tag:blogger.com,1999:blog-5888658295182480819.post107793784397659710..comments2022-04-05T09:43:19.308-03:00Comments on Alaska Ataca a Kamtchatka: Assessing AbstractionsMatías Giovanninihttp://www.blogger.com/profile/17772004856076119446noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-5888658295182480819.post-37536095417080985552012-07-03T01:54:51.285-03:002012-07-03T01:54:51.285-03:00It clearly is not an issue with the pattern matchi...It clearly is not an issue with the pattern matching compiler; the tuple is allocated and built just to be destructured and used well after that. Whether this optimization should be performed on the RTL or as a peephole pass over the generated instructions, I don't know. I don't know what the "real-word" (them fighting words, Dijkstra was famously wary of those) impact of this Matías Giovanninihttps://www.blogger.com/profile/17772004856076119446noreply@blogger.comtag:blogger.com,1999:blog-5888658295182480819.post-64875375188084372342012-07-03T01:42:34.359-03:002012-07-03T01:42:34.359-03:00Pattern matching is compiled away (and, in the pro...Pattern matching is compiled away (and, in the process, optimizes things such as `match 2,3 with (x,y) -> ...`) in an early stage of OCaml's compilation, that is common to both bytecode and native compilers. Inlining is performed later, in the native compiler only.<br /><br />This means that the pattern matching compiler won't get access to additional optimizations that will be made gaschenoreply@blogger.com