-
I've set up quickly through the bundle configs and I am able to get the input bar to pop up. Unfortunately after I type out anything I get errors within the console log and nothing populates within t…
-
Assuming a symbolic data object object `u = DenseData(name='u', shape=...)`, auto/sympy-generated new instances of `u(x, y)`, such as `u(x, y)` and `u(x+h, y)`, all cache on the original `u` object. H…
-
There is an issue with `transformer` in some cases.
The "new" version of Forward/Backward (PR 176) now requires `sympy solve` as some time indices are not directly accessible. As this fgunction is v…
-
The current design of Devito function symbols is centered around the `SymbolicData` objects that always have an associated `.data` property. The original rationale for this is that we want control ove…
-
@ggorman @navjotk @mlange05
This is a general concern about the generation part. We do have a really nice tool however there is a few flags that will hit us hard if we don't look at it now. I will wo…
-
not much to say. Just that we should always use absolute imports. For example, instead of doing
`
import logger
`
we should have:
`
import devito.logger
`
-
In trying to install devito on the yemoja cluster I realised the following:
- Our `requirements.txt` does not list numpy as a dependency
- Travis.yaml does not download/install numpy in any special wa…
-
I am opening an issue because I can't figure out how to comment on code already in master. Since we moved the compilation to `compiler.py`, the OpenMP flag is being set from the env variable directly …
-
@navjotk
the indexes x,y in __prepare_stencil are flipped (y,x) compared to the rest. It will create problem for non square domains
-
I realized that all inputs to TTI_codegen and Acoustic_codegen are copied e.g :
`self.m = DenseData(name="m", shape=self.model.vp.shape, dtype=self.dtype)`
`self.m.data[:] = self.model.vp**(-2)`
Ii…