I'm happy with this fix! However I think I would like to see if I can
just speed up thunk? and other arity accessors. Do you have a test
case that you can use to show the speed problem?
Andy
On Sat 14 Nov 2015 03:54, Christopher Allan Webber <cweb...@dustycloud.org>
writes:
> Here's a fix to this bug. Tests pass, and performance appears to be
> back here.
>
> I've assigned copyright to the FSF for Guile so it should be fine to
> commit!
>
>
>>From 79e3b5286a2699f9b302bd3abf8a6b884b13a4f4 Mon Sep 17 00:00:00 2001
> From: Christopher Allan Webber <cweb...@dustycloud.org>
> Date: Fri, 13 Nov 2015 20:42:31 -0600
> Subject: [PATCH] Remove thunk / arity check in make-promise
>
> * libguile/promises.c (s_scm_make_promise): Remove arity check in
> make-promise. This was causing considerably slowdown with the new elf
> code, causing considerable number of bytevector reading calls on
> every call to (make-promise). Removing this check fixes a performance
> regression in the new compiler.
> ---
> libguile/promises.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/libguile/promises.c b/libguile/promises.c
> index dcd0ac3..858b6f3 100644
> --- a/libguile/promises.c
> +++ b/libguile/promises.c
> @@ -1,4 +1,4 @@
> -/* Copyright (C)
> 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004,2005,2006,2007,2008,2009,2010,2011
> +/* Copyright (C)
> 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004,2005,2006,2007,2008,2009,2010,2011,2015
> * Free Software Foundation, Inc.
> *
> * This library is free software; you can redistribute it and/or
> @@ -77,7 +77,6 @@ SCM_DEFINE (scm_make_promise, "make-promise", 1, 0, 0,
> "@end lisp\n")
> #define FUNC_NAME s_scm_make_promise
> {
> - SCM_VALIDATE_THUNK (1, thunk);
> SCM_RETURN_NEWSMOB2 (scm_tc16_promise,
> SCM_UNPACK (thunk),
> SCM_UNPACK (scm_make_recursive_mutex ()));
