Low Power Synthesis, Problem with sleep_mode

[Arturi]

Hello there!
Does anybody have tried to optimize power in PKS with sleep_mode?
I do get nice results using clock gating but sleep_mode makes the area
of the design to substantially increase and brings no power saving at
all. I suppose that the sleep-mode logic should be committed only when
timing is not a issue and power is saved.
In order to understand it better I built a simple test case with
multipliers where operand isolation is a simple task for the tool.
There I do get some
improvement in power
But in our design's code it doesn't help at all (actually it only
harms). I tried in two blocks, with no success, one is supposed to
have lots of arithmetic operations (video scaling). I tried different
flows even the one where I stop for power simulation and generate a
toggle count file from the gate level simulation of the netlist.
Anybody with experience on that? It can't be that the tool generates
the additional logic if no power saving comes along with it.
Is the tool that BAD or am I missing something??
Best Regards