* Re: [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
@ 2016-09-11 20:53 Bernd Edlinger
2016-09-11 21:11 ` Peter Bergner
0 siblings, 1 reply; 11+ messages in thread
From: Bernd Edlinger @ 2016-09-11 20:53 UTC (permalink / raw)
To: Andrew Pinski
Cc: Peter Bergner, GCC Patches, Jeff Law, Vladimir Makarov, Bill Schmidt
FYI: I have a patch for the aarch64 regression here:
https://gcc.gnu.org/bugzilla/attachment.cgi?id=39600
It passes bootstrap and reg-testing on x86_64-linux-gnu.
Also the mentioned aarch64 and powerpc test cases
pass manually in a cross-compiler, but I cannot do the
boot-strap on powerpc or aarch64 by myself.
So It would be nice if one of you would take the time
and try that patch.
Thanks
Bernd.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-11 20:53 [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem Bernd Edlinger
@ 2016-09-11 21:11 ` Peter Bergner
2016-09-12 13:24 ` Kyrill Tkachov
0 siblings, 1 reply; 11+ messages in thread
From: Peter Bergner @ 2016-09-11 21:11 UTC (permalink / raw)
To: Bernd Edlinger
Cc: Andrew Pinski, GCC Patches, Jeff Law, Vladimir Makarov, Bill Schmidt
On 9/11/16 3:35 PM, Bernd Edlinger wrote:
> FYI: I have a patch for the aarch64 regression here:
> https://gcc.gnu.org/bugzilla/attachment.cgi?id=39600
>
> It passes bootstrap and reg-testing on x86_64-linux-gnu.
Thanks for debugging and fixing this!
> Also the mentioned aarch64 and powerpc test cases
> pass manually in a cross-compiler, but I cannot do the
> boot-strap on powerpc or aarch64 by myself.
>
> So It would be nice if one of you would take the time
> and try that patch.
I can confirm that is bootstraps and regtests with no regressions
on powerpc64le-linux and that the pr77289.c test case does pass.
Peter
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-11 21:11 ` Peter Bergner
@ 2016-09-12 13:24 ` Kyrill Tkachov
2016-09-12 14:05 ` [PATCH] Fix aarch64 fallout of " Bernd Edlinger
0 siblings, 1 reply; 11+ messages in thread
From: Kyrill Tkachov @ 2016-09-12 13:24 UTC (permalink / raw)
To: Peter Bergner, Bernd Edlinger
Cc: Andrew Pinski, GCC Patches, Jeff Law, Vladimir Makarov, Bill Schmidt
On 11/09/16 21:53, Peter Bergner wrote:
> On 9/11/16 3:35 PM, Bernd Edlinger wrote:
>> FYI: I have a patch for the aarch64 regression here:
>> https://gcc.gnu.org/bugzilla/attachment.cgi?id=39600
>>
>> It passes bootstrap and reg-testing on x86_64-linux-gnu.
>
> Thanks for debugging and fixing this!
>
>
>> Also the mentioned aarch64 and powerpc test cases
>> pass manually in a cross-compiler, but I cannot do the
>> boot-strap on powerpc or aarch64 by myself.
>>
>> So It would be nice if one of you would take the time
>> and try that patch.
>
> I can confirm that is bootstraps and regtests with no regressions
> on powerpc64le-linux and that the pr77289.c test case does pass.
>
The patch also passes bootstrap and testing on aarch64-none-linux-gnu.
The ICEs I was seeing go away.
Thanks!
Kyrill
> Peter
>
>
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH] Fix aarch64 fallout of [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-12 13:24 ` Kyrill Tkachov
@ 2016-09-12 14:05 ` Bernd Edlinger
2016-09-12 21:54 ` Jeff Law
2016-09-13 20:03 ` Jeff Law
0 siblings, 2 replies; 11+ messages in thread
From: Bernd Edlinger @ 2016-09-12 14:05 UTC (permalink / raw)
To: Kyrill Tkachov, Peter Bergner
Cc: Andrew Pinski, GCC Patches, Jeff Law, Vladimir Makarov, Bill Schmidt
[-- Attachment #1: Type: text/plain, Size: 158 bytes --]
Hi,
the attached patch boot-straps and reg-tests cleanly
on x86_64-pc-linux-gnu, powerpc and aarch64 targets.
