 sergei
|
04d8e8a5dc
* On 64-bit targets, handle abs(int64) internally, using the same code as for abs(longint), i.e. without branching. Both generic and x86-specific pass 2 code is already suitable for different operand sizes, only type checking needs removal of excessive conversions to 32 bits.
|
11 gadi atpakaļ |
 svenbarth
|
34394d6925
Fix for Mantis #26180. Accept undefineddef as first parameter type of an Assert if it is used inside a generic.
|
11 gadi atpakaļ |
|
sergei
|
274c2fc82f
* handle_ln_const: if/else branches were swapped, correct values are: ln(0)=-Inf and ln(-#)=NaN.
|
11 gadi atpakaļ |
|
sergei
|
743326dc70
* abs(): when translated to generic fpc_abs_real() helper, typecast result back to original precision, similar to how it is done for sqr and sqrt. Fixes compilation on m68k after r27808 (m86k looks like the only target still needing this generic handling).
|
11 gadi atpakaļ |
|
sergei
|
37625211f5
* Fixed sqrt procedure selection on arm-wince after r27809.
|
11 gadi atpakaļ |
|
sergei
|
22e099d000
* sqrt(real): for targets with emulated floating point, invoke float64_sqrt or float32_sqrt from softfpu code. Testing on ARM CPU shows that float64_sqrt executes about twice faster and offers better accuracy than fpc_sqrt_real from genmath.inc.
|
11 gadi atpakaļ |
|
sergei
|
fc5f45f65c
* sqr(real) and sqrt(real): remove typeconv node inserted by initial call processing (see explanation in comments), allowing these functions to be evaluated using precision of argument. In particular, sqrt(single) and sqrt(double) now emit 'sqrtss' and 'sqrtsd' instructions on x86 targets with -Cfsse3. Non-x86 targets already have the necessary support in code generators.
|
11 gadi atpakaļ |
 florian
|
d88d644925
+ support for FMA intrinsic: if there is no hardware support, the compiler throws an error.
|
11 gadi atpakaļ |
|
svenbarth
|
256c852631
Fix for Mantis #23776.
|
11 gadi atpakaļ |
 Jonas Maebe
|
d452686c39
* moved pbestrealtype from symdef to symcpu
|
11 gadi atpakaļ |
|
Jonas Maebe
|
b57c95043f
+ support overriding tdef/tsym methods with target-specific functionality:
|
11 gadi atpakaļ |
|
Jonas Maebe
|
0edb7b2288
* fixed r27320 for darwin, as its mangled names get prefixed with '_'
|
11 gadi atpakaļ |
|
svenbarth
|
a08e8ff982
Fix for Mantis #25929
|
11 gadi atpakaļ |
|
florian
|
ab3f5744e0
* skip dec/inc type checking in unspecialized generic methods, resolves #25603
|
11 gadi atpakaļ |
|
florian
|
e841027a48
* fixes handling of typed files in iso mode
|
11 gadi atpakaļ |
 nickysn
|
e1332304ef
* when converting Hi() to a shr node, convert the right shr const parameter to
|
11 gadi atpakaļ |
|
Jonas Maebe
|
d2b8275b99
+ support for arbitrary encodings in readstr/writestr
|
11 gadi atpakaļ |
|
sergei
|
8ecad469a6
* Reverted r26256 since it produced wrong results if argument of inc/dec is actually unaligned. Added description of the issue and slightly increased node complexity at which address is cached, so operations on record/class fields do not take address.
|
11 gadi atpakaļ |
|
sergei
|
b376f56bf3
* Use typed pointer to store address of x when processing inc(x)/dec(x) with range checking enabled. This preserves alignment of x and prevents generating unaligned code on alignment-sensitive targets.
|
11 gadi atpakaļ |
|
florian
|
4d5119bf1c
* fixes several issues which cause warnings by the dfa code when using it to detect uninitialized variables
|
11 gadi atpakaļ |
|
florian
|
412dc6586f
* the writestr helper initializes the target string, so mark it as being written
|
11 gadi atpakaļ |
|
sergei
|
9ad98a2c4d
* Handle "SarInt64" always as internal procedure, introduced virtual method tinlinenode.first_sar that by default converts it into call to "fpc_sarint64" unless target CPU is 64-bit. This provides a point to insert target-specific optimizations.
|
11 gadi atpakaļ |
|
sergei
|
dac8cbcefc
* Handle assigned(x) expressions entirely in first pass by converting them to "x<>nil", indicate that procedure variables should not be called by setting nf_load_procvar flag (which is already used with typeconvnode for exactly the same purpose). This allows to generate better code when assigned(x) is used on right side of assigments. Mantis #24572.
|
12 gadi atpakaļ |
|
florian
|
a2a405581c
* don't call the ansistring val variant for a zero based array if not necessary to avoid ansistring usage in embedded applications
|
12 gadi atpakaļ |
 masta
|
b81f746e05
Inline Length(DynamicArray)
|
12 gadi atpakaļ |
|
Jonas Maebe
|
fb8fb29574
* record/use the code page of string constants typecasted to an ansistring
|
12 gadi atpakaļ |
|
sergei
|
87cfd86172
* Define SUPPORT_GET_FRAME for targets having "get_frame" as internal symbol (x86,arm and jvm, currently), removes need to enumerate these targets in every related conditional compilation directive and simplifies configuring this feature for other targets.
|
12 gadi atpakaļ |
|
sergei
|
cc8772cf92
* sizeof(x), typeof(x): if "x" is typenode, request GOT for PIC because a global VMT label will be loaded.
|
12 gadi atpakaļ |
|
florian
|
0968d095ed
* convert function like used new into a appropriate nodes only in pass_1 so proper error checking can be carried out, resolves #24495
|
12 gadi atpakaļ |
|
nickysn
|
a86cb5c600
+ far pointer fixes in the internal compiler handling of ptr(const,const)
|
12 gadi atpakaļ |
