From: David Daney <ddaney@caviumnetworks.com>
To: rdsandiford@googlemail.com, GCC Patches <gcc-patches@gcc.gnu.org>
Cc: wuzhangjin@gmail.com, Adam Nemet <anemet@caviumnetworks.com>,
rostedt@goodmis.org, linux-kernel@vger.kernel.org,
linux-mips@linux-mips.org, Thomas Gleixner <tglx@linutronix.de>,
Ralf Baechle <ralf@linux-mips.org>,
Nicholas Mc Guire <der.herr@hofr.at>
Subject: [PATCH] MIPS: Add option to pass return address location to _mcount. Was: [PATCH -v4 4/9] tracing: add static function tracer support for MIPS
Date: Fri, 23 Oct 2009 15:48:13 -0700 [thread overview]
Message-ID: <4AE232AD.4050308@caviumnetworks.com> (raw)
In-Reply-To: <87y6n36plp.fsf@firetop.home>
[-- Attachment #1: Type: text/plain, Size: 3509 bytes --]
Richard Sandiford wrote:
> David Daney <ddaney@caviumnetworks.com> writes:
>> Wu Zhangjin wrote:
>>> On Wed, 2009-10-21 at 11:24 -0400, Steven Rostedt wrote:
>> [...]
>>>>> +
>>>>> +NESTED(_mcount, PT_SIZE, ra)
>>>>> + RESTORE_SP_FOR_32BIT
>>>>> + PTR_LA t0, ftrace_stub
>>>>> + PTR_L t1, ftrace_trace_function /* please don't use t1 later, safe? */
>>>> Is t0 and t1 safe for mcount to use? Remember, mcount does not follow
>>>> the dynamics of C function ABI.
>>> So, perhaps we can use the saved registers(a0,a1...) instead.
>>>
>> a0..a7 may not always be saved.
>>
>> You can use at, v0, v1 and all the temporary registers. Note that for
>> the 64-bit ABIs sometimes the names t0-t4 shadow a4-a7. So for a 64-bit
>> kernel, you can use: $1, $2, $3, $12, $13, $14, $15, $24, $25, noting
>> that at == $1 and contains the callers ra. For a 32-bit kernel you can
>> add $8, $9, $10, and $11
>>
>> This whole thing seems a little fragile.
>>
>> I think it might be a good idea to get input from Richard Sandiford,
>> and/or Adam Nemet about this approach (so I add them to the CC).
>>
>> This e-mail thread starts here:
>>
>> http://www.linux-mips.org/archives/linux-mips/2009-10/msg00286.html
>>
>> and here:
>>
>> http://www.linux-mips.org/archives/linux-mips/2009-10/msg00290.html
>
> I'm not sure that the "search for a save of RA" thing is really a good idea.
> The last version of that seemed to be "assume that any register stores
> will be in a block that immediately precedes the move into RA", but even
> if that's true now, it might not be in future. And as Wu Zhangjin says,
> it doesn't cope with long calls, where the target address is loaded
> into a temporary register before the call.
>
> FWIW, I'd certainly be happy to make GCC pass an additional parameter
> to _mcount. The parameter could give the address of the return slot,
> or null for leaf functions. In almost all cases[*], there would be
> no overhead, since the move would go in the delay slot of the call.
>
> [*] Meaning when the frame is <=32k. ;) I'm guessing you never
> get anywhere near that, and if you did, the scan thing wouldn't
> work anyway.
>
> The new behaviour could be controlled by a command-line option,
> which would also give linux a cheap way of checking whether the
> feature is available.
>
How about this patch, I think it does what you suggest.
When we pass -pg -mmcount-raloc, the address of the return address
relative to sp is passed in $12 to _mcount. If the return address is
not saved, $12 will be zero. I think this will work as registers are
never saved with an offset of zero. $12 is a temporary register that is
not part of the ABI.
$12 is also used by libffi closure support, but I think its use there
will not interfere with _mcount.
