25

I heard chip designers being described as "triangle pushers," the idea being that somehow the logic on the chip was formulated by arranging triangles on the silicon in certain ways. How does this work? I don't understand how triangles can be arranged to create digital logic or why the shape of a triangle would be important.

Adam Haun
  • 21,331
  • 4
  • 50
  • 91
Tyler Durden
  • 2,186
  • 4
  • 34
  • 49
  • 20
    Been in that industry for 25+ years and never heard that term. – Oldfart Jan 29 '18 at 15:19
  • 15
    Who calls them that, exactly? – Finbarr Jan 29 '18 at 15:21
  • 8
    Did they used to give out little sample triangles to get you hooked? – Trevor_G Jan 29 '18 at 16:29
  • 10
    Don't you see the schematic in the logo for this site? It's obviously a _triangle_ that is _pushing_ against a wall. It's right there in red and white. – JPhi1618 Jan 29 '18 at 20:34
  • 1
    Well, I worked in the same RCA office with the guy who designed the industry's first computer layout software for ICs, and I don't recall this term being used. I do vaguely recall that the computer "drawing" was done largely with overlaid triangles, so I'd guess that's what's being referred to, if the term is at all real. – Hot Licks Jan 30 '18 at 13:43
  • 1
    It looks like some functions in computer graphics rendering are called "triangle pushers" since triangles are used for polygonal 3D rendering. I couldn't find any references to "triangle pusher" being used to describe a person. But the Internet does not actually know all. – Todd Wilcox Jan 30 '18 at 15:46
  • 1
    How about "tri-wranglers"? :D –  Jan 30 '18 at 21:16
  • I would have pulled out of the ether that it was due to the symbols for the `OR` logic gate. – Joshua Drake Jan 30 '18 at 21:49

3 Answers3

33

They are called "polygon pushers".

polygon pusher: n.

A chip designer who spends most of his or her time at the physical layout level (which requires drawing lots of multi-colored polygons). Also rectangle slinger.

Mark Amery
  • 103
  • 2
Mario
  • 8,145
  • 1
  • 14
  • 19
  • 3
    +1 for "polygon pusher." Never heard "triangle pusher" but I imagine the etymology is the same. – Shamtam Jan 29 '18 at 16:26
  • 2
    [Here is a layout of an NMOS NOR gate as you might see it in layout editing software.](https://commons.wikimedia.org/wiki/File:NOR_gate_layout.png) The different colors of the shapes correspond to different metal layers of the process. – trent Jan 29 '18 at 22:33
  • 4
    @trentcl that's CMOS, not NMOS. Note there are transistors on both sides of the output and no resistors. – John Dvorak Jan 30 '18 at 01:31
  • @JohnDvorak Yes, you are right. The Wikipedia article in which it's linked is misleading, and semiconductors aren't my field, so I didn't notice. Thanks! – trent Jan 30 '18 at 01:43
  • 2
    In what sense is it misleading? Feel free to edit it yourself. – John Dvorak Jan 30 '18 at 01:48
  • 2
    On [this page](https://en.wikipedia.org/wiki/NOR_gate), the picture was in the same row with the ones of which the text says "The diagrams above show" an NMOS gate. I probably had already edited it when you looked at the page. – trent Jan 30 '18 at 03:08
27

Early masks for the creation of layers on an IC were created by a photographic process that involved exposing the original photographic plate through a mechanically controlled triangular aperture. Hence triangle pusher.

The light source was fixed above the aperture, the plate was moved xy underneath. The point of a triangle was that additive it could give any orthogonal geometry required.

There were no laser printers back then.

