012cf8ceae
libFIRM provides a graph-based intermediate representation (IR), which operates exclusively in SSA-form, starting at IR construction until assembler code emission. It offers many analyses and optimizations, provides extensive debug support and includes a backend framework. * analyses: dominance, loop tree, execution frequency, control dependencies, call graph, rapid type, def-use, alias analysis, class hierarchy analysis * optimizations: dead code elimination (happens implicitly), constant folding, local common subexpression elimination, arithmetic identities (happens on the fly), unreachable code elimination, global common subexpression elimination, code placement, strength reduction, scalar replacement, if-conversion, load/store optimization, control flow optimizations, reassociation, partial condition evaluation, tail recursion elimination, inlining, procedure cloning * enhanced debugging support: extensive checkers, breakpoints on node creation, entity creation, graph dumping * lowering of intrinsics, double word arithmetics, bitfields * generic backend features: * novel SSA based register allocator * several SSA copy coalescing and spilling algorithms * algorithms for instruction and basic block scheduling * working ia32 backend, unfinished backends for MIPS, ARM, PPC32 * handwritten recursive descent C89/C99 frontend available (lang/cparser)
24 lines
1.4 KiB
Text
24 lines
1.4 KiB
Text
libFIRM provides a graph-based intermediate representation (IR), which
|
|
operates exclusively in SSA-form, starting at IR construction until
|
|
assembler code emission. It offers many analyses and optimizations,
|
|
provides extensive debug support and includes a backend framework.
|
|
|
|
* analyses: dominance, loop tree, execution frequency, control
|
|
dependencies, call graph, rapid type, def-use, alias analysis,
|
|
class hierarchy analysis
|
|
* optimizations: dead code elimination (happens implicitly), constant
|
|
folding, local common subexpression elimination, arithmetic identities
|
|
(happens on the fly), unreachable code elimination, global common
|
|
subexpression elimination, code placement, strength reduction,
|
|
scalar replacement, if-conversion, load/store optimization, control
|
|
flow optimizations, reassociation, partial condition evaluation,
|
|
tail recursion elimination, inlining, procedure cloning
|
|
* enhanced debugging support: extensive checkers, breakpoints on node
|
|
creation, entity creation, graph dumping
|
|
* lowering of intrinsics, double word arithmetics, bitfields
|
|
* generic backend features:
|
|
* novel SSA based register allocator
|
|
* several SSA copy coalescing and spilling algorithms
|
|
* algorithms for instruction and basic block scheduling
|
|
* working ia32 backend, unfinished backends for MIPS, ARM, PPC32
|
|
* handwritten recursive descent C89/C99 frontend available (lang/cparser)
|