PFGW 3.8.3 (with gwnum v28.7) Released - Page 26

rogue · 2017-04-04, 09:34

I cannot reproduce this. I need to do some more digging.

ET_ · 2017-04-06, 12:57

Quote:

Originally Posted by rogue

I cannot reproduce this. I need to do some more digging.

Thank you Mark. I suspect the Linux version was compiled with all optimizations on, using Haswell or more advanced instructions. Anyway, I can wait while still using previous versions of pfgw.

henryzz · 2017-04-06, 20:55

Quote:

Originally Posted by ET_

Thank you Mark. I suspect the Linux version was compiled with all optimizations on, using Haswell or more advanced instructions. Anyway, I can wait while still using previous versions of pfgw.

The GWNUM potion should detect the cpu and have no issue. It is probably the rest of it that now has too recent instructions.

fivemack · 2017-05-17, 14:32

I just downloaded Pfgw_Linux_3.8.3_20161203.zip from SourceForge.

It gives an 'illegal instruction' error on my Ivy Bridge machine; running under gdb I see

Code:

Dump of assembler code for function __gmpz_sizeinbase:
...
   0x0000000000482f72 <+34>:	sub    %rax,%rcx
=> 0x0000000000482f75 <+37>:	blsr   %esi,%eax
   0x0000000000482f7a <+42>:	movslq %esi,%rsi
...

where blsr ('clear lowest set bit of register') is a BMI1 instruction existing only in Haswell and later.

ET_ · 2017-05-17, 15:11

Quote:

Originally Posted by fivemack

I just downloaded Pfgw_Linux_3.8.3_20161203.zip from SourceForge.

It gives an 'illegal instruction' error on my Ivy Bridge machine; running under gdb I see

Code:

Dump of assembler code for function __gmpz_sizeinbase:
...
   0x0000000000482f72 <+34>:	sub    %rax,%rcx
=> 0x0000000000482f75 <+37>:	blsr   %esi,%eax
   0x0000000000482f7a <+42>:	movslq %esi,%rsi
...

where blsr ('clear lowest set bit of register') is a BMI1 instruction existing only in Haswell and later.

Well, at least my shoot on the dark went on target

henryzz · 2017-05-17, 17:59

So not a gwnum issue but gmplib compiled with too recent optimizations.

Batalov · 2017-06-04, 01:54

Quote:

Originally Posted by fivemack

I just downloaded Pfgw_Linux_3.8.3_20161203.zip from SourceForge.

It gives an 'illegal instruction' error on my Ivy Bridge machine; ...

Same here (on a Broadwell).

Is the binary going to be rebuilt or is it going to continue being on a biennial release schedule?

(There is a 2015 binary, a 2017 (a broken one), ...next one in 2019?)

rogue · 2017-06-04, 18:35

Steven Harvey does the linux builds for pfgw. If anyone else wants to do it, I can help them.

ET_ · 2017-07-24, 11:32

Quote:

Originally Posted by rogue

Steven Harvey does the linux builds for pfgw. If anyone else wants to do it, I can help them.

Yes, please

rogue · 2017-07-24, 13:13

Quote:

Originally Posted by ET_

Yes, please

I'll contact you via e-mail.

CRGreathouse · 2018-01-16, 14:32

This is still affecting me. Has anyone been able to build a linux version which works for older processors?

2017-05-17, 14:32	#279
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 3·2,141 Posts	I just downloaded Pfgw_Linux_3.8.3_20161203.zip from SourceForge. It gives an 'illegal instruction' error on my Ivy Bridge machine; running under gdb I see Code: Dump of assembler code for function __gmpz_sizeinbase: ... 0x0000000000482f72 <+34>: sub %rax,%rcx => 0x0000000000482f75 <+37>: blsr %esi,%eax 0x0000000000482f7a <+42>: movslq %esi,%rsi ... where blsr ('clear lowest set bit of register') is a BMI1 instruction existing only in Haswell and later.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
A possible bug in LLR/PFGW while using GWNUM (no bug in P95)	Batalov	Software	77	2015-04-14 09:01
PFGW 3.2.0 has been Released	rogue	Software	94	2010-09-14 21:39
PFGW 3.2.3 has been Released	rogue	Software	10	2009-10-28 07:07
PFGW 3.2.2 has been Released	rogue	Software	20	2009-08-23 12:14
PFGW 3.2.1 has been released	rogue	Software	5	2009-08-10 01:43

2017-04-04, 09:34	#276
rogue "Mark" Apr 2003 Between here and the 1100011010010₂ Posts	I cannot reproduce this. I need to do some more digging.

2017-05-17, 17:59	#281
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 13376₈ Posts	So not a gwnum issue but gmplib compiled with too recent optimizations.

2017-06-04, 18:35	#283
rogue "Mark" Apr 2003 Between here and the 1100011010010₂ Posts	Steven Harvey does the linux builds for pfgw. If anyone else wants to do it, I can help them.

2018-01-16, 14:32	#286
CRGreathouse Aug 2006 13533₈ Posts	This is still affecting me. Has anyone been able to build a linux version which works for older processors?