2x2 matrix multiply

Hi,

i have made one project in simulink (.mdl)
just a  2x2 matrix multiply

i used xilinx blockset toolbox for making 2x2 matirx multiply with the 
help of muliplier, add/subs, convert,gateway blocks etc.
actually i am making a 32 bit width 2x2 matrix multiplier..

In model block of 2x2 matrix muliply ,the arrangement (connection of all 
individual blocks) is same like i mentioned below.

in--->gateway in-->convert(32 bit, signed, decimal at 
16)-->mult-->add-->convert-->gateway out--> out and display.

compilation is perfectely happening and results is coming and accurate.

when i generate VHD code of it and sysnthesis and i go for the systhesis 
report, it is generating everthing good (i.e. LUT,FF, SLICE etc used) 
but with having high IOB (44 %)

for the same example when i reduce number of bits to 16 bits, signed and 
decimal at 8 and did the same procedure again now everything remain same 
but IOB used  got decrease(22 %)



for the same example when i reduce number of bits to 8 bits, signed and 
decimal at 4 and did the same procedure again still everything remain 
same but IOB used  further decrease.(11 %)

though IOB are decreasing but still they are utilizing more resources in 
FPGA

why this is happening?
is there any way to improve my mdl block diagram so i can sought out the 
problem of IOB utilization