* extend tcallnode with the ability to pass a tspecializationcontext so that tcallcandidates can do a final specialization
* the final procdef is registered at the end of tcallnode.pass_typecheck
git-svn-id: trunk@31763 -
types until it finds a non-aggregate type, and then returns "true" if
that aggregate (record or array) contains one element
* it now also returns the type of that element directly rather than a
tfieldvarsym, since we only care about the type and in case of an array
there is no field
git-svn-id: trunk@30205 -
returned on the x87 fpu stack on x86-64
o also handle the fact that this means that we'll generated float
loads/stores with OS_128/OS_F128 in case the surrounding record
is padded to a multiple of its alignment
git-svn-id: trunk@28985 -
becoming X86_64_SSE_CLASS/X86_64_INTEGER_CLASS/... This is not defined in
the ABI because there sizeof(extended)=16 and hence the class of the upper
eightbyte of the extended value can never be merged with the class of
another field (except in a union, but then the class of the lower
eightbyte will also be overwritten).
Handle it the same as when the lower eightbyte class of an extended in an
aggregate gets changed into something else: pass everything in memory
(part of mantis #26993)
git-svn-id: trunk@28984 -
Currently it is implemented only for x86-CPUs supporting the FMA extension. While it would
be possible to implement it in software or simulate fma(<single>,<single>,<single>) using
double operations, it makes no sense in my opinion to do so.
git-svn-id: trunk@27564 -
o made all (non-abstract) tdef and tsym constructors virtual
o added c*def/c*sym classref types for every (non-abstract) t*def/t*sym
class
o added cpusym unit for every architecture that derives a tcpu*def/tcpu*sym
class from the base classes, and initialises the c*def/c*sym classes with
them. This is done so that the llvm target will be able to derive from
the tcpu*def/sym classes without umpteen ifdefs, and it also means that
the WPO can devirtualise everything because the c* variables are only
initialised with one class type
o replaced all t*def/t*sym constructor calls with c*def/c*sym constructor
calls
git-svn-id: trunk@27361 -
gcc and clang always allocate them in multiple of 4 bytes too (except in
case of _Bool, which is 1 byte -> do the same for PasBool) and this is
one possible interpretation of the ABI. This also frees us from having to
perform special adjustments afterwards to ensure that smaller values are
sign/zero-extended to 32 bits
git-svn-id: branches/hlcgllvm@26993 -
* Additional two minor changes: reset po_delphi_nested_cc on result, so it is independent of current {$modeswitch nestedprocvar} state, and removed call to alloc_proc_symbol (which belongs to pass 2).
git-svn-id: trunk@26222 -
llvm-targeting compiler, but use llvmnode instead
o after this commit it possible to build an llvm/x86-64 compiler with
very limited functionality (you can only compile dummy system units
containing procedures --not functions!-- with simple arithmetic
and array expressions; parameters are partially supported, making
calls is not). Use a make command line similar to the following:
make OPT="-Fullvm -Fillvm -dllvm" PPC_TARGET=x86_64 clean all
The resuling compiler can generate (some) valid code for at least
Mac OS X/x86-64 and should also be able to target Linux/x86-64.
Win64 support may work by simply adding it to the list of supported
targets in as_llvm_info in llvm/agllvm.pas.
git-svn-id: branches/hlcgllvm@26058 -
using it in inline asm causes an internal error and removing it didn't show
any regressions after running the testsuite on i386-linux.
git-svn-id: trunk@25827 -
cgobj.pas, tcg:
* g_save_registers: add the amount of used address registers to size as well
* g_save_registers: save all used address registers
* g_restore_registers: restore all stored address registers
m68k/cpubase.pas:
* rename saved_standard_address_registers to saved_address_registers
all other platform's cpubase.{inc,pas} (except alpha, ia64 and vis which are not up to date):
* add a saved_address_registers variable with one entry of RS_INVALID
At least a "make fullcycle" did complete.
git-svn-id: trunk@25664 -
* Split generation of regular and TLSIE GOT entries. Although code for both is similar to some extent, mixing them in a single method turns it into spaghetti.
