* [Patch,Aarch64] Make frame grow downwards
@ 2013-06-28 11:46 Christophe Lyon
2013-08-30 15:21 ` Christophe Lyon
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Christophe Lyon @ 2013-06-28 11:46 UTC (permalink / raw)
To: gcc-patches
[-- Attachment #1: Type: text/plain, Size: 666 bytes --]
Hi,
Following the discussion on
http://gcc.gnu.org/ml/gcc/2013-05/msg00208.html
here is a patch to change the frame direction.
Passed 'make check' with no regression on aarch64-none-linux-gnu and
with a few UNSUPPORTED -> PASS on aarch64-none-elf (pr34225, pr38616,
pr38902, pr40971, pr46440, pr47766, pr49307, all of which are guarded
by dg-require-effective-target fstack_protector).
OK for trunk?
Thanks,
Christophe.
2013-06-28 Christophe Lyon <christophe.lyon@linaro.org>
* config/aarch64/aarch64.h (FRAME_GROWS_DOWNWARD): Define to 1.
* config/aarch64/aarch64.c (aarch64_initial_elimination_offset):
Add get_frame_size () when eliminating frame pointer.
[-- Attachment #2: aarch64-frame.txt --]
[-- Type: text/plain, Size: 2148 bytes --]
diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
index 527b00d..d8f2801 100644
--- a/gcc/config/aarch64/aarch64.c
+++ b/gcc/config/aarch64/aarch64.c
@@ -1943,7 +1943,7 @@ aarch64_save_or_restore_callee_save_registers (HOST_WIDE_INT offset,
Establish the stack frame by decreasing the stack pointer with a
properly calculated size and, if necessary, create a frame record
filled with the values of LR and previous frame pointer. The
- current FP is also set up is it is in use. */
+ current FP is also set up if it is in use. */
void
aarch64_expand_prologue (void)
@@ -4031,7 +4031,7 @@ aarch64_initial_elimination_offset (unsigned from, unsigned to)
return offset - crtl->outgoing_args_size;
if (from == FRAME_POINTER_REGNUM)
- return cfun->machine->frame.saved_regs_size;
+ return cfun->machine->frame.saved_regs_size + get_frame_size ();
}
if (to == STACK_POINTER_REGNUM)
@@ -4040,6 +4040,7 @@ aarch64_initial_elimination_offset (unsigned from, unsigned to)
{
HOST_WIDE_INT elim = crtl->outgoing_args_size
+ cfun->machine->frame.saved_regs_size
+ + get_frame_size ()
- cfun->machine->frame.fp_lr_offset;
elim = AARCH64_ROUND_UP (elim, STACK_BOUNDARY / BITS_PER_UNIT);
return elim;
diff --git a/gcc/config/aarch64/aarch64.h b/gcc/config/aarch64/aarch64.h
index a08797b..5048258 100644
--- a/gcc/config/aarch64/aarch64.h
+++ b/gcc/config/aarch64/aarch64.h
@@ -434,7 +434,7 @@ enum reg_class
#define INDEX_REG_CLASS CORE_REGS
#define BASE_REG_CLASS POINTER_REGS
-/* Register pairs used to eliminate unneeded registers that point intoi
+/* Register pairs used to eliminate unneeded registers that point into
the stack frame. */
#define ELIMINABLE_REGS \
{ \
@@ -475,7 +475,7 @@ extern enum aarch64_processor aarch64_tune;
/* Stack layout; function entry, exit and calling. */
#define STACK_GROWS_DOWNWARD 1
-#define FRAME_GROWS_DOWNWARD 0
+#define FRAME_GROWS_DOWNWARD 1
#define STARTING_FRAME_OFFSET 0
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Patch,Aarch64] Make frame grow downwards
2013-06-28 11:46 [Patch,Aarch64] Make frame grow downwards Christophe Lyon
@ 2013-08-30 15:21 ` Christophe Lyon
2013-11-12 4:40 ` Christophe Lyon
2013-11-12 17:38 ` Richard Earnshaw
2 siblings, 0 replies; 4+ messages in thread
From: Christophe Lyon @ 2013-08-30 15:21 UTC (permalink / raw)
To: gcc-patches
[-- Attachment #1: Type: text/plain, Size: 1320 bytes --]
Hi,
Following our discussions about changing the frame direction on
aarch64, I have written a small test case using a large frame which
shows similar code generation with and without my patch. There
are small differences in register allocation which lead to larger
prologue/epilogue with my patch, but I think this could (should) be
improved independently.
