[Bug target/15822] New: Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[Bug target/15822] New: Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[gcc-bugzilla at gcc dot gnu dot org]

I noticed a couple of new testsuite failures (some of which are
regressions) on alpha-dec-osf5.1b between 20040524 and 20040601:

+FAIL: gcc.c-torture/execute/compare-fp-1.c execution,  -O0 
+FAIL: gcc.c-torture/execute/compare-fp-1.c execution,  -O1 
+FAIL: gcc.c-torture/execute/compare-fp-1.c execution,  -O2 
+FAIL: gcc.c-torture/execute/compare-fp-1.c compilation,  -O3 -fomit-frame-pointer 
+UNRESOLVED: gcc.c-torture/execute/compare-fp-1.c execution,  -O3 -fomit-frame-pointer 
+FAIL: gcc.c-torture/execute/compare-fp-1.c compilation,  -O3 -g 
+UNRESOLVED: gcc.c-torture/execute/compare-fp-1.c execution,  -O3 -g 
+FAIL: gcc.c-torture/execute/compare-fp-1.c execution,  -Os 
+FAIL: gcc.c-torture/execute/compare-fp-4.c execution,  -O0 
+FAIL: gcc.c-torture/execute/compare-fp-4.c execution,  -O1 
+FAIL: gcc.c-torture/execute/compare-fp-4.c execution,  -O2 
+FAIL: gcc.c-torture/execute/compare-fp-4.c execution,  -O3 -fomit-frame-pointer 
+FAIL: gcc.c-torture/execute/compare-fp-4.c execution,  -O3 -g 
+FAIL: gcc.c-torture/execute/compare-fp-4.c execution,  -Os 

+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O0 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O1 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O2 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O3 -fomit-frame-pointer 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O3 -fomit-frame-pointer -funroll-loops 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -O3 -g 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-4.c execution,  -Os 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O0 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O1 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O2 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O3 -fomit-frame-pointer 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O3 -fomit-frame-pointer -funroll-loops 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -O3 -g 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-5.c execution,  -Os 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O0 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O1 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O2 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O3 -fomit-frame-pointer 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O3 -fomit-frame-pointer -funroll-loops 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -O3 -g 
+FAIL: gcc.c-torture/execute/ieee/fp-cmp-8.c execution,  -Os 

While the compare-fp-[14].c tests are new, the fp-cmp-[458].c tests used to
pass.  This can also be obverved on alphaev67-unknown-linux-gnu and was
introduced between

http://gcc.gnu.org/ml/gcc-testresults/2004-05/msg01301.html (20040528)

and

http://gcc.gnu.org/ml/gcc-testresults/2004-05/msg01335.html (20040529)

>From the ChangeLog, most likely this patch (which introduced the new
failing tests above) is the culprit:

2004-05-28  Paolo Bonzini  <bonzini@gnu.org>
	    Roger Sayle  <roger@eyesopen.com>

	PR rtl-optimization/15649
	Add LTGT_EXPR and improve pretty-printing of unordered
	comparisons.

Environment:
System: OSF1 bartok V5.1 2650 alpha
Machine: alpha
	
host: alpha-dec-osf5.1b
build: alpha-dec-osf5.1b
target: alpha-dec-osf5.1b
configured with: /vol/gnu/src/gcc/gcc-dist/configure --prefix=/vol/gcc --with-local-prefix=/vol/gcc --disable-nls --host alpha-dec-osf5.1b --build alpha-dec-osf5.1b --target alpha-dec-osf5.1b --disable-libmudflap

How-To-Repeat:
Bootstrap and regtest mainline on any alpha target.