It is very lightly tested, I would bootstrap and regression test with
some new test cases if it were deemed acceptable.
2009-10-23 David Daney <ddaney@caviumnetworks.com>
* doc/invoke.texi (mmcount-raloc): Document new command line option.
* config/mips/mips.opt (config/mips/mips.opt): New option.
* config/mips/mips-protos.h (mips_function_profiler): Declare new
function.
* config/mips/mips.c (struct mips_frame_info): Add ra_fp_offset member.
(mips_for_each_saved_gpr_and_fpr): Set ra_fp_offset.
(mips_raloc_in_delay_slot_p): New function.
(mips_function_profiler): Moved from FUNCTION_PROFILER, and rewritten.
* config/mips/mips.h (FUNCTION_PROFILER): Body of macro moved to
mips_function_profiler.
[-- Attachment #2: mcount.patch --]
[-- Type: text/plain, Size: 8422 bytes --]
Index: gcc/doc/invoke.texi
===================================================================
--- gcc/doc/invoke.texi (revision 153502)
+++ gcc/doc/invoke.texi (working copy)
@@ -709,7 +709,7 @@ Objective-C and Objective-C++ Dialects}.
-mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
-mfp-exceptions -mno-fp-exceptions @gol
-mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
--mrelax-pic-calls -mno-relax-pic-calls}
+-mrelax-pic-calls -mno-relax-pic-calls -mmcount-raloc}
@emph{MMIX Options}
@gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
@@ -14192,6 +14192,20 @@ an assembler and a linker that supports
directive and @code{-mexplicit-relocs} is in effect. With
@code{-mno-explicit-relocs}, this optimization can be performed by the
assembler and the linker alone without help from the compiler.
+
+@item -mmcount-raloc
+@itemx -mno-mcount-raloc
+@opindex mmcount-raloc
+@opindex mno-mcount-raloc
+Emit (do not emit) code to pass the offset (from the stack pointer) of
+the return address save location to @code{_mcount}. The default is
+@option{-mno-mcount-raloc}.
+
+@option{-mmcount-raloc} can be used in conjunction with @option{-pg}
+and a specially coded @code{_mcount} function to record function exit
+by replacing the return address on the stack with a pointer to the
+function exit profiling function.
+
@end table
@node MMIX Options
Index: gcc/config/mips/mips.opt
===================================================================
--- gcc/config/mips/mips.opt (revision 153502)
+++ gcc/config/mips/mips.opt (working copy)
@@ -208,6 +208,10 @@ mlong64
Target Report RejectNegative Mask(LONG64)
Use a 64-bit long type
+mmcount-raloc
+Target Report Var(TARGET_RALOC)
+Pass the offset from sp of the ra save location to _mcount in $12
+
mmemcpy
Target Report Mask(MEMCPY)
Don't optimize block moves
Index: gcc/config/mips/mips-protos.h
===================================================================
--- gcc/config/mips/mips-protos.h (revision 153502)
+++ gcc/config/mips/mips-protos.h (working copy)
@@ -344,5 +344,6 @@ extern bool mips_eh_uses (unsigned int);
extern bool mips_epilogue_uses (unsigned int);
