Bug #20132 “atlas3: FTBFS with gcc-4.0 [i386, amd64]” : Bugs : atlas3 package : Ubuntu

Revision history for this message

In Debian Bug tracker #315908, Daniel Schepler (schepler-math) wrote on 2005-07-15: Now FTBFS on i386 as well

#1

package atlas3
severity 315908
retitle 315908 atlas3: FTBFS with gcc-4.0 [i386, amd64]
thanks

As I predicted, this bug is now causing atlas3 to fail to build on
i386 as well.
--
Daniel Schepler "Please don't disillusion me. I
<email address hidden> haven't had breakfast yet."
-- Orson Scott Card

Revision history for this message

In Debian Bug tracker #315908, Matthias Klose (doko-cs) wrote on 2005-08-15: FTBFS with gcc-4.0 is severity serious

#2

severity 315908 serious
thanks

Revision history for this message

In Debian Bug tracker #315908, Russ Allbery (rra-debian) wrote on 2005-08-17: Re: Bug#315908: atlas3: FTBFS with gcc-4.0 [i386, amd64]

#3

Download full text (7.7 KiB)

The following patch fixes the basic FTBFS problem with gcc 4.0.
Unfortunately, after this patch, I still get the following test suite
failure:

bin/Linux_base_shared/xsl2blastst -R all -n 1019 -m 1019 -F 200

---------------------------------- GBMV -----------------------------------
TST# T M N KL KU ALPHA LDA INCX BETA INCY TIME MFLOP SpUp TEST
==== = ==== ==== ==== ==== ===== ==== ==== ===== ==== ===== ===== ==== ====
0 N 1019 1019 1 1 1.0 3 1 1.0 1 0.00 112.1 1.00 ----
0 N 1019 1019 1 1 1.0 3 1 1.0 1 0.00 112.4 1.00 PASS

------------------------------- GEMV --------------------------------
TST# TR M N ALPHA LDA INCX BETA INCY TIME MFLOP SpUp TEST
==== == ==== ==== ===== ==== ==== ===== ==== ====== ===== ===== =====
1 N 1019 1019 1.0 1019 1 1.0 1 0.01 181.9 1.00 -----
1 N 1019 1019 1.0 1019 1 1.0 1 0.01 182.2 1.00 PASS

------------------------------- SBMV --------------------------------
TST# UP N K ALPHA LDA INCX BETA INCY TIME MFLOP SpUp TEST
==== == ==== ==== ===== ==== ==== ===== ==== ====== ===== ===== =====
2 L 1019 1 1.0 2 1 1.0 1 0.00 91.9 1.00 -----
2 L 1019 1 1.0 2 1 1.0 1 0.00 92.3 1.00 PASS

-------------------------- SPMV ---------------------------
TST# UP N ALPHA INCX BETA INCY TIME MFLOP SpUp TEST
==== == ==== ===== ==== ===== ==== ====== ===== ===== =====
3 L 1019 1.0 1 1.0 1 0.01 255.8 1.00 -----
3 L 1019 1.0 1 1.0 1 0.01 256.1 1.00 PASS

----------------------------- SYMV -----------------------------
TST# UP N ALPHA LDA INCX BETA INCY TIME MFLOP SpUp TEST
==== == ==== ===== ==== ==== ===== ==== ====== ===== ===== =====
4 L 1019 1.0 1019 1 1.0 1 0.01 340.4 1.00 -----
4 L 1019 1.0 1019 1 1.0 1 0.01 341.0 1.00 PASS

----------------------------- TBMV ------------------------------
TST# UPLO TRAN DIAG N K LDA INCX TIME MFLOP SpUp TEST
==== ==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
5 L N N 1019 1 2 1 0.00 25.8 1.00 -----
5 L N N 1019 1 2 1 0.00 25.7 1.00 PASS

------------------------ TPMV -------------------------
TST# UPLO TRAN DIAG N INCX TIME MFLOP SpUp TEST
==== ==== ==== ==== ==== ==== ====== ====== ===== =====
6 L N N 1019 1 0.00 291.9 1.00 -----
6 L N N 1019 1 0.00 293.7 1.01 PASS

-------------------------- TRMV ----------------------------
ERROR: resid=392.406769, normD=12.385225, normA=259.826324, normX=0.499337, eps=1.192093e-07
   resid=392.406769
TST# UPLO TRAN DIAG N LDA INCX TIME MFLOP SpUp TEST
==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
   7 L N N 1019 1019 1 0.00 591.7 1.00 -----
   7 L N N 1019 1019 1 0.01 175.5 0.30 FAIL

----------------------------- TBSV ------------------------------
TST# UPLO TRAN DIAG N K LDA INCX TIME MFLOP SpUp TEST
==== ==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
8 L N N 1019 1 2 1 0.00 ...

