connected hyperspace where architecture can be fluid, flexible and vivid, yet the aspect of materiality requires more attention.
Action-designed structures begin to move beyond the utopian proposals of the 20th century’s manifestos and hold a place in the world of realized designs. The AA Athens Visiting School aims to bring users closer to the built environment while revisiting habits of designing, building and experiencing space through materiality. Understanding materiality and form as a ‘unified whole’, the programme integrates manufacturing techniques through the experimentation fabrication of prototypes at a 1:1 scale.
Prominent Features of the workshop/ skills developed
Participants become part of an active learning environment where the large tutor to student ratio allows for personalized tutorials and debates.
The toolset of the Athens VS includes but is not limited to Processing and Grasshopper for Rhinoceros, as well as design analysis software.
Participants gain hands-on experience on digital fabrication.
Design seminars and a series of lectures support the key objectives of the programme, disseminating fundamental computational techniques, relevant critical thinking, theoretical understanding and professional awareness.
Applications
1) You can make an application by completing the online application found under ‘Links and Downloads’ on the AA Visiting School page. If you are not able to make an online application, email visitingschool@aaschool.ac.uk for instructions to pay by bank transfer. 2) Once you complete the online application and make a full payment, you are registered to the programme. A CV or a portfolio is NOT required.
The deadline for applications is 28 June.
Location AKTO College – Athens Campus 11Α Evelpidon Street (Pedion Areos) Athens, 113 62, Greece
Fees
The AA Visiting School requires a fee of £695 per participant, which includes a £60 Visiting membership fee. Fees do not include flights or accommodation, but accommodation options can be advised.
Eligibility The workshop is open to current Undergrad and Graduate architecture and design students, PhD candidates and young professionals. Software Requirements: Adobe Creative Suite, Rhino 5.
For more information, please visit:
http://www.aaschool.ac.uk/STUDY/VISITING/athens
http://ai.aaschool.ac.uk/athens/
For inquiries, please contact:
alexandros.kallegias@aaschool.ac.uk…
r graphics get saved as 24x24 pixel images before they are put into the grasshopper application, which means the icons look like crap when you zoom in. This is the aforementioned problem that needs to be addressed in GH2. There have historically been two approaches to this issue:
Provide pixel images with several sizes.
Render vector graphics directly.
Option 1 is common for apps that do not have variable levels of zoom, such as Windows Explorer. When explorer shows file icons it either shows them in 16x16, 32x32, 48x48, 96x96, or these days, various HUGE sizes. As a result *.ico files allow you put in different images for all these target sizes. Since Grasshopper has variable zoom levels, this is not an ideal solution. Also, it requires a lot more work per icon.
Option 2 is becoming more and more popular as increased graphics speed now allows for the real-time rendering of vector graphics. Yet, you still need a renderer that knows how to draw vector geometry crisply at low sizes. All vector renderers I know just interpolate the geometry linearly and if a line happens to end up 'between pixels' it's just fuzzy.
I don't have hard and fast rules for the icons, but I try to adhere to at least these:
Keep a border of 2 pixels free around the icon content. So basically only use the inner 20x20 pixels rather than the 24x24 you're allowed. This is needed because the drop shadow needs to go there.
Only draw silhouette edges around shapes, not inner creases. Typically a 1-pixel line will do. I prefer to use a dark version of the fill colour rather than black for edges.
Loose curves can be drawn in 1 or 2 pixel thicknesses, depending on how important the curve is.
Try to avoid text in your icons (not always possible).
Stick to 1 colour family per icon, preferably per icon family. You can add highlights with another colour if you must, but too many hues make an icon hard to read (for the example the [Voronoi] icon, it has red, green and blue and it's a bit of a mess, on the other hand [Colour Wheel] has the full spectrum and seems to work quite well...).
Very roughly speaking, if there's both black and red geometry in an icon, it means the red is component input and the black is component output.
