I want to export some point clouds to a readable format (ie PCD), which implies encoding point colors into a single integer value, which is usually done with bit operators:

I was surprised to discover that it doesn't seem feasible at all in he jit paradigm, while it work as expected with lists, as shown in my example patch. What seem even weirder to me is that shifting ('>>' and '<<') and logical bitwise operators ('&' and '|') are listed in the jit.op help file but they doesn't seem to give expected result. The same operators are also not recognized in jit.gen.

So where is the error? Am I using [jit.<<] and [jit.|] the wrong way? Are they supposed to exis at all, given that they have no equivalent in gen?

comment

jit.gen

codebox

jit.pwindow

jit.expr

jit.op.<<

jit.<<

jit.cellblock

message

route

jit.matrixinfo

jit.matrix

jit.unpack

button

jit.bfg

expr

unpack

number

Hi,
I want to export some point clouds to a readable format (ie PCD), which implies encoding point colors into a single integer value, which is usually done with bit operators:
rgb = (r << 16) | (g << 8) | b
and to convert back:
r = (rgb >> 16) & 255
g = (rgb >>8) & 255
b = rgb & 255

I was surprised to discover that it doesn't seem feasible at all in he jit paradigm, while it work as expected with lists, as shown in my example patch. What seem even weirder to me is that shifting ('>>' and '<<') and logical bitwise operators ('&' and '|') are listed in the jit.op help file but they doesn't seem to give expected result. The same operators are also not recognized in jit.gen.

So where is the error? Am I using [jit.<<]  and [jit.|] the wrong way? Are they supposed to exis at all, given that they have no equivalent in gen?

Hi,
I want to export some point clouds to a readable format (ie PCD), which implies encoding point colors into a single integer value, ... 

bitwise-operators-not-working-in-ji

If you keep your matrice type as char, you only have 8 bits to operate the bit shifting within. So, you really can't do 8-bit bit shifting, or 16!

Use type long - showing here the coding part:

If you keep your matrice type as char, you only have 8 bits to operate the bit shifting within. So, you really can't do 8-bit bit shifting, or 16!
Use type long - showing here the coding part:


Yes! I just figured this out! I wasn't aware of (or forgot about?) the long type, so I tried to fake ints using float32 but obviously bitshifting didnt work. So I did a much slower workaround using / and % (see patch below). Your solution is obviously much more elegant, thanks!

However this doesn't explain why bitwise operators are not available in jit.gen. Is it because of implementation or hardware limitation, something like gen always working with floats internally regardless the input type?

If you keep your matrice type as char, you only have 8 bits to operate  the bit shifting within. So, you really can't do 8-bit bit shifting, or  16!

I’d say it’s the latte, but shifting on floats is quite an unusual operation….

Bitwise operators not working in ji