The following sequence would drop the first store when eccBytes=4:

    sb x0, 0(t0)
    nop
    sb x0, 4(t0)
    nop
    sb x0, 1(t0)

Because the first and second store are to different ECC granules, the
hazard check correctly allowed the second one to proceed, but the third
was merged with the second, even though it conflicted with the first.
So, don't allow the third to be merged with the second, since the second
stored to a different ECC granule.

This reverts commit 7d94074b, which
works fine with optimistic behavioral RAMs but not real ones.

Fix ram model source reuse

This reverts commit 39b7e930.

Now that the RAMModel can properly tolerate overlapping responses
in the face of source reuse, we can re-enable the regression test.

maskrom: retain data for d channel is not ready

If a multibeat response comes back, the source might be reused.
If response reordering has made the multibeat response invalid,
we need to remember this even if the valid bit is cleared on reuse.

bump riscv-tools

Critical paths

The D$ corrects via writeback, so which word the error was in doesn't
matter, as the entire line is corrected.

This makes sure that all the base classes call instantiate()

diplomacy: provide default clock/reset for LazyRawModuleImp

Critical paths

We were using it to compute the next PC on flush vs. replay (which require
PC+4 and PC, respectively).  This fix gets rid of the adder altogether by
reusing the M-stage PC in the flush case, which by construction holds PC+4.

It was counting conflict misses but not cold misses.

Handle parity errors with a pipeline flush rather than a faster
frontend replay, reducing a critical path.

The deleted code was a holdover from Hwacha's vector fences.

It takes a while for the pipeline to compute the stall signal, so avoid
using it until the last logic levels in the clock cycle.

Don't build verilog twice for emulator and emulator-debug

Fix I$ miss perfctr

Since we aren't using chisel2, the output is the same either way.

TLXbar: Allow more masters and slaves and issue a warning.

Closes #920

The old version was counting prefetches, too.

Icache prefetch

Fix some critical paths

TLFuzzer: Allow Ordered clients to be created as well by the fuzzer