- 28 Feb, 2016 7 commits
-
-
Christoph Mallon authored
-
Christoph Mallon authored
-
Christoph Mallon authored
-
Christoph Mallon authored
-
Christoph Mallon authored
-
Christoph Mallon authored
-
Christoph Mallon authored
-
- 27 Feb, 2016 2 commits
-
-
yb9976 authored
-
- 26 Feb, 2016 8 commits
-
-
yb9976 authored
This fixes backend/subregister.c.
-
yb9976 authored
-
yb9976 authored
-
Christoph Mallon authored
-
yb9976 authored
Except for global CSE, comparing nodes of different block always results in 'not equal', so we should consider them for hashing beforehand.
-
yb9976 authored
-
Christoph Mallon authored
Simply directly check the magnitude of the tarval instead of converting it to a long. * On 64 bit hosts 32 bit operations could not match immediates with bit 31 set. E.g. v32 & 0xFFFFFFF0. * On 32 bit hosts 64 bit operations with 32 bit non-negative constants were erroneously matched and caused silent miscompilation. E.g. v64 & 0xFFFFFFF0. * On 32 bit hosts 64 bit operations with constants, which would fit into 32 bits, but have higher bits set, were not matched, because get_tarval_long() failed. E.g. v64 & 0xFFFFFFFFFFFFFFF0.
-
Christoph Mallon authored
-
- 23 Feb, 2016 1 commit
-
-
yb9976 authored
This fixes opt/boolopt4.c.
-
- 17 Feb, 2016 1 commit
-
-
Christoph Mallon authored
amd64: Change the parameter 'can_match_ip_relative' of match_immediate_32() to 'must_match_ip_relative'. Either a user must have a ip-relative address or it must not.
-
- 16 Feb, 2016 2 commits
-
-
yb9976 authored
This fixes opt/fehler318.c.
-
Andreas Seltenreich authored
-
- 15 Feb, 2016 9 commits
-
-
Andreas Seltenreich authored
-
yb9976 authored
The computation of dominance might invalidate out edges. This fixes opt/fehler319.c.
-
yb9976 authored
Dominance is computed explicitly later.
-
Christoph Mallon authored
-
Matthias Braun authored
-
Andreas Seltenreich authored
Otherwise repeatedly invoking be_main() leads to a double free.
-
Matthias Braun authored
They previously used be_stack_layout but that structure will be remove soon.
-
Matthias Braun authored
-
Matthias Braun authored
-
- 14 Feb, 2016 8 commits
-
-
Matthias Braun authored
We can just as well make up names on the fly while printing the entities for debugging.
-
Matthias Braun authored
-
Matthias Braun authored
The soon to be introduced spillslot entities can be at negative offset, so the previously used -1 as convention for invalid offsets is not a good idea anymore.
-
Matthias Braun authored
-
Matthias Braun authored
-
Matthias Braun authored
-
Matthias Braun authored
-
Matthias Braun authored
-
- 13 Feb, 2016 2 commits
-
-
Matthias Braun authored
-
Matthias Braun authored
-