summaryrefslogtreecommitdiffhomepage
path: root/src
Commit message (Collapse)AuthorAge
* Factor out common code.Christoph Mallon2017-04-26
|
* Correctly use 'false' instead of an enum constant which happens to have the ↵Christoph Mallon2017-04-04
| | | | value 0.
* Correct accidentally inverted logic of 'target.use_frame_pointer'.Christoph Mallon2017-03-20
| | | | '-O2' activated, instead of deactivated, the use of the frame pointer, because the use of the variable was flipped in two other places.
* Use new libfirm target APIMatthias Braun2017-02-20
| | | | | This moves the target and platform knowledge and initialization into libfirm.
* Rename long_double_x87_80bit_float -> x87_long_doubleMatthias Braun2017-02-17
|
* No need to initialize default_exe_outputMatthias Braun2017-02-17
| | | | Target initialization code will set it anyway.
* Rename init_gen_firm -> init_firm_optMatthias Braun2017-02-17
|
* Improve optimization level parsingMatthias Braun2017-02-17
| | | | | | | - Use an enum for the different optimization levels. - Set them slightly later in the initialization process (so we do not need to re-set them) - Deduce preprocessor defines from level instead of setting flags.
* Target: set biggest and long double alignment correctlyMatthias Braun2017-02-13
|
* predefs: Define ILP32Matthias Braun2017-02-13
|
* Avoid __pic__ preprocessor define if pic is disabledMatthias Braun2017-02-13
|
* Unifying name manglingMatthias Braun2017-02-13
| | | | No need to have an own mangling scheme for each target.
* Ignore fastcall/stdcall on non-win32 32bit targetsMatthias Braun2017-02-13
| | | | | | This is only a valid calling convention on win32 32bit targets, on other targets the linker cannot even deal with the name mangling containing '@' characters.
* Dump types also after lowering, not just before.Christoph Mallon2017-01-29
|
* Adapt to libfirm irarch changesMatthias Braun2017-01-28
|
* Adapt to libfirm va_arg changesMatthias Braun2017-01-28
|
* Allow statement expressions ending in a label containing an expression ↵Christoph Mallon2017-01-24
| | | | | | | | | statement to have this expression as result. Technically ({ bla: 0; }) does not end in an expression statement, but in a label statement. Due to an implementation detail in GCC (labels do not have an inner statement, but their inner statement is made their successor) the statement expression does end in an expression statement. This fixes C/gnu99/stmtexpr.c.
* Use big endian for MIPS.Christoph Mallon2017-01-12
|
* Add target `mips`.Christoph Mallon2017-01-12
|
* Adapt to libfirm new_int_mode() function.Matthias Braun2016-12-30
|
* Do not create Pin nodes in the const_code irgMatthias Braun2016-12-30
|
* Mark method types of functions with unspecified parameters as variadic.Manuel Mohr2016-12-22
|
* Claim to be older GCC version 4.2.0.Manuel Mohr2016-08-25
| | | | | | | This is the same version that Clang pretends to be. It prevents being served certain macro definitions by newer glibc headers (e.g., math.h) that refer to builtin functions that we do not implement, and which are not present in glibc as functions. This fixes #176.
* Remove unreachable codeSebastian Buchwald2016-07-28
|
* Handle definitely leaving a compound expression.Christoph Mallon2016-07-28
| | | | This fixes stmt_expr_goto.c.
* Resolve warning about possibly uninitialized local variable.Christoph Mallon2016-07-24
|
* Add support for __builtin_isnanSebastian Buchwald2016-07-15
|
* Update submodule 'libfirm'.Christoph Mallon2016-07-12
| | | | Now the additional properties are directly set when constructing a method type.
* Update submodule 'libfirm'.Christoph Mallon2016-07-12
| | | | Now the calling convention is directly set when constructing a method type.
* Check for '__unix__' instead of just '__linux__'.Christoph Mallon2016-07-06
| | | | All Unices should have 'mmap()'.
* Accept 'long float' as rare pre-ANSI spelling of 'double'.Christoph Mallon2016-07-06
|
* Regroup float characteristics for better readability.Christoph Mallon2016-06-13
|
* Factorise code to create an initializer for a complex value.Christoph Mallon2016-06-12
|
* Adapt to removal of atomic entity functions.Christoph Mallon2016-06-12
|
* Adapt to automatic mode inference of 'Div' and 'Mod'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'Sub'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'Add'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'Shl', 'Shr' and 'Shrs'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'Mul'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'And', 'Eor' and 'Or'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'Minus'.Christoph Mallon2016-06-07
|
* Adapt to automatic mode inference of 'Not'.Christoph Mallon2016-06-07
|
* Remove the parameter 'for_closure' from 'create_method_type()'.Christoph Mallon2016-05-22
| | | | The only caller passes 'false'.
* Adapt to API change of 'new_type_method()'.Christoph Mallon2016-05-22
|
* Adapt to API change of id_unique().Christoph Mallon2016-05-17
|
* Adapt to simplified handling of array types.Christoph Mallon2016-05-17
|
* Fix indentationSebastian Buchwald2016-05-12
|
* driver: Do not issue a warning, that the option is ignored, when repeating ↵Christoph Mallon2016-05-10
| | | | the compilation mode.
* Factorise code to parse argument lists.Christoph Mallon2016-05-09
| | | | | This improves the error message in case of too few/many arguments. Also '__builtin_constant_p', '__assume' and '__builtin_classfy_type' use 'parse_assignment_expression()' instead of 'parse_expression()' now, which seems more correct.
* Split '-Wparentheses' into 'Wparentheses-assignment', ↵Christoph Mallon2016-05-01
| | | | 'Wparentheses-comparison', 'Wparentheses-else', 'Wparentheses-logical' and 'Wparentheses-shift'.