Is it OK for trunk?
Thanks
Bernd.
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: patch-pr77289.diff --]
[-- Type: text/x-patch; name="patch-pr77289.diff", Size: 4027 bytes --]
2016-09-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
PR rtl-optimization/77289
* lra-constraints.c (get_final_hard_regno): Removed.
(get_hard_regno): Add new parameter final_p.
(get_reg_class): Directly call lra_get_elimination_hard_regno.
(operands_match_p): Adjust call to get_hard_regno.
(uses_hard_regs_p): Likewise.
(process_alt_operands): Likewise.
Index: gcc/lra-constraints.c
===================================================================
--- gcc/lra-constraints.c (revision 240073)
+++ gcc/lra-constraints.c (working copy)
@@ -182,27 +182,15 @@ get_try_hard_regno (int regno)
return ira_class_hard_regs[rclass][0];
}
-/* Return the final hard regno which will be after elimination.
- We do this because the final hard regno could have a different class. */
-static int
-get_final_hard_regno (int regno)
-{
- if (! HARD_REGISTER_NUM_P (regno))
- regno = lra_get_regno_hard_regno (regno);
- if (regno < 0)
- return regno;
- return lra_get_elimination_hard_regno (regno);
-}
-
/* Return the hard regno of X after removing its subreg. If X is not
a register or a subreg of a register, return -1. If X is a pseudo,
- use its assignment. We do not process register eliminiations while
- matching constraints. See PR77289. */
+ use its assignment. If FINAL_P return the final hard regno which will
+ be after elimination. */
static int
-get_hard_regno (rtx x)
+get_hard_regno (rtx x, bool final_p)
{
rtx reg;
- int offset, hard_regno;
+ int hard_regno;
reg = x;
if (SUBREG_P (x))
@@ -213,11 +201,12 @@ static int
hard_regno = lra_get_regno_hard_regno (hard_regno);
if (hard_regno < 0)
return -1;
- offset = 0;
+ if (final_p)
+ hard_regno = lra_get_elimination_hard_regno (hard_regno);
if (SUBREG_P (x))
- offset += subreg_regno_offset (hard_regno, GET_MODE (reg),
- SUBREG_BYTE (x), GET_MODE (x));
- return hard_regno + offset;
+ hard_regno += subreg_regno_offset (hard_regno, GET_MODE (reg),
+ SUBREG_BYTE (x), GET_MODE (x));
+ return hard_regno;
}
/* If REGNO is a hard register or has been allocated a hard register,
@@ -229,11 +218,11 @@ get_reg_class (int regno)
{
int hard_regno;
- if ((hard_regno = regno) >= FIRST_PSEUDO_REGISTER)
+ if (! HARD_REGISTER_NUM_P (hard_regno = regno))
hard_regno = lra_get_regno_hard_regno (regno);
if (hard_regno >= 0)
{
- hard_regno = get_final_hard_regno (hard_regno);
+ hard_regno = lra_get_elimination_hard_regno (hard_regno);
return REGNO_REG_CLASS (hard_regno);
}
if (regno >= new_regno_start)
@@ -694,7 +683,7 @@ operands_match_p (rtx x, rtx y, int y_hard_regno)
{
int j;
- i = get_hard_regno (x);
+ i = get_hard_regno (x, false);
if (i < 0)
goto slow;
@@ -1713,7 +1702,7 @@ uses_hard_regs_p (rtx x, HARD_REG_SET set)
if (REG_P (x))
{
- x_hard_regno = get_final_hard_regno (REGNO (x));
+ x_hard_regno = get_hard_regno (x, true);
return (x_hard_regno >= 0
&& overlaps_hard_reg_set_p (set, mode, x_hard_regno));
}
@@ -1840,7 +1829,7 @@ process_alt_operands (int only_alternative)
op = no_subreg_reg_operand[nop] = *curr_id->operand_loc[nop];
/* The real hard regno of the operand after the allocation. */
- hard_regno[nop] = get_hard_regno (op);
+ hard_regno[nop] = get_hard_regno (op, true);
operand_reg[nop] = reg = op;
biggest_mode[nop] = GET_MODE (op);
@@ -2001,7 +1990,7 @@ process_alt_operands (int only_alternative)
lra_assert (nop > m);
this_alternative_matches = m;
- m_hregno = get_hard_regno (*curr_id->operand_loc[m]);
+ m_hregno = get_hard_regno (*curr_id->operand_loc[m], false);
/* We are supposed to match a previous operand.
