[Bug target/29746] New: gcc fails to bootstrap on sh4-*-linux-gnu

[Bug target/29746] New: gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

The bootstrap on sh4-unknown-linux-gnu fails during stage2 with

../../TMP/trunk/gcc/ifcvt.c: In function 'noce_try_store_flag_constants':
../../TMP/trunk/gcc/ifcvt.c:1036: error: insn does not satisfy its constraints:
(insn 871 901 872 42 ../../TMP/trunk/gcc/ifcvt.c:977 (set (reg:SI 147 t)
        (eq:SI (reg:SI 1 r1)
            (const_int 1 [0x1]))) 1 {cmpeqsi_t} (nil)
    (nil))
../../TMP/trunk/gcc/ifcvt.c:1036: internal compiler error: in
copyprop_hardreg_forward_1, at regrename.c:1614


-- 
           Summary: gcc fails to bootstrap on sh4-*-linux-gnu
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: kkojima at gcc dot gnu dot org
 GCC build triplet: sh4-*-linux-gnu
  GCC host triplet: sh4-*-linux-gnu
GCC target triplet: sh4-*-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #1 from kkojima at gcc dot gnu dot org  2006-11-07 02:35 -------
Created an attachment (id=12556)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12556&action=view)
a reduced test case


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #2 from kkojima at gcc dot gnu dot org  2006-11-07 02:37 -------
I've attached a reduced test case which fails also on the cross compiler
with -O2:

foo.c:69: error: insn does not satisfy its constraints:
(insn 232 235 233 9 (set (reg:SI 147 t)
        (eq:SI (reg:SI 8 r8 [orig:169 diff ] [169])
            (const_int 1 [0x1]))) 1 {cmpeqsi_t} (nil)
    (nil))

The .greg dump for it has a cbranchdi4_i insn

(jump_insn:HI 61 222 155 8 (parallel [
            (set (pc)
                (if_then_else (eq (reg/v:DI 8 r8 [orig:169 diff ] [169])
                        (reg:DI 1 r1))
                    (label_ref:SI 82)
                    (pc)))
            (clobber (scratch:SI))
            (clobber (reg:SI 147 t))
        ]) 6 {cbranchdi4_i} (nil)
    (expr_list:REG_BR_PROB (const_int 5000 [0x1388])
        (nil)))

which is transformed to insns

(insn 232 235 233 9 (set (reg:SI 147 t)
        (eq:SI (reg:SI 8 r8 [orig:169 diff ] [169])
            (const_int 1 [0x1]))) 1 {cmpeqsi_t} (nil)
    (nil))

(jump_insn 233 232 237 9 (set (pc)
        (if_then_else (ne (reg:SI 147 t)
                (const_int 0 [0x0]))
            (label_ref 82)
            (pc))) 207 {branch_true} (nil)
    (expr_list:REG_BR_PROB (const_int 10000 [0x2710])
        (nil)))

at .flow2.  With the change of cbranchdi4_i insn_and_split like as

--- ORIG/trunk/gcc/config/sh/sh.md      2006-11-05 10:02:45.000000000 +0900
+++ LOCAL/trunk/gcc/config/sh/sh.md     2006-11-07 09:45:27.000000000 +0900
@@ -713,7 +713,7 @@
 (define_insn_and_split "cbranchdi4_i"
   [(set (pc)
        (if_then_else (match_operator 0 "comparison_operator"
-                       [(match_operand:DI 1 "arith_operand" "r,r")
+                       [(match_operand:DI 1 "arith_operand" "r,z")
                         (match_operand:DI 2 "arith_operand" "rN,i")])
                      (label_ref (match_operand 3 "" ""))
                      (pc)))

the reduced test case doesn't fail but the original ifcvt.c fails
with -O2:

../../TMP/trunk/gcc/ifcvt.c:1036: error: unable to find a register to spill in
class 'R0_REGS'
../../TMP/trunk/gcc/ifcvt.c:1036: error: this is the insn:
(jump_insn:HI 324 339 655 42 ../../TMP/trunk/gcc/ifcvt.c:977 (parallel [
            (set (pc)
                (if_then_else (eq (reg/v:DI 172 [ diff ])
                        (const_int 1 [0x1]))
                    (label_ref:SI 345)
                    (pc)))
            (clobber (scratch:SI))
            (clobber (reg:SI 147 t))
        ]) 6 {cbranchdi4_i} (nil)
    (expr_list:REG_UNUSED (reg:SI 147 t)
        (expr_list:REG_UNUSED (scratch:SI)
            (expr_list:REG_BR_PROB (const_int 5000 [0x1388])
                (nil)))))
../../TMP/trunk/gcc/ifcvt.c:1036: internal compiler error: in spill_failure, at
reload1.c:1938


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #3 from kkojima at gcc dot gnu dot org  2006-11-07 02:39 -------
Created an attachment (id=12558)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12558&action=view)
a reduced test case for the spill failure


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #4 from kkojima at gcc dot gnu dot org  2006-11-07 02:40 -------
If the comparison with a constant is removed from cbranchdi4_i,
the 2nd test case and the original ifcvt.c are compiled successfully.

