Configuration FrameStrobe Widths

[RobB720]

Hi,

Looking at the generated code the Frame_Select block instances have FrameStrobe inputs and outputs:
.FrameStrobe_I(FrameAddressRegister[MaxFramesPerCol-1:0]),
.FrameStrobe_O(FrameSelect[1MaxFramesPerCol+MaxFramesPerCol-1:1MaxFramesPerCol]),

FrameAddressRegister is defined by the parameter MaxFramesPerCol.

However, at the top of eFPGA_top FrameAddressRegister is defined as a wire in terms of FrameBitsPerRow.
wire[FrameBitsPerRow-1:0] FrameAddressRegister;

In my design, I have the parameters from my configuration:
parameter FrameBitsPerRow=32,
parameter MaxFramesPerCol=116,

So we therefore have mismatched bus widths in the code. Can this code be fixed?

Best regards,
Rob.

[KelvinChung2000]

I have done the fix in KelvinChung2000@98b8d5d
I checked it and changed the generated top wrapper to what you intended. But I am unsure if any side effects might cause problems.

If this works fine, please let me know, and I will merge this directly with the master.

[RobB720]

Just got around to following this up - thanks the fix looks good.

However, the FrameStrobe_I[MaxFramesPerCol-1:0] bus on Frame_Select block is connected to the FrameAddressRegister bus in eFPGA_top. This is the driven by eFPGA_Config, inside which FrameAddressRegister width is set by the parameter FrameBitsPerRow - i.e. 32 bits wide.

This code seems to force the FrameStrobe buses to be a maximum of 32 bits wide and we want 116 bits.

If we allow the FrameAddressRegister to be 116 wide by changing the parameters, will this break the eFPGA_Config RTL and the configuration mechanism?

Best regards,
Rob.

[KelvinChung2000]

I have designed the current version of FABulous so that if you change the parameter at the very top level, the parameter should propagate correctly to all the sub-modules. However, I am unsure if the configuration module is designed to handle the varying parameters. I have not looked into that side of the project.

[RobB720]

One idea we have had is to insert a decoder between the eFPGA_config block and the Frame_Select blocks. We could then have a 7-bit field within the FrameAddress register that is decoded to a 116-bit one hot bus that is then selectively driven by the Frame_Select blocks.

Would this work? Are we correct presume the FrameStrobe buses driven by the Frame_Select blocks are one-hot?

Could the configuration software be changed to encode the Frame_Select value?