I attached a frequency counter to the clock 4, 5, and 6 outputs on the XEM3005 breakout board and experimented with the FrontPanel PLL control.
I observe the following with these source options.
DIV2CLK/2 only produces the frequency indicated by the application when DIV2N is a multiple of 4.
DIV2CLK/4 only produces the frequency indicated by the application when DIV2N is a multiple of 8.
DIV1CLK/2 only produces the frequency indicated by the application when DIV1N is a multiple of 4.
DIV1CLK/3 only produces the frequency indicated by the application when DIV1N is equal to 6 (and only 6).
DIV1CLK/3 produces what looks like DIV1CLK/4 when DIV1N is a multiple of 8.
The frequencies shown by FrontPanel don’t change when DIV1N or DIV2N are changed and the above source options are used. The display is simply the REF or VCO frequency divided by 2, 3, or 4, as the case may be. But the frequency counter reports changes when DIV1N or DIV2N are changed. My observations seem to line up with a section of the Cypress document for the CY22150.
When DIV1N is divisible by four, then CLKSRC(0,1,0) is guaranteed to be rising edge phase-aligned with CLKSRC(0,0,1).
When DIV1N is six, then CLKSRC(0,1,1) is guaranteed to be rising edge phase-aligned with CLKSRC(0,0,1).
When DIV2N is divisible by four, then CLKSRC(1,0,1) is guaranteed to be rising edge phase-aligned with CLKSRC(1,0,0).
When DIV2N is divisible by eight, then CLKSRC(1,1,0) is guaranteed to be rising edge phase-aligned with CLKSRC(1,0,0).
Perhaps the FrontPanel application could indicate N/A when these source options are chosen and the DIV1N or DIV2N values don’t meet these constraints.