Bug 151648
Summary: | Missing support for Alpha arch | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Product: | [Fedora] Fedora | Reporter: | Sergey Tikhonov <tsv> | ||||||||
Component: | libaio | Assignee: | Jeff Moyer <jmoyer> | ||||||||
Status: | CLOSED RAWHIDE | QA Contact: | |||||||||
Severity: | medium | Docs Contact: | |||||||||
Priority: | medium | ||||||||||
Version: | 3 | ||||||||||
Target Milestone: | --- | ||||||||||
Target Release: | --- | ||||||||||
Hardware: | alpha | ||||||||||
OS: | Linux | ||||||||||
Whiteboard: | |||||||||||
Fixed In Version: | Doc Type: | Bug Fix | |||||||||
Doc Text: | Story Points: | --- | |||||||||
Clone Of: | Environment: | ||||||||||
Last Closed: | 2005-04-04 13:48:00 UTC | Type: | --- | ||||||||
Regression: | --- | Mount Type: | --- | ||||||||
Documentation: | --- | CRM: | |||||||||
Verified Versions: | Category: | --- | |||||||||
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||||||
Cloudforms Team: | --- | Target Upstream Version: | |||||||||
Embargoed: | |||||||||||
Attachments: |
|
Description
Sergey Tikhonov
2005-03-21 13:07:47 UTC
Created attachment 112174 [details]
Proposed updated libaio.h file
Created attachment 112175 [details]
Proposed implementation of syscall-alpha.h
Created attachment 112176 [details]
Proposed updated syscall.h file
Have you tested these patches? I'm not sure I have any alpha systems to play with. I tried to run testes that are bundled with the package and most of them were successful. I did the same on AMD64 machine and there were fewer passed tests. Is there any chance for my patches to be included into mainstream? I am working on AlphaCore (FedoraCore on Alpha) project and would like to have our patches to be in sync with fedora. My patches are not affect other arches at all. Thank you. Yes, I'll review them and get them committed to the tree. In the future, please post patches, not new versions of existing files. Thanks. +#ifdef __alpha +#define SYMVER(compat_sym, orig_sym, ver_sym) \ + __asm__(".set " SYMSTR(compat_sym) " = " SYMSTR(orig_sym) "@LIBAIO_" SYMST R(ver_sym)); + +#define DEFSYMVER(compat_sym, orig_sym, ver_sym) \ + __asm__(".set " SYMSTR(compat_sym) " = " SYMSTR(orig_sym) "@@LIBAIO_" SYMS TR(ver_sym)); +#else Why are you not using the .symver directive? It works for every other architecture, why is alpha different? You are right. It is not necessary and should be removed. Thank you for pointing out and for your time. I've checked your changes into the CVS repository here: :pserver:anonymous.com:/usr/local/CVS You can simply do a checkout of libaio. Please check these sources out, do a build, and verify that the aio-stress benchmark operates properly given the following parameters: # aio-stress -O -v -t 2 -s 512 testfile1 testfile2 Replace 512 with whatever size (in megabytes) you want to use for each file. Make sure you have enough disk space for them. ;) The aio-stress.c file can be found here: http://ftp.belnet.be/linux/SuSe/people/mason/utils/aio-stress.c Let me know if you have any further questions. Once you've completed these tests successfully, I'll build this version of the library (0.3.104) into rawhide. Thanks again! Jeff Here is the output of test program. I used 20Mb as file size and hope for the test it should not be matter. Thank you. [root@falcon src]# LD_LIBRARY_PATH=`pwd` ./aio-stress -O -v -t 2 -s 20 testfile1 testfile2 file size 20MB, record size 64KB, depth 64, ios per iteration 8 max io_submit 16, buffer alignment set to 8KB threads 2 files 2 contexts 1 context offset 2MB verification on adding file testfile1 thread 0 adding file testfile2 thread 1 write on testfile2 (3.33 MB/s) 20.00 MB in 6.01s thread 1 write totals (3.33 MB/s) 20.00 MB in 6.01s write on testfile1 (3.33 MB/s) 20.00 MB in 6.01s thread 0 write totals (3.33 MB/s) 20.00 MB in 6.01s write throughput (6.65 MB/s) 40.00 MB in 6.01s min transfer 20.00MB read on testfile2 (31.13 MB/s) 20.00 MB in 0.64s thread 1 read totals (26.93 MB/s) 20.00 MB in 0.74s read on testfile1 (30.43 MB/s) 20.00 MB in 0.66s thread 0 read totals (30.29 MB/s) 20.00 MB in 0.66s read throughput (51.71 MB/s) 40.00 MB in 0.77s min transfer 20.00MB random write on testfile2 (6.04 MB/s) 20.00 MB in 3.31s thread 1 random write totals (6.01 MB/s) 20.00 MB in 3.33s random write on testfile1 (5.23 MB/s) 17.50 MB in 3.35s thread 0 random write totals (5.22 MB/s) 17.50 MB in 3.35s random write throughput (11.17 MB/s) 37.50 MB in 3.36s min transfer 17.50MB random read on testfile2 (68.92 MB/s) 20.00 MB in 0.29s thread 1 random read totals (66.50 MB/s) 20.00 MB in 0.30s random read on testfile1 (64.93 MB/s) 1.00 MB in 0.02s thread 0 random read totals (12.69 MB/s) 1.00 MB in 0.08s random read throughput (56.59 MB/s) 21.00 MB in 0.37s min transfer 1.00MB pthread_join: Success Just checked .spec file. Please, don't forget to add "alpha alphaev6" into list of ExclusiveArch: Thank you. Good catch! Thanks! I'll get this pushed soon. -Jeff Thanks, I hope my other alpha specific patches gets accepted that quickly. :) Okay, I over-wrote the tag for 0.3.104 to include the exclusivearch fixes, and I built this into rawhide. This bug is filed against FC3. Is your port a port of FC3, or is it sufficient to push this fix into the FC4 stream? Thanks, Jeff Our port of FC3 is done, so please put it into FC4 stream (I am working on it right now). Thank you. Okay, it is in rawhide. Created redmine issue http://projects.theforeman.org/issues/23083 from this bug |