If we do, we win if that one did. If we do
not, count both of the operands as losers.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] Fix aarch64 fallout of [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-12 14:05 ` [PATCH] Fix aarch64 fallout of " Bernd Edlinger
@ 2016-09-12 21:54 ` Jeff Law
2016-09-12 22:06 ` Bernd Edlinger
2016-09-13 20:03 ` Jeff Law
1 sibling, 1 reply; 11+ messages in thread
From: Jeff Law @ 2016-09-12 21:54 UTC (permalink / raw)
To: Bernd Edlinger, Kyrill Tkachov, Peter Bergner
Cc: Andrew Pinski, GCC Patches, Vladimir Makarov, Bill Schmidt
On 09/12/2016 07:43 AM, Bernd Edlinger wrote:
> Hi,
>
> the attached patch boot-straps and reg-tests cleanly
> on x86_64-pc-linux-gnu, powerpc and aarch64 targets.
>
> Is it OK for trunk?
Unfortunately, there's no explanation for why the original code was
wrong. Yes, I can clearly see you're trying to fix an ICE, but what
about the code's behaviour was wrong?
jeff
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] Fix aarch64 fallout of [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-12 21:54 ` Jeff Law
@ 2016-09-12 22:06 ` Bernd Edlinger
0 siblings, 0 replies; 11+ messages in thread
From: Bernd Edlinger @ 2016-09-12 22:06 UTC (permalink / raw)
To: Jeff Law, Kyrill Tkachov, Peter Bergner
Cc: Andrew Pinski, GCC Patches, Vladimir Makarov, Bill Schmidt
On 09/12/16 23:45, Jeff Law wrote:
> On 09/12/2016 07:43 AM, Bernd Edlinger wrote:
>> Hi,
>>
>> the attached patch boot-straps and reg-tests cleanly
>> on x86_64-pc-linux-gnu, powerpc and aarch64 targets.
>>
>> Is it OK for trunk?
> Unfortunately, there's no explanation for why the original code was
> wrong. Yes, I can clearly see you're trying to fix an ICE, but what
> about the code's behaviour was wrong?
>
> jeff
>
Sorry,
everything is just folding, and continuing Peter's cleanup,
except here:
@@ -1840,7 +1829,7 @@ process_alt_operands (int only_alternative)
op = no_subreg_reg_operand[nop] = *curr_id->operand_loc[nop];
/* The real hard regno of the operand after the allocation. */
- hard_regno[nop] = get_hard_regno (op);
+ hard_regno[nop] = get_hard_regno (op, true);
operand_reg[nop] = reg = op;
biggest_mode[nop] = GET_MODE (op);
note the comment above. This restores the state before Peter's patch
here, but only here, not in the operand matching.
Bernd.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] Fix aarch64 fallout of [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-12 14:05 ` [PATCH] Fix aarch64 fallout of " Bernd Edlinger
2016-09-12 21:54 ` Jeff Law
@ 2016-09-13 20:03 ` Jeff Law
1 sibling, 0 replies; 11+ messages in thread
From: Jeff Law @ 2016-09-13 20:03 UTC (permalink / raw)
To: Bernd Edlinger, Kyrill Tkachov, Peter Bergner
Cc: Andrew Pinski, GCC Patches, Vladimir Makarov, Bill Schmidt
On 09/12/2016 07:43 AM, Bernd Edlinger wrote:
> Hi,
>
> the attached patch boot-straps and reg-tests cleanly
> on x86_64-pc-linux-gnu, powerpc and aarch64 targets.
