I synthesized a small device to test the block-ram inference.
I got a message from XST :
The small RAM will be implemented on LUTs in order to maximize performance and save block RAM resources. If you want to force its implementation on block, use option/constraint ram_style.
However, I don't know where to find this option/constraint either in ISE (11.1 in my case) or in constraint files...
I don't want to use VHDL attributes directly in my code.
In your project directory, you'll find a file called "your-design.xst". You can add the following at the end of the list (or anywhere after "run"):
-ram_style block # ( | auto | distributed )
-rom_style block # ( | auto | distributed )
These should make sure you're going to get BRAM mapping instead of distributed RAM (which means LUT-based memory).
This option could also come in handy:
-auto_bram_packing yes # ( | no )
Remember that each of those must be on a line of their own, and that you'll need to remove the "#" and whatever comes after that.
If you're using the ISE GUI, go to
Synthesis -> Process Properties -> HDL options
and choose the above options there.