boltzgen, branch v0.1.2 Symbolic generation of LBM kernels Improve lattice, model selection error reporting 2019-11-02T16:32:41+00:00 Adrian Kummerlaender 2019-11-02T16:32:41+00:00 05e74fb112f5b5f645b649c587d18052c7b7f9df 






Import `sympy.ccode` inside templates instead of as argument
2019-11-02T16:29:56+00:00

Adrian Kummerlaender

2019-11-02T16:29:56+00:00

f233beddfc74d5933f46684adab5298e03c08871












Restructure LBM model / lattice distinction
2019-11-02T16:18:32+00:00

Adrian Kummerlaender

2019-11-02T16:18:32+00:00

24847cbb2567f508a7c30b39c6fb7ba6379d1adc












Call symbolic generator inside code templates
2019-10-31T12:13:00+00:00

Adrian Kummerlaender

2019-10-31T12:13:00+00:00

e2b00f4ec963060be98939c7b0d12d6c00e50a02

This paves the way for dropping in other LBM collision models.
As a side benefit the default momenta calulcation is now fully inlined where possible.





This paves the way for dropping in other LBM collision models.
As a side benefit the default momenta calulcation is now fully inlined where possible.







Move C++ example to boltzgen_examples repository
2019-10-30T15:12:20+00:00

Adrian Kummerlaender

2019-10-30T15:12:20+00:00

c82b38122cc3ab7717cb0ba9ec530b4658bd03e4












Don't try to reuse population layout for moment array indexing
2019-10-29T19:33:30+00:00

Adrian Kummerlaender

2019-10-29T19:33:30+00:00

09f45c1d1da37bf4f6fa4094eb2d2ea18e8aaf21












Return cell id as string expression when required
2019-10-29T18:53:05+00:00

Adrian Kummerlaender

2019-10-29T18:53:05+00:00

1ce3b58dabba59741343cbd9e7e4c9f58f10f91b












Add support for generating custom templates in boltzgen's context
2019-10-29T18:38:32+00:00

Adrian Kummerlaender

2019-10-29T18:38:32+00:00

61d77cf8aa815b456d873ff3c01e54ad21a8fce9












Rename target module to memory
2019-10-29T18:30:50+00:00

Adrian Kummerlaender

2019-10-29T18:30:50+00:00

a0438d434a8dde45e6fdab38d44181f0cd0cb2c6












Move further generator arguments into its constructor
2019-10-29T18:25:38+00:00

Adrian Kummerlaender

2019-10-29T18:25:38+00:00

a6dcb57cff9a1dd9de7e5fafdc87230489be87b9












Extract cell indexing function
2019-10-29T15:05:04+00:00

Adrian Kummerlaender

2019-10-29T15:05:04+00:00

d801f538a090838a15e74282239369b73723c4f4












Unify AOS, SOA specific cell preshift between targets
2019-10-29T09:56:41+00:00

Adrian Kummerlaender

2019-10-29T09:56:41+00:00

dbd9a340a7809a770d52d10154712278431acdc3

SOA and AOS should not be target specific, neighbor offset calculation /
bijection between gid and cell coordinates should be customizable.





SOA and AOS should not be target specific, neighbor offset calculation /
bijection between gid and cell coordinates should be customizable.







Set default order for custom ndindex overload
2019-10-28T21:33:53+00:00

Adrian Kummerlaender

2019-10-28T21:29:50+00:00

15c0cba693646269c04f245ca52f405ddfdb4a07












Use order-accepting ndindex method for generating cell indices
2019-10-28T21:00:05+00:00

Adrian Kummerlaender

2019-10-28T21:00:05+00:00

72e7097073b03e6584f603b905b5a4fc236d7def












Add basic setup.py
2019-10-28T20:52:34+00:00

Adrian Kummerlaender

2019-10-28T20:52:34+00:00

091811762b7f2cbb6575139276ea16bf54d3456b

No guarantee for correctness - I mostly fiddled this together in order
to use common nixpkgs python package functions  for including boltzgen
in other shell environments.





No guarantee for correctness - I mostly fiddled this together in order
to use common nixpkgs python package functions  for including boltzgen
in other shell environments.







Optionally generate cell-list-based OpenCL dispatch functions
2019-10-27T21:22:24+00:00

Adrian Kummerlaender

2019-10-27T21:22:24+00:00

536b25e2c2b742c17d023d2b3386fed4dc60a339

Requires different function naming as OpenCL 1.2 doesn't support overloads.

