 florian
|
08817f1aed
* fixes to deadstore optimization
|
%!s(int64=4) %!d(string=hai) anos |
 marco
|
1edc9cd01b
* patch from mantis 37205, add OS X sysctl constants
|
%!s(int64=4) %!d(string=hai) anos |
 Jeppe Johansen
|
d12dd1bdae
* CortexM4: Add some missing system registers
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
e6f01065ec
+ Aarch64: use LSE if available for atomic intrinsics
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
f3d8efc8ec
* Aarch64: support LSE instructions
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
f879d76491
* forgotten " added
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
f38f9ff930
* Aarch64: cpu capabilites defines properly named
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
9e7d80a8bd
* Aarch64: the M1 is an ARM v8.4-a CPU
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
3999d037e3
+ write controller unit wiht -ix as well
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
bf65bad5c5
+ Aarch64: cpu_capabilities support
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
f0023a3b04
* Aarch64: patch by J. Gareth Moreton: fix constant writing, resolves #38695
|
%!s(int64=4) %!d(string=hai) anos |
 svenbarth
|
e53d18d876
* Windows on ARM64 adheres to the usual Aarch64 ABI, thus references need to be copied on the caller side as well
|
%!s(int64=4) %!d(string=hai) anos |
|
svenbarth
|
2885d3ab56
* fix for Mantis #37426: don't allow an instance function of a type helper to be used on the type instead of a concrete value
|
%!s(int64=4) %!d(string=hai) anos |
|
marco
|
787798e96f
* split propsheetheader in -A and -W variants.
|
%!s(int64=4) %!d(string=hai) anos |
|
Jeppe Johansen
|
d712c64236
* thumb2: Optimize fillchar a bit more with a wider inner loop chunk size
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
682e29c4ff
+ enable colored output on darwin, resolves #38693
|
%!s(int64=4) %!d(string=hai) anos |
|
Jeppe Johansen
|
8b779975aa
* thumb2: Make fillchar smaller, faster and more reliable. Some gas versions remove the "mov r0,r0" making the calculated branch break
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
7811b915a9
* x86-64: we can spill replace 32 bit operations if the operation does not modify a register
|
%!s(int64=4) %!d(string=hai) anos |
 pierre
|
8ea95e9b95
Use value $00000F00 instead of $F0000000 for IF_FPMASK, and adapt all floating point constants, to avoid IF_VFPv4 having the same value as IF_PASS2
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
cc64d9eb4e
* do not swap left/right code generation for assignment nodes if
|
%!s(int64=4) %!d(string=hai) anos |
|
pierre
|
c8c6e647be
Avoid keeping optimized out node in callcleanupblock.statements or callinitblock.statements
|
%!s(int64=4) %!d(string=hai) anos |
 Mattias Gaertner
|
b5a8164233
pastojs: specialzie anonymous function
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
07a8e6c1d6
+ some more change information entries added
|
%!s(int64=4) %!d(string=hai) anos |
 michael
|
a8df728548
* Apparently, taking code from freeclx is not OK
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
bc4a22e24d
+ Xtensa: tcpuinlinenode.second_prefetch
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
aa1a68000b
* prefetch instructions read only the operand
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
47557e0ad2
* cpubase.cgsize2subreg should handle all valid sizes, resolves #38557
|
%!s(int64=4) %!d(string=hai) anos |
|
florian
|
45b3d8e0bf
* check for cache entry before even trying to open the file containing the debug info
|
%!s(int64=4) %!d(string=hai) anos |
 lacak
|
89fc5b7f8d
fcl-db: Introduce TArrayField. Only essential parts (added new object properties according to Delphi documentation). Some code taken from FreeCLX project (https://sourceforge.net/projects/freeclx/)
|
%!s(int64=4) %!d(string=hai) anos |
 nickysn
|
8ec71bc810
* fixed memory leak, introduced accidentally in r48998
|
%!s(int64=4) %!d(string=hai) anos |
