public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
* [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough
@ 2017-06-27  7:37 Yao Qi
  2017-07-17  9:04 ` Alan Hayward
  0 siblings, 1 reply; 5+ messages in thread
From: Yao Qi @ 2017-06-27  7:37 UTC (permalink / raw)
  To: gdb-patches

Nowadays, regcache_cpy is used where src is read-only and dst is not
read-only, so the regcache_cpy can be simplified to handle this case only.
As a result, regcache::cpy_no_passthrough, which is about two read-only
regcache copy, is no longer used, remove it as well.

Regression tested on x86_64-linux, both native and gdbserver.

gdb:

2017-06-27  Yao Qi  <yao.qi@linaro.org>

	* regcache.c (regcache_cpy): Simplify it.
	(regcache::cpy_no_passthrough): Remove it.
	* regcache.h (cpy_no_passthrough): Remove it.
---
 gdb/regcache.c | 31 ++-----------------------------
 gdb/regcache.h |  2 --
 2 files changed, 2 insertions(+), 31 deletions(-)

diff --git a/gdb/regcache.c b/gdb/regcache.c
index 7eeb737..e8f92d6 100644
--- a/gdb/regcache.c
+++ b/gdb/regcache.c
@@ -388,36 +388,9 @@ regcache_cpy (struct regcache *dst, struct regcache *src)
   gdb_assert (src != NULL && dst != NULL);
   gdb_assert (src->m_descr->gdbarch == dst->m_descr->gdbarch);
   gdb_assert (src != dst);
-  gdb_assert (src->m_readonly_p || dst->m_readonly_p);
+  gdb_assert (src->m_readonly_p && !dst->m_readonly_p);
 
-  if (!src->m_readonly_p)
-    regcache_save (dst, do_cooked_read, src);
-  else if (!dst->m_readonly_p)
-    dst->restore (src);
-  else
-    dst->cpy_no_passthrough (src);
-}
-
-/* Copy/duplicate the contents of a register cache.  Unlike regcache_cpy,
-   which is pass-through, this does not go through to the target.
-   Only values values already in the cache are transferred.  The SRC and DST
-   buffers must not overlap.  */
-
-void
-regcache::cpy_no_passthrough (struct regcache *src)
-{
-  gdb_assert (src != NULL);
-  gdb_assert (src->m_descr->gdbarch == m_descr->gdbarch);
-  /* NOTE: cagney/2002-05-17: Don't let the caller do a no-passthrough
-     move of data into a thread's regcache.  Doing this would be silly
-     - it would mean that regcache->register_status would be
-     completely invalid.  */
-  gdb_assert (m_readonly_p && src->m_readonly_p);
-
-  memcpy (m_registers, src->m_registers,
-	  m_descr->sizeof_cooked_registers);
-  memcpy (m_register_status, src->m_register_status,
-	  m_descr->sizeof_cooked_register_status);
+  dst->restore (src);
 }
 
 struct regcache *
diff --git a/gdb/regcache.h b/gdb/regcache.h
index b416d5e..03c042a 100644
--- a/gdb/regcache.h
+++ b/gdb/regcache.h
@@ -369,8 +369,6 @@ private:
 
   void restore (struct regcache *src);
 
-  void cpy_no_passthrough (struct regcache *src);
-
   enum register_status xfer_part (int regnum, int offset, int len, void *in,
 				  const void *out,
 				  decltype (regcache_raw_read) read,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough
  2017-06-27  7:37 [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough Yao Qi
@ 2017-07-17  9:04 ` Alan Hayward
  2017-07-17 11:32   ` Yao Qi
  0 siblings, 1 reply; 5+ messages in thread
From: Alan Hayward @ 2017-07-17  9:04 UTC (permalink / raw)
  To: Yao Qi; +Cc: gdb-patches, nd


> On 27 Jun 2017, at 08:37, Yao Qi <qiyaoltc@gmail.com> wrote:
> 
> Nowadays, regcache_cpy is used where src is read-only and dst is not
> read-only, so the regcache_cpy can be simplified to handle this case only.
> As a result, regcache::cpy_no_passthrough, which is about two read-only
> regcache copy, is no longer used, remove it as well.
> 

I like the simplification.

But, I don’t think it’s clear that regcache_cpy now only handles the single case
of read-only src with !read-only dest. It only becomes clear after reading the
gdb_asserts.

Would it be better to remove regcache_cpy completely?
There are only three places that use it, and they would call (for example)
get_current_regcache ()->restore (scratch) instead.
That would make the caller code clearer, and would remove an outside the class
regcache_ function.