Drop shadows are pixel effects, applied to the 24x24 image. They have a blurring radius of 2 pixels, a horizontal offset of 1 pixel to the right, a vertical offset of 1 pixel to the bottom and they are 65% black.
When you use high contrast shapes (for example black edges on a light background) the anti-aliasing provided by vector renderers such as Xara or Illustrator won't be enough to make it look smooth. I'd recommend avoiding high contrast if at all possible, but if not possible then draw a 1-pixel line around the dark bits in 95% transparent black. This effectively extends the anti-aliasing range from 1.5 to 2.5 pixels and it helps make things looks smoother.
--
David Rutten
david@mcneel.com…
s levels of detail by subdividing a 6 sided cube mesh and projecting its vertices according to a referenced height map. This is one of the standard conventions for building full sizes planets. At the lowest level (0) the mesh planet is made of 6 pieces(each 32x32 resolution). The next level down (1) is made of 24 pieces... 6 divided by 4 = 24. Level (2) is 96 quads etc etc. The script will generate each quad at its sub-division level and compare edge vertices to neighboring quads. It will then make sure any shared vertices are in fact at the same projected vector. This ensures a planet quad with edge vertices that match.
The problems comes in texturing each quad.
If I build the quad as a nurb surface from points I can place the texture easily because each surface UV maps squarely to my texture map (which is also square).
If I build the quad as a mesh I cannot just apply the square texture to the mesh UVs. This is because when you unwrap the UVs from a mesh they will not unwrap like a nurb surface's UVs. Therefore to get the correct mapping I would have to manipulate each UV back to an evenly aligned array (which is 1024 points in a 32x32 resolution UV). Maya and blender have 'relax uv' and 'align UV' functions but they don't do the trick and manual corrections are out of the question. So why not skip the mesh method and use the nurb method?
I did this and there is a trade off. The nurb will accept the material texture I want with no other work on my end but when I export the object as an .obj rhino creates its own mesh to describe the nurb(with various unsatisfactory setting options). This works great up to a point because at some level the interpreted mesh will have vertices that do no match at the edges, ie .. creating visible seams in the mesh. The picture below is the nearly seamless planet at LOD(1) made of 24 quads, each with 32x32 vertice resolution and a 512x512 jpg texture running in Unity3d 5. It works but at close level there are seams. This will be resolved simply by having the next LOD(x) instantiate before getting close enough to see the seam but at core nerd level I want the seamless mesh.
So, I can make the seamless mesh but I can not realistically texture map it. I can also make the nurb surface from points and texture it at the expense of the edge vertices matching. I am at the split in the road but I want to have my cake and eat it too. Thoughts, comments, trolls...?
Thanks for reading =)
Footnote: For you pros I am not using seamless noise across the map I am using grasshopper to sew up my otherwise non perfect edges.
Other programs in the pipeline:
-WorldMachine 2
-Wilbur
-Photoshop
-Unity3d…
well, very similar input data must result in wildly different hashes. For example, imagine we have an algorithm which computes hashes of text, and the hashes it computes are all numbers between 0 and 999. We then apply this algorithm to a piece of text:
"When Spring comes back with rustling shade" = 385
So far so good. Now imagine we change the text slightly, for example by removing a single "l":
"When Spring comes back with rusting shade" = 973
Minor change -> very different hash. There are of course way more unique texts than there are numbers between 0 and 999. This must therefore mean that a lot of text will result in the same hash. For example "When Spring brings back blue days and fair." may also result in a hash of 385. Because of the pigeonhole principle, there is nothing to be done about this.
Now for the tricky bit. Hashes are often used to validate executable code. Say your friend James at MI6 sends you a small program that will allow you to eavesdrop on Angela Merkel, and -over the phone- he tells you the hashcode for that application. You can then hash the application yourself, verify that it indeed results in the same hashcode and then you know you can trust the executable.
But now Jack from the FBI intercepts the email and adds a few sneaky lines of code to the original application allowing him to determine from your internet search history with up to 95% accuracy whether you like extra cheese on your pizza. The application has now been tampered with, it can no longer be trusted and you should be able to figure this out as it will no longer result in the same hash code.