The OpenCL kernel code generated using this commit was successfully tested
on an actual GPU. Time to set up some automatic validation.





Requires different function naming as OpenCL 1.2 doesn't support overloads.

The OpenCL kernel code generated using this commit was successfully tested
on an actual GPU. Time to set up some automatic validation.







Verify precision parameter
2019-10-27T21:09:27+00:00

Adrian Kummerlaender

2019-10-27T21:09:27+00:00

1b9ac6e7aee3cf63495a65c2d7dbf79a0be23d7d












Add README
2019-10-27T19:11:36+00:00

Adrian Kummerlaender

2019-10-27T19:11:36+00:00

89f60f39a226bef5ccd8c52dbd57891c3a4d74c7












Accept cell id as parameter in OpenCL functions
2019-10-27T18:50:21+00:00

Adrian Kummerlaender

2019-10-27T18:50:21+00:00

a99edaaa0e9a95354f68614cd2f4ab801179b946

It is more flexible to place OpenCL thread ID dependent dispatching in a separate function.





It is more flexible to place OpenCL thread ID dependent dispatching in a separate function.







Add bounce back boundary condition
2019-10-27T18:40:56+00:00

Adrian Kummerlaender

2019-10-27T18:40:56+00:00

dcd162aef328cde66082a3333740ec6f58298a4c












Tidy up shell environment
2019-10-27T18:18:20+00:00

Adrian Kummerlaender

2019-10-27T18:18:20+00:00

6e99286c9a756272e972fe21212b094de95d36f6












Use Mako defines to generate momenta boundaries
2019-10-27T18:11:26+00:00

Adrian Kummerlaender

2019-10-27T18:11:26+00:00

4315b5c62fe862dcacbe2a8958732c8a66bdb6e1












Extract target-dependent floating point type name
2019-10-27T15:11:17+00:00

Adrian Kummerlaender

2019-10-27T15:11:17+00:00

30f262287bd43015b155b0e882e6478cfae2780c












Move layout implementations into separate folder
2019-10-27T15:10:33+00:00

Adrian Kummerlaender

2019-10-27T15:03:07+00:00

6938758f6c1754f0ee49d0709dd0ca376a146010












Disable bytecode caching
2019-10-27T15:02:31+00:00

Adrian Kummerlaender

2019-10-27T15:02:31+00:00

74c60dcbe56862d73b000d569423cb298fb06686












Separate functions into separate template files
2019-10-27T13:05:21+00:00

Adrian Kummerlaender

2019-10-27T13:05:21+00:00

18c54d79699db7554faa851c87d7113db67a8a08

Selection of the desired templates is possible via a new `functions` parameter.





Selection of the desired templates is possible via a new `functions` parameter.







Add extra toggle for OpenMP in C++ test function
2019-10-26T21:00:50+00:00

Adrian Kummerlaender

2019-10-26T21:00:50+00:00

7fa72d8718d96727bcfd60cc3bcb1609526d3c9b

Yields ~160 MLUPs on a Xeon E3-1241 for D2Q9 double precision lid driven cavity.
Obviously not anywhere near what is possible on GPUs but respectable for a CPU implementation.
Especially considering how simple it is.





Yields ~160 MLUPs on a Xeon E3-1241 for D2Q9 double precision lid driven cavity.
Obviously not anywhere near what is possible on GPUs but respectable for a CPU implementation.
Especially considering how simple it is.







Change C++ test function to LDC with optional VTK output
2019-10-26T20:44:59+00:00

Adrian Kummerlaender

2019-10-26T20:44:59+00:00

2a976c2c60565ea3f904feaf4ea573b2769e3084












Generate primitive velocity momenta BC for C++ target
2019-10-26T20:44:09+00:00

Adrian Kummerlaender

2019-10-26T20:44:09+00:00

820ed4b674a199c252e1b77ee7013f330ef284bb












Support passing additional string arguments to the generator
2019-10-26T13:01:39+00:00

Adrian Kummerlaender

2019-10-26T13:01:39+00:00

1bf577b1c5e606ac2c0553857297ce8c0c04ccb7












Implement AOS layout for the OpenCL target
2019-10-26T12:21:17+00:00

Adrian Kummerlaender

2019-10-26T12:21:17+00:00

71e0d07ec0cdf528eb5d962ce97bd6c55da52eae












Add optional flag to disable CSE
2019-10-26T12:11:25+00:00

Adrian Kummerlaender

2019-10-26T12:11:25+00:00

769bc4c97d5686ab614bb134abddef87f3f41519












Add BGK relaxation time argument
2019-10-26T11:42:11+00:00

Adrian Kummerlaender

2019-10-26T11:42:11+00:00

9ddfd0f54ddb95b181421c851826e0bb462c249d

Still thinking how various collision models (BGK, MRT and so on) may be
nicely abstracted in this setting.





