Add initial support for LoongArch64 #4500
Conversation
| @@ -765,6 +765,14 @@ void registerPredefinedTargetVersions() { | |||
| case llvm::Triple::wasm64: | |||
| VersionCondition::addPredefinedGlobalIdent("WebAssembly"); | |||
| break; | |||
| case llvm::Triple::loongarch32: | |||
There was a problem hiding this comment.
you have the 32-bit version here, but in most other locations you only support 64bit, e.g. in TargetABI::getTarget() and getTargetCPU
There was a problem hiding this comment.
you have the 32-bit version here, but in most other locations you only support 64bit, e.g. in
TargetABI::getTarget()andgetTargetCPU
This is because 32-bit LoongArch is still tentative, and there is no real hardware for this specification.
I have added a tentative implementation for LoongArch32 now, but since the 32-bit ABI specification is not stable, I will not include an implementation for it in TargetABI::getTarget() .
liushuyu
force-pushed
the
la64
branch
2 times, most recently
from
7a53134 to
a47e2e4
Compare
|
New changes since the last push:
|
|
All CI tests have successfully passed. |
| IndirectByvalRewrite indirectByvalRewrite{}; | ||
|
|
||
| public: | ||
| auto returnInArg(TypeFunction *tf, bool) -> bool override { |
There was a problem hiding this comment.
nitpick: is there any benefit to this auto .... -> bool over bool ... ? We don't really use this syntax in LDC's codebase anywhere else.
There was a problem hiding this comment.
nitpick: is there any benefit to this
auto .... -> booloverbool ...? We don't really use this syntax in LDC's codebase anywhere else.
When I run clang-tidy on my changes, it tells me to use this expression instead of bool func_name(args...).
I do notice you have a .clang-tidy configuration in the project root. Perhaps that configuration makes clang-tidy to add this specific lint? (I am using Clang 16 on my system).
|
The PR looks good to me. Tests are missing, perhaps good to have at least one test for the triple and ABI ? |
Will do in a bit. |
Done. Let's see if CI will pick it up. |
|
CI passed without issue and the new test was picked up by CIs that running with LLVM 16 |
| case "LoongArch32": | ||
| case "LoongArch64": | ||
| case "LoongArch_HardFloat": | ||
| case "LoongArch_SoftFloat": |
There was a problem hiding this comment.
[We'd get this with the next frontend merge automatically, after your upstream PR has been merged. But is okay (we usually guard LDC-specific frontends mods in version (IN_LLVM) for future merge conflicts).]
|
Hi, DMD upstream has merged the LoongArch support for the D runtime. Do you want me to do a merge, or do you want to do it yourself? |
|
not adding it to release notes, because runtime support is lacking and will only come with dlang 2.106 |
|
Thank u ;) |
This pull request ports the LDC compiler to LoongArch64 CPU architecture.
LoongArch64 is a new CPU architecture and has been supported by LLVM/Clang since LLVM 16.
You can find more information about this architecture here: https://loongson.github.io/LoongArch-Documentation/README-EN.html.
The port for D runtime is also in progress at dlang/dmd#15628.