IRC channel logs
2025-08-10.log
back to list of logs
<sham1>Need to be more sneeky than that <Arsen>ACTION uses trunk emacs on gentoo <cow_2001>(use-modules (sxml ssax)) and then (sxpath '()) => ;;; <stdin>:21:1: warning: possibly unbound variable `sxpath' <ieure>cow_2001, (use-modules (sxml xpath)) <cow_2001>i don't see it mentioned anywhere in this manual <cow_2001>3.0.9 version C-s sxml xpath C-s C-s C-s => nothing! <ieure>cow_2001, No idea. I looked at the source. <identity>3.0.10 is latest stable, but iirc it had some problems with 32-bit architectures, so it is not in Guix <dsmith>rg reveals sxpath is in (sxml xpath) <dsmith>(use-modules (sxml xpath)) (sxpath '()) <dsmith>$1 = #<procedure 7f4d07687d40 at sxml/xpath.scm:453:2 (nodeset)> <ttz>Hi, does Guile really perform JIT compulation? is there a place where the process is documented? Does anyone know if there is a way to hint at the compiler to use AVX instructions? <ttz>Isn't the Guile compiler supposed to be able to remove bound check based on flow analysis? <ttz>when disassembling (define (dummy n) (if (and (< n 4) (< n 2)) #t #f)) I see both checks are being performed <dsmith>So when n is 3, don't you need both comparisons? <dsmith>Perhaps you meant (and (< n 2) (< n 4)) ? <dsmith>Though that also probably still needs both comparisons.. <ttz`>What I wanted to simplify is a procedure iterating on two vectors <ttz`>First a check that their length is equal is performed, then at each iteration of the loop that the index is in bound. <ttz`>In the bytecode I see that f32vector-ref generates a bound check upon accessing each vector. It seems to me the rest of the code tend to prove they are not necessary,