 florian
|
e0a1bc6675
+ optimize (a and b) or (c and not(b)) into c xor ((c xor a) and b)
|
4 years ago |
 Jonas Maebe
|
8ba4c4bcf0
* factored out the creation of loadnodes for special variables/parameters
|
4 years ago |
|
Jonas Maebe
|
5d51a788f7
- revert r47208
|
4 years ago |
|
florian
|
57dc416d96
+ replace appropriate if statements by min/max intrinsics
|
4 years ago |
 pierre
|
bf5b85e5e5
Also look for function name of 'result' is not found in load_result_node
|
4 years ago |
|
florian
|
6fc930a0fc
* better complexity calculation of tvecnodes
|
4 years ago |
|
Jonas Maebe
|
a88288bab3
- reverted r46457, accidentally committed (breaks -Oooptcse)
|
5 years ago |
|
Jonas Maebe
|
f26735fc77
* execute the tempinitcode whent the tempcreate node gets processed rather
|
5 years ago |
|
florian
|
ff34fb1645
* better complexity handling of mul, slash, div, mod
|
5 years ago |
|
florian
|
5da913de0e
* float operations can have an exception as a side effect, resolves #37398
|
5 years ago |
|
florian
|
8738b4b47f
+ might_have_sideeffects checks also if the tempinitcode of a tempref node might have side effects
|
5 years ago |
|
florian
|
b3817f6d92
* slash and divn might have an exception as side effect
|
5 years ago |
|
florian
|
b80d051249
* make fparainit and fparacopyback public
|
5 years ago |
|
florian
|
e3870079b2
* remove tempref mess for simple inlined functions
|
5 years ago |
|
florian
|
0f06a7205b
* do not create full copies of objects when object checking is on, resolves #36863
|
5 years ago |
|
florian
|
424c8a0ac0
* simple record accesses have not side effects
|
5 years ago |
|
florian
|
9819df6fe1
* calculate complexity of an rtti node correctly
|
5 years ago |
|
florian
|
5a87855b06
+ might_have_sideeffects takes care of vo_volatile now (it is not exposed yet)
|
5 years ago |
|
florian
|
d479069f8c
* might_have_sideeffects takes care of absolute
|
5 years ago |
|
Jonas Maebe
|
e1d9be897a
* fixed (harmless) wrong typecast
|
5 years ago |
|
florian
|
f6c16323fa
* unified loadnf_load_self_pointer into loadnf_load_addr
|
5 years ago |
|
pierre
|
7eb0be6127
Fix invalid typecast revealed by -CriotR by changing order of terms in boolean expression
|
5 years ago |
|
florian
|
851accbb4b
* cse adjusted
|
5 years ago |
|
Jonas Maebe
|
d23c7980aa
* properly perform (or not) range/overflow checking when performing constant
|
5 years ago |
|
florian
|
b7f58dd8c8
* load nodes for variables in registers do not generated instructions, so their complexity is zero
|
5 years ago |
|
florian
|
31e5b2d9a4
* more precise estimation of subscript node complexity
|
5 years ago |
|
florian
|
e80eee5da5
* generate better code for length(<dyn. type>)=0, resolves #36060
|
5 years ago |
|
Jonas Maebe
|
14a7429e19
* factored out inlinenode check whether it may have a side effect
|
6 years ago |
|
Jonas Maebe
|
1b6425176b
* synchronised with trunk till r42049
|
6 years ago |
|
Jonas Maebe
|
281b3ad276
* fix case completeness and unreachable code warnings in compiler that would
|
6 years ago |
