time since our last comparison against #KBLAS. https://t.co/jwnccYvz8w

1

3

0

KAUST ECRC @KAUST_ECRC

over 7 years ago

Releasing #KBLAS with support to low-rank #matrix approximations: a major building block for future high performance linear algebra libraries on #nvidia #GPUs. Try it out at https://t.co/j9jpVUT54k. #SC18 #HPC @KAUST_News @NVIDIADC @amcharara

KAUST_ECRC's tweet photo. Releasing #KBLAS with support to low-rank #matrix approximations: a major building block for future high performance linear algebra libraries on #nvidia #GPUs. Try it out at https://t.co/j9jpVUT54k. #SC18 #HPC @KAUST_News @NVIDIADC @amcharara https://t.co/fTMfOEh1HS

0

2

0

Hatem Ltaief @HatemLtaief

almost 8 years ago

Time is up! Giving this week two talks at @europar2018 in Torino, Italy about #HiCMA and #KBLAS. Joint work w/ @kadir__akbudak @amcharara @cray_inc @KAUST_ECRC

HatemLtaief's tweet photo. Time is up! Giving this week two talks at @europar2018 in Torino, Italy about #HiCMA and #KBLAS. Joint work w/ @kadir__akbudak @amcharara @cray_inc @KAUST_ECRC https://t.co/ZT6IFMmZI7

0

6

1

0

KAUST ECRC @KAUST_ECRC

about 8 years ago

We have a PhD defense coming up: Ali Charara @amcharara from @cemseKAUST, the architect of #KBLAS! #GPU #HPC #ML @KAUST_News

0

3

2

0

KAUST ECRC @KAUST_ECRC

about 8 years ago

The #KBLAS software project is freely available at https://t.co/lgXhcXyJOU. Congrats to our PhD student @amcharara! @KAUST_News https://t.co/CI6onY5Wq1

Hatem Ltaief @HatemLtaief

about 8 years ago

Leveraging #KBLAS library to support #batch tile low-rank GEMM: a building block toward supporting data-sparse applications on @NvidiaAI #GPUs. #HPC #BigData convergence. Paper accepted @europar2018. Joint work w/ @amcharara from @cemseKAUST and D. Keyes @KAUST_HPC @HPC_Guru

HatemLtaief's tweet photo. Leveraging #KBLAS library to support #batch tile low-rank GEMM: a building block toward supporting data-sparse applications on @NvidiaAI #GPUs. #HPC #BigData convergence. Paper accepted @europar2018. Joint work w/ @amcharara from @cemseKAUST and D. Keyes @KAUST_HPC @HPC_Guru https://t.co/S5UBNUDneI

0

6

1

0

1

0

Hatem Ltaief @HatemLtaief

about 8 years ago

Leveraging #KBLAS library to support #batch tile low-rank GEMM: a building block toward supporting data-sparse applications on @NvidiaAI #GPUs. #HPC #BigData convergence. Paper accepted @europar2018. Joint work w/ @amcharara from @cemseKAUST and D. Keyes @KAUST_HPC @HPC_Guru

0

6

1

0

Bilel Hadri @mnoukhiya

over 8 years ago

Prof. David Keyes Director @KAUST_ECRC @KAUST_News talks about Algorithms Adaptations to Extreme Scale Computing #HPC_Saudi #HICMA #KBLAS @nvidia #GPU #KSVD @cray_inc #LibSci #MOAO @Obs_Paris @KAUST_HPC

mnoukhiya's tweet photo. Prof. David Keyes Director @KAUST_ECRC @KAUST_News talks about Algorithms Adaptations to Extreme Scale Computing #HPC_Saudi #HICMA #KBLAS @nvidia #GPU #KSVD @cray_inc #LibSci #MOAO @Obs_Paris @KAUST_HPC https://t.co/goBPU0FpeF

0

14

8

0

KAUST ECRC @KAUST_ECRC

over 8 years ago

Our PhD student @amcharara presenting #KBLAS for batched linear algebra operations on #GPU #HPC @HatemLtaief @KAUST_News

0

5

3

0

KAUST ECRC @KAUST_ECRC

over 8 years ago

Introducing new batched tile low-rank #GEMM on #GPU by @KAUST_CEMSE PhD student @amcharara, a much needed BLAS operation for H-matrix! #KBLAS #HiCMA @KAUST_HPC