>
> Is it OK for trunk?
>
>
> Thanks
> Bernd.
>
>
> patch-pr77289.diff
>
>
> 2016-09-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
>
> PR rtl-optimization/77289
> * lra-constraints.c (get_final_hard_regno): Removed.
> (get_hard_regno): Add new parameter final_p.
> (get_reg_class): Directly call lra_get_elimination_hard_regno.
> (operands_match_p): Adjust call to get_hard_regno.
> (uses_hard_regs_p): Likewise.
> (process_alt_operands): Likewise.
A bit painful to follow as there's cleanup mixed in with the bugfix.
Thanks for pointing out the key change in the follow-up, that helped
considerably.
OK.
jeff
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
@ 2016-08-31 4:23 Peter Bergner
2016-09-09 22:57 ` Jeff Law
0 siblings, 1 reply; 11+ messages in thread
From: Peter Bergner @ 2016-08-31 4:23 UTC (permalink / raw)
To: Vladimir Makarov; +Cc: GCC Patches, Bill Schmidt
PR77289 exposes a latent problem with LRA constraint matching. In the buggy
test cases, LRA performs a speculative register elimination before checking
operands for matching constraints. With the elimination, the operands
appear to match. However, when we call check_rtl() which attempts to
recognize the instruction, the reg above it not eliminated leading to a
"insn does not satisfy its constraints" ICE.
This patch fixes the problem by not performing register elimination during
constraint matching. operands_match_p() uses get_hard_reg() to grab a
REG's hard reg number, so I have removed get_hard_reg()'s call to
get_final_hard_regno() which performs the register elimination, which
fixes the bug. uses_hard_regs_p() is the other caller of get_hard_regno()
and it still needs register elimination to be called, so I have changed
it to call get_final_hard_regno() instead. Since uses_hard_regs_p()
may call get_final_hard_regno() with a pseudo, I have added support
for mapping those to hard reg numbers before performing the register
elimination.
This has passed bootstrap and regtesting with no regressions.
Ok for mainline?
Peter
gcc/
PR rtl-optimization/77289
* lra-constraints.c (get_final_hard_regno): Add support for non hard
register numbers. Remove support for subregs.
(get_hard_regno): Use SUBREG_P. Don't call get_final_hard_regno().
(get_reg_class): Delete removed get_final_hard_regno() argument.
(uses_hard_regs_p): Call get_final_hard_regno().
gcc/testsuite/
PR rtl-optimization/77289
* gcc.target/powerpc/pr77289.c: New test.
Index: gcc/lra-constraints.c
===================================================================
--- gcc/lra-constraints.c (revision 239866)
+++ gcc/lra-constraints.c (working copy)
@@ -182,21 +182,22 @@ get_try_hard_regno (int regno)
return ira_class_hard_regs[rclass][0];
}
-/* Return final hard regno (plus offset) which will be after
- elimination. We do this for matching constraints because the final
- hard regno could have a different class. */
+/* Return the final hard regno which will be after elimination.
