tree e7cb09036874dd2aa6e152cf9f76b72bdc1de7ad
parent c4788bff8266497aafd526f887722982127a2e11
author Philip Reames <preames@rivosinc.com> 1759352428 -0700
committer GitHub <noreply@github.com> 1759352428 -0700
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJo3ZZsCRC1aQ7uu5UhlAAAwV8QACMqByVduTRAxgn3cmJOwFA0
 hKwDL50z7JQ+1KpVhZxjKuXGL55jaNjcTDDBKQv6P6PhkSmnA5p3cbw8s7D/Z+Zj
 GDRLcLrrSzIvXMSYpPso4G50O6TT8l/eQ5wBQ6n77CC5kB+C2364KfeW7VWhcew8
 jgMjVMmKVQWLtDcYyjxQKtVbUJbhyd17+ssSI+SLVt1yWR5z0nfFbMjOQ2psvWLF
 tUhgpKyHB9XuYMPWOnBW5vF6ljxmcir0zuaCoXl0001kDRc1mMiZJDNykobLj667
 yGYatcJOXap54oA6boQMq8b2cI3ZckuhgNkDdUIOL0+MbSvDylsRKHR3Tzs6TGYB
 94DEM3Sf3aLlKG+M+Z/VBVl/5XYoSmiHB+wo8Dxa6992rYedQwgYbcUiQQB1sml2
 CF0gMrnY6P4DSORUFvrYFMcdJR/v2mjoKq6SHaeRDKkaZcJxSCIIk7rgXfmVeQjr
 K5GAFsl7S3vnRdVcLuABor/uVu6VwmQVkg76T0Ys4EoUei0emOrxt1gQtyn/WeoJ
 6BWbyD5PKjByrv/0Q0sHVZ4z2o5PEklptEK+hw/VTud9gjapYqbM7nOr3/ptXhc+
 UQw7rvFy8tn3pFMdRGX1mTqJWBlG8Mqrq6grorcZHbqUBSmaua0qqFuEhvLSNTvU
 dMK9J+bfJUQCdA8DDfYZ
 =NOR8
 -----END PGP SIGNATURE-----
 

[RISCV] Allow non-canonicalized splats in isProfitableToSinkOperands (#161586)

This isn't an optimization change - IR transforms should have remove the
operands and replaced them with poison. However, I noticed the
non-canonical splat structure in a couple of llvm-reduce outputs. This
results in us creating extremely atypical IR which is quite misleading
about the true cause of what's going on. (Because the non-canonical
splat doesn't get sunk, we then prone whatever was actually holding it
outside the loop in the original example, eliminating insight as to the
true cause of whatever issue we're debugging.)