extern void mips_final_prescan_insn (rtx, rtx *, int);
extern int mips_trampoline_code_size (void);
+extern void mips_function_profiler (FILE *);
#endif /* ! GCC_MIPS_PROTOS_H */
Index: gcc/config/mips/mips.c
===================================================================
--- gcc/config/mips/mips.c (revision 153502)
+++ gcc/config/mips/mips.c (working copy)
@@ -319,6 +319,9 @@ struct GTY(()) mips_frame_info {
HOST_WIDE_INT acc_sp_offset;
HOST_WIDE_INT cop0_sp_offset;
+ /* Similar, but the value passed to _mcount. */
+ HOST_WIDE_INT ra_fp_offset;
+
/* The offset of arg_pointer_rtx from the bottom of the frame. */
HOST_WIDE_INT arg_pointer_offset;
@@ -9616,6 +9619,9 @@ mips_for_each_saved_gpr_and_fpr (HOST_WI
for (regno = GP_REG_LAST; regno >= GP_REG_FIRST; regno--)
if (BITSET_P (cfun->machine->frame.mask, regno - GP_REG_FIRST))
{
+ /* Record the ra offset for use by mips_function_profiler. */
+ if (regno == RETURN_ADDR_REGNUM)
+ cfun->machine->frame.ra_fp_offset = offset + sp_offset;
mips_save_restore_reg (word_mode, regno, offset, fn);
offset -= UNITS_PER_WORD;
}
@@ -16088,6 +16094,86 @@ mips_trampoline_init (rtx m_tramp, tree
emit_insn (gen_add3_insn (end_addr, addr, GEN_INT (TRAMPOLINE_SIZE)));
emit_insn (gen_clear_cache (addr, end_addr));
}
+
+/* Return true if the loading of $12 can be done in the jal/jalr delay
+ slot. We can do this only if the jal will expand to a single
+ instruction and only a single instruction is required to load the
+ value to $12. */
+
+static bool mips_raloc_in_delay_slot_p(void)
+{
+ return (SMALL_OPERAND_UNSIGNED(cfun->machine->frame.ra_fp_offset)
+ && !TARGET_USE_GOT);
+}
+
+/* Implement FUNCTION_PROFILER. */
+
+void mips_function_profiler (FILE *file)
+{
+ int emit_small_ra_offset;
+
+ emit_small_ra_offset = 0;
+
+ if (TARGET_MIPS16)
+ sorry ("mips16 function profiling");
+ if (TARGET_LONG_CALLS)
+ {
+ /* For TARGET_LONG_CALLS use $3 for the address of _mcount. */
+ if (Pmode == DImode)
+ fprintf (file, "\tdla\t%s,_mcount\n", reg_names[3]);
+ else
+ fprintf (file, "\tla\t%s,_mcount\n", reg_names[3]);
+ }
+ if (TARGET_RALOC)
+ {
+ /* If TARGET_RALOC load $12 with the offset of the ra save
+ location. */
+ if (mips_raloc_in_delay_slot_p())
+ emit_small_ra_offset = 1;
+ else
+ {
+ if (Pmode == DImode)
+ fprintf (file, "\tdli\t%s,%d\t\t# offset of ra\n", reg_names[12],
+ cfun->machine->frame.ra_fp_offset);
+ else
+ fprintf (file, "\tli\t%s,%d\t\t# offset of ra\n", reg_names[12],
+ cfun->machine->frame.ra_fp_offset);
+ }
+ }
+ mips_push_asm_switch (&mips_noat);
+ fprintf (file, "\tmove\t%s,%s\t\t# save current return address\n",
+ reg_names[AT_REGNUM], reg_names[RETURN_ADDR_REGNUM]);
+ /* _mcount treats $2 as the static chain register. */
+ if (cfun->static_chain_decl != NULL)
+ fprintf (file, "\tmove\t%s,%s\n", reg_names[2],
+ reg_names[STATIC_CHAIN_REGNUM]);
+ if (!TARGET_NEWABI)
+ {
+ fprintf (file,
+ "\t%s\t%s,%s,%d\t\t# _mcount pops 2 words from stack\n",
+ TARGET_64BIT ? "dsubu" : "subu",
