Asic

See merge request projectn/noop!75

* refer to the RAS prediction hints in RV32I in the RISC-V manual

This reverts commit 2dfc501e.

Revert "Merge branch 'fix-mmio' into 'master'"

See merge request projectn/noop!73

This reverts merge request !72

noop: move the source of lsuMMIO from dcache to dtlb

See merge request projectn/noop!72

Fix debugfmt

See merge request projectn/noop!68

* if we disable dcache, the compilation will fail

device,AXI4Timer: add programmable freq and inc

See merge request projectn/noop!69

* This can help to generate desired frequency of timer interrupt, no
  matter what frequency of the clock is.

Debian

See merge request projectn/noop!66

* According to The RISC-V Instruction Set Manual Volume II:
Privileged Architecture, for instruction-fetch access or
page-fault exceptions on systems with variable-length instructions,
m/stval will contain the virtual address of the portion
of the instruction that caused the fault while m/sepc will
point to the beginning of the instruction.

* To run more programs on debian, we use 256MB memory described in dts.
  But if the hardware still provides 128MB memory, the memory access
  will wrap around when Linux allocate a physical page excceed 128MB,
  which causes wrong memory access.

* after 4 hours, the command will be sent repeatly

* An instruciton fetch will set the A bit of the PTE, even this
  instruction will be flushed later due to branch mis-prediction.
  Although this will not cause correctness issue, it will introduce
  inconsistent behavior compared with NEMU.
* To run DiffTest while testing debian, a solution is to disable
  updating A bit and D bit in both NOOP and NEMU.

* this can reduce the idle time in Linux

* this is because we can not control the mip.mtip bit in NEMU