The fluorescence of GFP is lost when the protein is fully denatured, as it would most likely be on SDS-PAGE. The protein might possibly refold if the SDS is soaked out of the gel and replaced with a buffer, allowing the fluorescence to recover, at least partially.
If you don't heat the samples for SDS-PAGE, GFP may retain enough of its structure to fluoresce, but the migration on the gel will be anomalous, giving a molecular weight that is higher than the true value.
See this application note for some useful details:
Dear Adam, Thank you so much for your answer. I assumed that exactly as you describe the GFP tagged proteins should behave in SDS-PAGE. Perhaps I will try to use gel filtration on a plane for similar purposes.
If you want to detect your gpf-labeled proteins, maybe you can carry out a western blot after the electrophoresis and then use an antibody against the GFP tag.
You won't be able to detect GFP fluorescence but you will detect your labeled proteins.