The following patch fixes the basic FTBFS problem with gcc 4.0.
Unfortunately, after this patch, I still get the following test suite
failure:

bin/Linux_base_shared/xsl2blastst -R all -n 1019 -m 1019 -F 200

---------------------------------- GBMV -----------------------------------
TST# T    M    N   KL   KU ALPHA  LDA INCX  BETA INCY  TIME MFLOP SpUp TEST
==== = ==== ==== ==== ==== ===== ==== ==== ===== ==== ===== ===== ==== ====
   0 N 1019 1019    1    1   1.0    3    1   1.0    1  0.00 112.1 1.00 ----
   0 N 1019 1019    1    1   1.0    3    1   1.0    1  0.00 112.4 1.00 PASS

------------------------------- GEMV --------------------------------
TST# TR    M    N ALPHA  LDA INCX  BETA INCY   TIME MFLOP  SpUp  TEST
==== == ==== ==== ===== ==== ==== ===== ==== ====== ===== ===== =====
   1  N 1019 1019   1.0 1019    1   1.0    1   0.01 181.9  1.00 -----
   1  N 1019 1019   1.0 1019    1   1.0    1   0.01 182.2  1.00 PASS

------------------------------- SBMV --------------------------------
TST# UP    N    K ALPHA  LDA INCX  BETA INCY   TIME MFLOP  SpUp  TEST
==== == ==== ==== ===== ==== ==== ===== ==== ====== ===== ===== =====
   2  L 1019    1   1.0    2    1   1.0    1   0.00  91.9  1.00 -----
   2  L 1019    1   1.0    2    1   1.0    1   0.00  92.3  1.00 PASS

-------------------------- SPMV ---------------------------
TST# UP    N ALPHA INCX  BETA INCY   TIME MFLOP  SpUp  TEST
==== == ==== ===== ==== ===== ==== ====== ===== ===== =====
   3  L 1019   1.0    1   1.0    1   0.01 255.8  1.00 -----
   3  L 1019   1.0    1   1.0    1   0.01 256.1  1.00 PASS

----------------------------- SYMV -----------------------------
TST# UP    N ALPHA  LDA INCX  BETA INCY   TIME MFLOP  SpUp  TEST
==== == ==== ===== ==== ==== ===== ==== ====== ===== ===== =====
   4  L 1019   1.0 1019    1   1.0    1   0.01 340.4  1.00 -----
   4  L 1019   1.0 1019    1   1.0    1   0.01 341.0  1.00 PASS

----------------------------- TBMV ------------------------------
TST# UPLO TRAN DIAG    N    K  LDA INCX   TIME  MFLOP  SpUp  TEST
==== ==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
   5    L    N    N 1019    1    2    1   0.00   25.8  1.00 -----
   5    L    N    N 1019    1    2    1   0.00   25.7  1.00 PASS

------------------------ TPMV -------------------------
TST# UPLO TRAN DIAG    N INCX   TIME  MFLOP  SpUp  TEST
==== ==== ==== ==== ==== ==== ====== ====== ===== =====
   6    L    N    N 1019    1   0.00  291.9  1.00 -----
   6    L    N    N 1019    1   0.00  293.7  1.01 PASS

-------------------------- TRMV ----------------------------
ERROR:   resid=392.406769, normD=12.385225, normA=259.826324, normX=0.499337, eps=1.192093e-07
   resid=392.406769
TST# UPLO TRAN DIAG    N  LDA INCX   TIME  MFLOP  SpUp  TEST
==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
   7    L    N    N 1019 1019    1   0.00  591.7  1.00 -----
   7    L    N    N 1019 1019    1   0.01  175.5  0.30 FAIL

