Skip to contentSkip to conversation

radeon: Use only a single work queue thread for crt

PATCH NAK Andi Kleen 6 participants 7 messages Aug 3, 2019 – Sep 7, 2020

Subsystems (2)
lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

PATCH 1/1 radeon: Use only a single work queue thread for crt

From: Andi Kleen <[email protected]>

I got tired of seeing a lot of radeon-crt kernel threads in ps on my workstation, one for each CPU and one for each display, which never use any CPU time. Surely a single kernel thread is enough to handle the display.

---
 drivers/gpu/drm/radeon/radeon_display.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/radeon/radeon_display.c +1 -1
index bd52f15e6330..fb0ca688f6fe 100644
@@ -682,7 +682,7 @@ static void radeon_crtc_init(struct drm_device *dev, int index)
 
 	drm_mode_crtc_set_gamma_size(&radeon_crtc->base, 256);
 	radeon_crtc->crtc_id = index;
-	radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc", WQ_HIGHPRI, 0);
+	radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc", WQ_HIGHPRI|WQ_UNBOUND, 0);
 	rdev->mode_info.crtcs[index] = radeon_crtc;
 
 	if (rdev->family >= CHIP_BONAIRE) {
Signature · 6 lines
2.20.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Reviewed-by on v1 1/1
#1 lore raw

Dear Andi,

Am 03.08.19 um 02:09 schrieb Andi Kleen:

Quoted 16 lines

From: Andi Kleen <[email protected]>

I got tired of seeing a lot of radeon-crt kernel threads in ps on my workstation, one for each CPU and one for each display, which never use any CPU time. Surely a single kernel thread is enough to handle the display.

---
  drivers/gpu/drm/radeon/radeon_display.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/radeon/radeon_display.c +0 -0
index bd52f15e6330..fb0ca688f6fe 100644

drivers/gpu/drm/radeon/radeon_display.c static void radeon_crtc_init(struct drm_device *dev, int index) · line 688

Quoted 8 lines
drivers/gpu/drm/radeon/radeon_display.c +1 -1
@@ -682,7 +682,7 @@ static void radeon_crtc_init(struct drm_device *dev, int index)
 
  	drm_mode_crtc_set_gamma_size(&radeon_crtc->base, 256);
  	radeon_crtc->crtc_id = index;
-	radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc", WQ_HIGHPRI, 0);
+	radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc", WQ_HIGHPRI|WQ_UNBOUND, 0);
  	rdev->mode_info.crtcs[index] = radeon_crtc;
 
  	if (rdev->family >= CHIP_BONAIRE) {

Kind regards,

Paul
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#2 lore raw
In reply to Andi Kleen

Am 03.08.19 um 02:09 schrieb Andi Kleen:

Quoted 5 lines

From: Andi Kleen <[email protected]>

I got tired of seeing a lot of radeon-crt kernel threads in ps on my workstation, one for each CPU and one for each display, which never use any CPU time. Surely a single kernel thread is enough to handle the display.

NAK, radeon blocks inside the kernel thread and those need to run in parallel or otherwise the hardware can hang.

Christian.

Quoted 11 lines
---
  drivers/gpu/drm/radeon/radeon_display.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/radeon/radeon_display.c +0 -0
index bd52f15e6330..fb0ca688f6fe 100644

drivers/gpu/drm/radeon/radeon_display.c static void radeon_crtc_init(struct drm_device *dev, int index) · line 688

Quoted 8 lines
drivers/gpu/drm/radeon/radeon_display.c +1 -1
@@ -682,7 +682,7 @@ static void radeon_crtc_init(struct drm_device *dev, int index)
 
  	drm_mode_crtc_set_gamma_size(&radeon_crtc->base, 256);
  	radeon_crtc->crtc_id = index;
-	radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc", WQ_HIGHPRI, 0);
+	radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc", WQ_HIGHPRI|WQ_UNBOUND, 0);
  	rdev->mode_info.crtcs[index] = radeon_crtc;
 
  	if (rdev->family >= CHIP_BONAIRE) {

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw
Quoted 11 lines · Christian König

Am 03.08.19 um 02:09 schrieb Andi Kleen:

Quoted 4 lines

From: Andi Kleen <[email protected]>

I got tired of seeing a lot of radeon-crt kernel threads in ps on my workstation, one for each CPU and one for each display, which never use

any CPU time.

Quoted 1 line

Surely a single kernel thread is enough to handle the display.

NAK, radeon blocks inside the kernel thread and those need to run in parallel or otherwise the hardware can hang.

Shouldn't WQ_UNBOUND create a new worker thread whenever all current workers go to sleep/block (and the total number of worker threads is below 'max_active') ?

Jan

Quoted 23 lines

Christian.

