Lemberg is a time-predictable VLIW processor optimized for
performance.

Lemberg can execute up to four instructions per cycle in a four-stage
pipeline. The processor includes one to four ALUs, a jump- and a
memory-unit, and optionally also a floating-point unit for single and
double precision floating-point numbers. Instructions are cached in a
method cache, and data may be cached in a stack cache, direct-mapped
cache or a fully associative cache.

Lemberg is time-predictable in the sense that the worst case execution
time of program fragments can be computed easily. The caches have been
designed to reduce the overestimation for worst-case execution time
analysis as much as possible. Branch prediction is optional and was
designed with analyzability in mind.

The processor in its largest configuration runs at 66 MHz on an Altera
DE2-70 board, where it consumes about 45K logic cells and 62KB of
on-chip memory, including the floating-point unit and
caches. Naturally, smaller configurations require fewer resources and
usually can run at higher clock frequencies.

Along with the hardware implementation come an assembler, a back-end
for LLVM, and a basic port of the newlib libc implementation.