----------------------------- TBSV ------------------------------
TST# UPLO TRAN DIAG    N    K  LDA INCX   TIME  MFLOP  SpUp  TEST
==== ==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
   8    L    N    N 1019    1    2    1   0.00   44.1  1.00 -----
   8    L    N    N 1019    1    2    1   0.00   44.5  1.01 PASS

------------------------ TPSV -------------------------
TST# UPLO TRAN DIAG    N INCX   TIME  MFLOP  SpUp  TEST
==== ==== ==== ==== ==== ==== ====== ====== ===== =====
   9    L    N    N 1019    1   0.00  283.2  1.00 -----
   9    L    N    N 1019    1   0.00  286.1  1.01 PASS

-------------------------- TRSV ----------------------------
ERROR:   resid=nan, normD=nan, normA=1.739231, normX=0.499337, eps=1.192093e-07
TST# UPLO TRAN DIAG    N  LDA INCX   TIME  MFLOP  SpUp  TEST
==== ==== ==== ==== ==== ==== ==== ====== ====== ===== =====
  10    L    N    N 1019 1019    1   0.08   13.0  1.00 -----
  10    L    N    N 1019 1019    1   0.01  172.5 13.28 FAIL

------------------------------ GER -----------------------------
TST#     M     N ALPHA INCX INCY   LDA   TIME  MFLOP  SpUp  TEST
==== ===== ===== ===== ==== ==== ===== ====== ====== ===== =====
  11  1019  1019   1.0    1    1  1019   0.02   98.2  1.00 -----
  11  1019  1019   1.0    1    1  1019   0.02   98.2  1.00 PASS

----------------------- SPR ------------------------
TST# UPLO     N ALPHA INCX   TIME  MFLOP  SpUp  TEST
==== ==== ===== ===== ==== ====== ====== ===== =====
  12    L  1019   1.0    1   0.01  101.1  1.00 -----
  12    L  1019   1.0    1   0.01  101.1  1.00 PASS

--------------------------- SYR --------------------------
TST# UPLO     N ALPHA INCX   LDA   TIME  MFLOP  SpUp  TEST
==== ==== ===== ===== ==== ===== ====== ====== ===== =====
  13    L  1019   1.0    1  1019   0.01   96.5  1.00 -----
  13    L  1019   1.0    1  1019   0.01   96.5  1.00 PASS

-------------------------- SPR2 -------------------------
TST# UPLO     N ALPHA INCX INCY   TIME  MFLOP  SpUp  TEST
==== ==== ===== ===== ==== ==== ====== ====== ===== =====
  14    L  1019   1.0    1    1   0.01  175.6  1.00 -----
  14    L  1019   1.0    1    1   0.01  175.6  1.00 PASS

---------------------------- SYR2 -----------------------------
TST# UPLO     N ALPHA INCX INCY   LDA   TIME  MFLOP  SpUp  TEST
==== ==== ===== ===== ==== ==== ===== ====== ====== ===== =====
  15    L  1019   1.0    1    1  1019   0.01  171.8  1.00 -----
  15    L  1019   1.0    1    1  1019   0.01  172.7  1.01 PASS

16 tests run, 14 passed

Unfortunately, it's beyond my knowledge of the package to figure out
what's going wrong here or how to fix it.

Anyway, here is the patch, which at least moves things a little bit
farther along.  In the camm_dpa.h change, I removed the bx register
clobber because ebx was already being saved and restored in the assembly
and therefore I don't believe the register was actually being clobbered.

diff -ruNp atlas3-3.6.0/bin/uumtst.c atlas3-3.6.0.new/bin/uumtst.c
--- atlas3-3.6.0/bin/uumtst.c	2003-12-22 06:06:43.000000000 -0800
+++ atlas3-3.6.0.new/bin/uumtst.c	2005-08-16 16:14:21.000000000 -0700
@@ -59,13 +59,14 @@ static void test_lauum(const enum ATLAS_
       ATL_lauum(Maj_, Uplo_, N_, A_, lda_)
 #endif
 
