Hi Damian --
I'm heading out the door, and taking a stab at your question without
having had a chance to read your code very carefully. Hopefully, I've got
the thrust of the question right and keep you moving without letting
another day slip by:
Chapel distinguished between multidimensional arrays:
var A: [1..m, 1..n] real; // a 2D array
and arrays of arrays (arrays whose elements are arrays):
var B: [1..m] [1..n] real;
The former says "give me a 2D array of reals" while the latter says "give
me a 1D array whose elements are 1D arrays of reals". The former must be
indexed as:
A[i,j] (or A(i,j))
while the latter must be indexed as:
B[i][j] (or B(i)(j))
If you find that this is not the case, please let us know -- I'm not aware
of any bugs along these lines at present.
I think the only reason that this isn't called out specifically in the
language spec is that it feels like a natural composition of concepts
rather than something specific. But perhaps a note of clarification could
help. I do think that the arrays.chpl primer in the release touches on
this at least briefly:
https://github.com/chapel-lang/chapel/blob/master/test/release/examples/primers/arrays.chpl
One thing that can sometimes muddy the waters is that arrays can also be
indexed by tuple indices, so one can do things like:
const ij = (i,j);
...A[ij]...
Moreover, when iterating over multidimensional domains, the indices
generated are tuples. Thus, given:
const D = {1..m, 1..n};
One can iterate like this:
forall ij in D ... // ij is a 2-tuple
or this:
forall (i,j) in D ... // de-tuple into scalars i and j
Running out the door now,
-Brad
On Mon, 13 Apr 2015, Damian McGuckin wrote:
>
> Hi all,
>
> Looking at the following code to play with boundcode condition (masks)
> taken, and hacked, from the LULESU benchmark.
>
> Apologies if my use of tabs causes poor readability.
>
> /* Compile-time constants */
>
> param
> XI_M = 0x003,
> XI_M_SYMM = 0x001,
> XI_M_FREE = 0x002,
>
> XI_P = 0x00c,
> XI_P_SYMM = 0x004,
> XI_P_FREE = 0x008,
>
> ETA_M = 0x030,
> ETA_M_SYMM = 0x010,
> ETA_M_FREE = 0x020,
>
> ETA_P = 0x0c0,
> ETA_P_SYMM = 0x040,
> ETA_P_FREE = 0x080,
>
> ZETA_M = 0x300,
> ZETA_M_SYMM = 0x100,
> ZETA_M_FREE = 0x200,
>
> ZETA_P = 0xc00,
> ZETA_P_SYMM = 0x400,
> ZETA_P_FREE = 0x800;
>
> // example of single dimension array of tuples - all Good
>
> proc eBC1(mask : int) : int
> {
> var amask : [1..6] (int, int, int) =
> [
> ( 0x0f, ZETA_M_SYMM, ZETA_M_FREE ),
> ( 0x33, ETA_M_SYMM, ETA_M_FREE ),
> ( 0x99, XI_M_SYMM, XI_M_FREE ),
> ( 0xf0, ZETA_P_SYMM, ZETA_P_FREE ),
> ( 0xcc, ETA_P_SYMM, ZETA_P_FREE ),
> ( 0x66, XI_P_SYMM, XI_P_FREE )
> ];
> var eBCe : int;
>
> for (tmask, tsymm, tfree) in amask
> {
> if ((mask & tmask) == tmask) then eBCe |= tsymm;
> }
> return eBCe;
> }
>
> // array of arrays - x[i, j] indexing - does NOT compile inside LULESH
> // compiles successfully outside LULESH, i.e. if you just use this code
> // in a file by itself.
>
> proc eBC2(mask : int) : int
> {
> var amask : [1..6, 1..3] int =
> [
> [ 0x0f, ZETA_M_SYMM, ZETA_M_FREE ],
> [ 0x33, ETA_M_SYMM, ETA_M_FREE ],
> [ 0x99, XI_M_SYMM, XI_M_FREE ],
> [ 0xf0, ZETA_P_SYMM, ZETA_P_FREE ],
> [ 0xcc, ETA_P_SYMM, ZETA_P_FREE ],
> [ 0x66, XI_P_SYMM, XI_P_FREE ]
> ];
> var eBCe : int;
>
> for i in 1..6
> {
> var tmask : int = amask[i, 1];
> var tsymm : int = amask[i, 2];
>
> if ((mask & tmask) == tmask) then eBCe |= tsymm;
> }
> return eBCe;
> }
>
> // array of arrays - C-style array indexing - does compile inside LULESH
>
> proc eBC3(mask : int) : int
> {
> var amask : [1..6] [1..3] int =
> [
> [ 0x0f, ZETA_M_SYMM, ZETA_M_FREE ],
> [ 0x33, ETA_M_SYMM, ETA_M_FREE ],
> [ 0x99, XI_M_SYMM, XI_M_FREE ],
> [ 0xf0, ZETA_P_SYMM, ZETA_P_FREE ],
> [ 0xcc, ETA_P_SYMM, ZETA_P_FREE ],
> [ 0x66, XI_P_SYMM, XI_P_FREE ]
> ];
> var eBCe : int;
>
> for i in 1..6
> {
> var tmask : int = amask[i][1];
> var tsymm : int = amask[i][2];
>
> if ((mask & tmask) == tmask) then eBCe |= tsymm;
> }
> Return eBCe;
> }
>
> II notice that if I try use an array indexing scheme like
>
> amask[i, j]
>
> and use it with the compilation configuration options in the LULESH
> benchmark, the compiler tells me there are problems.
>
> I have to resort to something like
>
> amask[i][j]
>
> While I can live with either, such issues are not discussed, or even
> hinted at, in the language specification. I also cannot find notes or a
> discussion elsewhere to provide guidance. That said, my search keywords
> might be poor choices.
>
> Regards - Damian
>
> Pacific Engineering Systems International, 277-279 Broadway, Glebe NSW 2037
> Ph:+61-2-8571-0847 .. Fx:+61-2-9692-9623 | unsolicited email not wanted here
> Views & opinions here are mine and not those of any past or present employer
>
> ------------------------------------------------------------------------------
> BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
> Develop your own process in accordance with the BPMN 2 standard
> Learn Process modeling best practices with Bonita BPM through live exercises
> http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
> source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
> _______________________________________________
> Chapel-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/chapel-users
>
------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Chapel-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-users