-- 
           Summary: Several alpha testsuite regressions: gcc.c-
                    torture/execute/ieee/fp-cmp-[458].c
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ro at techfak dot uni-bielefeld dot de
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: alpha-dec-osf5.1b
  GCC host triplet: alpha-dec-osf5.1b
GCC target triplet: alpha-dec-osf5.1b


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[ro at gcc dot gnu dot org]

------- Additional Comments From ro at gcc dot gnu dot org  2004-06-04 16:01 -------
Your patch

2004-05-28  Paolo Bonzini  <bonzini@gnu.org>
	    Roger Sayle  <roger@eyesopen.com>

	PR rtl-optimization/15649
	Add LTGT_EXPR and improve pretty-printing of unordered
	comparisons.

is most likely causing those testsuite regressions.  Please fix.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bonzini at gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-06-04 16:29 -------
Subject: Bug 15822

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	bonzini@gcc.gnu.org	2004-06-04 16:29:35

Modified files:
	gcc            : dojump.c ChangeLog 

Log message:
	2004-06-04  Paolo Bonzini  <bonzini@gnu.org>
	
	PR target/15822
	* dojump.c (do_jump): Fix uninitialized variable tcode1.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/dojump.c.diff?cvsroot=gcc&r1=1.17&r2=1.18
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.3840&r2=2.3841



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-06-04 16:38 -------
Is this fixed after Paolo's patch?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
           Keywords|                            |wrong-code
            Summary|Several alpha testsuite     |[3.5 Regression] Several
                   |regressions: gcc.c-         |alpha testsuite regressions:
                   |torture/execute/ieee/fp-cmp-|gcc.c-
                   |[458].c                     |torture/execute/ieee/fp-cmp-
                   |                            |[458].c
   Target Milestone|---                         |3.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[ro at techfak dot uni-bielefeld dot de]

------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2004-06-04 16:51 -------
Subject: Re:  [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

A quick manual test (just recompiling dojump.c and relinking cc1) confirms
that fp-cmp-4.c passes now at -O0.  I'll do a full bootstrap to check.

The compare-fp-1.x0 test, on the other hand, fails with

Floating point exception

but adding -mieee lets it pass.  So I suppose that those tests should go
into gcc.c-torture/execute/ieee instead of just gcc.c-torture/execute.

	Rainer


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[bonzini at gnu dot org]

------- Additional Comments From bonzini at gnu dot org  2004-06-04 18:34 -------
Subject: Re:  [3.5 Regression] Several alpha testsuite regressions:
 gcc.c-torture/execute/ieee/fp-cmp-[458].c

> but adding -mieee lets it pass.  So I suppose that those tests should go
> into gcc.c-torture/execute/ieee instead of just gcc.c-torture/execute.

"The resulting code is less efficient but is able to correctly support 
denormalized numbers and exceptional IEEE values such as not-a-number 
and plus/minus infinity."  (from gcc.info).

This is definitely the case, I'm committing the patch to move these tests.

Paolo

2004-06-04  Paolo Bonzini  <bonzini@gnu.org>

         PR target/15822
         * gcc.c-torture/execute/compare-fp-1.c,
         gcc.c-torture/execute/compare-fp-2.c,
         gcc.c-torture/execute/compare-fp-3.c,
         gcc.c-torture/execute/compare-fp-4.c,
         gcc.c-torture/execute/compare-fp-3.x,
         gcc.c-torture/execute/compare-fp-4.x: Moved...
         * gcc.c-torture/execute/ieee/compare-fp-1.c,
         gcc.c-torture/execute/ieee/compare-fp-2.c,
         gcc.c-torture/execute/ieee/compare-fp-3.c,
         gcc.c-torture/execute/ieee/compare-fp-4.c,
         gcc.c-torture/execute/ieee/compare-fp-3.x,
         gcc.c-torture/execute/ieee/compare-fp-4.x: ...here.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-06-04 18:34 -------
Subject: Bug 15822

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	bonzini@gcc.gnu.org	2004-06-04 18:34:22

Modified files:
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.c-torture/execute/ieee: compare-fp-1.c 
	                                          compare-fp-2.c 
	                                          compare-fp-3.c 
	                                          compare-fp-3.x 
	                                          compare-fp-4.c 
	                                          compare-fp-4.x 
Removed files:
	gcc/testsuite/gcc.c-torture/execute: compare-fp-1.c 
	                                     compare-fp-2.c 
	                                     compare-fp-3.c 
	                                     compare-fp-3.x 
	                                     compare-fp-4.c 
	                                     compare-fp-4.x 