+ reg_names[STACK_POINTER_REGNUM],
+ reg_names[STACK_POINTER_REGNUM],
+ Pmode == DImode ? 16 : 8);
+ }
+ if (emit_small_ra_offset)
+ mips_push_asm_switch (&mips_noreorder);
+ if (TARGET_LONG_CALLS)
+ fprintf (file, "\tjalr\t%s\n", reg_names[3]);
+ else
+ fprintf (file, "\tjal\t_mcount\n");
+ if (emit_small_ra_offset)
+ {
+ fprintf (file, "\tori\t%s,%s,%d\t\t# offset of ra\n",
+ reg_names[12], reg_names[0], cfun->machine->frame.ra_fp_offset);
+ mips_pop_asm_switch (&mips_noreorder);
+ }
+ mips_pop_asm_switch (&mips_noat);
+ /* _mcount treats $2 as the static chain register. */
+ if (cfun->static_chain_decl != NULL)
+ fprintf (file, "\tmove\t%s,%s\n", reg_names[STATIC_CHAIN_REGNUM],
+ reg_names[2]);
+}
\f
/* Initialize the GCC target structure. */
#undef TARGET_ASM_ALIGNED_HI_OP
Index: gcc/config/mips/mips.h
===================================================================
--- gcc/config/mips/mips.h (revision 153502)
+++ gcc/config/mips/mips.h (working copy)
@@ -2372,44 +2372,7 @@ typedef struct mips_args {
/* Output assembler code to FILE to increment profiler label # LABELNO
for profiling a function entry. */
-#define FUNCTION_PROFILER(FILE, LABELNO) \
-{ \
- if (TARGET_MIPS16) \
- sorry ("mips16 function profiling"); \
- if (TARGET_LONG_CALLS) \
- { \
- /* For TARGET_LONG_CALLS use $3 for the address of _mcount. */ \
- if (Pmode == DImode) \
- fprintf (FILE, "\tdla\t%s,_mcount\n", reg_names[GP_REG_FIRST + 3]); \
- else \
- fprintf (FILE, "\tla\t%s,_mcount\n", reg_names[GP_REG_FIRST + 3]); \
- } \
- mips_push_asm_switch (&mips_noat); \
- fprintf (FILE, "\tmove\t%s,%s\t\t# save current return address\n", \
- reg_names[AT_REGNUM], reg_names[RETURN_ADDR_REGNUM]); \
- /* _mcount treats $2 as the static chain register. */ \
- if (cfun->static_chain_decl != NULL) \
- fprintf (FILE, "\tmove\t%s,%s\n", reg_names[2], \
- reg_names[STATIC_CHAIN_REGNUM]); \
- if (!TARGET_NEWABI) \
- { \
- fprintf (FILE, \
- "\t%s\t%s,%s,%d\t\t# _mcount pops 2 words from stack\n", \
- TARGET_64BIT ? "dsubu" : "subu", \
- reg_names[STACK_POINTER_REGNUM], \
- reg_names[STACK_POINTER_REGNUM], \
- Pmode == DImode ? 16 : 8); \
- } \
- if (TARGET_LONG_CALLS) \
- fprintf (FILE, "\tjalr\t%s\n", reg_names[GP_REG_FIRST + 3]); \
- else \
- fprintf (FILE, "\tjal\t_mcount\n"); \
- mips_pop_asm_switch (&mips_noat); \
- /* _mcount treats $2 as the static chain register. */ \
- if (cfun->static_chain_decl != NULL) \
- fprintf (FILE, "\tmove\t%s,%s\n", reg_names[STATIC_CHAIN_REGNUM], \
- reg_names[2]); \
-}
+#define FUNCTION_PROFILER(FILE, LABELNO) mips_function_profiler ((FILE))
/* The profiler preserves all interesting registers, including $31. */
#define MIPS_SAVE_REG_FOR_PROFILING_P(REGNO) false
next prev parent reply other threads:[~2009-10-23 22:50 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <cover.1256135456.git.wuzhangjin@gmail.com>
2009-10-21 14:34 ` [PATCH -v4 1/9] tracing: convert trace_clock_local() as weak function Wu Zhangjin