Review

Quoted 7 lines
---
  drivers/gpu/drm/radeon/radeon_display.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/drivers/gpu/drm/radeon/radeon_display.c

drivers/gpu/drm/radeon/radeon_display.c static void radeon_crtc_init(struct drm_device *dev,

Quoted 4 lines

index bd52f15e6330..fb0ca688f6fe 100644

drivers/gpu/drm/radeon/radeon_display.c +0 -0
@@ -682,7 +682,7 @@ static void radeon_crtc_init(struct drm_device *dev,

int index)

drivers/gpu/drm/radeon/radeon_display.c static void radeon_crtc_init(struct drm_device *dev, int index)

Quoted 4 lines
drivers/gpu/drm/radeon/radeon_display.c +0 -1
      drm_mode_crtc_set_gamma_size(&radeon_crtc->base, 256);
      radeon_crtc->crtc_id = index;
-     radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc",

WQ_HIGHPRI, 0);

drivers/gpu/drm/radeon/radeon_display.c static void radeon_crtc_init(struct drm_device *dev, · line 684

Quoted 1 line
drivers/gpu/drm/radeon/radeon_display.c +1 -0
+     radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc",

WQ_HIGHPRI|WQ_UNBOUND, 0);

Quoted 9 lines

drivers/gpu/drm/radeon/radeon_display.c static void radeon_crtc_init(struct drm_device *dev, int index) · line 688

Quoted 3 lines
drivers/gpu/drm/radeon/radeon_display.c +0 -0
      rdev->mode_info.crtcs[index] = radeon_crtc;
 
      if (rdev->family >= CHIP_BONAIRE) {

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#4 lore raw

Am 06.09.20 um 18:02 schrieb Jan Vesely:

Quoted 12 lines · On Sun, Sep 6, 2020 at 4:54 AM Christian König…

Am 03.08.19 um 02:09 schrieb Andi Kleen:
> From: Andi Kleen <[email protected] <mailto:[email protected]>> >
> I got tired of seeing a lot of radeon-crt kernel threads in ps on my > workstation, one for each CPU and one for each display, which never use any CPU time.
> Surely a single kernel thread is enough to handle the display.

Quoted 5 lines

NAK, radeon blocks inside the kernel thread and those need to run in parallel or otherwise the hardware can hang.

Shouldn't WQ_UNBOUND create a new worker thread whenever all current workers go to sleep/block (and the total number of worker threads is below 'max_active') ?

If that is guaranteed then this is indeed sufficient, but the sentence "Surely a single kernel thread is enough to handle the display." sounded like it creates only a single one.

Christian.

Quoted 37 lines

Jan

    Christian.

    >
    > Signed-off-by: Andi Kleen <[email protected]
    <mailto:[email protected]>>
    > ---
    >   drivers/gpu/drm/radeon/radeon_display.c | 2 +-
    >   1 file changed, 1 insertion(+), 1 deletion(-)
    >
    > diff --git a/drivers/gpu/drm/radeon/radeon_display.c
    b/drivers/gpu/drm/radeon/radeon_display.c
    > index bd52f15e6330..fb0ca688f6fe 100644
    > --- a/drivers/gpu/drm/radeon/radeon_display.c
    > +++ b/drivers/gpu/drm/radeon/radeon_display.c
    > @@ -682,7 +682,7 @@ static void radeon_crtc_init(struct
    drm_device *dev, int index)
    >
    >  drm_mode_crtc_set_gamma_size(&radeon_crtc->base, 256);
    >       radeon_crtc->crtc_id = index;
    > -     radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc",
    WQ_HIGHPRI, 0);
    > +     radeon_crtc->flip_queue = alloc_workqueue("radeon-crtc",
    WQ_HIGHPRI|WQ_UNBOUND, 0);
    >       rdev->mode_info.crtcs[index] = radeon_crtc;
    >
    >       if (rdev->family >= CHIP_BONAIRE) {

    _______________________________________________
    amd-gfx mailing list
    [email protected] <mailto:[email protected]>
    https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Quoted 4 lines

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

maintainer
#5 lore raw
In reply to Christian König
Quoted 10 lines · Christian König

Am 03.08.19 um 02:09 schrieb Andi Kleen:

Quoted 5 lines

From: Andi Kleen <[email protected]>

I got tired of seeing a lot of radeon-crt kernel threads in ps on my workstation, one for each CPU and one for each display, which never use any CPU time. Surely a single kernel thread is enough to handle the display.

NAK, radeon blocks inside the kernel thread and those need to run in parallel or otherwise the hardware can hang.

Do we need one per cpu? or is one per crtc enough?

Dave.
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

maintainer
#6 lore raw

Am 07.09.20 um 01:47 schrieb Dave Airlie:

Quoted 11 lines
Quoted 8 lines · Christian König

Am 03.08.19 um 02:09 schrieb Andi Kleen:

Quoted 5 lines

From: Andi Kleen <[email protected]>

I got tired of seeing a lot of radeon-crt kernel threads in ps on my workstation, one for each CPU and one for each display, which never use any CPU time. Surely a single kernel thread is enough to handle the display.

NAK, radeon blocks inside the kernel thread and those need to run in parallel or otherwise the hardware can hang.

Do we need one per cpu? or is one per crtc enough?

One per CRTC is enough, but the patch description sounds like it creates only a global one.

Christian.

Quoted 2 lines

Dave.

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx