At 18:33 31/01/2008, you wrote:
>The big change from version 3.5.4 is that the internal virtual
>machine was reworked to use operands in registers rather than
>pulling operands from a stack. The virtual machine stack has
>now been removed. The removal of the VM stack will help prevent
>future stack overflow bugs and will also facilitate new optimizations
>in future releases.
>
>There should be no user-visible changes to the operation of SQLite
>in this release, except that the output of EXPLAIN looks different.
>
>In order to make this change, about 8.5% of the core SQLite code
>had to be reworked. We thought this might introduce instability.
>But we have done two weeks of intensive testing, during which time
>we have increased the statement test coverage to 99% and during
>which we have found and fixed lots of minor bugs (mostly things
>like leaking memory following a malloc failure). But for all of
>that testing, we have not detected a single bug in the new
>register-based VM. And for that reason, we believe the new
>VM, and hence version 3.5.5, is stable and ready for production
>use.
The new VM was designed with a possible (low probable) port to
VHDL/Verilog for Sqlite in FPGA? Changing from a VM stack to VM
register based make easier to port.
Perhaps it's a very crazy idea...
--------------------------------------------------------------------------------
One reason that life is complex is that it has a real part and an
imaginary part
-Andrew Koenig
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