2009-10-21 14:34 ` [PATCH -v4 2/9] MIPS: add mips_timecounter_read() to get high precision timestamp Wu Zhangjin
2009-10-22 13:03 ` pajko
2009-10-21 14:34 ` [PATCH -v4 3/9] tracing: add MIPS specific trace_clock_local() Wu Zhangjin
2009-10-21 14:46 ` Steven Rostedt
2009-10-21 15:11 ` Wu Zhangjin
2009-10-21 14:34 ` [PATCH -v4 4/9] tracing: add static function tracer support for MIPS Wu Zhangjin
2009-10-21 15:24 ` Steven Rostedt
2009-10-22 17:47 ` Wu Zhangjin
2009-10-22 17:59 ` Steven Rostedt
2009-10-22 18:34 ` Wu Zhangjin
2009-10-22 18:34 ` David Daney
2009-10-22 19:13 ` Wu Zhangjin
2009-10-22 20:30 ` Adam Nemet
2009-10-22 20:52 ` Steven Rostedt
2009-10-22 21:09 ` Frederic Weisbecker
2009-10-22 21:29 ` Adam Nemet
2009-10-22 21:55 ` Steven Rostedt
2009-10-23 1:09 ` Wu Zhangjin
2009-10-22 22:17 ` Richard Sandiford
2009-10-23 9:32 ` Wu Zhangjin
2009-10-23 22:48 ` David Daney [this message]
2009-10-24 9:12 ` [PATCH] MIPS: Add option to pass return address location to _mcount. Was: " Richard Sandiford
2009-10-24 15:53 ` Wu Zhangjin
2009-10-23 7:21 ` Wu Zhangjin
2009-10-21 14:34 ` [PATCH -v4 5/9] tracing: enable HAVE_FUNCTION_TRACE_MCOUNT_TEST " Wu Zhangjin
2009-10-21 14:35 ` [PATCH -v4 6/9] tracing: add an endian argument to scripts/recordmcount.pl Wu Zhangjin
2009-10-21 15:26 ` Steven Rostedt
2009-10-21 14:35 ` [PATCH -v4 7/9] tracing: add dynamic function tracer support for MIPS Wu Zhangjin
2009-10-21 14:35 ` [PATCH -v4 8/9] tracing: not trace mips_timecounter_init() in MIPS Wu Zhangjin
2009-10-21 14:35 ` [PATCH -v4 9/9] tracing: add function graph tracer support for MIPS Wu Zhangjin
2009-10-21 15:21 ` Wu Zhangjin
2009-10-21 16:14 ` Steven Rostedt
2009-10-21 16:12 ` Steven Rostedt
2009-10-21 16:37 ` David Daney
2009-10-21 16:46 ` Steven Rostedt
2009-10-21 17:07 ` David Daney
2009-10-21 17:23 ` Steven Rostedt
2009-10-21 17:48 ` David Daney
2009-10-21 18:09 ` Steven Rostedt
2009-10-21 18:17 ` Nicholas Mc Guire
2009-10-21 18:34 ` Steven Rostedt
2009-10-21 18:25 ` David Daney
2009-10-22 11:38 ` Wu Zhangjin
2009-10-22 13:17 ` Steven Rostedt
2009-10-22 13:31 ` Wu Zhangjin
2009-10-22 15:20 ` Steven Rostedt
2009-10-22 15:59 ` David Daney
2009-10-22 16:11 ` Steven Rostedt
2009-10-22 16:16 ` David Daney
2009-10-22 18:00 ` Steven Rostedt
2009-10-22 17:39 ` Wu Zhangjin
2009-10-22 17:58 ` Steven Rostedt
2009-10-22 11:37 ` pajko
2009-10-25 10:48 ` Wu Zhangjin
2009-10-25 13:37 ` Patrik Kluba
2009-10-25 14:22 ` Wu Zhangjin
2009-10-25 15:55 ` Richard Sandiford
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4AE232AD.4050308@caviumnetworks.com \
--to=ddaney@caviumnetworks.com \
--cc=anemet@caviumnetworks.com \
--cc=der.herr@hofr.at \
--cc=gcc-patches@gcc.gnu.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mips@linux-mips.org \
--cc=ralf@linux-mips.org \
--cc=rdsandiford@googlemail.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=wuzhangjin@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome