accel/tcg: Precompute curr_cflags into cpu->tcg_cflags

The primary motivation is to remove a dozen insns along
the fast-path in tb_lookup.  As a byproduct, this allows
us to completely remove parallel_cpus.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

6 files changed, 11 insertions, 10 deletions

diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c
index 931da96c2b..bdfa036ac8 100644
--- a/accel/tcg/cpu-exec.c
+++ b/accel/tcg/cpu-exec.c
@@ -267,8 +267,6 @@ void cpu_exec_step_atomic(CPUState *cpu)
             mmap_unlock();
         }
 
-        /* Since we got here, we know that parallel_cpus must be true.  */
-        parallel_cpus = false;
         cpu_exec_enter(cpu);
         /* execute the generated code */
         trace_exec_tb(tb, pc);
@@ -296,7 +294,6 @@ void cpu_exec_step_atomic(CPUState *cpu)
      * the execution.
      */
     g_assert(cpu_in_exclusive_context(cpu));
-    parallel_cpus = true;
     cpu->running = false;
     end_exclusive();
 }
diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index 42973fb062..847d2079d2 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -114,8 +114,7 @@ void mttcg_start_vcpu_thread(CPUState *cpu)
     char thread_name[VCPU_THREAD_NAME_SIZE];
 
     g_assert(tcg_enabled());
-
-    parallel_cpus = (current_machine->smp.max_cpus > 1);
+    tcg_cpu_init_cflags(cpu, current_machine->smp.max_cpus > 1);
 
     cpu->thread = g_malloc0(sizeof(QemuThread));
     cpu->halt_cond = g_malloc0(sizeof(QemuCond));
diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index 4a66055e0d..018b54c508 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -269,7 +269,7 @@ void rr_start_vcpu_thread(CPUState *cpu)
     static QemuThread *single_tcg_cpu_thread;
 
     g_assert(tcg_enabled());
-    parallel_cpus = false;
+    tcg_cpu_init_cflags(cpu, false);
 
     if (!single_tcg_cpu_thread) {
         cpu->thread = g_malloc0(sizeof(QemuThread));
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 6144d9df87..6cdcaa2855 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -41,6 +41,14 @@
 
 /* common functionality among all TCG variants */
 
+void tcg_cpu_init_cflags(CPUState *cpu, bool parallel)
+{
+    uint32_t cflags = cpu->cluster_index << CF_CLUSTER_SHIFT;
+    cflags |= parallel ? CF_PARALLEL : 0;
+    cflags |= icount_enabled() ? CF_USE_ICOUNT : 0;
+    cpu->tcg_cflags = cflags;
+}
+
 void tcg_cpus_destroy(CPUState *cpu)
 {
     cpu_thread_signal_destroyed(cpu);
diff --git a/accel/tcg/tcg-accel-ops.h b/accel/tcg/tcg-accel-ops.h
index 48130006de..6a5fcef889 100644
--- a/accel/tcg/tcg-accel-ops.h
+++ b/accel/tcg/tcg-accel-ops.h
@@ -17,5 +17,6 @@
 void tcg_cpus_destroy(CPUState *cpu);
 int tcg_cpus_exec(CPUState *cpu);
 void tcg_handle_interrupt(CPUState *cpu, int mask);
+void tcg_cpu_init_cflags(CPUState *cpu, bool parallel);
 
 #endif /* TCG_CPUS_H */
diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
index 0b0bfd35ab..f32df8b240 100644
--- a/accel/tcg/translate-all.c
+++ b/accel/tcg/translate-all.c
@@ -224,7 +224,6 @@ static void *l1_map[V_L1_MAX_SIZE];
 TCGContext tcg_init_ctx;
 __thread TCGContext *tcg_ctx;
 TBContext tb_ctx;
-bool parallel_cpus;
 
 static void page_table_config_init(void)
 {
@@ -1867,9 +1866,6 @@ TranslationBlock *tb_gen_code(CPUState *cpu,
         cflags = (cflags & ~CF_COUNT_MASK) | 1;
     }
 
-    cflags &= ~CF_CLUSTER_MASK;
-    cflags |= cpu->cluster_index << CF_CLUSTER_SHIFT;
-
     max_insns = cflags & CF_COUNT_MASK;
     if (max_insns == 0) {
         max_insns = CF_COUNT_MASK;