Skip to contentSkip to conversation

ARM: shmobile: Multicluster APMU prototype support

PATCH In review Magnus Damm 2 participants 3 messages Apr 8, 2014 – Apr 11, 2014

Subsystems (4)
Arm (32-bit) Arm sub-architectures SuperH Arm/RISC-V/Renesas architecture
lore
Share & tools
lore

Magnus Damm

author maintainer
#0 lore raw

PATCH 1/2 ARM: shmobile: Multicluster APMU prototype support

From: Magnus Damm <[email protected]>

Allow APMU multicluster operation when enabled via the kernel command line. Default is still single cluster operation. Use together with appropriate scheduler modifcations or static CPU affinity.

Signed-off-by Magnus Damm <[email protected]>

---

Applies on top of renesas-devel-v3.14-20140408

 arch/arm/mach-shmobile/platsmp-apmu.c |   16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)
arch/arm/mach-shmobile/platsmp-apmu.c +14 -2
@@ -18,6 +18,19 @@
 #include <asm/smp_plat.h>
 #include <mach/common.h>
 
+/* only enable the cluster that includes the boot CPU by default */
+static bool enable_multicluster = false;
+
+static __init int apmu_setup(char *opt)
+{
+	if (!opt)
+		return -EINVAL;
+	if (!strncmp(opt, "multicluster", 12))
+		enable_multicluster = true;
+	return 0;
+}
+early_param("apmu", apmu_setup);
+
 static struct {
 	void __iomem *iomem;
 	int bit;

@@ -100,8 +113,7 @@ static void apmu_parse_cfg(void (*fn)(st
 	bool is_allowed;
 
 	for (k = 0; k < ARRAY_SIZE(apmu_config); k++) {
-		/* only enable the cluster that includes the boot CPU */
-		is_allowed = false;
+		is_allowed = enable_multicluster;
 		for (bit = 0; bit < ARRAY_SIZE(apmu_config[k].cpus); bit++) {
 			id = apmu_config[k].cpus[bit];
 			if (id >= 0) {
#1 lore raw

Hi Magnus,

arch/arm/mach-shmobile/platsmp-apmu.c line 28

Quoted 13 lines · Magnus Damm

+++ work/arch/arm/mach-shmobile/platsmp-apmu.c 2014-04-08 19:01:55.000000000 +0900

arch/arm/mach-shmobile/platsmp-apmu.c +8 -0
@@ -18,6 +18,19 @@
 #include <asm/smp_plat.h>
 #include <mach/common.h>
 
+/* only enable the cluster that includes the boot CPU by default */
+static bool enable_multicluster = false;
+
+static __init int apmu_setup(char *opt)
+{
+       if (!opt)
+               return -EINVAL;
+       if (!strncmp(opt, "multicluster", 12))

Why do you ignore the characters after "multicluster", i.e. why not simply "!strcmp(opt, "multicluster)"?

arch/arm/mach-shmobile/platsmp-apmu.c line 32

Quoted 4 lines
arch/arm/mach-shmobile/platsmp-apmu.c +4 -0
+               enable_multicluster = true;
+       return 0;
+}
+early_param("apmu", apmu_setup);

This can use some documentation under Documentation/.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#2 lore raw
In reply to Magnus Damm

Hi Geert,

Quoted 8 lines · Geert Uytterhoeven

Hi Magnus,

Quoted 5 lines · Magnus Damm

+++ work/arch/arm/mach-shmobile/platsmp-apmu.c 2014-04-08 19:01:55.000000000 +0900

arch/arm/mach-shmobile/platsmp-apmu.c +0 -0
@@ -18,6 +18,19 @@
 #include <asm/smp_plat.h>
 #include <mach/common.h>
Quoted 11 lines

arch/arm/mach-shmobile/platsmp-apmu.c line 28

Quoted 8 lines
arch/arm/mach-shmobile/platsmp-apmu.c +8 -0
+/* only enable the cluster that includes the boot CPU by default */
+static bool enable_multicluster = false;
+
+static __init int apmu_setup(char *opt)
+{
+       if (!opt)
+               return -EINVAL;
+       if (!strncmp(opt, "multicluster", 12))

Why do you ignore the characters after "multicluster", i.e. why not simply "!strcmp(opt, "multicluster)"?

Good point, thanks.

Quoted 6 lines

arch/arm/mach-shmobile/platsmp-apmu.c line 32

Quoted 4 lines
arch/arm/mach-shmobile/platsmp-apmu.c +4 -0
+               enable_multicluster = true;
+       return 0;
+}
+early_param("apmu", apmu_setup);

This can use some documentation under Documentation/.

If this was destined for upstream merge I totally agree, but in this particular case it is just prototype-level code. The proper solution involves upstream scheduler modifications and perhaps also MCPM once we care about cluster-level power management.

Cheers,

/ magnus