Log message:
	2004-06-04  Paolo Bonzini  <bonzini@gnu.org>
	
	PR target/15822
	* gcc.c-torture/execute/compare-fp-1.c,
	gcc.c-torture/execute/compare-fp-2.c,
	gcc.c-torture/execute/compare-fp-3.c,
	gcc.c-torture/execute/compare-fp-4.c,
	gcc.c-torture/execute/compare-fp-3.x,
	gcc.c-torture/execute/compare-fp-4.x: Moved...
	* gcc.c-torture/execute/ieee/compare-fp-1.c,
	gcc.c-torture/execute/ieee/compare-fp-2.c,
	gcc.c-torture/execute/ieee/compare-fp-3.c,
	gcc.c-torture/execute/ieee/compare-fp-4.c,
	gcc.c-torture/execute/ieee/compare-fp-3.x,
	gcc.c-torture/execute/ieee/compare-fp-4.x: ...here.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.3810&r2=1.3811
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/compare-fp-1.c.diff?cvsroot=gcc&r1=1.1&r2=NONE
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/compare-fp-2.c.diff?cvsroot=gcc&r1=1.1&r2=NONE
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/compare-fp-3.c.diff?cvsroot=gcc&r1=1.1&r2=NONE
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/compare-fp-3.x.diff?cvsroot=gcc&r1=1.1&r2=NONE
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/compare-fp-4.c.diff?cvsroot=gcc&r1=1.1&r2=NONE
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/compare-fp-4.x.diff?cvsroot=gcc&r1=1.1&r2=NONE
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-1.c.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-2.c.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-3.c.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-3.x.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-4.c.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-4.x.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-06-04 19:38 -------
Fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[ro at techfak dot uni-bielefeld dot de]

------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2004-06-07 17:24 -------
Subject: Re:  [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

After moving the new tests to gcc.c-torture/execute/ieee, all but two
failures are gone on alpha-dec-osf{4.0f, 5.1b}:

FAIL: gcc.c-torture/execute/ieee/compare-fp-1.c compilation,  -O3 -fomit-frame-pointer 
UNRESOLVED: gcc.c-torture/execute/ieee/compare-fp-1.c execution,  -O3 -fomit-frame-pointer 
FAIL: gcc.c-torture/execute/ieee/compare-fp-1.c compilation,  -O3 -g 
UNRESOLVED: gcc.c-torture/execute/ieee/compare-fp-1.c execution,  -O3 -g 

/vol/gnu/src/gcc/gcc-dist/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-1.c: In function `main':
/vol/gnu/src/gcc/gcc-dist/gcc/testsuite/gcc.c-torture/execute/ieee/compare-fp-1.c:32: internal compiler error: Segmentation fault

It can be reproduced with

cc1 -O3 compare-fp-1.i

The stacktrace is as follows:

Program received signal SIGSEGV, Segmentation fault.
0x00000001202fb63c in register_operand (op=0x4282a0, mode=DImode) at /vol/gnu/src/gcc/gcc-dist/gcc/recog.c:1029
(gdb) where
#0  0x00000001202fb63c in register_operand (op=0x4282a0, mode=DImode) at /vol/gnu/src/gcc/gcc-dist/gcc/recog.c:1029
#1  0x00000001201c0410 in alpha_emit_setcc (code=UNORDERED) at /vol/gnu/src/gcc/gcc-dist/gcc/config/alpha/alpha.c:3312
#2  0x000000012031f80c in gen_sunordered (operand0=0x4282a0) at insn-emit.c:5216
#3  0x00000001202e49cc in emit_store_flag (target=0x402a60, code=CONST_DOUBLE, op0=0x4282a0, op1=0x1c090, mode=SFmode, unsignedp=0, normalizep=1) at /vol/gnu/src/gcc/gcc-dist/gcc/expmed.c:4569
#4  0x000000012020105c in do_store_flag (exp=0x4282a0, target=0x402a60, mode=DImode, only_cheap=0) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:9891
#5  0x00000001201ee4b0 in expand_expr_real_1 (exp=0x427360, target=0x42, tmode=VOIDmode, modifier=EXPAND_NORMAL, alt_rtl=0x0) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:8309
#6  0x00000001201fd644 in expand_expr_real (exp=0x4282a0, target=0x1c090, tmode=4358816, modifier=66, alt_rtl=0x408d20) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:6384
#7  0x00000001201ffdf0 in store_expr (exp=0x4282a0, target=0x1c090, want_value=9) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.h:533
#8  0x0000000120201fc8 in expand_assignment (to=0x42, from=0x301540, want_value=114832) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:3985
#9  0x00000001201eb1f0 in expand_expr_real_1 (exp=0x3014f0, target=0x42, tmode=VOIDmode, modifier=EXPAND_NORMAL, alt_rtl=0x0) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:8924
#10 0x00000001201fd7dc in expand_expr_real (exp=0x4282a0, target=0x1c090, tmode=4358816, modifier=66, alt_rtl=0x408d20) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:6378
#11 0x00000001201ea2fc in expand_expr_real_1 (exp=0x36b230, target=0x42, tmode=VOIDmode, modifier=EXPAND_NORMAL, alt_rtl=0x0) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.h:533
#12 0x00000001201fd644 in expand_expr_real (exp=0x4282a0, target=0x1c090, tmode=4358816, modifier=66, alt_rtl=0x408d20) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:6384
#13 0x00000001201f021c in expand_expr_real_1 (exp=0x421bd0, target=0x42, tmode=VOIDmode, modifier=EXPAND_NORMAL, alt_rtl=0x11fffbe20) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.h:533
#14 0x00000001201fd644 in expand_expr_real (exp=0x4282a0, target=0x1c090, tmode=4358816, modifier=66, alt_rtl=0x408d20) at /vol/gnu/src/gcc/gcc-dist/gcc/expr.c:6384
#15 0x000000012011d370 in expand_expr_stmt_value (exp=0x4282a0, want_value=66, maybe_last=-1347440721) at /vol/gnu/src/gcc/gcc-dist/gcc/stmt.c:2128
#16 0x000000012029090c in tree_rest_of_compilation (fndecl=0x4282a0, nested_p=9) at /vol/gnu/src/gcc/gcc-dist/gcc/tree-optimize.c:561
#17 0x000000012007e368 in c_expand_body (fndecl=0x4282a0) at /vol/gnu/src/gcc/gcc-dist/gcc/c-decl.c:6374
#18 0x0000000120292a30 in cgraph_expand_function (node=0x4282a0) at /vol/gnu/src/gcc/gcc-dist/gcc/cgraphunit.c:794
#19 0x0000000120293c64 in cgraph_optimize () at /vol/gnu/src/gcc/gcc-dist/gcc/cgraphunit.c:1668
#20 0x00000001200c7388 in c_objc_common_finish_file () at /vol/gnu/src/gcc/gcc-dist/gcc/c-objc-common.c:240
#21 0x0000000120065c08 in finish_file () at /vol/gnu/src/gcc/gcc-dist/gcc/c-lang.c:213
#22 0x00000001200bcb24 in c_common_parse_file (set_yydebug=0) at /vol/gnu/src/gcc/gcc-dist/gcc/c-opts.c:1240
#23 0x00000001200e655c in toplev_main (argc=114832, argv=0x4282a0) at /vol/gnu/src/gcc/gcc-dist/gcc/toplev.c:1654
#24 0x00000001200da408 in main (argc=0, argv=0x7) at /vol/gnu/src/gcc/gcc-dist/gcc/main.c:35

	Rainer


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-06-07 17:26 -------
Reopening for the last few failures.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|FIXED                       |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|wrong-code                  |ice-on-valid-code


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[bonzini at gnu dot org]

------- Additional Comments From bonzini at gnu dot org  2004-06-08 16:03 -------
Created an attachment (id=6493)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=6493&action=view)
reduced testcase, segfaults at -O1 or higher


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[bonzini at gnu dot org]

------- Additional Comments From bonzini at gnu dot org  2004-06-08 21:29 -------
Created an attachment (id=6495)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=6495&action=view)
even more reduced testcase, independent of patch


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
Attachment #6493 is|0                           |1
           obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[bonzini at gnu dot org]

------- Additional Comments From bonzini at gnu dot org  2004-06-08 21:38 -------
int main() {
  return __builtin_isunordered (1.0,4.0);
}

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mark at codesourcery dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[bonzini at gnu dot org]

------- Additional Comments From bonzini at gnu dot org  2004-06-08 21:41 -------
This fails at -O0 as well:

  int main() {
    return __builtin_isunordered (1.0,4.0);
  }

so this really egregious bug predates my patch: basically, __builtin_is*
functions are broken whenever their arguments are both constants, because we do
not fold them and the alpha expander breaks.  This is quite serious; I do not
know to what extent it applies for 3.4.1, but the patch below seems quite safe.
 I will test it tomorrow on i686-pc-linux-gnu, I would appreciate full testing
on alpha as well.

