diff --git a/doc.zih.tu-dresden.de/docs/archive/HardwareAltix.md b/doc.zih.tu-dresden.de/docs/archive/HardwareAltix.md index 7e163a400d67d6025996229bb20d3d547beffc88..7912181e49c8c113b601c419d64cd859c4163b69 100644 --- a/doc.zih.tu-dresden.de/docs/archive/HardwareAltix.md +++ b/doc.zih.tu-dresden.de/docs/archive/HardwareAltix.md @@ -1,5 +1,3 @@ - - # HPC Component SGI Altix The SGI Altix 4700 is a shared memory system with dual core Intel @@ -7,34 +5,36 @@ Itanium 2 CPUs (Montecito) operated by the Linux operating system SuSE SLES 10 with a 2.6 kernel. Currently, the following Altix partitions are installed at ZIH: -\|\*Name \*\|\*Total Cores \*\|**Compute Cores**\|**Memory per Core**\| -\| Mars \|384 \|348 \|1 GB\| \|Jupiter \|512 \|506 \|4 GB\| \|Saturn -\|512 \|506 \|4 GB\| \|Uranus \|512 \|506 \|4 GB\| \|Neptun \|128 \|128 -\|1 GB\| +|Name|Total Cores|Compute Cores|Memory per Core| +|:----|:----|:----|:----| +| Mars |384 |348 |1 GB| +|Jupiter |512 |506 |4 GB| +|Saturn |512 |506 |4 GB| +|Uranus |512 |506|4 GB| +|Neptun |128 |128 |1 GB| -\<P> The jobs for these partitions (except \<TT>Neptun\</TT>) are -scheduled by the [Platform LSF](Platform LSF) batch system running on -`mars.hrsk.tu-dresden.de`. The actual placement of a submitted job may +The jobs for these partitions (except Neptun) are scheduled by the [Platform LSF](PlatformLSF.md) +batch system running on `mars.hrsk.tu-dresden.de`. The actual placement of a submitted job may depend on factors like memory size, number of processors, time limit. -## Filesystems All partitions share the same CXFS filesystems `/work` and `/fastfs`. ... [more information](FileSystems) +## Filesystems + +All partitions share the same CXFS filesystems `/work` and `/fastfs`. ## ccNuma Architecture -The SGI Altix has a ccNUMA architecture, which stands for Cache Coherent -Non-Uniform Memory Access. It can be considered as a SM-MIMD (*shared -memory - multiple instruction multiple data*) machine. The SGI ccNuma -system has the following properties: +The SGI Altix has a ccNUMA architecture, which stands for Cache Coherent Non-Uniform Memory Access. +It can be considered as a SM-MIMD (*shared memory - multiple instruction multiple data*) machine. +The SGI ccNuma system has the following properties: -- Memory is physically distributed but logically shared -- Memory is kept coherent automatically by hardware. -- Coherent memory: memory is always valid (caches hold copies) -- Granularity is L3 cacheline (128 B) -- Bandwidth of NumaLink4 is 6.4 GB/s +- Memory is physically distributed but logically shared +- Memory is kept coherent automatically by hardware. +- Coherent memory: memory is always valid (caches hold copies) +- Granularity is L3 cacheline (128 B) +- Bandwidth of NumaLink4 is 6.4 GB/s -The ccNuma is a compromise between a distributed memory system and a -flat symmetric multi processing machine (SMP). Altough the memory is -shared, the access properties are not the same. +The ccNuma is a compromise between a distributed memory system and a flat symmetric multi processing +machine (SMP). Altough the memory is shared, the access properties are not the same. ## Compute Module @@ -79,13 +79,9 @@ properties: | L3 cache | 9 MB, 12 clock latency | | front side bus | 128 bit x 200 MHz | -The theoretical peak performance of all Altix partitions is hence about -13.1 TFLOPS. - -The processor has hardware support for efficient software pipelining. -For many scientific applications it provides a high sustained -performance exceeding the performance of RISC CPUs with similar peak -performance. On the down side is the fact that the compiler has to -explicitely discover and exploit the parallelism in the application. +The theoretical peak performance of all Altix partitions is hence about 13.1 TFLOPS. -<span class="twiki-macro COMMENT"></span> +The processor has hardware support for efficient software pipelining. For many scientific +applications it provides a high sustained performance exceeding the performance of RISC CPUs with +similar peak performance. On the down side is the fact that the compiler has to explicitely discover +and exploit the parallelism in the application. diff --git a/doc.zih.tu-dresden.de/docs/archive/SystemAltix.md b/doc.zih.tu-dresden.de/docs/archive/SystemAltix.md index 6f26ccc8ce43edbe1d3ad1e7c6e0402ea2db7c99..504d983cf142662f2c615775d91a29ddf15b9bf5 100644 --- a/doc.zih.tu-dresden.de/docs/archive/SystemAltix.md +++ b/doc.zih.tu-dresden.de/docs/archive/SystemAltix.md @@ -1,29 +1,25 @@ # SGI Altix -**`%RED%This page is deprecated! The SGI Atlix is a former system! [[Compendium.Hardware][(Current hardware)]]%ENDCOLOR%`** +**This page is deprecated! The SGI Atlix is a former system!