Laytan Laats
f80e73e036
few llvm 20 changes
2025-03-19 21:22:55 +01:00
gingerBill
4c3c3a0631
Add new LLVM 20 passes
2025-03-18 16:21:08 +00:00
gingerBill
d209af5094
Update to LLVM 20.1.0
2025-03-18 15:39:18 +00:00
gingerBill
d15be94a0f
Remove debug print
2025-03-11 15:50:46 +00:00
gingerBill
6e1c21411d
Fix object name generation
2025-03-11 15:50:31 +00:00
gingerBill
ee84487106
Merge branch 'master' of https://github.com/odin-lang/Odin
2025-03-01 15:19:51 +00:00
gingerBill
211680569f
Fix for -use-single-module
2025-03-01 15:19:46 +00:00
Laytan Laats
66540b75f5
fix hidden linkage applied to foreign symbols
2025-03-01 16:06:32 +01:00
gingerBill
fd6d7d412d
Use more predictable object name
2025-02-25 14:55:33 +00:00
gingerBill
83f1c879ec
Canonicalize generated procedures (hasher/equal/map)
2025-02-24 15:18:27 +00:00
gingerBill
2c86fadee1
More name canonicalization for globals
2025-02-24 14:28:24 +00:00
gingerBill
f0f395c510
Fix canonicalization
2025-02-24 14:24:06 +00:00
gingerBill
46a8991d0f
Canonicalize generated procedures
2025-02-24 14:19:02 +00:00
gingerBill
ebda946d61
Move temporary array out of CheckerInfo
2025-02-22 18:12:43 +00:00
gingerBill
5489a88983
Change typeid definition to be based around the canonical type hash
...
`typeid` used to be a fancy index with extra metadata stored on it. Now it is direct hash of the type.
This is safe to do in practice since any possible collisions are checked at compile time AND the chances of having a 1% collision are around 1 in 600K (see the Birthday Paradox).
Therefore accessing a `^Type_Info` is now a hash table lookup with linear probing. The table is twice the size than necessary so prevent too much probing due to an overly dense hash table.
2025-02-20 14:10:45 +00:00
gingerBill
4a29d9bb84
Simplify type info table construction
2025-02-17 16:29:42 +00:00
gingerBill
9b26bb2e6a
Begin work on hash types
2025-02-17 13:10:38 +00:00
Laytan Laats
1053ec3051
make corrected linkage with -use-separate-modules apply hidden visibility
...
Fixes #4798
The DLL was using the type info of the host/exe, causing crashes.
This PR tries fixing by applying hidden visibility to these corrected
symbols which makes sure that the DLL can't see the type table of the host/exe.
2025-02-06 20:45:58 +01:00
Laytan Laats
9dc17f4c47
optimize fix
2025-01-24 19:33:57 +01:00
Laytan Laats
b2aaf90f88
fix separate modules with objc code
2025-01-24 19:23:49 +01:00
gingerBill
bca08d3b85
Make -no-dynamic-literals the default now
2025-01-05 13:33:06 +00:00
gingerBill
a25a9e6ebe
Check for LLVM_VERSION_MAJOR >= 18
2024-09-09 14:47:44 +01:00
gingerBill
0dddd2b97e
Add internal flag
...
`-internal-fast-isel`
2024-09-09 14:39:53 +01:00
Laytan
ca6ef95b03
add support for linux_riscv64 and freestanding_riscv64
2024-08-20 14:06:40 +02:00
Laytan Laats
4d1d754cae
fix specific_union_variant in map_keyed_by_union not converting to union type
2024-07-29 02:32:13 +02:00
gingerBill
bc5b41938e
Fix #3964
2024-07-23 02:40:51 +01:00
gingerBill
7d643bcae3
Make linkage weak in certain places
2024-07-15 12:30:32 +01:00
gingerBill
549311fac9
Fix global variables being "missing" with -use-separate-modules
2024-07-15 12:21:42 +01:00
gingerBill
c5decd3eae
Fix possible race and correct linkage _after_ generation
2024-07-15 11:49:07 +01:00
gingerBill
a867378418
Default to -o:minimal again
2024-07-14 15:55:26 +01:00
gingerBill
34c6868e78
Merge pull request #3895 from laytan/fix-optimization-mode-attribute
...
remove misleading @(optimization_mode) values and make "none" inhibit optimizations
2024-07-10 13:22:57 +01:00
gingerBill
87ac68fcf2
Add -internal-cached
2024-07-08 23:39:14 +01:00
Laytan Laats
1a20b78633
remove misleading @(optimization_mode) values and make "none" inhibit optimizations
2024-07-08 21:06:57 +02:00
gingerBill
2c9ef6907a
Clean up timings messages showing used module count
2024-07-08 15:30:33 +01:00
gingerBill
2a219fa830
Correct -use-separate-module behaviour
2024-07-08 15:13:40 +01:00
gingerBill
8491e2491c
Experiment with different uses of -use-separate-modules
2024-07-08 14:48:59 +01:00
gingerBill
7dd4cccce7
Use a temporary directory for -use-separate-modules
...
Windows only currently
2024-07-08 14:21:07 +01:00
gingerBill
0f664893dd
Add sort for global types and procedures
2024-07-08 13:59:25 +01:00
Laytan Laats
b79d7e6917
disallow non-global foreign import of variables on wasm
2024-06-28 01:34:09 +02:00
Jeroen van Rijn
ed7d34beb0
Avoid sporadic wasm crash
2024-06-28 00:39:51 +02:00
gingerBill
1dc90103bd
Make verification ignorable with a define flag
2024-06-11 12:30:24 +01:00
gingerBill
0b02c67cdf
Minor clean up for backend
2024-06-11 12:19:52 +01:00
gingerBill
9b0e87544a
Unify LLVMVerifyFunction invocations into on place
2024-06-11 12:07:22 +01:00
gingerBill
9ef43fc782
Add @(rodata)
2024-06-06 15:16:34 +01:00
Laytan Laats
e3e04ffa22
llvm-18: enable sroa and static map calls
2024-05-07 16:52:46 +02:00
Laytan Laats
bb58926b7a
llvm 18: general unix and darwin specifics
2024-05-07 16:52:46 +02:00
Laytan Laats
25f1d0906d
compiler: improve target features support
2024-05-02 00:59:52 +02:00
gingerBill
2416380f34
Enforce as global constant
2024-04-18 12:56:18 +01:00
Laytan Laats
9647cb74ad
debug info fixes/refactor
...
This fixes (on my end) #3340 , #3117 , #2945 , #2922 , and #2762
A general refactor of debug info generation in order to fix issues and
increase stability.
What I believe is the root cause of a bunch of issues is that we use the
temporary metadata/forward declarations too much (/ hold onto them for
too long). It seems to cause problems with the reference counting inside
LLVM.
This PR reduces the use of these forward declarations to a minimum, it
creates it, fills in the fields, and resolves it, instead of waiting
until the end of generating code.
Some smaller issues I came across have also been solved.
2024-04-01 16:28:44 +02:00
gingerBill
433109ff52
Replace gb_exit(1) with exit_with_errors() where appropriate
2024-03-19 16:29:45 +00:00