--- gcc-save/gcc/fold-const.c	2004-05-28 20:30:47.000000000 +0200
+++ gcc/gcc/fold-const.c	2004-06-08 20:00:36.000000000 +0200
@@ -7365,6 +7572,14 @@
     case GT_EXPR:
     case LE_EXPR:
     case GE_EXPR:
+    case UNLE_EXPR:
+    case UNLT_EXPR:
+    case UNGE_EXPR:
+    case UNGT_EXPR:
+    case UNEQ_EXPR:
+    case LTGT_EXPR:
+    case UNORDERED_EXPR:
+    case ORDERED_EXPR:
       /* If one arg is a real or integer constant, put it last.  */
       if (tree_swap_operands_p (arg0, arg1, true))
 	return fold (build2 (swap_tree_comparison (code), type, arg1, arg0));
@@ -9964,8 +10361,22 @@
   tree tem;
   int invert;
 
-  /* From here on, the only cases we handle are when the result is
-     known to be a constant.
+  /* Two real constants can be compared explicitly from the tree
+     code.  */
+  if (TREE_CODE (op0) == REAL_CST && TREE_CODE (op1) == REAL_CST)
+    {
+      REAL_VALUE_TYPE cst0 = TREE_REAL_CST (arg0);
+      REAL_VALUE_TYPE cst1 = TREE_REAL_CST (arg0);
+      int result = real_compare (code, &cst0, &cst1);
+      tem = result ? integer_one_node : integer_zero_node;
+      if (TREE_CODE (type) == BOOLEAN_TYPE)
+        return lang_hooks.truthvalue_conversion (tem);
+      else
+	return tem;
+    }
+
+  /* The only cases we handle here are when the result is known to be a
+     constant.
 
      To compute GT, swap the arguments and do LT.
      To compute GE, do LT and invert the result.
@@ -9980,9 +10391,6 @@
       code = swap_tree_comparison (code);
     }
 
-  /* Note that it is safe to invert for real values here because we
-     will check below in the one case that it matters.  */
-
   tem = NULL_TREE;
   invert = 0;
   if (code == NE_EXPR || code == GE_EXPR)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[ro at techfak dot uni-bielefeld dot de]

------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2004-06-09 19:18 -------
Subject: Re:  [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

bonzini at gnu dot org writes:

>  I will test it tomorrow on i686-pc-linux-gnu, I would appreciate full testing
> on alpha as well.

Mainline bootstrap on alpha-dec-osf{4.0f, 5.1b} in progress ...

	Rainer


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[ro at techfak dot uni-bielefeld dot de]

------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2004-06-09 19:43 -------
Subject: Re:  [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

The patch doesn't compile on mainline at all:

/vol/gnu/src/gcc/gcc-dist/gcc/fold-const.c: In function `fold_relational_const':
/vol/gnu/src/gcc/gcc-dist/gcc/fold-const.c:9936: error: `arg0' undeclared (first use in this function)
/vol/gnu/src/gcc/gcc-dist/gcc/fold-const.c:9936: error: (Each undeclared identifier is reported only once
/vol/gnu/src/gcc/gcc-dist/gcc/fold-const.c:9936: error: for each function it appears in.)
make[2]: *** [fold-const.o] Error 1

	Rainer


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822

[Bug target/15822] [3.5 Regression] Several alpha testsuite regressions: gcc.c-torture/execute/ieee/fp-cmp-[458].c

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-06-09 21:05 -------
Fixed by:
        * expmed.c (emit_store_flag): Cope with FLOAT_STORE_FLAG_VALUE.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15822