slight optimisation to cur_tcb on riscv

1 files changed, 10 insertions, 7 deletions

diff --git a/arch/riscv64/kernel/cpu.c b/arch/riscv64/kernel/cpu.c
index 38cf262..4273d02 100644
--- a/arch/riscv64/kernel/cpu.c
+++ b/arch/riscv64/kernel/cpu.c
@@ -16,12 +16,15 @@ void cpu_assign(struct tcb *t)
 
 id_t cpu_id()
 {
-	/* yes, slightly weird situation where cur_tcb() call cpu_id() which
-	 * gets the current tcb and returns the cpu id in the tcb, so that that
-	 * id can be used to get the tcb we want. I might try to work out
-	 * something slightly smarter, although this is likely not going to have
-	 * basically any kind of importance for perfomance. */
-	struct tcb *t;
-	__asm__ volatile ("mv %0, tp\n" : "=r" (t) ::);
+	struct tcb *t = cur_tcb();
 	return t->cpu_id;
 }
+
+/* override cur_tcb() in common/tcb.c since on risc-v this is apparently the
+ * optimal strategy. */
+/* @todo should this be the default for all arches? */
+struct tcb *cur_tcb()
+{
+	register struct tcb *t __asm__ ("tp");
+	return t;
+}