Eli Friedman

DAGCombine carry folding makes code worse in given testcase

1

Testcase derived from boringssl: [boringssl-bcm-testcase.txt](https://github.com/llvm/llvm-project/files/9339270/boringssl-bcm-testcase.txt). The code is sort of complicated, but I'm not sure how to simplify it. I've found that if I revert D64190 (e9aed963ce36) and D57302 (bddb8c359739),...

riscv popcount lowering shouldn't call __muldi3/__mulsi3

6

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...

riscv 64-bit popcount uses inefficient constant materialization

4

Consider: ``` int a(unsigned long long x) { return __builtin_popcountll(x); } ``` Targeting rv64, this generates: ``` a: srli a1, a0, 1 lui a2, 349525 addiw a2, a2, 1365 slli...