If not, then there should be a comment above regcache_cpy stating the
restrictions.

> Regression tested on x86_64-linux, both native and gdbserver.
> 
> gdb:
> 
> 2017-06-27  Yao Qi  <yao.qi@linaro.org>
> 
> 	* regcache.c (regcache_cpy): Simplify it.
> 	(regcache::cpy_no_passthrough): Remove it.
> 	* regcache.h (cpy_no_passthrough): Remove it.
> ---
> gdb/regcache.c | 31 ++-----------------------------
> gdb/regcache.h |  2 --
> 2 files changed, 2 insertions(+), 31 deletions(-)
> 
> diff --git a/gdb/regcache.c b/gdb/regcache.c
> index 7eeb737..e8f92d6 100644
> --- a/gdb/regcache.c
> +++ b/gdb/regcache.c
> @@ -388,36 +388,9 @@ regcache_cpy (struct regcache *dst, struct regcache *src)
>   gdb_assert (src != NULL && dst != NULL);
>   gdb_assert (src->m_descr->gdbarch == dst->m_descr->gdbarch);
>   gdb_assert (src != dst);
> -  gdb_assert (src->m_readonly_p || dst->m_readonly_p);
> +  gdb_assert (src->m_readonly_p && !dst->m_readonly_p);
> 
> -  if (!src->m_readonly_p)
> -    regcache_save (dst, do_cooked_read, src);
> -  else if (!dst->m_readonly_p)
> -    dst->restore (src);
> -  else
> -    dst->cpy_no_passthrough (src);
> -}
> -
> -/* Copy/duplicate the contents of a register cache.  Unlike regcache_cpy,
> -   which is pass-through, this does not go through to the target.
> -   Only values values already in the cache are transferred.  The SRC and DST
> -   buffers must not overlap.  */
> -
> -void
> -regcache::cpy_no_passthrough (struct regcache *src)
> -{
> -  gdb_assert (src != NULL);
> -  gdb_assert (src->m_descr->gdbarch == m_descr->gdbarch);
> -  /* NOTE: cagney/2002-05-17: Don't let the caller do a no-passthrough
> -     move of data into a thread's regcache.  Doing this would be silly
> -     - it would mean that regcache->register_status would be
> -     completely invalid.  */
> -  gdb_assert (m_readonly_p && src->m_readonly_p);
> -
> -  memcpy (m_registers, src->m_registers,
> -	  m_descr->sizeof_cooked_registers);
> -  memcpy (m_register_status, src->m_register_status,
> -	  m_descr->sizeof_cooked_register_status);
> +  dst->restore (src);
> }
> 
> struct regcache *
> diff --git a/gdb/regcache.h b/gdb/regcache.h
> index b416d5e..03c042a 100644
> --- a/gdb/regcache.h
> +++ b/gdb/regcache.h
> @@ -369,8 +369,6 @@ private:
> 
>   void restore (struct regcache *src);
> 
> -  void cpy_no_passthrough (struct regcache *src);
> -
>   enum register_status xfer_part (int regnum, int offset, int len, void *in,
> 				  const void *out,
> 				  decltype (regcache_raw_read) read,
> -- 
> 1.9.1
> 

Alan.


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough
  2017-07-17  9:04 ` Alan Hayward
@ 2017-07-17 11:32   ` Yao Qi
  2017-07-18  9:01     ` Alan Hayward
  0 siblings, 1 reply; 5+ messages in thread
From: Yao Qi @ 2017-07-17 11:32 UTC (permalink / raw)
  To: Alan Hayward; +Cc: gdb-patches, nd

Alan Hayward <Alan.Hayward@arm.com> writes:

> If not, then there should be a comment above regcache_cpy stating the
> restrictions.

I updated the comments to regcache_dup and regcache_cpy.

-- 
Yao (齐尧)
From 4ff91e90cad73f9420f8ca37fc7e22f465ebcb88 Mon Sep 17 00:00:00 2001
From: Yao Qi <yao.qi@linaro.org>
Date: Mon, 19 Jun 2017 22:45:12 +0100
Subject: [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough

Nowadays, regcache_cpy is used where src is read-only and dst is not
read-only, so the regcache_cpy can be simplified to handle this case only.
As a result, regcache::cpy_no_passthrough, which is about two read-only
regcache copy, is no longer used, remove it as well.

gdb:

2017-07-17  Yao Qi  <yao.qi@linaro.org>

	* regcache.c (regcache_cpy): Simplify it.
	(regcache::cpy_no_passthrough): Remove it.
	* regcache.h (cpy_no_passthrough): Remove it.
	(regcache_dup, regcache_cpy): Update comments.