* Several fixes to content and dynamic relocations of TLSIE entries.
git-svn-id: trunk@25180 -
differentiate between 64 bit and 128 bit (R_SUBMMWHOLE) SSE vector regs,
and support spilling/assembling for R_SUBQ SSE registers (8 bytes)
(mantis #23962)
We currently never use the full 128 bit of an SSE register, and
spilling for those hasn't been implemented yet either (R_SUBMMWHOLE
SSE regs are spilled into a 4-byte temp currently -> can overwrite data)
git-svn-id: trunk@23700 -
paramgr.pas, tparamanager:
+ add new method "handle_common_ret_in_param"
* extract common code (safecall + record constructor handling) from "ret_in_param" to "handle_common_ret_in_param" and call the latter in "ret_in_param"
[aarch64,arm,avr,i386,x86_64]/cpupara.pas, tCPUparamanager.ret_in_param:
* call "handle_common_ret_in_param" instead of implementing the same check again and again
ncgcal.pas, tcgcallnode.handle_return_value:
* move the check for (record) constructors to "tparamanager.handle_common_ret_in_param"
git-svn-id: trunk@23520 -
* For unresolved weak symbols, provide dynamic binding only if they are referenced via GOT or PLT (ld-compatible behavior).
* Made more TElfExeOutput properties/methods usable by descendant classes.
* x86_64/cpuelf.pas: some refactoring, handle DTPOFF relocations, prohibit TPOFF relocations in shared objects.
* i386/cpuelf.pas: set symref_from_text flags for copy relocations to work correctly.
git-svn-id: trunk@23375 -
that the type of the parameters can be determined automatically
o added compilerproc declarations for all helpers called in the compiler
via their assembler name, so we can look up the corresponding procdef
git-svn-id: trunk@23325 -
* Methods ReportNonDSOReloc and ReportRelocOverflow moved to TElfExeOutput, they are applicable to all target-specific descendants.
git-svn-id: trunk@23094 -
* Moved target-dependent constants into ElfTarget records, eliminates a lot of $ifdef's in ogelf.pas.
+ Added TElfTarget.loadsection hook and pass unknown sections to it.
git-svn-id: trunk@23090 -
* Changed signature of GOTRelocPass1 to pass relocation index instead of relocation itself, because sometimes (e.g. when doing TLS relaxation) we need to process several subsequent relocations at once.
git-svn-id: trunk@23077 -
x86_64 is the most elaborated, has some degree of indirect function (GNU_IFUNC) and TLS support, so it is even able to link with static libc/pthreads code (tw14265) and produce an executable that can launch (but still fails due to invalid DWARF unwind info).
i386 produces working shared libraries if they are compiled with -Cg, without one your mileage may vary. tw14265 does not link yet due to missing COMDAT group support.
git-svn-id: trunk@22775 -
assembler writers on x86-64 after r22181, the shift in positions
caused every occurrence of RIP to become EIP in the assembler code)
git-svn-id: trunk@22234 -
like field reordering (possible problems cracker classes) or using ebp as normal register (broken
stack traces from dump_stack)
+ niln is also valid in a cse domain
* parameters passed by reference shall have a complexity >1
* load nodes from outer scopes shall have a complexity >1
* better cse debugging
+ more node types added to cse
* consider parameters passed by reference in cse
* take care of cse in parameters in simple cases
git-svn-id: trunk@22050 -
(tf_safecall_clearstack is set for it) -> also use pocalls_cdecl for it
and adjust the safecall tests not to try parameter scenarios that are
unsupported for cdecl
git-svn-id: trunk@21948 -
order to minimise memory losses due to alignment padding. Not yet enabled
by default at any optimization level, but can be (de)activated separately
via -Oo(no)orderfields
o added separate tdef.structalignment method that returns the alignment
of a type when it appears in a record/object/class (factors out
AIX-specific double alignment in structs)
o changed the handling of the offset of a delegate interface
implemented via a field, by taking the field offset on demand
rather than at declaration time (because the ordering optimization
causes the offsets of fields to be unknown until the entire
declaration has been parsed)
git-svn-id: trunk@21947 -
getintparaloc + adapted all call sites of getintparaloc. This
led to a number of additional, related changes:
o corrected the type information for some getintparaloc parameters
o don't allocate some intparalocs in cases they aren't used
o changed "const tvardata" parameter into "constref tvardata" for
fpc_variant_copy_overwrite to make pass-by-reference semantics
explicit
o moved a number of routines that now have to call find_system_type()
from cgobj to hlcgobj so that cgobj doesn't have to start depending
on the symtable unit
o added versions of the cpureg alloc/dealloc methods to hlcgobj that
call through to their cgobj counter parts, so we can call save/restore
the cpu registers before/after calling system helpers from hlcgobj
(not implemented in hlcgobj itself, because all basic register
allocator functionality is still part of cgobj/cgcpu)
git-svn-id: trunk@21696 -
* set tcgpara.def for the function return location (field introduced for and
already used by the JVM code generator, required for future hlcg
functionality)
git-svn-id: trunk@21691 -