** -The SGI Altix is shared memory system for large parallel jobs using up -to 2000 cores in parallel ( [information on the -hardware](HardwareAltix)). It's partitions are Mars (login), Jupiter, -Saturn, Uranus, and Neptun (interactive). +The SGI Altix is shared memory system for large parallel jobs using up to 2000 cores in parallel ( +[information on the hardware](HardwareAltix.md)). It's partitions are Mars (login), Jupiter, Saturn, +Uranus, and Neptun (interactive). ## Compiling Parallel Applications -This installation of the Message Passing Interface supports the MPI 1.2 -standard with a few MPI-2 features (see `man mpi` ). There is no command -like `mpicc`, instead you just have to use the normal compiler (e.g. -`icc`, `icpc`, or `ifort`) and append `-lmpi` to the linker command -line. Since the include files as well as the library are in standard -directories there is no need to append additional library- or -include-paths. +This installation of the Message Passing Interface supports the MPI 1.2 standard with a few MPI-2 +features (see `man mpi` ). There is no command like `mpicc`, instead you just have to use the normal +compiler (e.g. `icc`, `icpc`, or `ifort`) and append `-lmpi` to the linker command line. Since the +include files as well as the library are in standard directories there is no need to append +additional library- or include-paths. -- Note for C++ programmers: You need to link with - `-lmpi++abi1002 -lmpi` instead of `-lmpi`. -- Note for Fortran programmers: The MPI module is only provided for - the Intel compiler and does not work with gfortran. +- Note for C++ programmers: You need to link with `-lmpi++abi1002 -lmpi` instead of `-lmpi`. +- Note for Fortran programmers: The MPI module is only provided for the Intel compiler and does not + work with gfortran. -Please follow these following guidelines to run your parallel program -using the batch system on Mars. +Please follow these following guidelines to run your parallel program using the batch system on +Mars. ## Batch system @@ -42,24 +38,27 @@ user's job. Normally a job can be submitted with these data: ### LSF -The batch sytem on Atlas is LSF. For general information on LSF, please -follow [this link](PlatformLSF). +The batch sytem on Atlas is LSF. For general information on LSF, please follow +[this link](PlatformLSF.md). ### Submission of Parallel Jobs -The MPI library running on the Altix is provided by SGI and highly -optimized for the ccNUMA architecture of this machine. However, -communication within a partition is faster than across partitions. Take -this into consideration when you submit your job. +The MPI library running on the Altix is provided by SGI and highly optimized for the ccNUMA +architecture of this machine. However, communication within a partition is faster than across +partitions. Take this into consideration when you submit your job. Single-partition jobs can be started like this: - <span class='WYSIWYG_HIDDENWHITESPACE'> </span>bsub -R "span[hosts=1]" -n 16 mpirun -np 16 a.out<span class='WYSIWYG_HIDDENWHITESPACE'> </span> +```Bash +bsub -R "span[hosts=1]" -n 16 mpirun -np 16 a.out< +``` Really large jobs with over 256 CPUs might run over multiple partitions. Cross-partition jobs can be submitted via PAM like this - <span class='WYSIWYG_HIDDENWHITESPACE'> </span>bsub -n 1024 pamrun a.out<span class='WYSIWYG_HIDDENWHITESPACE'> </span> +```Bash +bsub -n 1024 pamrun a.out +``` ### Batch Queues @@ -70,5 +69,3 @@ Cross-partition jobs can be submitted via PAM like this | `intermediate` | `all` | `min. 64, max. 255` | `12h` | `120h` | | `large` | `all` | `min.256, max.1866` | `12h` | `24h` | | `ilr` | `selected users` | `min. 1, max. 768` | `12h` | `24h` | - --- Main.UlfMarkwardt - 2013-02-27