I have attached the sample C source and the code generated with
(frane64.s.O2) and without (frame64.s.O2.ref) my patch when compiling
at -O2.
What do you think?
Thanks,
Christophe.
On 28 June 2013 13:41, Christophe Lyon <christophe.lyon@linaro.org> wrote:
> Hi,
> Following the discussion on
> http://gcc.gnu.org/ml/gcc/2013-05/msg00208.html
> here is a patch to change the frame direction.
>
> Passed 'make check' with no regression on aarch64-none-linux-gnu and
> with a few UNSUPPORTED -> PASS on aarch64-none-elf (pr34225, pr38616,
> pr38902, pr40971, pr46440, pr47766, pr49307, all of which are guarded
> by dg-require-effective-target fstack_protector).
>
> OK for trunk?
>
> Thanks,
>
> Christophe.
> 2013-06-28 Christophe Lyon <christophe.lyon@linaro.org>
>
> * config/aarch64/aarch64.h (FRAME_GROWS_DOWNWARD): Define to 1.
> * config/aarch64/aarch64.c (aarch64_initial_elimination_offset):
> Add get_frame_size () when eliminating frame pointer.
[-- Attachment #2: frame64.c --]
[-- Type: text/x-csrc, Size: 347 bytes --]
#define SIZE1 100000
extern int func2(int);
int func1(int arg1, int arg2)
{
int var1;
int array1[SIZE1];
int var2 = 0;
for (var1 = arg1; var1 < arg2; var1++) {
array1[var1] = func2(var1);
}
for (var1 = arg1; var1 < arg2; var1++) {
var2 += array1[var1];
}
var2 += (array1[arg1] + array1[arg2-1]) / 2;
return var2;
}
[-- Attachment #3: frame64.s.O2 --]
[-- Type: application/octet-stream, Size: 969 bytes --]
.cpu generic
.file "frame64.c"
.text
.align 2
.global func1
.type func1, %function
func1:
sub sp, sp, #397312
sub sp, sp, #2688
stp x29, x30, [sp, -64]!
add x29, sp, 0
stp x19, x20, [sp,16]
stp x21, x22, [sp,32]
str x23, [sp,48]
cmp w0, w1
mov w22, w0
mov w20, w1
bge .L6
add x23, x29, 2752
mov w19, w0
sub x21, x23, #2688
.L4:
mov w0, w19
bl func2
str w0, [x21,w19,sxtw 2]
add w19, w19, 1
cmp w19, w20
bne .L4
mov w2, w22
mov w1, 0
sub x4, x23, #2688
.L5:
ldr w3, [x4,w2,sxtw 2]
add w2, w2, 1
cmp w2, w20
add w1, w1, w3
bne .L5
.L2:
sub x23, x23, #2688
sub w20, w20, #1
ldr w2, [x23,w22,sxtw 2]
ldr w0, [x23,w20,sxtw 2]
ldp x21, x22, [sp,32]
ldp x19, x20, [sp,16]
ldr x23, [sp,48]
ldp x29, x30, [sp], 64
add w0, w2, w0
add sp, sp, 2688
add w0, w0, w0, lsr 31
add w0, w1, w0, asr 1
add sp, sp, 397312
ret
.L6:
mov w1, 0
add x23, x29, 2752
b .L2
.size func1, .-func1
.ident "GCC: (GNU) 4.9.0 20130523 (experimental)"
[-- Attachment #4: frame64.s.O2.ref --]
[-- Type: application/octet-stream, Size: 869 bytes --]
.cpu generic
.file "frame64.c"
.text
.align 2
.global func1
.type func1, %function
func1:
sub sp, sp, #397312
sub sp, sp, #2688
stp x29, x30, [sp, -48]!
