A little C compiler https://git.lemon.rip/antcc/atom?h=trunk 2026-03-24T10:55:13+00:00 2026-03-24T10:55:13+00:00 lemon lsof@mailbox.org 2026-03-24T10:55:13+00:00 urn:sha1:5c209b4df903c5f71cdc8320adbb34f1499f9232 Do it in 2 passes: 1. gather variables, 2. step thru RPO transforming loads/stores to each variable. More elegant solution to f9e3a52eaa. More efficient since all variable usages are visited in CFG order, which matters due to block sealing. 2026-03-22T10:21:01+00:00 lemon lsof@mailbox.org 2026-03-22T10:20:14+00:00 urn:sha1:adf244459ef6946c2050764906357a4ec4010b3f Move the alloca to the beginning of the block and do a 2nd pass on those odd cases. 2026-03-22T09:49:30+00:00 lemon lsof@mailbox.org 2026-03-22T09:45:22+00:00 urn:sha1:f9e3a52eaa6ae91388aa247182da6df3cc8d5a05 mem2reg was assuming alloca's appeared before all their uses. When this didn't happen it broke the logic re. sealed blocks. Normally this doesn't happen in most natural code. Even with gotos, RPO would assure this in most cases, the exception I found is a goto that jumps into a loop past a variable declaration that is earlier in the loop. Then even RPO would keep the alloca itself past the first uses, breaking things. 2026-03-22T09:49:26+00:00 lemon lsof@mailbox.org 2026-03-21T21:20:34+00:00 urn:sha1:79874c83bf76a5b3efd3d558933b90d9b53b829e The motivation is for aarch64 msub/madd instrs, for isel to produce. But it should be useful for other things too. 2026-03-18T10:33:41+00:00 lemon lsof@mailbox.org 2026-03-18T10:33:41+00:00 urn:sha1:1d9e19fb3bb941cdc28e9d4c3063d3e213fd8312 Looks nicer 2026-03-17T17:30:37+00:00 lemon lsof@mailbox.org 2026-03-17T17:30:10+00:00 urn:sha1:04930d578e65d560253d0c24af43e0ecd06117c8 2026-03-17T12:22:00+00:00 lemon lsof@mailbox.org 2026-03-17T12:22:00+00:00 urn:sha1:a8d6f8bf30c07edb775e56889f568ca20240bedf