[PATCH] mpg123: Add support for FPU-less targets

[PATCH] mpg123: Add support for FPU-less targets

[robert.rosengren]

From: Robert Rosengren <robert.rosengren@axis.com>

Support added to configure mpg123 for FPU-less targets. Building for
fixed-point arithmetic increases performance on such devices.

Signed-off-by: Robert Rosengren <robert.rosengren@axis.com>
---
 meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb | 1 +
 1 file changed, 1 insertion(+)

diff --git a/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb b/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb
index c9bbcd30ff..35cad6ffc4 100644
--- a/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb
+++ b/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb
@@ -40,6 +40,7 @@ EXTRA_OECONF = " \
     --with-audio='${AUDIOMODS}' \
     ${@bb.utils.contains('TUNE_FEATURES', 'neon', '--with-cpu=neon', '', d)} \
     ${@bb.utils.contains('TUNE_FEATURES', 'altivec', '--with-cpu=altivec', '', d)} \
+    ${@bb.utils.contains('TARGET_FPU', 'soft', '--with-cpu=generic_nofpu', '', d)} \
 "
 # Fails to build with thumb-1 (qemuarm)
 #| {standard input}: Assembler messages:
-- 
2.20.1

Re: [OE-core] [PATCH] mpg123: Add support for FPU-less targets

[Khem Raj]

is --with-cpu=generic_nofpu applicable for all soft fpu machines or
just arm ? I wonder if it will improve or regress other nofpu
machines. Do you have any data

On Tue, Feb 9, 2021 at 6:16 AM <robert.rosengren@axis.com> wrote:
>
> From: Robert Rosengren <robert.rosengren@axis.com>
>
> Support added to configure mpg123 for FPU-less targets. Building for
> fixed-point arithmetic increases performance on such devices.
>
> Signed-off-by: Robert Rosengren <robert.rosengren@axis.com>
> ---
>  meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb b/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb
> index c9bbcd30ff..35cad6ffc4 100644
> --- a/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb
> +++ b/meta/recipes-multimedia/mpg123/mpg123_1.26.4.bb
> @@ -40,6 +40,7 @@ EXTRA_OECONF = " \
>      --with-audio='${AUDIOMODS}' \
>      ${@bb.utils.contains('TUNE_FEATURES', 'neon', '--with-cpu=neon', '', d)} \
>      ${@bb.utils.contains('TUNE_FEATURES', 'altivec', '--with-cpu=altivec', '', d)} \
> +    ${@bb.utils.contains('TARGET_FPU', 'soft', '--with-cpu=generic_nofpu', '', d)} \
>  "
>  # Fails to build with thumb-1 (qemuarm)
>  #| {standard input}: Assembler messages:
> --
> 2.20.1
>
>
> 
>

Re: [OE-core] [PATCH] mpg123: Add support for FPU-less targets

[robert.rosengren]

Actually, I used a MIPS-based device lacking FPU and ran simple gstreamer/gst-play to test a mp3. A very non-scientific measurement on that device using nmon moved from roughly 100% CPU load and stuttering audio to a doable ~15-20% CPU load with the fixed point configuration. (of course all such measurements depends heavily on audio latency, queues and so on of a device hardware.)

On 2/9/21 8:10 PM, Khem Raj wrote:
> is --with-cpu=generic_nofpu applicable for all soft fpu machines or
> just arm ? I wonder if it will improve or regress other nofpu
> machines. Do you have any data
> 

Re: [OE-core] [PATCH] mpg123: Add support for FPU-less targets

[Khem Raj]

On Wed, Feb 10, 2021 at 2:54 AM Robert Rosengren <robertr@axis.com> wrote:
>
> Actually, I used a MIPS-based device lacking FPU and ran simple gstreamer/gst-play to test a mp3. A very non-scientific measurement on that device using nmon moved from roughly 100% CPU load and stuttering audio to a doable ~15-20% CPU load with the fixed point configuration. (of course all such measurements depends heavily on audio latency, queues and so on of a device hardware.)
>

OK thanks, this makes sense.

> On 2/9/21 8:10 PM, Khem Raj wrote:
> > is --with-cpu=generic_nofpu applicable for all soft fpu machines or
> > just arm ? I wonder if it will improve or regress other nofpu
> > machines. Do you have any data
> >