+ We do this because the final hard regno could have a different class. */
static int
-get_final_hard_regno (int hard_regno, int offset)
+get_final_hard_regno (int regno)
{
- if (hard_regno < 0)
- return hard_regno;
- hard_regno = lra_get_elimination_hard_regno (hard_regno);
- return hard_regno + offset;
+ if (! HARD_REGISTER_NUM_P (regno))
+ regno = lra_get_regno_hard_regno (regno);
+ if (regno < 0)
+ return regno;
+ return lra_get_elimination_hard_regno (regno);
}
-/* Return hard regno of X after removing subreg and making
- elimination. If X is not a register or subreg of register, return
- -1. For pseudo use its assignment. */
+/* Return the hard regno of X after removing its subreg. If X is not
+ a register or a subreg of a register, return -1. If X is a pseudo,
+ use its assignment. We do not process register eliminiations while
+ matching constraints. See PR77289. */
static int
get_hard_regno (rtx x)
{
@@ -204,19 +205,19 @@ get_hard_regno (rtx x)
int offset, hard_regno;
reg = x;
- if (GET_CODE (x) == SUBREG)
+ if (SUBREG_P (x))
reg = SUBREG_REG (x);
if (! REG_P (reg))
return -1;
- if ((hard_regno = REGNO (reg)) >= FIRST_PSEUDO_REGISTER)
+ if (! HARD_REGISTER_NUM_P (hard_regno = REGNO (reg)))
hard_regno = lra_get_regno_hard_regno (hard_regno);
if (hard_regno < 0)
return -1;
offset = 0;
- if (GET_CODE (x) == SUBREG)
+ if (SUBREG_P (x))
offset += subreg_regno_offset (hard_regno, GET_MODE (reg),
SUBREG_BYTE (x), GET_MODE (x));
- return get_final_hard_regno (hard_regno, offset);
+ return hard_regno + offset;
}
/* If REGNO is a hard register or has been allocated a hard register,
@@ -232,7 +233,7 @@ get_reg_class (int regno)
hard_regno = lra_get_regno_hard_regno (regno);
if (hard_regno >= 0)
{
- hard_regno = get_final_hard_regno (hard_regno, 0);
+ hard_regno = get_final_hard_regno (hard_regno);
return REGNO_REG_CLASS (hard_regno);
}
if (regno >= new_regno_start)
@@ -1712,7 +1713,7 @@ uses_hard_regs_p (rtx x, HARD_REG_SET se
if (REG_P (x))
{
- x_hard_regno = get_hard_regno (x);
+ x_hard_regno = get_final_hard_regno (REGNO (x));
return (x_hard_regno >= 0
&& overlaps_hard_reg_set_p (set, mode, x_hard_regno));
}
Index: gcc/testsuite/gcc.target/powerpc/pr77289.c
===================================================================
--- gcc/testsuite/gcc.target/powerpc/pr77289.c (revision 0)
+++ gcc/testsuite/gcc.target/powerpc/pr77289.c (working copy)
@@ -0,0 +1,31 @@
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_vsx_ok } */
+/* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power7" } } */
+/* { dg-options "-O3 -mcpu=power7 -funroll-loops -ffast-math -mlra -mupdate -fno-auto-inc-dec" } */
+
+/* PR 77289: LRA ICEs due to invalid constraint checking. */
+
+void dummy0 (float *);
+float bar0 (float);
+void
+foo0 (long a, long b)
+{
+ float c[0];
+ b = 0;
+ for (; b < a; b++)
+ c[b] = bar0 (c[b]);
+ dummy0 (c);
+}
+
+void dummy1 (long *);
+long bar1 (long);
+void
+foo1 (long a, long b)
+{
+ long array[128];
+ long *c = array;
+ for (b=0; b < a; b++)
+ c[b] = bar1 (c[b]);
+ dummy1 (c);
+}
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-08-31 4:23 Peter Bergner
@ 2016-09-09 22:57 ` Jeff Law
2016-09-10 6:04 ` Peter Bergner
0 siblings, 1 reply; 11+ messages in thread
From: Jeff Law @ 2016-09-09 22:57 UTC (permalink / raw)
To: Peter Bergner, Vladimir Makarov; +Cc: GCC Patches, Bill Schmidt
On 08/30/2016 10:23 PM, Peter Bergner wrote:
> PR77289 exposes a latent problem with LRA constraint matching. In the buggy
> test cases, LRA performs a speculative register elimination before checking
> operands for matching constraints. With the elimination, the operands
> appear to match. However, when we call check_rtl() which attempts to
> recognize the instruction, the reg above it not eliminated leading to a
> "insn does not satisfy its constraints" ICE.