But wait! Some hashing algorithms are more secure than others. MD5 is now officially considered to be 'hacked' and it is no longer recommended for doing naughty spying. Specifically, Jack will be able to inject his own code in such a way that it does not result in a different hash. Instead, the SHA family of hashers are to be used, as it is not yet known how to trick these hashers.
This is where the problem comes in, because apparently the US government has forcefully disabled the use of MD5 for all purposes. This is a shame because I use it to quickly compare bitmap icons for identicalness so I only have to store an icon in memory once. There is no security hole due to this, because I'm not hashing secure data. MD5 is somewhat faster than SHA, and since I have to hash several hundred icons on Grasshopper start, I opted for the faster one.
(Very) long story short; you're hosed. Grasshopper uses MD5; USgov does not like; Grasshopper does not run on USgov computers.
I'll do some testing to see if I can switch to SHA and then we can see whether or not that solves the problem. This however will take a while as I'm going on a business trip next week and have yet to prepare my presentations.
--
David Rutten
david@mcneel.com…
Added by David Rutten at 12:06pm on March 31, 2014
ed file and code below:
Color ColorAt(Mesh mesh, int faceIndex, double t0, double t1, double t2, double t3) { // int rc = -1; var color = Rhino.Display.Color4f.Black;
if( mesh.VertexColors.Count != 0) { // test to see if face exists if( faceIndex >= 0 && faceIndex < mesh.Faces.Count ) { /// Barycentric quad coordinates for the point on the mesh /// face mesh.Faces[FaceIndex].
/// If the face is a triangle /// disregard T[3] (it should be set to 0.0).
/// If the face is /// a quad and is split between vertexes 0 and 2, then T[3] /// will be 0.0 when point is on the triangle defined by vi[0], /// vi[1], vi[2]
/// T[1] will be 0.0 when point is on the /// triangle defined by vi[0], vi[2], vi[3].
/// If the face is a /// quad and is split between vertexes 1 and 3, then T[2] will /// be -1 when point is on the triangle defined by vi[0], /// vi[1], vi[3]
/// and m_t[0] will be -1 when point is on the /// triangle defined by vi[1], vi[2], vi[3].
MeshFace face = mesh.Faces[faceIndex];
// Collect data for barycentric evaluation. Color p0, p1, p2;
if(face.IsTriangle) { p0 = mesh.VertexColors[face.A]; p1 = mesh.VertexColors[face.B]; p2 = mesh.VertexColors[face.C]; } else { if( t3 == 0 ) { // point is on subtriangle {0,1,2} p0 = mesh.VertexColors[face.A]; p1 = mesh.VertexColors[face.B]; p2 = mesh.VertexColors[face.C]; } else if( t1 == 0 ) { // point is on subtriangle {0,2,3} p0 = mesh.VertexColors[face.A]; p1 = mesh.VertexColors[face.C]; p2 = mesh.VertexColors[face.D]; //t0 = t0; t1 = t2; t2 = t3; } else if( t2 == -1 ) { // point is on subtriangle {0,1,3} p0 = mesh.VertexColors[face.A]; p1 = mesh.VertexColors[face.B]; p2 = mesh.VertexColors[face.D]; //t0 = t0; //t1 = t1; t2 = t3; } else { // point must be on remaining subtriangle {1,2,3} p0 = mesh.VertexColors[face.B]; p1 = mesh.VertexColors[face.C]; p2 = mesh.VertexColors[face.D]; t0 = t1; t1 = t2; t2 = t3; } }
/** double r = t0 * p0.FractionRed() + t1 * p1.FractionRed() + t2 * p2.FractionRed(); double g = t0 * p0.FractionGreen() + t1 * p1.FractionGreen() + t2 * p2.FractionGreen(); double b = t0 * p0.FractionBlue() + t1 * p1.FractionBlue() + t2 * p2.FractionBlue();
ON_Color color; color.SetFractionalRGB(r, g, b);
unsigned int abgr = (unsigned int)color; rc = (int) ABGR_to_ARGB(abgr); **/ var c0 = new Rhino.Display.Color4f(p0); var c1 = new Rhino.Display.Color4f(p1); var c2 = new Rhino.Display.Color4f(p2); float s0 = (float) t0; float s1 = (float) t1; float s2 = (float) t2;
float R = s0 * c0.R + s1 * c1.R + s2 * c2.R; float G = s0 * c0.G + s1 * c1.G + s2 * c2.G; float B = s0 * c0.B + s1 * c1.B + s2 * c2.B; color = new Rhino.Display.Color4f(R, G, B, 1); } } return color.AsSystemColor(); }
…
ay how many valid permutations exist.
