The cpu features are fixed once the probe of cpu features are done.
And the function cpu_has_feature() does be used in some hot path.
The checking of the cpu features for each time of invoking of
cpu_has_feature() seems suboptimal. This tries to reduce this
overhead of this check by using jump label.
The generated assemble code of the following c program:
if (cpu_has_feature(CPU_FTR_XXX))
xxx()
Before:
lis r9,-16230
lwz r9,12324(r9)
lwz r9,12(r9)
andi. r10,r9,512
beqlr-
After:
nop if CPU_FTR_XXX is enabled
b xxx if CPU_FTR_XXX is not enabled
Signed-off-by: Kevin Hao <haoke...@gmail.com>
---
arch/powerpc/include/asm/cpufeatures.h | 20 ++++++++++++++++++++
arch/powerpc/include/asm/cputable.h | 8 ++++++++
arch/powerpc/kernel/cputable.c | 18 ++++++++++++++++++
arch/powerpc/kernel/setup_32.c | 1 +
arch/powerpc/kernel/setup_64.c | 1 +
5 files changed, 48 insertions(+)
diff --git a/arch/powerpc/include/asm/cpufeatures.h
b/arch/powerpc/include/asm/cpufeatures.h
index 37650db5044f..405a97fe6ef9 100644
--- a/arch/powerpc/include/asm/cpufeatures.h
+++ b/arch/powerpc/include/asm/cpufeatures.h
@@ -3,6 +3,25 @@
#include <asm/cputable.h>
+#ifdef CONFIG_JUMP_LABEL
+#include <linux/jump_label.h>
+
+extern struct static_key_true cpu_feat_keys[MAX_CPU_FEATURES];
+
+static inline int cpu_has_feature(unsigned long feature)
+{
+ int i;
+
+ if (CPU_FTRS_ALWAYS & feature)
+ return 1;
+
+ if (!(CPU_FTRS_POSSIBLE & feature))
+ return 0;
+
+ i = __builtin_ctzl(feature);
+ return static_branch_likely(&cpu_feat_keys[i]);
+}
+#else
static inline int cpu_has_feature(unsigned long feature)
{
return (CPU_FTRS_ALWAYS & feature) ||
@@ -10,5 +29,6 @@ static inline int cpu_has_feature(unsigned long feature)
& cur_cpu_spec->cpu_features
& feature);
}
+#endif
#endif /* __ASM_POWERPC_CPUFEATURE_H */
diff --git a/arch/powerpc/include/asm/cputable.h
b/arch/powerpc/include/asm/cputable.h
index ae4b6ef341cd..2ebee2894102 100644
--- a/arch/powerpc/include/asm/cputable.h
+++ b/arch/powerpc/include/asm/cputable.h
@@ -114,6 +114,12 @@ extern void do_feature_fixups(unsigned long value, void
*fixup_start,
extern const char *powerpc_base_platform;
+#ifdef CONFIG_JUMP_LABEL
+extern void cpu_feat_keys_init(void);
+#else
+static inline void cpu_feat_keys_init(void) { }
+#endif
+
/* TLB flush actions. Used as argument to cpu_spec.flush_tlb() hook */
enum {
TLB_INVAL_SCOPE_GLOBAL = 0, /* invalidate all TLBs */
@@ -124,6 +130,8 @@ enum {
/* CPU kernel features */
+#define MAX_CPU_FEATURES (8 * sizeof(((struct cpu_spec
*)0)->cpu_features))
+
/* Retain the 32b definitions all use bottom half of word */
#define CPU_FTR_COHERENT_ICACHE ASM_CONST(0x00000001)
#define CPU_FTR_L2CR ASM_CONST(0x00000002)
diff --git a/arch/powerpc/kernel/cputable.c b/arch/powerpc/kernel/cputable.c
index 7d80bfdfb15e..7d4fe69a61ed 100644
--- a/arch/powerpc/kernel/cputable.c
+++ b/arch/powerpc/kernel/cputable.c
@@ -15,6 +15,7 @@
#include <linux/threads.h>
#include <linux/init.h>
#include <linux/export.h>
+#include <linux/jump_label.h>
#include <asm/oprofile_impl.h>
#include <asm/cputable.h>
@@ -2195,3 +2196,20 @@ struct cpu_spec * __init identify_cpu(unsigned long
offset, unsigned int pvr)
return NULL;
}
+
+#ifdef CONFIG_JUMP_LABEL
+DEFINE_STATIC_KEY_TRUE_ARRAY(cpu_feat_keys, MAX_CPU_FEATURES);
+EXPORT_SYMBOL_GPL(cpu_feat_keys);
+
+void __init cpu_feat_keys_init(void)
+{
+ int i;
+
+ for (i = 0; i < MAX_CPU_FEATURES; i++) {
+ unsigned long f = 1ul << i;
+
+ if (!(cur_cpu_spec->cpu_features & f))
+ static_branch_disable(&cpu_feat_keys[i]);
+ }
+}
+#endif
diff --git a/arch/powerpc/kernel/setup_32.c b/arch/powerpc/kernel/setup_32.c
index f0868f510b3b..93756175a13c 100644
--- a/arch/powerpc/kernel/setup_32.c
+++ b/arch/powerpc/kernel/setup_32.c
@@ -115,6 +115,7 @@ notrace void __init machine_init(u64 dt_ptr)
lockdep_init();
jump_label_init();
+ cpu_feat_keys_init();
/* Enable early debugging if any specified (see udbg.h) */
udbg_early_init();
diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index f0802a0b4a20..4cf3894d91fa 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c
@@ -251,6 +251,7 @@ void __init early_setup(unsigned long dt_ptr)
lockdep_init();
jump_label_init();
+ cpu_feat_keys_init();
/* -------- printk is now safe to use ------- */
--
2.1.0
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