+static void ATL_L2GE(const enum CBLAS_ORDER, const int, const TYPE*,
+                     const int, TYPE*, const int);
+
 static void ATL_U2GE
    (const enum CBLAS_ORDER Order, const int N, const TYPE *U, const int ldu,
     TYPE *C, const int ldc)
 {
    int j;
-   static void ATL_L2GE(const enum CBLAS_ORDER, const int, const TYPE*,
-                        const int, TYPE*, const int);
 
    if (Order == CblasRowMajor) ATL_L2GE(CblasColMajor, N, U, ldu, C, ldc);
    else
diff -ruNp atlas3-3.6.0/config.c atlas3-3.6.0.new/config.c
--- atlas3-3.6.0/config.c	2005-08-16 13:21:51.000000000 -0700
+++ atlas3-3.6.0.new/config.c	2005-08-16 10:59:45.000000000 -0700
@@ -694,7 +694,7 @@ int IsGoodGcc(enum OSTYPE OS, enum MACHT
       iret = 0;
    else if ( MachIsX86(mach) ) /* don't use gcc 2.96 - 3.0 */
    {
-      if ((major == 2 && minor < 96) || (major == 3 && minor > 1))
+      if ((major == 2 && minor < 96) || (major == 3 && minor > 1) || major > 3)
          iret = 1;
    }
    else if (mach == IA64Itan || mach == IA64Itan2 || MachIsUS(mach) ||
Binary files atlas3-3.6.0/error_Linux_sse2_shared.tgz and atlas3-3.6.0.new/error_Linux_sse2_shared.tgz differ
diff -ruNp atlas3-3.6.0/include/contrib/camm_dpa.h atlas3-3.6.0.new/include/contrib/camm_dpa.h
--- atlas3-3.6.0/include/contrib/camm_dpa.h	2003-12-22 06:07:04.000000000 -0800
+++ atlas3-3.6.0.new/include/contrib/camm_dpa.h	2005-08-16 13:06:58.000000000 -0700
@@ -1619,7 +1619,7 @@ MY_FUNCTION(aconst TYPE *a,int lda,
 #if defined(SCPLX) || (defined(DCPLX) && defined(ATL_SSE2))
 	 ,"m" (w)
 #endif
-	 :"ax","bx","cx","dx","si","di");
+	 :"ax","cx","dx","si","di");
 
 
 }

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Revision history for this message

Debian Bug Importer (debzilla) wrote on 2005-08-21:

#4

Automatically imported from Debian bug report #315908 http://bugs.debian.org/315908

Revision history for this message

Debian Bug Importer (debzilla) wrote on 2005-08-21:

#5

Message-ID: <email address hidden>
Date: Sun, 26 Jun 2005 18:30:48 -0700
From: Daniel Schepler <email address hidden>
To: Debian Bug Tracking System <email address hidden>
Subject: atlas3: FTBFS with gcc-4.0 [amd64]

Package: atlas3
Severity: normal
Version: 3.6.0-19

When I try to build atlas3 with the default gcc set to gcc-4.0, I get:

...
ATLAS has detected that you are running on an x86 architecture
(eg., Athlon, P4, PIII, etc), but do not have a pre-2.96 gcc installed.
Using gcc 3/2.96 will result in a performance loss of 10-20% IN THE BEST CASE
over a gcc previous to 2.96. If your install goes slightly wrong, performance
drops of 50% are more normal with these compilers. Full details of this
problem can be found at:
http://www.cs.utk.edu/~rwhaley/ATLAS/gcc30.html
In the meantime, we suggest you abondon this ATLAS install, and install a gcc
that delivers decent x86 performance, and restart the install then. You
do not need to be root to install, and full instructions, along with all
needed files, are given in the ATLAS errata file at:
http://math-atlas.sourceforge.net/errata.html#UnixGccInst

   Stop ATLAS install? [y]: couldn't execute "kill": no such file or directory
    while executing
"exec kill -TERM $i"
    invoked from within
"expect -re "ready to continue?" {send "\n" ; exp_continue} \
        "Enter number at top left of screen" {send "\n" ; exp_continue} \
        "ENTER TO CONTINUE" {..."
    (file "debian/config.expect" line 6)
make: *** [bin/Linux_base_static/Make.inc] Error 1

It seems config.c doesn't know about gcc 4.0; changing line 697 of
config.c to accept the case (major >= 4) as well, the package then
compiles fine until it runs into compiler errors in uumtst.c. (This
will probably affect i386 as well.)

