antcc/ir/optmem.c, branch trunkA little C compilerhttps://git.lemon.rip/antcc/atom?h=trunk2025-12-19T11:59:24+00:00ir: move some filluses() to ir.c, rename optmem.c -> mem2reg.c2025-12-19T11:59:24+00:00lemonlsof@mailbox.org2025-12-19T11:59:24+00:00urn:sha1:575b24a9f023f7950eefea6d85431281f04cc1dcbitset: better implementation of bsiter() and stuff2025-12-16T17:46:08+00:00lemonlsof@mailbox.org2025-12-16T17:33:59+00:00urn:sha1:b3762f81b1a276f06bace301d56c9e8f6538058d
Also changed the type to size_t for portability
mem2reg: fix obvious inefficiency2025-12-16T07:16:21+00:00lemonlsof@mailbox.org2025-12-16T06:17:09+00:00urn:sha1:a00214e6e569e87f82bb7bcce9df7b5365884ba1
deltrivialphis() was iterating over every variably instead of just
looking at the variable being examined. And I'd been wondering why
mem2reg was such a bottleneck for a testcase like sqlite3 amalgamation..
it's easy to miss the forest for the trees.
rename arraylength macro -> countof2025-12-11T19:43:24+00:00lemonlsof@mailbox.org2025-12-11T19:43:24+00:00urn:sha1:88652eeb10cd9381aafb2d55e9474bb0799630b1ir: barebones IR passes checked contracts2025-11-21T16:35:55+00:00lemonlsof@mailbox.org2025-11-21T16:07:28+00:00urn:sha1:87f9753fb776a1fa6e59baef759e4687fb9a1ac7change op names to match 285063eba442025-11-21T15:49:53+00:00lemonlsof@mailbox.org2025-11-21T15:49:53+00:00urn:sha1:ec4cfe9db9afc1d1c633a922174f5bb0685b0c32rename IR classes to reflect bitsize2025-11-21T10:03:23+00:00lemonlsof@mailbox.org2025-11-21T10:03:23+00:00urn:sha1:285063eba442e2a8ac29fd42e0d17d996bcc5d00mem2reg: implement marker algorithm from Braun et al2025-11-21T09:32:13+00:00lemonlsof@mailbox.org2025-11-21T09:32:13+00:00urn:sha1:b930d1adc1847547e17f307ea01cea8c587cadc9mem2reg: store pending phis implicitly2025-11-21T08:59:48+00:00lemonlsof@mailbox.org2025-11-21T08:59:48+00:00urn:sha1:9062212c63d0f80e00dc24aecd6c8f034b996217mem2reg: fix edgecase..2025-11-19T10:22:58+00:00lemonlsof@mailbox.org2025-11-19T10:22:58+00:00urn:sha1:34ff9d31d2ead5bdc2a22518c4496b8070fd679b