GitLab

Squashed commit of the following:

commit 60953ed4
Merge: 3380ebc2 b1c08ad7
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 23 09:04:24 2018 +0100

    Merge branch '2018.1' into evaluate-optimizations

commit 3380ebc2
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 23 08:49:49 2018 +0100

    Extend optimization 42 target period span

    * extended range from 0.5 to 10ns (2GHz - 100MHz)
    * increased step size to 0.25ns => 38 measurements
    * fixed bug in optimization level 1/2 settings
    * improved O3 level to even more aggressive settings
    * added sweep, remapping and constant propagation to initial netlist
      optimization

commit 44e5ca91
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Sun Feb 18 08:04:15 2018 +0100

    Fix bug in optimization 42 loop

commit ea32698b
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Sat Feb 17 15:45:01 2018 +0100

    Fix bug in evaluate optimization 42

    * files would be prematurely closed in loop

commit d81319fa
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 16 21:06:56 2018 +0100

    Move flatten hierarchy to O2

commit 7acd060d
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 16 21:05:51 2018 +0100

    Increase default target frequency in Import to 1GHz

commit 2dbe1061
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 16 08:27:27 2018 +0100

    Re-activate normal evaluation

    * fixed optimization 42 implementation: normal runs (O != 42) are
      executed in single run mode
    * fixed bug with SAIF reading
    * fixed common_ip bug

commit 095f9417
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 16 07:06:39 2018 +0100

    Remove clock pinning and time limit for opt 42

commit 061d43bb
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Tue Feb 13 16:00:24 2018 +0100

    Reactivate clock pinning

    * also fixes some output bugs

commit 4ae807ea
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Fri Feb 9 06:34:28 2018 +0100

    Fix bugs in optimzation debug mode 42

commit bc7c9676
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Thu Feb 8 18:10:45 2018 +0100

    Fix typo

commit 02f83925
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Thu Feb 8 18:05:54 2018 +0100

    Fix O-level iterations

    * O-level iterations included only the synthesis phase, not PnR; thus
      PnR was always done on O0 synthesis results
    * fixed - PnR is now included in loop

commit ecc9e235
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Thu Feb 8 17:36:12 2018 +0100

    WIP: Implement special synthesis optimization evaluation level

    * level 42 activates a sweep search that iterates over
        - clock periods 0.5 - 5 (0.125 steps)
        - optimization levels 3 - 0
    * it also logs the runtimes of each step and outputs them in a
      synthesis_runtimes.tsv file
    * resulting WNS is logged in synthesis_results.tsv for each step
    * useful for analysis of evaluation runs
    * careful - regular evaluation is currently broken by the changes

commit 5d081bae
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Thu Feb 8 17:35:49 2018 +0100

    Fix bug in optimization level passing from parser

commit 47bb3dad
Merge: a7138abf 3d912101
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Thu Feb 8 10:55:45 2018 +0100

    Merge branch '2018.1' of git:tapasco/tapasco into evaluate-optimizations

commit a7138abf
Author: Jens Korinth <jk@esa.cs.tu-darmstadt.de>
Date:   Wed Feb 7 11:10:21 2018 +0100

    WIP: define optimization levels during evaluation

    * evaluation results are too conservative w.r.t. f_max
    * adding new --optimization flag for import task that defines gcc-like
      opt levels, which activate different synth + pnr options
    * Tcl does not work yet

* memory allocation/deallocation and copy operations are no in
  tapasco_memory.h/c fragment in arch/common
* removed code from axi4mm Arch implementation and linking against
  default fragment instead

Merge remote-tracking branch 'origin/2017.2'

* fixed new issues; old scalastyle problems persist
* postponed fixing old issues

* GitLab pipelines notified me of the problem, awesome

* check all Job instances for empty filters, e.g., Platforms,
  Architectures, Kernels, and issue error message accordingly
* helpful error message contains available instances
* job is not executed, if check fails

Adds support for Gitlab CI to run the scala tests

See merge request !14

Adds support for older kernel Versions < 4.8

See merge request !13

* assumption: control slave is first in address map; following a
  non-control slave / custom
* added function to compute number of slaves for a PE coordinate
* status core now checks number of slaves an increase slot ids acc.

* error in parser caused it to fail when run without options

* remove support for 2016.2, 2016.3

* XDC set to unmanaged to avoid problems
* /uArch/irq_0 and /InterruptControl/irq are now connected by default
* renamed constant zero properly

* XDC for system clock pin accidentally had ILA instance in it
* set '-unmanaged' flag to avoid the problem in the future