Hi, I designed a microblaze using cache and MIG (DDR3) according to xilinx example. My microblaze uses "maximum frequency" profile, and runs at 200MHz. DDR3 runs at 400MHz with 2:1 divider, and has 256MB.
I use system debugger to download bitstream and elf (standalone application) to run it. But I found the application needs a very long time before running to main(). I stepped the assembly code, and found that startup code cosumes much time (about more than one minute) on initializing .BSS section (about 70MB) to zeros.