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>

diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c
index 931da96c2b0998dd03a3e5c8786f3de0e56494d9..bdfa036ac87484ea5b5aef53938e6ade50116e69 100644 (file)
--- 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 42973fb062b8f97a3880bd714c8c1a505b252468..847d2079d21f1eedcb3e214ccd1ffec845ce39f6 100644 (file)
--- 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 4a66055e0d7b62a3edf1d1ae7ca1990b348f4622..018b54c508fac1c7f5c7f8ff0c83baa044c2a9b0 100644 (file)
--- 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 6144d9df8779525720a411defb01ca6607b3b213..6cdcaa2855c8e90cf1a8224ff04202d6c2c7b959 100644 (file)
--- 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 48130006de6ba188fce7124f59f98153ea2bd125..6a5fcef889808c681ddb48060dd4dbf34bc34036 100644 (file)
--- 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 0b0bfd35ab80f207ade78df621064fbfee93ec14..f32df8b24042b2aeabc801945b81290b3562ff4e 100644 (file)
--- 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;

diff --git a/include/exec/exec-all.h b/include/exec/exec-all.h
index a262dd580497b7368d4294585ff2a71344a0938d..6b036cae8f653108eac8050c557d9f2554cf5b5f 100644 (file)
--- a/include/exec/exec-all.h
+++ b/include/exec/exec-all.h
@@ -513,8 +513,6 @@ struct TranslationBlock {
     uintptr_t jmp_dest[2];
 };
 
-extern bool parallel_cpus;
-
 /* Hide the qatomic_read to make code a little easier on the eyes */
 static inline uint32_t tb_cflags(const TranslationBlock *tb)
 {
@@ -524,10 +522,7 @@ static inline uint32_t tb_cflags(const TranslationBlock *tb)
 /* current cflags for hashing/comparison */
 static inline uint32_t curr_cflags(CPUState *cpu)
 {
-    uint32_t cflags = deposit32(0, CF_CLUSTER_SHIFT, 8, cpu->cluster_index);
-    cflags |= parallel_cpus ? CF_PARALLEL : 0;
-    cflags |= icount_enabled() ? CF_USE_ICOUNT : 0;
-    return cflags;
+    return cpu->tcg_cflags;
 }
 
 /* TranslationBlock invalidate API */

diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index c005d3dc2d814bd3568173cea6edf1a6598b4bd3..c68bc3ba8aff0d204d678f617ef3aaf3195f0b6c 100644 (file)
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -282,6 +282,7 @@ struct qemu_work_item;
  *   to a cluster this will be UNASSIGNED_CLUSTER_INDEX; otherwise it will
  *   be the same as the cluster-id property of the CPU object's TYPE_CPU_CLUSTER
  *   QOM parent.
+ * @tcg_cflags: Pre-computed cflags for this cpu.
  * @nr_cores: Number of cores within this CPU package.
  * @nr_threads: Number of threads within this CPU.
  * @running: #true if CPU is currently running (lockless).
@@ -412,6 +413,7 @@ struct CPUState {
     /* TODO Move common fields from CPUArchState here. */
     int cpu_index;
     int cluster_index;
+    uint32_t tcg_cflags;
     uint32_t halted;
     uint32_t can_do_io;
     int32_t exception_index;

diff --git a/linux-user/main.c b/linux-user/main.c
index 81f48ff54ed4bf8cab297310bd7190a01b0e4aca..4f4746dce8253200699b909e1e8afd1d08f5b93c 100644 (file)
--- a/linux-user/main.c
+++ b/linux-user/main.c
@@ -205,6 +205,7 @@ CPUArchState *cpu_copy(CPUArchState *env)
     /* Reset non arch specific state */
     cpu_reset(new_cpu);
 
+    new_cpu->tcg_cflags = cpu->tcg_cflags;
     memcpy(new_env, env, sizeof(CPUArchState));
 
     /* Clone all break/watchpoints.

diff --git a/linux-user/sh4/signal.c b/linux-user/sh4/signal.c
index cc89a48ff8bc78a62e2fc6f2498b97d303c1dc0e..29c1ee30e6d6b3f8c0a90c207f4fcf337f01dd49 100644 (file)
--- a/linux-user/sh4/signal.c
+++ b/linux-user/sh4/signal.c
@@ -82,9 +82,11 @@ static abi_ulong get_sigframe(struct target_sigaction *ka,
     return (sp - frame_size) & -8ul;
 }
 
-/* Notice when we're in the middle of a gUSA region and reset.
-   Note that this will only occur for !parallel_cpus, as we will
-   translate such sequences differently in a parallel context.  */
+/*
+ * Notice when we're in the middle of a gUSA region and reset.
+ * Note that this will only occur when #CF_PARALLEL is unset, as we
+ * will translate such sequences differently in a parallel context.
+ */
 static void unwind_gusa(CPUSH4State *regs)
 {
     /* If the stack pointer is sufficiently negative, and we haven't

diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 389ec09764734b2dc71d111181caf9d6fb68337e..9522f603aad6f1f7762f2f3edc9ca25bea2c22c1 100644 (file)
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -6481,6 +6481,16 @@ static int do_fork(CPUArchState *env, unsigned int flags, abi_ulong newsp,
         /* Grab a mutex so that thread setup appears atomic.  */
         pthread_mutex_lock(&clone_lock);
 
+        /*
+         * If this is our first additional thread, we need to ensure we
+         * generate code for parallel execution and flush old translations.
+         * Do this now so that the copy gets CF_PARALLEL too.
+         */
+        if (!(cpu->tcg_cflags & CF_PARALLEL)) {
+            cpu->tcg_cflags |= CF_PARALLEL;
+            tb_flush(cpu);
+        }
+
         /* we create a new CPU instance. */
         new_env = cpu_copy(env);
         /* Init regs that differ from the parent.  */
@@ -6521,14 +6531,6 @@ static int do_fork(CPUArchState *env, unsigned int flags, abi_ulong newsp,
         sigprocmask(SIG_BLOCK, &sigmask, &info.sigmask);
         cpu->random_seed = qemu_guest_random_seed_thread_part1();
 
-        /* If this is our first additional thread, we need to ensure we
-         * generate code for parallel execution and flush old translations.
-         */
-        if (!parallel_cpus) {
-            parallel_cpus = true;
-            tb_flush(cpu);
-        }
-
         ret = pthread_create(&info.thread, &attr, clone_func, &info);
         /* TODO: Free new CPU state if thread creation failed.  */