[PATCH 1/2] [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences

[PATCH 1/2] [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences

[Daniel Cederman]

The sequence
  st
  fdivd / fsqrtd
  std
was generated in some cases with -mfix-ut699 when there was
a st before the div/sqrt. This sequence could trigger the b2bst errata.

Now the following safe sequence is generated instead:
  st
  nop
  fdivd / fsqrtd
  std

gcc/ChangeLog:

2017-11-27  Martin Aberg  <maberg@gaisler.com>

	* config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
        to prevent b2bst errata sequence.
        (sqrtdf2_fix): Likewise.
---
 gcc/config/sparc/sparc.md | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/gcc/config/sparc/sparc.md b/gcc/config/sparc/sparc.md
index 0e45509..18d77bb 100644
--- a/gcc/config/sparc/sparc.md
+++ b/gcc/config/sparc/sparc.md
@@ -6221,10 +6221,10 @@ visl")
 	(div:DF (match_operand:DF 1 "register_operand" "e")
 		(match_operand:DF 2 "register_operand" "e")))]
   "TARGET_FPU && sparc_fix_ut699"
-  "fdivd\t%1, %2, %0\n\tstd\t%0, [%%sp-8]\n\tnop"
+  "fdivd\t%1, %2, %0\n\tnop\n\tstd\t%0, [%%sp-8]\n\tnop"
   [(set_attr "type" "fpdivd")
    (set_attr "fptype" "double")
-   (set_attr "length" "3")])
+   (set_attr "length" "4")])
 
 (define_insn "divsf3"
   [(set (match_operand:SF 0 "register_operand" "=f")
@@ -6473,10 +6473,10 @@ visl")
   [(set (match_operand:DF 0 "register_operand" "=e")
 	(sqrt:DF (match_operand:DF 1 "register_operand" "e")))]
   "TARGET_FPU && sparc_fix_ut699"
-  "fsqrtd\t%1, %0\n\tstd\t%0, [%%sp-8]\n\tnop"
+  "fsqrtd\t%1, %0\n\tnop\n\tstd\t%0, [%%sp-8]\n\tnop"
   [(set_attr "type" "fpsqrtd")
    (set_attr "fptype" "double")
-   (set_attr "length" "3")])
+   (set_attr "length" "4")])
 
 (define_insn "sqrtsf2"
   [(set (match_operand:SF 0 "register_operand" "=f")
-- 
2.9.3

[PATCH 0/2] [SPARC] Fixes for UT699 errata workaround

[Daniel Cederman]

Hello,

This patch series adds two fixes for the UT699 errata workaround.

Daniel Cederman (2):
  [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences
  [SPARC] Recognize the load when accessing the GOT

 gcc/config/sparc/sparc.c  | 8 +++++++-
 gcc/config/sparc/sparc.md | 8 ++++----
 2 files changed, 11 insertions(+), 5 deletions(-)

-- 
2.9.3

[PATCH 2/2] [SPARC] Recognize the load when accessing the GOT

[Daniel Cederman]

Needed for the UT699 errata workaround to function correctly when
compiling with -fPIC.

gcc/ChangeLog:

2017-11-27  Daniel Cederman  <cederman@gaisler.com>

	* config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
        movsi_pic_gotdata_op instruction as a load for the UT699 errata
        workaround.
---
 gcc/config/sparc/sparc.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/gcc/config/sparc/sparc.c b/gcc/config/sparc/sparc.c
index 6e91114..8b11c1a 100644
--- a/gcc/config/sparc/sparc.c
+++ b/gcc/config/sparc/sparc.c
@@ -1321,7 +1321,8 @@ sparc_do_work_around_errata (void)
 	       && NONJUMP_INSN_P (insn)
 	       && (set = single_set (insn)) != NULL_RTX
 	       && GET_MODE_SIZE (GET_MODE (SET_SRC (set))) <= 4
-	       && mem_ref (SET_SRC (set)) != NULL_RTX
+	       && (mem_ref (SET_SRC (set)) != NULL_RTX
+		   || INSN_CODE (insn) == CODE_FOR_movsi_pic_gotdata_op)
 	       && REG_P (SET_DEST (set))
 	       && REGNO (SET_DEST (set)) < 32)
 	{
@@ -1359,6 +1360,11 @@ sparc_do_work_around_errata (void)
 			       && REGNO (src) != REGNO (x)))
 		       && !reg_mentioned_p (x, XEXP (dest, 0)))
 		insert_nop = true;
+
+	      /* GOT accesses uses LD.  */
+	      else if (INSN_CODE (next) == CODE_FOR_movsi_pic_gotdata_op
+		       && !reg_mentioned_p (x, XEXP (XEXP (src, 0), 1)))
+		insert_nop = true;
 	    }
 	}
 
-- 
2.9.3

Re: [PATCH 1/2] [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences

[Eric Botcazou]

> 2017-11-27  Martin Aberg  <maberg@gaisler.com>
> 
> 	* config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
>         to prevent b2bst errata sequence.
>         (sqrtdf2_fix): Likewise.

OK for mainline and 7 branch, thanks.

-- 
Eric Botcazou

Re: [PATCH 2/2] [SPARC] Recognize the load when accessing the GOT

[Eric Botcazou]

> 2017-11-27  Daniel Cederman  <cederman@gaisler.com>
> 
> 	* config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
>         movsi_pic_gotdata_op instruction as a load for the UT699 errata
>         workaround.

OK for mainline, 7 and 6 branches, thanks.

-- 
Eric Botcazou

Re: [PATCH 1/2] [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences

[Eric Botcazou]

> The sequence
>   st
>   fdivd / fsqrtd
>   std
> was generated in some cases with -mfix-ut699 when there was
> a st before the div/sqrt. This sequence could trigger the b2bst errata.
> 
> Now the following safe sequence is generated instead:
>   st
>   nop
>   fdivd / fsqrtd
>   std
> 
> gcc/ChangeLog:
> 
> 2017-11-27  Martin Aberg  <maberg@gaisler.com>
> 
> 	* config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
>         to prevent b2bst errata sequence.
>         (sqrtdf2_fix): Likewise.

But isn't that the change I already rejected back in July?
  https://gcc.gnu.org/ml/gcc-patches/2017-07/msg00496.html

If so, what has changed since then?  If nothing, then please revert.

-- 
Eric Botcazou

Re: [PATCH 1/2] [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences

[Daniel Cederman]

> But isn't that the change I already rejected back in July?

You are right, that was a mistake from our side to submit it again. We 
had rediscovered the sequence using our scanning script and I forgot 
that the sequence was harmless in this case. We will update our scripts 
and revert the patch.

The person here with write access, danielh, is currently traveling, so 
we wont be able to revert it until next week. If it is urgent, could I 
ask you to revert it for us?

Thanks for checking the patches in detail and for improving the formating.

/Daniel C

Re: [PATCH 1/2] [SPARC] Prevent -mfix-ut699 from generating b2bst errata sequences

[Eric Botcazou]

> You are right, that was a mistake from our side to submit it again. We
> had rediscovered the sequence using our scanning script and I forgot
> that the sequence was harmless in this case. We will update our scripts
> and revert the patch.

OK, thanks in advance.

> The person here with write access, danielh, is currently traveling, so
> we wont be able to revert it until next week. If it is urgent, could I
> ask you to revert it for us?

It's not that urgent of course but I've reverted it.

-- 
Eric Botcazou