Himbaechel GUI #1295
Himbaechel GUI #1295
Conversation
mmicko
force-pushed
the
micko/himbaechel_gui
branch
from
6d7e0cb
to
e3b9859
Compare
| @@ -1,5 +1,5 @@ | |||
| #!/usr/bin/env bash | |||
| set -ex | |||
| yosys -p "synth_xilinx -flatten -abc9 -nobram -arch xc7 -top top; write_json blinky.json" blinky.v | |||
| nextpnr-himbaechel --device xc7a35tcsg324-1 -o xdc=arty.xdc --json blinky.json -o fasm=blinky.fasm --router router2 | |||
| ../../../../../nextpnr-himbaechel --device xc7a35tcsg324-1 -o xdc=arty.xdc --json blinky.json -o fasm=blinky.fasm --router router2 | |||
There was a problem hiding this comment.
this breaks if someone does an out of tree build
himbaechel/himbaechel_api.h
Outdated
| // Graphics | ||
| virtual uint32_t gfxAttributes() { return 0; } | ||
|
|
||
| virtual void gfxTileBel(std::vector<GraphicElement> &g, int x, int y, int z, int w, int h, |
There was a problem hiding this comment.
can you perhaps think of some slightly more descriptive names for these API functions, and also add some documentation about them, so I can review them better?
There was a problem hiding this comment.
sure, will update these, but probably not today
himbaechel/himbaechel_api.h
Outdated
| @@ -102,6 +109,20 @@ struct HimbaechelAPI | |||
| virtual bool isClusterStrict(const CellInfo *cell) const; | |||
| virtual bool getClusterPlacement(ClusterId cluster, BelId root_bel, | |||
| std::vector<std::pair<CellInfo *, BelId>> &placement) const; | |||
|
|
|||
| // Graphics | |||
| virtual uint32_t gfxAttributes() { return 0; } | |||
There was a problem hiding this comment.
I really don't like this function, can you think of a better way here, like perhaps a struct that the architecture configures?
mmicko
force-pushed
the
micko/himbaechel_gui
branch
from
e3b9859
to
ddf3cbe
Compare
|
@gatecat Documentation is not added yet, but main question I have is if using tilewire as additional data is fine (need to bump db version in that case as well) ? not sure if there is other way we can actually use to determine which wire to draw. |
mmicko
force-pushed
the
micko/himbaechel_gui
branch
from
5d504c6
to
2f03e88
Compare
mmicko
force-pushed
the
micko/himbaechel_gui
branch
from
2f03e88
to
4749794
Compare
|
|
||
| NEXTPNR_NAMESPACE_BEGIN | ||
|
|
||
| enum GfxTileWireId |
There was a problem hiding this comment.
this will need to be in a uarch-specific namespace, otherwise there will be conflicts
| @@ -857,3 +865,18 @@ def write_bba(self, filename): | |||
| bba.ref('chip_info') | |||
| self.serialise(bba) | |||
| bba.pop() | |||
|
|
|||
| def read_gfx_h(self, filename): | |||
| @@ -144,6 +144,7 @@ class TileWireData: | |||
| index: int | |||
| name: IdString | |||
| wire_type: IdString | |||
| tile_wire: int | |||
There was a problem hiding this comment.
this should be given a name that makes it's very clear that it's gui-specific, otherwise it will be very unclear to anyone looking at the structures
| BelId bel(decal.tile, decal.index); | ||
| GraphicElement::style_t style = decal.active ? GraphicElement::STYLE_ACTIVE : GraphicElement::STYLE_INACTIVE; | ||
| uarch->drawBel(ret, style, getBelType(bel), getBelLocation(bel)); | ||
| } else if (decal.type == DecalId::TYPE_WIRE) { |
There was a problem hiding this comment.
if a wire spans multiple tiles, as it can in the himbaechel deduplication scheme, this would need to deal with all the constituent tile wires not just the index one
|
Is this close to merging? I am pretty desperate to see things, and |
This extends Himbaechel API with additional methods for drawing bels, wires and pips. There is also configuration attribute to control if some of these are used and also do change in coordinate system (y axis).
Was mostly mimic what we do on ECP5 and XO2/3, and was not sure about equivalent for tile_wire (mapped to GfxTileWireId) so used flags for now, since it anyway keeps some additional custom info.
First few commits are mostly cleanup and fixes of some old standing GUI issues.
Would add at least basic things in Xilinx and Gowin when we agree on API level. Please do comment and suggest changes.