--- ORIG/trunk/gcc/config/sh/sh.md      2006-11-05 10:02:45.000000000 +0900
+++ TMP/trunk/gcc/config/sh/sh.md       2006-11-07 09:34:22.000000000 +0900
@@ -713,11 +713,11 @@
 (define_insn_and_split "cbranchdi4_i"
   [(set (pc)
        (if_then_else (match_operator 0 "comparison_operator"
-                       [(match_operand:DI 1 "arith_operand" "r,r")
-                        (match_operand:DI 2 "arith_operand" "rN,i")])
+                       [(match_operand:DI 1 "arith_operand" "r")
+                        (match_operand:DI 2 "arith_operand" "rN")])
                      (label_ref (match_operand 3 "" ""))
                      (pc)))
-   (clobber (match_scratch:SI 4 "=X,&r"))
+   (clobber (match_scratch:SI 4 "=X"))
    (clobber (reg:SI T_REG))]
   "TARGET_CBRANCHDI4"
   "#"


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #5 from kkojima at gcc dot gnu dot org  2006-11-07 02:46 -------
I'd like to add Joern to the list.


-- 

kkojima at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |amylaar at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #6 from amylaar at gcc dot gnu dot org  2006-11-07 15:57 -------
Created an attachment (id=12563)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12563&action=view)
current patch for software floating point

In the framework of the software floating point patches, I make the early
emitting of cbranchdi as separate branches the default.

However, this patch set has still some issues - there are patches to generic
code,
and I also see ICE regressions for some of the tls tests on sh-elf.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #7 from amylaar at gcc dot gnu dot org  2006-11-07 17:05 -------
Created an attachment (id=12564)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12564&action=view)
I have a new patch for sched-deps.c now.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #8 from kkojima at gcc dot gnu dot org  2006-11-08 23:53 -------
I don't see ICEs for the tls tests on my sh-elf build with
your patches in #6 and #7.  Does the new patch in #7 fix them?

I've confirmed that the trunk bootstraps successfully with
the patch in #4 and a one-liner to suppress a bogus compile
time warning.  Should I send it to the list as a workaround
this PR?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #9 from amylaar at gcc dot gnu dot org  2006-11-09 16:26 -------
Created an attachment (id=12576)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12576&action=view)
Amended sched-deps.c patch

It occurend to me that the previous patch would not do the right thing when
the set is above an existing SCHED_GROUP_P group, with at least one other
instruction intervening, while the use is below the group.
In that case, all the dependencies of the old SCHED_GROUP_P group need to
be transferred to the new group leader.


-- 

amylaar at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #12564|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #10 from amylaar at gcc dot gnu dot org  2006-11-09 16:33 -------
(In reply to comment #8)
> I don't see ICEs for the tls tests on my sh-elf build with
> your patches in #6 and #7.  Does the new patch in #7 fix them?

Yes, but I am still seeing these changes:

> FAIL: g++.dg/opt/reg-stack2.C (internal compiler error)
> FAIL: g++.dg/opt/reg-stack2.C (test for excess errors)
189a189,190
> FAIL: g++.dg/opt/reg-stack2.C (internal compiler error)
> FAIL: g++.dg/opt/reg-stack2.C (test for excess errors)
202d202
< FAIL: ext/pb_ds/regression/hash_data_map_rand.cc execution test



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #11 from amylaar at gcc dot gnu dot org  2006-11-09 17:28 -------
Created an attachment (id=12580)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12580&action=view)
current software floating point patch

I am testing this patch now.


-- 

amylaar at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #12563|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #12 from amylaar at gcc dot gnu dot org  2006-11-09 17:37 -------
(In reply to comment #8)
> I don't see ICEs for the tls tests on my sh-elf build with
> your patches in #6 and #7.  Does the new patch in #7 fix them?
> 
> I've confirmed that the trunk bootstraps successfully with
> the patch in #4 and a one-liner to suppress a bogus compile
> time warning.  Should I send it to the list as a workaround
> this PR?

The problem looks like another instance of the scheduler bug.
Could you try the original cbranchdi_i pattern with the patch
from comment #9?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #13 from kkojima at gcc dot gnu dot org  2006-11-09 22:42 -------
I've tested the patch in #9 with the trunk (rev. 118619) and
got the same ICE in #2.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #14 from amylaar at gcc dot gnu dot org  2006-11-10 15:48 -------
Created an attachment (id=12587)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12587&action=view)
testsuite patch

This patch is needed additionally for the patch from #11.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[amylaar at gcc dot gnu dot org]

------- Comment #15 from amylaar at gcc dot gnu dot org  2006-11-10 17:26 -------
Created an attachment (id=12589)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12589&action=view)
current softfp patch

I've been able to reproduce the 'does not satisfy contraint' ICE using
-O2 xxx.i -w -mno-expand-cbranchdi -m2 .

I have amended expand_cbranchdi4 so that it properly uses the scratch register.


-- 

amylaar at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #12580|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #16 from kkojima at gcc dot gnu dot org  2006-11-11 11:08 -------
I'm testing the new softfp patch with bootstrap on sh4-unknown-linux-gnu.
It isn't finished yet but already passes the compilation of ifcvt.c
at stage2.  No new failures on the regtest for x86 cross sh4-linux.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[patchapp at dberlin dot org]

------- Comment #17 from patchapp at dberlin dot org  2007-01-07 00:10 -------
Subject: Bug number PR target/29746

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2007-01/msg00494.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #18 from kkojima at gcc dot gnu dot org  2007-02-06 23:12 -------
Subject: Bug 29746

Author: kkojima
Date: Tue Feb  6 23:12:35 2007
New Revision: 121671

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=121671
Log:
        PR target/29746
        * config/sh/sh.c (expand_cbranchdi4): Use scratch register
        properly.
        (sh_initialize_trampoline): Add parentheses to avoid a warning.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/sh/sh.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746

[Bug target/29746] gcc fails to bootstrap on sh4-*-linux-gnu

[kkojima at gcc dot gnu dot org]

------- Comment #19 from kkojima at gcc dot gnu dot org  2007-02-06 23:29 -------
Fixed.


-- 

kkojima at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29746