Bug#722248: [Pkg-haskell-maintainers] Bug#722248: ghci not provided on arm
Le mardi 10 septembre 2013 18:23:42 Colin Watson a écrit : Thanks for looking into this further. I think that inlining the cache flush will mean that every chunk of bytecode assembled by the interpreter will flush the instruction cache every time it's called. It will be even slower than it's necessary for it to be. That in itself would probably be tolerable. Much more importantly, though, this patch misses the point of flushing the instruction cache. The point is that we need to flush the cache in order for the processor to reliably read back the code that we just wrote out; flushing the cache in that very code is not a viable approach to that, because the cache-flushing code itself might not be read. So I'm afraid I don't think this is going to work properly. It might work some of the time, of course, depending on cache locality, but I can't see how it would work reliably. My apologize, I didn't realize this was the generated code. I don't know if it's because of the language but I can't make much sense of what the code is doing. I assumed this code was some jit compiler and that the jump was executed when a function has just been compiled in order to actually call that function. If it's part of the generated function indeed it's very suboptimal and won't work as you very well explained. My apologize but I don't think I can bring this further without investing quite some time to understand the code. I would suggest testing whether some reasonable stack actually builds with this; I don't know how much testing you did but it e.g. isn't enough to just start ghci. Perhaps see if haskell-conduit runs its doctests successfully during build. You're right, I should have done so. I don't know anything about ghc and I didn't know how to test it. I should have searched a bit on google, probably I would have found some code sample easily but I guess I lacked motivation. Thanks for being so quick in looking at the patch. Best regards, Thomas signature.asc Description: This is a digitally signed message part.
Bug#722248: [Pkg-haskell-maintainers] Bug#722248: ghci not provided on arm
On Tue, Sep 10, 2013 at 06:35:56PM +0200, Joachim Breitner wrote: Am Montag, den 09.09.2013, 14:37 +0200 schrieb Thomas Preud'homme: please reenable ghci on arm as it prevents many package from building. I improved on the patch proposed upstream and it works on porterbox. Maybe you could give it a try in experimental. See the patch in the attached file. you looked into that issue last. Can you evaluate that patch? Thanks for looking into this further. I think that inlining the cache flush will mean that every chunk of bytecode assembled by the interpreter will flush the instruction cache every time it's called. It will be even slower than it's necessary for it to be. That in itself would probably be tolerable. Much more importantly, though, this patch misses the point of flushing the instruction cache. The point is that we need to flush the cache in order for the processor to reliably read back the code that we just wrote out; flushing the cache in that very code is not a viable approach to that, because the cache-flushing code itself might not be read. So I'm afraid I don't think this is going to work properly. It might work some of the time, of course, depending on cache locality, but I can't see how it would work reliably. Other things we need to do after this is fixed: We should include my R_ARM_V4BX relocation patch too in order to support armel. See: http://lists.debian.org/debian-haskell/2013/05/msg00054.html I would suggest testing whether some reasonable stack actually builds with this; I don't know how much testing you did but it e.g. isn't enough to just start ghci. Perhaps see if haskell-conduit runs its doctests successfully during build. Sorry, -- Colin Watson [cjwat...@debian.org] -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#722248: [Pkg-haskell-maintainers] Bug#722248: ghci not provided on arm
Dear Colin, Am Montag, den 09.09.2013, 14:37 +0200 schrieb Thomas Preud'homme: Dear Maintainer, please reenable ghci on arm as it prevents many package from building. I improved on the patch proposed upstream and it works on porterbox. Maybe you could give it a try in experimental. See the patch in the attached file. you looked into that issue last. Can you evaluate that patch? Greetings, Joachim -- Joachim nomeata Breitner Debian Developer nome...@debian.org | ICQ# 74513189 | GPG-Keyid: 4743206C JID: nome...@joachim-breitner.de | http://people.debian.org/~nomeata signature.asc Description: This is a digitally signed message part
Bug#722248: ghci not provided on arm
Source: ghc
Version: 7.6.3-4
Severity: wishlist
Tags: patch upstream
Dear Maintainer,
please reenable ghci on arm as it prevents many package from building. I
improved on the patch proposed upstream and it works on porterbox. Maybe
you could give it a try in experimental. See the patch in the attached
file.
Best regards.
-- System Information:
Debian Release: jessie/sid
APT prefers unstable
APT policy: (990, 'unstable'), (500, 'oldstable-updates'), (500, 'testing'),
(500, 'stable'), (1, 'experimental')
Architecture: amd64 (x86_64)
Foreign Architectures: i386
Kernel: Linux 3.10-1-amd64 (SMP w/4 CPU cores)
Locale: LANG=fr_FR.UTF-8, LC_CTYPE=fr_FR.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Description: Add ARM implementation for mkJumpToAddr
TODO: Put a short summary on the line above and replace this paragraph
with a longer explanation of this change. Complete the meta-information
with other relevant fields (see below for details). To make it easier, the
information below has been extracted from the changelog. Adjust it or drop
it.
.
ghc (7.6.3-4) unstable; urgency=low
.
[ Colin Watson ]
* Enable verbose build output (see
http://lists.debian.org/debian-devel/2013/06/msg00539.html).
.
[ Gianfranco Costamagna ]
.
* Switch to llvm (Closes: #711948)
* removed deprecated DM-Upload-Allowed
* removed some version checks, higher versions are already in oldstable.
* Added dpkg-buildflags as build-dep, fixing lintian warning.
Author: Joachim Breitner nome...@debian.org
Bug-Debian: http://bugs.debian.org/711948
---
The information above should follow the Patch Tagging Guidelines, please
checkout http://dep.debian.net/deps/dep3/ to learn about the format. Here
are templates for supplementary fields that you might want to add:
Origin: vendor|upstream|other, url of original patch
Bug: url in upstream bugtracker
Bug-Debian: http://bugs.debian.org/bugnumber
Bug-Ubuntu: https://launchpad.net/bugs/bugnumber
Forwarded: no|not-needed|url proving that it has been forwarded
Reviewed-By: name and email of someone who approved the patch
Last-Update: -MM-DD
--- ghc-7.6.3.orig/compiler/ghci/ByteCodeItbls.lhs
+++ ghc-7.6.3/compiler/ghci/ByteCodeItbls.lhs
@@ -242,6 +242,18 @@ mkJumpToAddr a
, fromIntegral ((w64 `shiftR` 32) .. 0x) ]
where w64 = fromIntegral (ptrToInt a) :: Word64
+#elif arm_TARGET_ARCH
+type ItblCode = Word32
+mkJumpToAddr a
+= [ 0xe92d0080 -- push{r7}
+ , 0xe3a0780f -- mov r7, #983040 ; 0xf
+ , 0xe2877002 -- add r7, r7, #2
+ , 0xe3a02000 -- mov r2, #0
+ , 0xef00 -- swi 0x0
+ , 0xe8bd0080 -- pop {r7}
+ , 0xe51ff004 -- ldr pc, [pc, #-4]# pc reads as current insn+8
+ , fromIntegral (ptrToInt a) ]
+
#else
type ItblCode = Word32
mkJumpToAddr a