But allow me to guesstimate a number for 20 components (no more, no less). Here are my starting assumptions:
Let's say the average input and output parameter count of any component is 2. So we have 20 components, each with 2 inputs and 2 outputs.
There are roughly 35 types of parameter, so the odds of connecting two parameters at random that have the same type are roughly 3%. However there are many conversions defined and often you want a parameter of type A to seed a parameter of type B. So let's say that 10% of random connections are in fact valid. (This assumption ignores the obvious fact that certain parameters (number, point, vector) are far more common than others, so the odds of connecting identical types are actually much higher than 3%)
Now even when data can be shared between two parameters, that doesn't mean that hooking them up will result in a valid operation (let's ignore for the time being that the far majority of combinations that are valid are also bullshit). So let's say that even when we manage to pick two parameters that can communicate, the odds of us ending up with a valid component combo are still only 1 in 2.
We will limit ourselves to only single connections between parameters. At no point will a single parameter seed more than one recipient and at no point will any parameter have more than one source. We do allow for parameters which do not share or receive data.
So let's start by creating the total number of permutations that are possible simply by positioning all 20 components from left to right. This is important because we're not allowed to make wires go from right to left. The left most component can be any one of 20. So we have 20 possible permutations for the first one. Then for each of those we have 19 options to fill the second-left-most slot. 20×19×18×17×...×3×2×1 = 20! ~2.5×1018.
We can now start drawing wires from the output of component #1 to the inputs of any of the other components. We can choose to share no outputs, output #1, output #2 or both with any of the downstream components (19 of them, with two inputs each). That's 2×(19×2) + (19×2)×(19×2-1) ~ 1500 possible connections we can make for the outputs of the first component. The second component is very similar, but it only has 18 possible targets and some of the inputs will already have been used. So now we have 2×(18×2-1) + (18×2-1)×(18×2-1) ~1300. If we very roughly (not to mention very incorrectly, but I'm too tired to do the math properly) extrapolate to the other 18 components where the number of possible connections decreases in a similar fashion thoughout, we end up with a total number of 1500×1300×1140×1007×891×789×697×...×83×51×24×1 which is roughly 6.5×1050. However note that only 10% of these wires connect compatible parameters and only 50% of those will connect compatible components. So the number of valid connections we can make is roughly 3×1049.
All we have to do now is multiply the total number of valid connection per permutation with the total number of possible permutations; 20! × 3×1049 which comes to 7×1067 or 72 unvigintillion as Wolfram|Alpha tells me.
Impressive as these numbers sound, remember that by far the most of these permutations result in utter nonsense. Nonsense that produces a result, but not a meaningful one.
EDIT: This computation is way off, see this response for an improved estimate.
--
David Rutten
david@mcneel.com
Poprad, Slovakia…
Added by David Rutten at 12:06pm on March 15, 2013
ting at multiple geometries in the same location. I simply sorted the list of values and used the Delete Consecutive component. This potentially rearranges the order of values but I don't think that matters in your case. I also threw in an Int component which actually seems to make a difference (try sidestepping it and you will see!).