>
> This patch fixes the problem by not performing register elimination during
> constraint matching. operands_match_p() uses get_hard_reg() to grab a
> REG's hard reg number, so I have removed get_hard_reg()'s call to
> get_final_hard_regno() which performs the register elimination, which
> fixes the bug. uses_hard_regs_p() is the other caller of get_hard_regno()
> and it still needs register elimination to be called, so I have changed
> it to call get_final_hard_regno() instead. Since uses_hard_regs_p()
> may call get_final_hard_regno() with a pseudo, I have added support
> for mapping those to hard reg numbers before performing the register
> elimination.
>
> This has passed bootstrap and regtesting with no regressions.
> Ok for mainline?
>
> Peter
>
> gcc/
> PR rtl-optimization/77289
> * lra-constraints.c (get_final_hard_regno): Add support for non hard
> register numbers. Remove support for subregs.
> (get_hard_regno): Use SUBREG_P. Don't call get_final_hard_regno().
> (get_reg_class): Delete removed get_final_hard_regno() argument.
> (uses_hard_regs_p): Call get_final_hard_regno().
>
> gcc/testsuite/
> PR rtl-optimization/77289
> * gcc.target/powerpc/pr77289.c: New test.
OK.
Jeff
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-09 22:57 ` Jeff Law
@ 2016-09-10 6:04 ` Peter Bergner
2016-09-10 8:34 ` Andrew Pinski
0 siblings, 1 reply; 11+ messages in thread
From: Peter Bergner @ 2016-09-10 6:04 UTC (permalink / raw)
To: Jeff Law; +Cc: Vladimir Makarov, GCC Patches, Bill Schmidt
On 9/9/16 5:51 PM, Jeff Law wrote:
> On 08/30/2016 10:23 PM, Peter Bergner wrote:
>> gcc/
>> PR rtl-optimization/77289
>> * lra-constraints.c (get_final_hard_regno): Add support for non hard
>> register numbers. Remove support for subregs.
>> (get_hard_regno): Use SUBREG_P. Don't call get_final_hard_regno().
>> (get_reg_class): Delete removed get_final_hard_regno() argument.
>> (uses_hard_regs_p): Call get_final_hard_regno().
>>
>> gcc/testsuite/
>> PR rtl-optimization/77289
>> * gcc.target/powerpc/pr77289.c: New test.
> OK.
Committed as revision 240065. Thanks!
Peter
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem
2016-09-10 6:04 ` Peter Bergner
@ 2016-09-10 8:34 ` Andrew Pinski
0 siblings, 0 replies; 11+ messages in thread
From: Andrew Pinski @ 2016-09-10 8:34 UTC (permalink / raw)
To: Peter Bergner; +Cc: Jeff Law, Vladimir Makarov, GCC Patches, Bill Schmidt
On Fri, Sep 9, 2016 at 6:37 PM, Peter Bergner <bergner@vnet.ibm.com> wrote:
> On 9/9/16 5:51 PM, Jeff Law wrote:
>>
>> On 08/30/2016 10:23 PM, Peter Bergner wrote:
>>>
>>> gcc/
>>> PR rtl-optimization/77289
>>> * lra-constraints.c (get_final_hard_regno): Add support for non hard
>>> register numbers. Remove support for subregs.
>>> (get_hard_regno): Use SUBREG_P. Don't call get_final_hard_regno().
>>> (get_reg_class): Delete removed get_final_hard_regno() argument.
>>> (uses_hard_regs_p): Call get_final_hard_regno().
>>>
>>> gcc/testsuite/
>>> PR rtl-optimization/77289
>>> * gcc.target/powerpc/pr77289.c: New test.
>>
>> OK.
>
>
> Committed as revision 240065. Thanks!
