[Chicken-hackers] [PATCH] Revert half of "Add some optimizer simplification rules"
Hi, This reverts the rewrite that unearthed #1648. >From 896601d3fd9fd9e9e1a0bac407dccbf0e4bed1e8 Mon Sep 17 00:00:00 2001 From: megane Date: Mon, 16 Sep 2019 12:04:49 +0300 Subject: [PATCH] Revert half of "Add some optimizer simplification rules" This causes uri-generic to fail to compile in similar way #1648 does. --- optimizer.scm | 32 1 file changed, 32 deletions(-) diff --git a/optimizer.scm b/optimizer.scm index fc2d3165..bd163710 100644 --- a/optimizer.scm +++ b/optimizer.scm @@ -830,38 +830,6 @@ args1) nargs) #t)) -(else (loop (cdr args) -(cons (car args) nargs) -ok))) - - ;; (let (( (##core#inline ...))) - ;; ( ... ...)) - ;; -> ( ... (##core#inline ...) ...) - ;; ...)) - ;; - is used only once. - `((let (var) (##core#inline (op) . args1) - (##core#call p . args2)) -(var op args1 p args2) -,(lambda (db may-rewrite var op args1 p args2) - (and may-rewrite ; give other optimizations a chance first -(= 1 (length (db-get-list db var 'references))) -(let loop ((args args2) (nargs '()) (ok #f)) - (cond ((null? args) - (and ok - (make-node - '##core#call p - (reverse nargs -((and (eq? '##core#variable - (node-class (car args))) - (eq? var - (car (node-parameters (car args) - (loop (cdr args) - (cons (make-node - '##core#inline - (list op) - args1) - nargs) - #t)) (else (loop (cdr args) (cons (car args) nargs) ok -- 2.17.1 ___ Chicken-hackers mailing list Chicken-hackers@nongnu.org https://lists.nongnu.org/mailman/listinfo/chicken-hackers
Re: [Chicken-hackers] [PATCH] Some simplification rules for nested ##core#inline forms
Peter Bex writes: > On Wed, Sep 04, 2019 at 11:59:31AM +0200, felix.winkelm...@bevuta.com wrote: >> The attached patch adds two optimization rules for certain uses of >> ##core#inline. It basically rewrites >> >> (let (( (##core#inline ...))) >> ( ... ...)) >> >> into >> >> ( ... (##core##inline ...) ...) > > It looks like there's a problem caused by this: > > https://salmonella-freebsd-x86-64.call-cc.org/master/clang/freebsd/x86-64/2019/09/16/salmonella-report/install/uri-generic.html > > uri-generic uses matchable, which presumably generates a lot of > code which looks like it should fit the pattern. > Here's a simplified version that does the OOM here: (module uri-generic (uri-relative-from) (import scheme) (define (uri-relative-from uabs base) (dif-segs-from uabs base)) (define (dif-segs-from sabs base) (if (null? base) sabs (dif-segs-from sabs base ___ Chicken-hackers mailing list Chicken-hackers@nongnu.org https://lists.nongnu.org/mailman/listinfo/chicken-hackers
Re: [Chicken-hackers] [PATCH] Some simplification rules for nested ##core#inline forms
On Wed, Sep 04, 2019 at 11:59:31AM +0200, felix.winkelm...@bevuta.com wrote: > The attached patch adds two optimization rules for certain uses of > ##core#inline. It basically rewrites > > (let (( (##core#inline ...))) > ( ... ...)) > > into > > ( ... (##core##inline ...) ...) It looks like there's a problem caused by this: https://salmonella-freebsd-x86-64.call-cc.org/master/clang/freebsd/x86-64/2019/09/16/salmonella-report/install/uri-generic.html uri-generic uses matchable, which presumably generates a lot of code which looks like it should fit the pattern. Interestingly, nothing else that depends on matchable breaks: https://salmonella-freebsd-x86-64.call-cc.org/master/clang/freebsd/x86-64/2019/09/16/salmonella-report/rev-dep-graphs/matchable.html (there are a few red ones but they're due to missing C library packages) Cheers, Peter signature.asc Description: PGP signature ___ Chicken-hackers mailing list Chicken-hackers@nongnu.org https://lists.nongnu.org/mailman/listinfo/chicken-hackers