2-I flattened the output of the mesh component before sending it to union. This ensures that the original mesh is booleaned once with all the components rather than individually with each of the 86 components.
Is this what the result should look like?
One suggestion for future postings: when referencing geometry in rhino, it often helps if you attach your rhino file as well so people don't have to guess where you are starting from.
If you have further questions, just ask ;-)
cbass…
mplex the models are. If we are running multi-room E+ studies, that will take far longer to calculate.
Rhino/Grasshopper = <1%
Generating Radiance .ill files = 88%
Processing .ill files into DA, etc. = ~2%
E+ = 10%
Parallelizing Grasshopper:
My first instinct is to avoid this problem by running GH on one computer only. Creating the batch files is very fast. The trick will be sending the radiance and E+ batch files to multiple computers. Perhaps a “round-robin” approach could send each iteration to another node on the network until all iterations are assigned. I have no idea how to do that but hope that it is something that can be executed within grasshopper, perhaps a custom code module. I think GH can set a directory for Radiance and E+ to save all final files to. We can set this to a local server location so all runs output to the same location. It will likely run slower than it would on the C:drive, but those losses are acceptable if we can get parallelization to work.
I’m concerned about post-processing of the Radiance/E+ runs. For starters, Honeybee calculates DA after it runs the .ill files. This doesn’t take very long, but it is a separate process that is not included in the original Radiance batch file. Any other data manipulation we intend to automatically run in GH will be left out of the batch file as well. Consolidating the results into a format that Design Explorer or Pollination can read also takes a bit of post-processing. So, it seems to me that we may want to split up the GH automation as follows:
Initiate
Parametrically generate geometry
Assign input values, material, etc.
Generate radiance/ E+ batch files for all iterations
Calculate
Calc separate runs of Radiance/E+ in parallel via network clusters. Each run will be a unique iteration.
Save all temp files to single server location on server
Post Processing
Run a GH script from a single computer. Translate .ill files or .idf files into custom metrics or graphics (DA, ASE, %shade down, net solar gain, etc.)
Collect final data in single location (excel document) to be read by Design Explorer or Pollination.
The above workflow avoids having to parallelize GH. The consequence is that we can’t parallelize any post-processing routines. This may be easier to implement in the short term, but long term we should try to parallelize everything.
Parallelizing EnergyPlus/Radiance:
I agree that the best way to enable large numbers of iterations is to set up multiple unique runs of radiance and E+ on separate computers. I don’t see the incentive to split individual runs between multiple processors because the modular nature of the iterative parametric models does this for us. Multiple unique runs will simplify the post-processing as well.
It seems that the advantages of optimizing matrix based calculations (3-5 phase methods) are most beneficial when iterations are run in series. Is it possible for multiple iterations running on different CPUs to reference the same matrices stored in a common location? Will that enable parallel computation to also benefit from reusing pre-calculated information?
Clustering computers and GPU based calculations:
Clustering unused computers seems like a natural next step for us. Our IT guru told me that we need come kind of software to make this happen, but that he didn’t know what that would be. Do you know what Penn State uses? You mentioned it is a text-only Linux based system. Can you please elaborate so I can explain to our IT department?
Accelerad is a very exciting development, especially for rpict and annual glare analysis. I’m concerned that the high quality GPU’s required might limit our ability to implement it on a large scale within our office. Does it still work well on standard GPU’s? The computer cluster method can tap into resources we already have, which is a big advantage. Our current workflow uses image-based calcs sparingly, because grid-based simulations gather the critical information much faster. The major exception is glare. Accelerad would enable luminance-based glare metrics, especially annual glare metrics, to be more feasible within fast-paced projects. All of that is a good thing.
So, both clusters and GPU-based calcs are great steps forward. Combining both methods would be amazing, especially if it is further optimized by the computational methods you are working on.
Moving forward, I think I need to explore if/how GH can send iterations across a cluster network of some kind and see what it will take to implement Accelerad. I assume some custom scripting will be necessary.…