* [PATCHv3] arm64: fpsimd: avoid restoring fpcr if the contents haven't changed
@ 2014-07-10 11:40 Will Deacon
2014-07-11 7:35 ` Ard Biesheuvel
0 siblings, 1 reply; 2+ messages in thread
From: Will Deacon @ 2014-07-10 11:40 UTC (permalink / raw)
To: linux-arm-kernel
Writing to the FPCR is commonly implemented as a self-synchronising
operation in the CPU, so avoid writing to the register when the saved
value matches that in the hardware already.
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
---
arch/arm64/include/asm/fpsimdmacros.h | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/include/asm/fpsimdmacros.h b/arch/arm64/include/asm/fpsimdmacros.h
index 768414d55e64..007618b8188c 100644
--- a/arch/arm64/include/asm/fpsimdmacros.h
+++ b/arch/arm64/include/asm/fpsimdmacros.h
@@ -40,6 +40,19 @@
str w\tmpnr, [\state, #16 * 2 + 4]
.endm
+.macro fpsimd_restore_fpcr state, tmp
+ /*
+ * Writes to fpcr may be self-synchronising, so avoid restoring
+ * the register if it hasn't changed.
+ */
+ mrs \tmp, fpcr
+ cmp \tmp, \state
+ b.eq 9999f
+ msr fpcr, \state
+9999:
+.endm
+
+/* Clobbers \state */
.macro fpsimd_restore state, tmpnr
ldp q0, q1, [\state, #16 * 0]
ldp q2, q3, [\state, #16 * 2]
@@ -60,7 +73,7 @@
ldr w\tmpnr, [\state, #16 * 2]
msr fpsr, x\tmpnr
ldr w\tmpnr, [\state, #16 * 2 + 4]
- msr fpcr, x\tmpnr
+ fpsimd_restore_fpcr x\tmpnr, \state
.endm
.altmacro
@@ -84,7 +97,7 @@
.macro fpsimd_restore_partial state, tmpnr1, tmpnr2
ldp w\tmpnr1, w\tmpnr2, [\state]
msr fpsr, x\tmpnr1
- msr fpcr, x\tmpnr2
+ fpsimd_restore_fpcr x\tmpnr2, x\tmpnr1
adr x\tmpnr1, 0f
ldr w\tmpnr2, [\state, #8]
add \state, \state, x\tmpnr2, lsl #4
--
2.0.0
^ permalink raw reply related [flat|nested] 2+ messages in thread
* [PATCHv3] arm64: fpsimd: avoid restoring fpcr if the contents haven't changed
2014-07-10 11:40 [PATCHv3] arm64: fpsimd: avoid restoring fpcr if the contents haven't changed Will Deacon
@ 2014-07-11 7:35 ` Ard Biesheuvel
0 siblings, 0 replies; 2+ messages in thread
From: Ard Biesheuvel @ 2014-07-11 7:35 UTC (permalink / raw)
To: linux-arm-kernel
On 10 July 2014 13:40, Will Deacon <will.deacon@arm.com> wrote:
> Writing to the FPCR is commonly implemented as a self-synchronising
> operation in the CPU, so avoid writing to the register when the saved
> value matches that in the hardware already.
>
> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> Reviewed-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Signed-off-by: Will Deacon <will.deacon@arm.com>
> ---
> arch/arm64/include/asm/fpsimdmacros.h | 17 +++++++++++++++--
> 1 file changed, 15 insertions(+), 2 deletions(-)
>
Hi Will,
This works fine as far as I can tell.
--
Ard.
> diff --git a/arch/arm64/include/asm/fpsimdmacros.h b/arch/arm64/include/asm/fpsimdmacros.h
> index 768414d55e64..007618b8188c 100644
> --- a/arch/arm64/include/asm/fpsimdmacros.h
> +++ b/arch/arm64/include/asm/fpsimdmacros.h
> @@ -40,6 +40,19 @@
> str w\tmpnr, [\state, #16 * 2 + 4]
> .endm
>
> +.macro fpsimd_restore_fpcr state, tmp
> + /*
> + * Writes to fpcr may be self-synchronising, so avoid restoring
> + * the register if it hasn't changed.
> + */
> + mrs \tmp, fpcr
> + cmp \tmp, \state
> + b.eq 9999f
> + msr fpcr, \state
> +9999:
> +.endm
> +
> +/* Clobbers \state */
> .macro fpsimd_restore state, tmpnr
> ldp q0, q1, [\state, #16 * 0]
> ldp q2, q3, [\state, #16 * 2]
> @@ -60,7 +73,7 @@
> ldr w\tmpnr, [\state, #16 * 2]
> msr fpsr, x\tmpnr
> ldr w\tmpnr, [\state, #16 * 2 + 4]
> - msr fpcr, x\tmpnr
> + fpsimd_restore_fpcr x\tmpnr, \state
> .endm
>
> .altmacro
> @@ -84,7 +97,7 @@
> .macro fpsimd_restore_partial state, tmpnr1, tmpnr2
> ldp w\tmpnr1, w\tmpnr2, [\state]
> msr fpsr, x\tmpnr1
> - msr fpcr, x\tmpnr2
> + fpsimd_restore_fpcr x\tmpnr2, x\tmpnr1
> adr x\tmpnr1, 0f
> ldr w\tmpnr2, [\state, #8]
> add \state, \state, x\tmpnr2, lsl #4
> --
> 2.0.0
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2014-07-11 7:35 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-07-10 11:40 [PATCHv3] arm64: fpsimd: avoid restoring fpcr if the contents haven't changed Will Deacon
2014-07-11 7:35 ` Ard Biesheuvel
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 NNTP newsgroup(s).