This patch caused many regressions on aarch64-linux-gnu:
gcc.c-torture/compile/20020604-1.c -O1 (test for excess errors)
gcc.c-torture/compile/complex-6.c -O2 (test for excess errors)
gcc.c-torture/compile/complex-6.c -O2 -flto -fno-use-linker-plugin
-flto-partition=none (test for excess errors)
gcc.c-torture/compile/complex-6.c -O3 -g (test for excess errors)
gcc.c-torture/compile/complex-6.c -Os (test for excess errors)
gcc.c-torture/compile/pr22422.c -O1 (test for excess errors)
gcc.c-torture/compile/pr22422.c -O2 (test for excess errors)
gcc.c-torture/compile/pr22422.c -O2 -flto -fno-use-linker-plugin
-flto-partition=none (test for excess errors)
gcc.c-torture/compile/pr22422.c -O3 -g (test for excess errors)
gcc.c-torture/compile/pr22422.c -Os (test for excess errors)
gcc.dg/ipa/ipcp-cstagg-5.c (test for excess errors)
gcc.dg/ipa/ipcp-cstagg-6.c (test for excess errors)
gcc.dg/torture/pr39074.c -Os (test for excess errors)
gcc.dg/torture/pr55882.c -O1 (test for excess errors)
gcc.dg/torture/pr55882.c -O2 (test for excess errors)
gcc.dg/torture/pr55882.c -O2 -flto -fno-use-linker-plugin
-flto-partition=none (test for excess errors)
gcc.dg/torture/pr55882.c -O3 -g (test for excess errors)
gcc.dg/torture/pr55882.c -Os (test for excess errors)
gcc.dg/torture/pr57569.c -O1 (test for excess errors)
gcc.dg/torture/pr57569.c -O2 (test for excess errors)
gcc.dg/torture/pr57569.c -O2 -flto -fno-use-linker-plugin
-flto-partition=none (test for excess errors)
gcc.dg/torture/pr57569.c -O2 -flto -fuse-linker-plugin
-fno-fat-lto-objects (test for excess errors)
gcc.dg/torture/pr57569.c -Os (test for excess errors)
gcc.dg/tree-ssa/ssa-fre-13.c (test for excess errors)
gcc.dg/tree-ssa/ssa-fre-42.c (test for excess errors)
An example ICEs are:
/home/jenkins/workspace/BuildThunderX_native_gcc_upstream/gcc/gcc/testsuite/gcc.c-torture/compile/20020604-1.c:91:1:
error: insn does not satisfy its constraints:
(insn 112 111 113 5 (set (reg:DI 7 x7 [196])
(plus:DI (ashift:DI (reg:DI 7 x7 [194])
(const_int 4 [0x4]))
(reg/f:DI 31 sp))) 189 {*add_lsl_di}
(nil))
/home/jenkins/workspace/BuildThunderX_native_gcc_upstream/gcc/gcc/testsuite/gcc.c-torture/compile/20020604-1.c:91:1:
internal compiler error: in extract_constrain_insn, at recog.c:2212
/home/jenkins/workspace/BuildThunderX_native_gcc_upstream/gcc/gcc/testsuite/gcc.c-torture/compile/pr22422.c:29:1:
error: insn does not satisfy its constraints:
(insn 7 6 12 2 (set (mem/f:DI (reg:DI 0 x0 [ b ]) [2 b_2(D)->a+0 S8 A64])
(reg/f:DI 31 sp))
/home/jenkins/workspace/BuildThunderX_native_gcc_upstream/gcc/gcc/testsuite/gcc.c-torture/compile/pr22422.c:25
50 {*movdi_aarch64}
(nil))
Notice how sp is being used there. That register is not valid in that location.
Thanks,
Andrew
>
> Peter
>
>
>
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2016-09-13 19:58 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-09-11 20:53 [PATCH, LRA] Fix PR rtl-optimization 77289, LRA matching constraint problem Bernd Edlinger
2016-09-11 21:11 ` Peter Bergner
2016-09-12 13:24 ` Kyrill Tkachov
2016-09-12 14:05 ` [PATCH] Fix aarch64 fallout of " Bernd Edlinger
2016-09-12 21:54 ` Jeff Law
2016-09-12 22:06 ` Bernd Edlinger
2016-09-13 20:03 ` Jeff Law
-- strict thread matches above, loose matches on Subject: below --
2016-08-31 4:23 Peter Bergner
2016-09-09 22:57 ` Jeff Law
2016-09-10 6:04 ` Peter Bergner
2016-09-10 8:34 ` Andrew Pinski
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).