diff --git a/gdb/regcache.c b/gdb/regcache.c
index 7eeb737..e8f92d6 100644
--- a/gdb/regcache.c
+++ b/gdb/regcache.c
@@ -388,36 +388,9 @@ regcache_cpy (struct regcache *dst, struct regcache *src)
   gdb_assert (src != NULL && dst != NULL);
   gdb_assert (src->m_descr->gdbarch == dst->m_descr->gdbarch);
   gdb_assert (src != dst);
-  gdb_assert (src->m_readonly_p || dst->m_readonly_p);
+  gdb_assert (src->m_readonly_p && !dst->m_readonly_p);
 
-  if (!src->m_readonly_p)
-    regcache_save (dst, do_cooked_read, src);
-  else if (!dst->m_readonly_p)
-    dst->restore (src);
-  else
-    dst->cpy_no_passthrough (src);
-}
-
-/* Copy/duplicate the contents of a register cache.  Unlike regcache_cpy,
-   which is pass-through, this does not go through to the target.
-   Only values values already in the cache are transferred.  The SRC and DST
-   buffers must not overlap.  */
-
-void
-regcache::cpy_no_passthrough (struct regcache *src)
-{
-  gdb_assert (src != NULL);
-  gdb_assert (src->m_descr->gdbarch == m_descr->gdbarch);
-  /* NOTE: cagney/2002-05-17: Don't let the caller do a no-passthrough
-     move of data into a thread's regcache.  Doing this would be silly
-     - it would mean that regcache->register_status would be
-     completely invalid.  */
-  gdb_assert (m_readonly_p && src->m_readonly_p);
-
-  memcpy (m_registers, src->m_registers,
-	  m_descr->sizeof_cooked_registers);
-  memcpy (m_register_status, src->m_register_status,
-	  m_descr->sizeof_cooked_register_status);
+  dst->restore (src);
 }
 
 struct regcache *
diff --git a/gdb/regcache.h b/gdb/regcache.h
index b416d5e..aa64a00 100644
--- a/gdb/regcache.h
+++ b/gdb/regcache.h
@@ -369,8 +369,6 @@ private:
 
   void restore (struct regcache *src);
 
-  void cpy_no_passthrough (struct regcache *src);
-
   enum register_status xfer_part (int regnum, int offset, int len, void *in,
 				  const void *out,
 				  decltype (regcache_raw_read) read,
@@ -415,13 +413,12 @@ private:
   regcache_cpy (struct regcache *dst, struct regcache *src);
 };
 
-/* Copy/duplicate the contents of a register cache.  By default, the
-   operation is pass-through.  Writes to DST and reads from SRC will
-   go through to the target.  See also regcache_cpy_no_passthrough.
-
-   regcache_cpy can not have overlapping SRC and DST buffers.  */
-
+/* Duplicate the contents of a register cache to a read-only register
+   cache.  The operation is pass-through.  */
 extern struct regcache *regcache_dup (struct regcache *regcache);
+
+/* Writes to DEST will go through to the target.  SRC is a read-only
+   register cache.  */
 extern void regcache_cpy (struct regcache *dest, struct regcache *src);
 
 extern void registers_changed (void);

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough
  2017-07-17 11:32   ` Yao Qi
@ 2017-07-18  9:01     ` Alan Hayward
  2017-07-18 11:49       ` Yao Qi
  0 siblings, 1 reply; 5+ messages in thread
From: Alan Hayward @ 2017-07-18  9:01 UTC (permalink / raw)
  To: Yao Qi; +Cc: gdb-patches, nd


> On 17 Jul 2017, at 12:32, Yao Qi <qiyaoltc@gmail.com> wrote:
> 
> Alan Hayward <Alan.Hayward@arm.com> writes:
> 
>> If not, then there should be a comment above regcache_cpy stating the
>> restrictions.
> 
> I updated the comments to regcache_dup and regcache_cpy.

Ok for me.