Still thinking how various collision models (BGK, MRT and so on) may be
nicely abstracted in this setting.







Fix cpp test function
2019-10-26T11:38:09+00:00

Adrian Kummerlaender

2019-10-26T11:38:09+00:00

1c58183e84b2f7024855dc3c5bac19c48ba45f69












Add python shebang
2019-10-26T11:34:56+00:00

Adrian Kummerlaender

2019-10-26T11:34:56+00:00

b2013631b880972f5632e88b3d1bee77935720fb












Add geometry argument
2019-10-26T11:30:17+00:00

Adrian Kummerlaender

2019-10-26T11:30:17+00:00

4a56c8448653ecc4fea9cee89e70b32dd5aadbc2












Add lattice argument
2019-10-26T10:51:49+00:00

Adrian Kummerlaender

2019-10-26T10:51:49+00:00

afbb5520cd2140c108adffd6170b5a83faf329f2












Generalize floating point precision argument
2019-10-26T10:40:35+00:00

Adrian Kummerlaender

2019-10-26T10:40:35+00:00

8e968727a9c2c5a1b3be858b1102e70651f34168












Add ctags to environment
2019-10-26T10:40:02+00:00

Adrian Kummerlaender

2019-10-26T10:40:02+00:00

58d926cb7329a0da941ab1f3451ec0282795811c












Extract offset helper into target and layout specific classes
2019-10-24T19:52:45+00:00

Adrian Kummerlaender

2019-10-24T19:52:45+00:00

73de5f16efc696cf0c88beec086eb9a4df9098dd












Handle missing memory layout argument
2019-10-24T12:43:06+00:00

Adrian Kummerlaender

2019-10-24T12:43:06+00:00

b3d131b94f9417c0c4cd6733433c86ca780dde5e












Add test template for C++, enable switching between AOS and SOA
2019-10-24T12:33:59+00:00

Adrian Kummerlaender

2019-10-24T12:33:59+00:00

f1c5164b327054fc540527e973ff7843d4c01996












Some cleanup, add `collect_moments` to C++ template
2019-10-23T20:03:46+00:00

Adrian Kummerlaender

2019-10-23T20:03:46+00:00

2e07d5fac4d3fedcd2cc9ccbab27f50c69f65983












Return error for unavailable templates
2019-10-22T08:55:07+00:00

Adrian Kummerlaender

2019-10-22T08:55:07+00:00

54efae7153dd9a6ef1768ff3a054c082e090b4ca












Add basic Generator class
2019-10-22T08:49:44+00:00

Adrian Kummerlaender

2019-10-22T08:49:44+00:00

c27ee89f265f73fa77ec043e4f2a52c5c3859965












Add basic argument parsing
2019-10-21T20:33:13+00:00

Adrian Kummerlaender

2019-10-21T20:33:13+00:00

8fb7705428a6524a436a5dd8806ad8076b7f4e08












Pull in C++ template from symlbm_playground's standalone branch
2019-10-21T20:24:51+00:00

Adrian Kummerlaender

2019-10-21T20:23:41+00:00

8f88aea2c963ebe63db83480fc9a4b48451220af












Pull in basics from symlbm_playground
2019-10-21T16:48:38+00:00

Adrian Kummerlaender

2019-10-21T16:42:24+00:00

82a44e0d64afb8818ea98d68dc08108885d503c2

It's time to extract the generator-part of my GPU LBM playground and turn it
into a nice reusable library. The goal is to produce a framework that can be
used to generate collision and streaming programs from symbolic descriptions.

i.e. it should be possible to select a LB model, the desired boundary
conditions as well as a data structure / streaming model and use this
information to automatically generate matching OpenCL / CUDA / C++
programs.





It's time to extract the generator-part of my GPU LBM playground and turn it
into a nice reusable library. The goal is to produce a framework that can be
used to generate collision and streaming programs from symbolic descriptions.

i.e. it should be possible to select a LB model, the desired boundary
conditions as well as a data structure / streaming model and use this
information to automatically generate matching OpenCL / CUDA / C++
programs.