ALTracer has quit [Read error: Connection reset by peer]
ALTracer has joined #scopehal
ALTracer has quit [Remote host closed the connection]
ALTracer has joined #scopehal
Yamakaja_ has joined #scopehal
octorian_ has joined #scopehal
nohit has quit [Ping timeout: 260 seconds]
Yamakaja has quit [Ping timeout: 248 seconds]
octorian_ has quit [Remote host closed the connection]
octorian has quit [Ping timeout: 265 seconds]
nohit has joined #scopehal
<dingwat>
is there a baked-in limit on acquisition length when using thunderscope or RAM dependent? Referencing the TS crowdsupply page which says 1Gpts..
octorian has joined #scopehal
<azonenberg>
dingwat: The maximum waveform side scopehal supports for analog waveforms is 2^30 samples, which results in a 2^32 byte memory buffer of float32
<azonenberg>
Vulkan doesn't let you allocate >4GB of memory in a single call
<azonenberg>
(the maximum digital waveform size is thus 4G points or 2^32 samples since we store them as int8)
<dingwat>
azonenberg: gotcha
<azonenberg>
There is a long term open wishlist item to figure out how to make a single waveform span multiple non-contiguous memory blocks
<azonenberg>
nobody is actively working on it, it's architecturally difficult
<d1b2>
<david.rysk> apparently that's implementation dependent and provided in VkPhysicalDeviceMaintenance3Properties::maxMemoryAllocationSize
<azonenberg>
the limit can be smaller but iirc 4GB is a technological limit
<azonenberg>
it uses a 32 bit size somewhere
<azonenberg>
and every implementation i've seen or tested on is actually 4GB
<azonenberg>
(now you may be limited to less by available free memory, but if the memory exists every implementation i've looked at will let yo uallocate it)