> 
> -- 
> Yao (齐尧)
> From 4ff91e90cad73f9420f8ca37fc7e22f465ebcb88 Mon Sep 17 00:00:00 2001
> From: Yao Qi <yao.qi@linaro.org>
> Date: Mon, 19 Jun 2017 22:45:12 +0100
> Subject: [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough
> 
> Nowadays, regcache_cpy is used where src is read-only and dst is not
> read-only, so the regcache_cpy can be simplified to handle this case only.
> As a result, regcache::cpy_no_passthrough, which is about two read-only
> regcache copy, is no longer used, remove it as well.
> 
> gdb:
> 
> 2017-07-17  Yao Qi  <yao.qi@linaro.org>
> 
> 	* regcache.c (regcache_cpy): Simplify it.
> 	(regcache::cpy_no_passthrough): Remove it.
> 	* regcache.h (cpy_no_passthrough): Remove it.
> 	(regcache_dup, regcache_cpy): Update comments.
> 
> diff --git a/gdb/regcache.c b/gdb/regcache.c
> index 7eeb737..e8f92d6 100644
> --- a/gdb/regcache.c
> +++ b/gdb/regcache.c
> @@ -388,36 +388,9 @@ regcache_cpy (struct regcache *dst, struct regcache *src)
>   gdb_assert (src != NULL && dst != NULL);
>   gdb_assert (src->m_descr->gdbarch == dst->m_descr->gdbarch);
>   gdb_assert (src != dst);
> -  gdb_assert (src->m_readonly_p || dst->m_readonly_p);
> +  gdb_assert (src->m_readonly_p && !dst->m_readonly_p);
> 
> -  if (!src->m_readonly_p)
> -    regcache_save (dst, do_cooked_read, src);
> -  else if (!dst->m_readonly_p)
> -    dst->restore (src);
> -  else
> -    dst->cpy_no_passthrough (src);
> -}
> -
> -/* Copy/duplicate the contents of a register cache.  Unlike regcache_cpy,
> -   which is pass-through, this does not go through to the target.
> -   Only values values already in the cache are transferred.  The SRC and DST
> -   buffers must not overlap.  */
> -
> -void
> -regcache::cpy_no_passthrough (struct regcache *src)
> -{
> -  gdb_assert (src != NULL);
> -  gdb_assert (src->m_descr->gdbarch == m_descr->gdbarch);
> -  /* NOTE: cagney/2002-05-17: Don't let the caller do a no-passthrough
> -     move of data into a thread's regcache.  Doing this would be silly
> -     - it would mean that regcache->register_status would be
> -     completely invalid.  */
> -  gdb_assert (m_readonly_p && src->m_readonly_p);
> -
> -  memcpy (m_registers, src->m_registers,
> -	  m_descr->sizeof_cooked_registers);
> -  memcpy (m_register_status, src->m_register_status,
> -	  m_descr->sizeof_cooked_register_status);
> +  dst->restore (src);
> }
> 
> struct regcache *
> diff --git a/gdb/regcache.h b/gdb/regcache.h
> index b416d5e..aa64a00 100644
> --- a/gdb/regcache.h
> +++ b/gdb/regcache.h
> @@ -369,8 +369,6 @@ private:
> 
>   void restore (struct regcache *src);
> 
> -  void cpy_no_passthrough (struct regcache *src);
> -
>   enum register_status xfer_part (int regnum, int offset, int len, void *in,
> 				  const void *out,
> 				  decltype (regcache_raw_read) read,
> @@ -415,13 +413,12 @@ private:
>   regcache_cpy (struct regcache *dst, struct regcache *src);
> };
> 
> -/* Copy/duplicate the contents of a register cache.  By default, the
> -   operation is pass-through.  Writes to DST and reads from SRC will
> -   go through to the target.  See also regcache_cpy_no_passthrough.
> -
> -   regcache_cpy can not have overlapping SRC and DST buffers.  */
> -
> +/* Duplicate the contents of a register cache to a read-only register
> +   cache.  The operation is pass-through.  */
> extern struct regcache *regcache_dup (struct regcache *regcache);
> +
> +/* Writes to DEST will go through to the target.  SRC is a read-only
> +   register cache.  */
> extern void regcache_cpy (struct regcache *dest, struct regcache *src);
> 
> extern void registers_changed (void);


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough
  2017-07-18  9:01     ` Alan Hayward
@ 2017-07-18 11:49       ` Yao Qi
  0 siblings, 0 replies; 5+ messages in thread
From: Yao Qi @ 2017-07-18 11:49 UTC (permalink / raw)
  To: Alan Hayward; +Cc: gdb-patches, nd

Alan Hayward <Alan.Hayward@arm.com> writes:

>> I updated the comments to regcache_dup and regcache_cpy.
>
> Ok for me.

I pushed it in.

-- 
Yao (齐尧)

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2017-07-18 11:49 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-06-27  7:37 [PATCH] Simplify regcache_cpy and remove regcache::cpy_no_passthrough Yao Qi
2017-07-17  9:04 ` Alan Hayward
2017-07-17 11:32   ` Yao Qi
2017-07-18  9:01     ` Alan Hayward
2017-07-18 11:49       ` Yao Qi

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).