[compiler-rt] Make sure __clzdi2 doesn't call itself recursively on sparc64 #136737
Conversation
…parc64 On 64-bit platforms, libgcc doesn't ship with __clzsi2, so __builtin_clz gets lowered to __clzdi2. A check already exists for GCC, but as of commit 8210ca0 clang also lowers __builtin_clz to __clzdi2 on sparc64. Update the check so that building __clzdi2 with clang/sparc64 also works.
koachan
requested review from
compnerd,
wzssyqa,
rorth,
topperc and
s-barannikov
compiler-rt/lib/builtins/clzdi2.c
Outdated
| (defined(__riscv) && __SIZEOF_POINTER__ >= 8)) | ||
| // On 64-bit architectures with neither a native clz instruction nor a native | ||
| // ctz instruction, gcc resolves __builtin_clz to __clzdi2 rather than | ||
| // ctz instruction, resolves __builtin_clz resolves to __clzdi2 rather than |
There was a problem hiding this comment.
| // ctz instruction, resolves __builtin_clz resolves to __clzdi2 rather than | |
| // ctz instruction, `__builtin_clz` resolves to `__clzdi2` rather than |
compiler-rt/lib/builtins/clzdi2.c
Outdated
| // __clzsi2, leading to infinite recursion. | ||
| // This is because on those platforms, libgcc doesn't ship with __clzsi2. |
There was a problem hiding this comment.
| // __clzsi2, leading to infinite recursion. | |
| // This is because on those platforms, libgcc doesn't ship with __clzsi2. | |
| // `__clzsi2` as libgcc does not ship with `__clzsi2`, leading to infinite recursion. |
| // ctz instruction, `__builtin_clz` resolves to `__clzdi2` rather than | ||
| // __clzsi2 as libgcc does not ship with `__clzsi2`, leading to infinite | ||
| // recursion. |
There was a problem hiding this comment.
That doesn't seem to be true on RISC-V/Mips (clang).
https://godbolt.org/z/z8Yd9dr5K
There was a problem hiding this comment.
Hmm, looking at the respective backends, seems like they are Expanding CTTZ instead of lowering it to a LibCall in absence of native instructions.
There was a problem hiding this comment.
I could rewrite the check so that the clang check is omitted only if compiling on SPARC, but wouldn't that be too complicated?
There was a problem hiding this comment.
I'm fine with any solution as long as it doesn't break/pessimize other targets.
I think RISC-V will eventually require the same change, not sure about Mips.
|
Note that |
Noted, will do another PR for it after this lands. |
On 64-bit platforms, libgcc doesn't ship with __clzsi2, so __builtin_clz gets lowered to __clzdi2. A check already exists for GCC, but as of commit 8210ca0 clang also lowers __builtin_clz to __clzdi2 on sparc64.
Update the check so that building __clzdi2 with clang/sparc64 also works.