[Qemu-devel] [PATCH v7 07/26] target/arm: optimize indirect branches

[Richard Henderson]

From: "Emilio G. Cota" <address@hidden>

Speed up indirect branches by jumping to the target if it is valid.

Softmmu measurements (see later commit for user-mode results):

Note: baseline (i.e. speedup == 1x) is QEMU v2.9.0.

- Impact on Boot time

| setup  | ARM debian jessie boot+shutdown time | stddev |
|--------+--------------------------------------+--------|
| v2.9.0 |                                 8.84 |   0.07 |
| +cross |                                 8.85 |   0.03 |
| +jr    |                                 8.83 |   0.06 |

-                            NBench, arm-softmmu (debian jessie guest). Host: 
Intel i7-4790K @ 4.00GHz

  1.3x 
+-+-------------------------------------------------------------------------------------------------------------+-+
       |                                                                        
                                         |
       |   cross                                                          ####  
                                         |
 1.25x 
+cross+jr..........................................................#++#.........................................+-+
       |                                                        ####      #  #  
                                         |
       |                                                     +++#  #      #  #  
                                         |
       |                                      +++            ****  #      #  #  
                                         |
  1.2x 
+-+...................................####............*..*..#......#..#.........................................+-+
       |                                  ****  #            *  *  #      #  #  
   ####                                  |
       |                                  *  *  #            *  *  #      #  #  
   #  #                                  |
 1.15x 
+-+................................*..*..#............*..*..#......#..#.....#..#................................+-+
       |                                  *  *  #            *  *  #      #  #  
   #  #                                  |
       |                                  *  *  #      ####  *  *  #      #  #  
   #  #                                  |
       |                                  *  *  #      #  #  *  *  #      #  #  
   #  #                         ####     |
  1.1x 
+-+................................*..*..#......#..#..*..*..#......#..#.....#..#.........................#..#...+-+
       |                                  *  *  #      #  #  *  *  #      #  #  
   #  #                         #  #     |
       |                                  *  *  #      #  #  *  *  #      #  #  
   #  #                         #  #     |
 1.05x 
+-+..........................####..*..*..#......#..#..*..*..#......#..#.....#..#......+++............*****..#...+-+
       |                        *****  #  *  *  #      #  #  *  *  #  *****  #  
   #  #   +++ |    ****###  *   *  #     |
       |                        *+++*  #  *  *  #      #  #  *  *  #  *+++*  #  
****  #  *****###  *  *  #  *   *  #     |
       |     *****###  +++####  *   *  #  *  *  #  *****  #  *  *  #  *   *  #  
*  *  #  * | *++#  *  *  #  *   *  #     |
    1x 
+-++-+*+++*-+#++****++#++*+-+*++#+-*++*++#-+*+++*-+#++*++*++#++*+-+*++#+-*++*++#-+*+++*-+#++*++*++#++*+-+*++#+-++-+
       |     *   *  #  *  *  #  *   *  #  *  *  #  *   *  #  *  *  #  *   *  #  
*  *  #  *   *  #  *  *  #  *   *  #     |
       |     *   *  #  *  *  #  *   *  #  *  *  #  *   *  #  *  *  #  *   *  #  
*  *  #  *   *  #  *  *  #  *   *  #     |
 0.95x 
+-+---*****###--****###--*****###--****###--*****###--****###--*****###--****###--*****###--****###--*****###---+-+
       ASSIGNMENT BITFIELD   FOURFP EMULATION   HUFFMAN   LU DECOMPOSITIONEURAL 
NNUMERIC SOSTRING SORT     hmean
  png: http://imgur.com/eOLmZNR

NB. 'cross' represents the previous commit.

Signed-off-by: Emilio G. Cota <address@hidden>
Message-Id: <address@hidden>
[rth: Replace gen_jr global variable with DISAS_EXIT state.]
Signed-off-by: Richard Henderson <address@hidden>
---
 target/arm/translate.c | 25 ++++++++++++++++---------
 target/arm/translate.h |  4 ++++
 2 files changed, 20 insertions(+), 9 deletions(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index facb52f..f879da6 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -1178,7 +1178,7 @@ static void gen_exception_internal_insn(DisasContext *s, 
int offset, int excp)
     gen_set_condexec(s);
     gen_set_pc_im(s, s->pc - offset);
     gen_exception_internal(excp);
-    s->is_jmp = DISAS_JUMP;
+    s->is_jmp = DISAS_EXC;
 }
 
 static void gen_exception_insn(DisasContext *s, int offset, int excp,
@@ -1187,14 +1187,14 @@ static void gen_exception_insn(DisasContext *s, int 
offset, int excp,
     gen_set_condexec(s);
     gen_set_pc_im(s, s->pc - offset);
     gen_exception(excp, syn, target_el);
-    s->is_jmp = DISAS_JUMP;
+    s->is_jmp = DISAS_EXC;
 }
 
 /* Force a TB lookup after an instruction that changes the CPU state.  */
 static inline void gen_lookup_tb(DisasContext *s)
 {
     tcg_gen_movi_i32(cpu_R[15], s->pc & ~1);
-    s->is_jmp = DISAS_JUMP;
+    s->is_jmp = DISAS_EXIT;
 }
 
 static inline void gen_hlt(DisasContext *s, int imm)
@@ -4146,19 +4146,23 @@ static inline bool use_goto_tb(DisasContext *s, 
target_ulong dest)
 #endif
 }
 
-static inline void gen_goto_tb(DisasContext *s, int n, target_ulong dest)
+static void gen_goto_ptr(void)
+{
+    TCGv addr = tcg_temp_new();
+    tcg_gen_extu_i32_tl(addr, cpu_R[15]);
+    tcg_gen_lookup_and_goto_ptr(addr);
+    tcg_temp_free(addr);
+}
+
+static void gen_goto_tb(DisasContext *s, int n, target_ulong dest)
 {
     if (use_goto_tb(s, dest)) {
         tcg_gen_goto_tb(n);
         gen_set_pc_im(s, dest);
         tcg_gen_exit_tb((uintptr_t)s->tb + n);
     } else {
-        TCGv addr = tcg_temp_new();
-
         gen_set_pc_im(s, dest);
-        tcg_gen_extu_i32_tl(addr, cpu_R[15]);
-        tcg_gen_lookup_and_goto_ptr(addr);
-        tcg_temp_free(addr);
+        gen_goto_ptr();
     }
 }
 
@@ -12091,11 +12095,14 @@ void gen_intermediate_code(CPUARMState *env, 
TranslationBlock *tb)
             gen_set_pc_im(dc, dc->pc);
             /* fall through */
         case DISAS_JUMP:
+            gen_goto_ptr();
+            break;
         default:
             /* indicate that the hash table must be used to find the next TB */
             tcg_gen_exit_tb(0);
             break;
         case DISAS_TB_JUMP:
+        case DISAS_EXC:
             /* nothing more to generate */
             break;
         case DISAS_WFI:
diff --git a/target/arm/translate.h b/target/arm/translate.h
index 629dab9..93de13f 100644
--- a/target/arm/translate.h
+++ b/target/arm/translate.h
@@ -139,6 +139,10 @@ static void disas_set_insn_syndrome(DisasContext *s, 
uint32_t syn)
  * custom end-of-TB code)
  */
 #define DISAS_BX_EXCRET 11
+/* For instructions which want an immediate exit to the main loop,
+ * as opposed to attempting to use lookup_and_goto_ptr.
+ */
+#define DISAS_EXIT 12
 
 #ifdef TARGET_AARCH64
 void a64_translate_init(void);
-- 
2.9.4