[llvm] [SDag][ARM][RISCV] Allow lowering CTPOP into a libcall (PR #101786)

[Nathan Chancellor via llvm-commits]

nathanchance wrote:

> > I could try to push for something like this in the kernel:
> 
> I don't mind sharing the headache 😅. And, who knows, maybe this patch will be welcomed?
> 
> > I do understand the kernel is being special here so I can pursue the addition of `__popcount{d,s}i2()` in the kernel first and come back if that is not successful.
> 
> Will be much appreciated, thank you! I'll continue trying to fix the issue on the compiler side.

Heh, I [sent a message to Linus yesterday](https://lore.kernel.org/20250425003342.GA795313@ax162/) and [the reaction](https://lore.kernel.org/CAHk-=whfT3A8K2Z+WbieGG5Hhc9QAT5s3qsbB19O0Roj2G5tfA@mail.gmail.com/) was generally what I expected.

His suggestion was effectively

```diff
diff --git a/lib/hweight.c b/lib/hweight.c
index c94586b62551..00b2ff970203 100644
--- a/lib/hweight.c
+++ b/lib/hweight.c
@@ -66,3 +66,8 @@ unsigned long __sw_hweight64(__u64 w)
 #endif
 }
 EXPORT_SYMBOL(__sw_hweight64);
+
+int __popcountsi2(unsigned int w) __alias(__sw_hweight32);
+EXPORT_SYMBOL(__popcountsi2);
+int __popcountdi2(__u64 w) __alias(__sw_hweight64);
+EXPORT_SYMBOL(__popcountdi2);
```

which does also resolve the error but I realized that neither of these suggestions I have proposed will work in their current form because (at least for the RISC-V configuration that I chcked) `__sw_hweight64` gets transformed into a call to `__popcountdi2` so trying to reuse either `__arch_hweight64` or `__sw_hweight64` to define `__popcountdi2` will just result in an infinite loop…

```
00000000000000be <__sw_hweight64>:
      be: 1141          addi    sp, sp, -0x10
      c0: e406          sd  ra, 0x8(sp)
      c2: e022          sd  s0, 0x0(sp)
      c4: 0800          addi    s0, sp, 0x10
      c6: 00000097      auipc   ra, 0x0
        00000000000000c6:  R_RISCV_CALL_PLT __popcountdi2
        00000000000000c6:  R_RISCV_RELAX    *ABS*
      ca: 000080e7      jalr    ra <__sw_hweight64+0x8>
      ce: 60a2          ld  ra, 0x8(sp)
      d0: 6402          ld  s0, 0x0(sp)
      d2: 0141          addi    sp, sp, 0x10
      d4: 8082          ret
```

https://github.com/llvm/llvm-project/pull/101786