-- System Information:
Debian Release: testing/unstable
Architecture: amd64 (x86_64)
Shell: /bin/sh linked to /bin/bash
Kernel: Linux 2.6.11-9-amd64-k8
Locale: LANG=en, LC_CTYPE=en (charmap=UTF-8) (ignored: LC_ALL set to en_US.UTF-8)

--
Daniel Schepler "Please don't disillusion me. I
<email address hidden> haven't had breakfast yet."
-- Orson Scott Card

Message-ID: <873br4k2lz.fsf@frobnitz.homelinux.net>
Date: Sun, 26 Jun 2005 18:30:48 -0700
From: Daniel Schepler <schepler@math.berkeley.edu>
To: Debian Bug Tracking System <submit@bugs.debian.org>
Subject: atlas3: FTBFS with gcc-4.0 [amd64]

Package: atlas3
Severity: normal
Version: 3.6.0-19

When I try to build atlas3 with the default gcc set to gcc-4.0, I get:

...
ATLAS has detected that you are running on an x86 architecture
(eg., Athlon, P4, PIII, etc), but do not have a pre-2.96 gcc installed.
Using gcc 3/2.96 will result in a performance loss of 10-20% IN THE BEST CASE
over a gcc previous to 2.96.  If your install goes slightly wrong, performance
drops of 50% are more normal with these compilers.  Full details of this
problem can be found at:
   http://www.cs.utk.edu/~rwhaley/ATLAS/gcc30.html
In the meantime, we suggest you abondon this ATLAS install, and install a gcc
that delivers decent x86 performance, and restart the install then.  You
do not need to be root to install, and full instructions, along with all
needed files, are given in the ATLAS errata file at:
   http://math-atlas.sourceforge.net/errata.html#UnixGccInst

Stop ATLAS install? [y]: couldn't execute "kill": no such file or directory
    while executing
"exec kill -TERM $i"
    invoked from within
"expect -re "ready to continue?" {send "\n" ; exp_continue} \
        "Enter number at top left of screen" {send "\n" ; exp_continue} \
        "ENTER TO CONTINUE" {..."
    (file "debian/config.expect" line 6)
make: *** [bin/Linux_base_static/Make.inc] Error 1

It seems config.c doesn't know about gcc 4.0; changing line 697 of
config.c to accept the case (major >= 4) as well, the package then
compiles fine until it runs into compiler errors in uumtst.c.  (This
will probably affect i386 as well.)

-- System Information:
Debian Release: testing/unstable
Architecture: amd64 (x86_64)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.11-9-amd64-k8
Locale: LANG=en, LC_CTYPE=en (charmap=UTF-8) (ignored: LC_ALL set to en_US.UTF-8)

-- 
Daniel Schepler              "Please don't disillusion me.  I
schepler@math.berkeley.edu    haven't had breakfast yet."
                                 -- Orson Scott Card

Revision history for this message

Debian Bug Importer (debzilla) wrote on 2005-08-21:

#6

Message-ID: <email address hidden>
Date: Fri, 15 Jul 2005 15:44:27 -0700
From: Daniel Schepler <email address hidden>
To: <email address hidden>
Cc: <email address hidden>
Subject: Now FTBFS on i386 as well

package atlas3
severity 315908
retitle 315908 atlas3: FTBFS with gcc-4.0 [i386, amd64]
thanks

As I predicted, this bug is now causing atlas3 to fail to build on
i386 as well.
--
Daniel Schepler "Please don't disillusion me. I
<email address hidden> haven't had breakfast yet."
-- Orson Scott Card

Revision history for this message

Debian Bug Importer (debzilla) wrote on 2005-08-21:

#7

Message-ID: <email address hidden>
Date: Mon, 15 Aug 2005 13:39:58 +0200
From: Matthias Klose <email address hidden>
To: <email address hidden>
Subject: FTBFS with gcc-4.0 is severity serious

severity 315908 serious
thanks

Revision history for this message

Debian Bug Importer (debzilla) wrote on 2005-08-21:

#8

Download full text (8.0 KiB)

Message-ID: <email address hidden>
Date: Tue, 16 Aug 2005 19:19:56 -0700
From: Russ Allbery <email address hidden>
To: <email address hidden>
Cc: <email address hidden>
Subject: Re: Bug#315908: atlas3: FTBFS with gcc-4.0 [i386, amd64]