News
coins-1.4.4.4 was released (2010/04/20)
Compiler option of the form-coins:lir-opt=xxx/yyy/zzz
was added to specify back end optimization. (see "How to use the Compiler Driver " in "How to use COINS ")
Bugs in the data flow analyzer, snapshot, etc. were corrected.
coins-1.4.4.3 was released (2009/05/15)
Target machine description x86sse2.tmd was added.Corrections and revisions for SSA optimization, globalReform, data flow analysis, HIR handling, etc. were done.
coins-1.4.4.2 was released (2008/09/17)
Target machine descriptions (x86_64.tmd, mips.tmd) and SSA optimization were corrected.Program comments and documentations (COINSdoc.en loop parallelization, etc.) were revised.
coins-1.4.4.1 was released (2008/06/23)
Target machine description of 64 bit x86 (x86_64.tmd) and arm (arm.tmd) were modified.coins-1.4.4 was released (2008/05/17)
Compile option gprof was added.Copyright notices were changed to
Copyright (C) 2007 Association for the COINS Compiler Infrastructure
coins-1.4.3.3 was released (2008/03/14)
Modifications for MachineParam and its subclasses, etc. were done.coins-1.4.3.2 was released (2008/03/07)
Modification for word-addressing machines was done. Comments starting with "//" were permitted. Bugs concerning asm-statement etc. were corrected.coins-1.4.3.1 was released (2007/12/19)
Bugs (on -O3, -O4 optimization and some case of type cast) were corrected.coins-1.4.3 was released (2007/11/21)
64 bit x86 (x86_64) code generator and HIR transformer based on global pattern matching were added.Correction of bugs and modification of test programs were done.
coins-1.4.2.2 was released (2007/04/02)
Optimization items for -O3 were changed.coins-1.4.2.1 was released (2007/03/22)
Bugs were corrected.coins-1.4.2 was released (2007/02/28)
More than 300 test programs were added.English documents were revised (see doc/international/spec.en in the archive file).
coins-1.4.0 was released (2006/12/01)
Features for embedded systems were added:- C language expansion
- asm statement
- pragma
- TMD expansion
- Macro facilities to simplify the description of TMD
- 2 cost parameters (execution time, code size) can be specified for each LIR pattern
- Common TMD
- Make full use of short address field instructions
- Profiling (execution count of each statement)
-
Loop parallelization
- Generation of parallelized machine code (as well as OpenMP program)