add x29, sp, 0
stp x19, x20, [sp,16]
str x21, [sp,32]
cmp w0, w1
mov w21, w0
mov w20, w1
bge .L6
mov w19, w0
.L4:
mov w0, w19
bl func2
add x1, x29, 48
str w0, [x1,w19,sxtw 2]
add w19, w19, 1
cmp w19, w20
bne .L4
mov w2, w21
mov w1, 0
.L5:
add x0, x29, 48
ldr w3, [x0,w2,sxtw 2]
add w2, w2, 1
cmp w2, w20
add w1, w1, w3
bne .L5
.L2:
add x3, x29, 48
sub w20, w20, #1
ldr w2, [x3,w21,sxtw 2]
ldr w0, [x3,w20,sxtw 2]
ldr x21, [sp,32]
ldp x19, x20, [sp,16]
ldp x29, x30, [sp], 48
add w0, w2, w0
add sp, sp, 2688
add w0, w0, w0, lsr 31
add w0, w1, w0, asr 1
add sp, sp, 397312
ret
.L6:
mov w1, 0
b .L2
.size func1, .-func1
.ident "GCC: (GNU) 4.9.0 20130523 (experimental)"
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Patch,Aarch64] Make frame grow downwards
2013-06-28 11:46 [Patch,Aarch64] Make frame grow downwards Christophe Lyon
2013-08-30 15:21 ` Christophe Lyon
@ 2013-11-12 4:40 ` Christophe Lyon
2013-11-12 17:38 ` Richard Earnshaw
2 siblings, 0 replies; 4+ messages in thread
From: Christophe Lyon @ 2013-11-12 4:40 UTC (permalink / raw)
To: gcc-patches, Richard Earnshaw, Patch Tracking
Ping?
The original post is here:
http://gcc.gnu.org/ml/gcc-patches/2013-06/msg01588.html
Thanks,
Christophe.
On 28 June 2013 13:41, Christophe Lyon <christophe.lyon@linaro.org> wrote:
> Hi,
> Following the discussion on
> http://gcc.gnu.org/ml/gcc/2013-05/msg00208.html
> here is a patch to change the frame direction.
>
> Passed 'make check' with no regression on aarch64-none-linux-gnu and
> with a few UNSUPPORTED -> PASS on aarch64-none-elf (pr34225, pr38616,
> pr38902, pr40971, pr46440, pr47766, pr49307, all of which are guarded
> by dg-require-effective-target fstack_protector).
>
> OK for trunk?
>
> Thanks,
>
> Christophe.
> 2013-06-28 Christophe Lyon <christophe.lyon@linaro.org>
>
> * config/aarch64/aarch64.h (FRAME_GROWS_DOWNWARD): Define to 1.
> * config/aarch64/aarch64.c (aarch64_initial_elimination_offset):
> Add get_frame_size () when eliminating frame pointer.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Patch,Aarch64] Make frame grow downwards
2013-06-28 11:46 [Patch,Aarch64] Make frame grow downwards Christophe Lyon
2013-08-30 15:21 ` Christophe Lyon
2013-11-12 4:40 ` Christophe Lyon
@ 2013-11-12 17:38 ` Richard Earnshaw
2 siblings, 0 replies; 4+ messages in thread
From: Richard Earnshaw @ 2013-11-12 17:38 UTC (permalink / raw)
To: Christophe Lyon; +Cc: gcc-patches
On 28/06/13 12:41, Christophe Lyon wrote:
> Hi,
> Following the discussion on
> http://gcc.gnu.org/ml/gcc/2013-05/msg00208.html
> here is a patch to change the frame direction.
