Le 2011-10-25 à 20:06:00, Epic Jefferson a écrit :

But when 64-bit audio support

The stuff that was fixed in pd versions 41, 42 and 43 are bugs about 64-bit mode of the CPU. This means that whether you use Win64 executables on Windows, or ELF64 executables on Linux, or the equivalent thing on OSX, those bugfixes are about the pointer size being different from the float size. Those bugs are only related to arrays ([table] and such).

pointer size is related to how much RAM you can use. With 16-bit pointers, your total RAM has to be 64k. With 24-bit pointers, it can be 16 megs. With 32-bit pointers, it can be up to 4 gigs. With 48-bit pointers, it can be up to 262144 gigs. 64-bit pointers could do many times more than that, but in practice, computers don't allow so many gigs of RAM.

The new stuff for pd 44 would be the optional use of 64-bit floats, which is the same thing as in the video, and has rather little to do with 64-bit pointers.

is mentioned and seeing the differences between max 5 and max 6 in terms of this supposed improvement.

If the vertical axis is a log plot (in decibels) then it looks a lot worse than it sounds like, because a log plot is made to stretch quiet details while squishing loud things.

The patch in the video does not say what the axes mean.

But there are true problems with indexing a table that long. This is why Pd already supports higher-resolution seeking. It does so using two 32-bit floats at once. For this, you send a float to the right-inlet of [tabread~] and then you feed a signal in the left-inlet. But it's harder to use than just a single [line~] with 64-bit floats.

And even more so, I have no clue what " 0.44 - support for using 64-bit double numbers as the base number, see pd-double" means.

«Pd-double» is a branch of pd currently developed to achieve 64-bit float. The name «double» usually stands for 64-bit float.

| Mathieu BOUCHARD ----- téléphone : +1.514.383.3801 ----- Montréal, QC