IEEE float radix in Simvision

From the same context menu, you need to use the "Trace Type" option to set the signal as Analog Linear, or Analog Sample + Hold.

You can see the original signal above. I can change the trace type to Analog Sample + Hold, Analog Linear gives an error. However, even if I change it to analog sample + hold, I still cannot view the signal value in floating-point format. What I see is below,

What am I missing here? 

Are you trying to view an element of an array? "mem[504]" suggests that's the case.

I suspect SimVision can only convert simple vectors to analogue, and not arrays or array elements.

That's right. "mem" is actually a 512 by 32-bit array. Thus, "mem[504]" is a 32-bit element. It's strange that SimVision cannot convert this directly to analogue. Is there a possible workaround for this?  

Actually, I see that Analog Sample + Hold works for arrays of int and bit[31:0], at least for my setup.

You might need to share some more details of your code to help us understand why it's not working for you.

Actually, I see that Analog Sample + Hold works for arrays of int and bit[31:0], at least for my setup.

You might need to share some more details of your code to help us understand why it's not working for you.

Interesting... In my case, "mem" is defined as "reg [31:0] mem [0:511]". 

"mem" is just the storage variable of an SRAM model. I do not know what else I should share, honestly. The image above shows the waveform to a greater extent. Analog sample + hold shows nothing useful. The value we see in the waveform is not even the correct floating-point value of the hex representation. I'm not sure if I should point out that I'm using Verilog, not SystemVerilog.

Also, I see that in your waveform, the value of the signals are in decimal format, not floating-point format, even though your trace type is analog sample + hold.

Ah, so what you missed is that you have an array of 32 words, each word is 512 bits. SimVision is still able to show a 512 bit word as an analogue value (I'm surprised but credit where credit is due).

However there is no obvious conversion from a 512 bit integral value into a IEEE floating point number. If you know which of the 512 bits are the mantissa and which are the exponent, you could use the expression calculator to write a conversion function and see the results in the waveform...

Oops, I just noticed that I mixed the order of the numbers in my previous comment. It should be correct now. In other words, word length is 32 bits, and there are 512 words. So it still makes no sense why SimVision cannot show the floating-point representation of a 32-bit value.

How should a 32-bit int be represented as a 64-bit IEEE float? I don't think there is a built-in definition for that so you would need to create a conversion expression for it.

There is also a 32-bit IEEE float definition, i.e. the single precision floating-point, which is what I'm trying to get. I do not know why you assumed I wanted a 64-bit IEEE float. It's strange how difficult it is to do such a simple task in SimVision. I just want the waveform window to show me the floating-point value of a 32-bit signal. For instance, instead of " 'h479710af ", I want it to show " 77345.3671875 ". It shouldn't convert an integer to float, just show the floating-point representation of a 32-bit signal.

I assumed 64 bit because "real" is 64 bit in SV and you're the first person to ask about 32 bit fp in the 16 years I've been supporting SimVision