llvm-project
llvm-project copied to clipboard
llvm
riscv popcount lowering shouldn't call __muldi3/__mulsi3
Consider:
int a(unsigned x) { return __builtin_popcount(x); }
With -march=rv32i, this generates a call to __mulsi3. This is likely to be slow; if we don't have "m", we should use shifts instead.
@llvm/issue-subscribers-backend-risc-v
Author: Eli Friedman (efriedma-quic)
int a(unsigned x) { return __builtin_popcount(x); }
With -march=rv32i, this generates a call to __mulsi3. This is likely to be slow; if we don't have "m", we should use shifts instead.
We are using multiply here because this is the default expanding of ISD::CTPOP.
https://github.com/llvm/llvm-project/blob/72c729f354d71697a1402720c90b57ff521b6739/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp#L8653-L8718
We may need custom lowering of ISD::CTPOP when there is no m and Zbb.
We are using multiply here because this is the default expanding of
ISD::CTPOP.https://github.com/llvm/llvm-project/blob/72c729f354d71697a1402720c90b57ff521b6739/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp#L8653-L8718
We may need custom lowering of
ISD::CTPOPwhen there is nomandZbb.
Can we check isOperationLegalOrCustom in the default handler?
We are using multiply here because this is the default expanding of
ISD::CTPOP. https://github.com/llvm/llvm-project/blob/72c729f354d71697a1402720c90b57ff521b6739/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp#L8653-L8718We may need custom lowering of
ISD::CTPOPwhen there is nomandZbb.Can we check isOperationLegalOrCustom in the default handler?
Yeah, I think so! But I don't get how to use shifts here...
x *= 0x01010101; is just x += x << 8; x += x << 16;. Maybe you can save a shift somewhere.
I think this should be generic optimization opportunity for converting multiplication with specific constants into shift & add. It would not be limited to the popcount use case.
In #79823, I mentioned that (x * 0x08040201) can be converted to { tmp = x + (x << 9); tmp += (tmp << 18); } and that GCC does perform such optimization.