>
> Passed 'make check' with no regression on aarch64-none-linux-gnu and
> with a few UNSUPPORTED -> PASS on aarch64-none-elf (pr34225, pr38616,
> pr38902, pr40971, pr46440, pr47766, pr49307, all of which are guarded
> by dg-require-effective-target fstack_protector).
>
> OK for trunk?
>
> Thanks,
>
> Christophe.
> 2013-06-28 Christophe Lyon <christophe.lyon@linaro.org>
>
> * config/aarch64/aarch64.h (FRAME_GROWS_DOWNWARD): Define to 1.
> * config/aarch64/aarch64.c (aarch64_initial_elimination_offset):
> Add get_frame_size () when eliminating frame pointer.
In general, when the files changed are same as the name of the port,
it's ok to write:
* aarch64.h (...): ...
For aarch64_initial_elimination_offset, I suggest "Update offset
calculations."
Otherwise OK.
R.
>
>
> aarch64-frame.txt
>
>
> diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
> index 527b00d..d8f2801 100644
> --- a/gcc/config/aarch64/aarch64.c
> +++ b/gcc/config/aarch64/aarch64.c
> @@ -1943,7 +1943,7 @@ aarch64_save_or_restore_callee_save_registers (HOST_WIDE_INT offset,
> Establish the stack frame by decreasing the stack pointer with a
> properly calculated size and, if necessary, create a frame record
> filled with the values of LR and previous frame pointer. The
> - current FP is also set up is it is in use. */
> + current FP is also set up if it is in use. */
>
> void
> aarch64_expand_prologue (void)
> @@ -4031,7 +4031,7 @@ aarch64_initial_elimination_offset (unsigned from, unsigned to)
> return offset - crtl->outgoing_args_size;
>
> if (from == FRAME_POINTER_REGNUM)
> - return cfun->machine->frame.saved_regs_size;
> + return cfun->machine->frame.saved_regs_size + get_frame_size ();
> }
>
> if (to == STACK_POINTER_REGNUM)
> @@ -4040,6 +4040,7 @@ aarch64_initial_elimination_offset (unsigned from, unsigned to)
> {
> HOST_WIDE_INT elim = crtl->outgoing_args_size
> + cfun->machine->frame.saved_regs_size
> + + get_frame_size ()
> - cfun->machine->frame.fp_lr_offset;
> elim = AARCH64_ROUND_UP (elim, STACK_BOUNDARY / BITS_PER_UNIT);
> return elim;
> diff --git a/gcc/config/aarch64/aarch64.h b/gcc/config/aarch64/aarch64.h
> index a08797b..5048258 100644
> --- a/gcc/config/aarch64/aarch64.h
> +++ b/gcc/config/aarch64/aarch64.h
> @@ -434,7 +434,7 @@ enum reg_class
> #define INDEX_REG_CLASS CORE_REGS
> #define BASE_REG_CLASS POINTER_REGS
>
> -/* Register pairs used to eliminate unneeded registers that point intoi
> +/* Register pairs used to eliminate unneeded registers that point into
> the stack frame. */
> #define ELIMINABLE_REGS \
> { \
> @@ -475,7 +475,7 @@ extern enum aarch64_processor aarch64_tune;
> /* Stack layout; function entry, exit and calling. */
> #define STACK_GROWS_DOWNWARD 1
>
> -#define FRAME_GROWS_DOWNWARD 0
> +#define FRAME_GROWS_DOWNWARD 1
>
> #define STARTING_FRAME_OFFSET 0
>
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2013-11-12 16:38 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-06-28 11:46 [Patch,Aarch64] Make frame grow downwards Christophe Lyon
2013-08-30 15:21 ` Christophe Lyon
2013-11-12 4:40 ` Christophe Lyon
2013-11-12 17:38 ` Richard Earnshaw
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).