RoyC
  • 8,418
  • 6
  • 25
  • 40
  • 3
    @oldfart obviously before your time then :) – Solar Mike Jan 29 '18 at 15:43
  • 3
    @Solar Mike No I remember rubylith ;). – RoyC Jan 29 '18 at 15:45
  • 1
    Can you elaborate how a triangular aperture would be of any benefit? It seems to me that it would give uneven exposure depending on orientation of the triangle relative to direction of travel. – Transistor Jan 29 '18 at 18:48
  • @Transistor I'm just guessing, but the aperture in optics is typically in the focal plane, not image plane. Therefore, the projected image is the Fourier Transform of a triangle. If the blades are rounded a bit you'll soften it more. This is fairly typical in old cameras (the "bookeh" comes out as hexagons). –  Jan 29 '18 at 19:29
  • 7
    Is this from your own experience, or can you provide any reference for the phrase "triangle pusher" in this sense? Google results for the phrase are unfortunately polluted with nail care products. I'd be interested to learn more about this old manufacturing process. – trent Jan 29 '18 at 22:40
  • 1
    I also searched Google before putting the answer up with no sensible result as you have pointed out. This is from personal experience. But goes back 40 yrs. I worked for an IC design consultancy. Our CAD files created on a Calma system would be broken down to triangles and then sent to a specialist photolith company to produce the mask masters and also coloured photo enlargements of each layer for manual checking. I have no direct connection of the term "triangle pusher" to this process it was not used in our company. But you have to admit that the coincidence is too significant to ignore. – RoyC Jan 29 '18 at 23:15
  • @trentcl If you have any better ideas I would be prepared to listen to them. – RoyC Jan 29 '18 at 23:21
  • 1
    @RoyC Nope, and I appreciate you sharing :) I had never heard "triangle pusher", though "polygon pusher" is common parlance for layout artists. I wonder if there might be some cross-contamination, perhaps "triangle pusher" started out in manufacturing and inspired the other phrase, if it did indeed come from a physical device with a triangular aperture. – trent Jan 29 '18 at 23:39
  • 4
    I've never heard of an optical pattern generator with a triangular aperture. I've only heard of rectangles, usually rotating. ([examples](http://www.cnf.cornell.edu/doc%5Cchoosing%20an%20optical%20pattern%20generator%20at%20the%20CNF%20v2.pdf)) Considering 99% of IC patterns will be rectilinear, the whole photolithography process is set up for masks with it (e.g. quadrupole illumination), although by that time most masks were e-beam. Making mask patterns out of triangles sounds strange to me, and twice as much work. – uhoh Jan 30 '18 at 04:29
  • 3
    How come this answer has upvotes? It seems to be basically wrong. – Fattie Jan 30 '18 at 17:59
  • 1
    Agree totally with @uhoh - The old mask maker I recall using in the mid-80s (running from a PDP-11 with paper tape to boot it up) was definitely rectangle-only, no triangles. – Jon Custer Jan 30 '18 at 20:54
  • @Fattie where exactly is this wrong. – RoyC Jan 30 '18 at 21:58
  • @Jon Custer modern ebeam mask writers use both rectangular and triangular beam shapes to achieve the required pattern. The advantage of a triangle is that you can create any shape polygon by combining triangles. – RoyC Jan 30 '18 at 22:01
  • @RoyC - sure, but your answer explicitly stated ‘early masks’, so I went for early mask makers. It was a big deal when non-rectangular machines were available. – Jon Custer Jan 31 '18 at 16:52
2

In computer graphics, a poly-pusher (short for polygon-pusher), is a system that uses a brute force approach to simply draw as many polygons as possible, as opposed to more intelligent systems that try to figure out things like which triangles are visible, or ray-tracing. Historically, brute-force has always won when compared to more complex systems. The winning philosophy seems to be, "do the simplest thing possible, as fast as possible.". Thus most modern-day graphics cards are "poly-pushers".

Not heard the term used for a hardware engineer, but it's possible that someone who believes in simple, fast, hardware, might be called a "poly-pusher", or maybe a "poly-pusher-pusher" :-)

Theagi
  • 67
  • 1
  • 2
    This was also the etymology I thought of as I'm sure I've heard it used in the context of NVIDIA's or ATI/AMD's graphics chips engineering teams. – Dai Jan 29 '18 at 21:00
  • @downvoters: please provide feedback as to why you down voted – gatorback Jan 30 '18 at 17:17
  • 1
    @gatorback Because this is a nonsense answer and the modern use of the word related to drawing polygons has _nothing_ to do with the use of the word in the ASIC industry 40 years ago. The only reason this even has a positive score is because it ended up on the hot network list, attracting kids from all over SE. This took it from 100% negative score to lots of upvotes. The final sentence even explicitly says "Never heard the term used for a hardware engineer". – pipe Jan 31 '18 at 12:35