[Qemu-devel] [RFC 21/28] target-xtensa: implement loop option

[Max Filippov <jcmvbkbc@gmail.com>]

See ISA, 4.3.2 for details.

Operations that change LEND SR value invalidate TBs at the old and at
the new LEND. LEND value at TB compilation time is considered constant
and loop instruction is generated based on this value.

Invalidation may be avoided for the TB at the old LEND address, since
looping code verifies actual LEND value.

Invalidation may be avoided for the TB at the new LEND address if
there's a way to associate LEND address with TB at compilation time and
later verify that it doesn't change.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 target-xtensa/cpu.h       |    3 ++
 target-xtensa/helpers.h   |    1 +
 target-xtensa/op_helper.c |   11 ++++++
 target-xtensa/translate.c |   80 +++++++++++++++++++++++++++++++++++++-------
 4 files changed, 82 insertions(+), 13 deletions(-)

diff --git a/target-xtensa/cpu.h b/target-xtensa/cpu.h
index 3ebccd1..7fade0c 100644
--- a/target-xtensa/cpu.h
+++ b/target-xtensa/cpu.h
@@ -106,6 +106,9 @@ enum {
 };
 
 enum {
+    LBEG = 0,
+    LEND = 1,
+    LCOUNT = 2,
     SAR = 3,
     SCOMPARE1 = 12,
     WINDOW_BASE = 72,
diff --git a/target-xtensa/helpers.h b/target-xtensa/helpers.h
index 4a50280..7e212a3 100644
--- a/target-xtensa/helpers.h
+++ b/target-xtensa/helpers.h
@@ -10,6 +10,7 @@ DEF_HELPER_1(rotw, void, i32)
 DEF_HELPER_2(window_check, void, i32, i32)
 DEF_HELPER_0(restore_owb, void)
 DEF_HELPER_1(movsp, void, i32)
+DEF_HELPER_1(wsr_lend, void, i32)
 DEF_HELPER_0(dump_state, void)
 
 #include "def-helper.h"
diff --git a/target-xtensa/op_helper.c b/target-xtensa/op_helper.c
index b5925dd..f0690ee 100644
--- a/target-xtensa/op_helper.c
+++ b/target-xtensa/op_helper.c
@@ -253,6 +253,17 @@ void HELPER(movsp)(uint32_t pc)
     }
 }
 
+void HELPER(wsr_lend)(uint32_t v)
+{
+    if (env->sregs[LEND] != v) {
+        tb_invalidate_phys_page_range(
+                env->sregs[LEND] - 1, env->sregs[LEND], 0);
+        env->sregs[LEND] = v;
+        tb_invalidate_phys_page_range(
+                env->sregs[LEND] - 1, env->sregs[LEND], 0);
+    }
+}
+
 void HELPER(dump_state)(void)
 {
     cpu_dump_state(env, stderr, fprintf, 0);
diff --git a/target-xtensa/translate.c b/target-xtensa/translate.c
index f3aecaa..e33c2fa 100644
--- a/target-xtensa/translate.c
+++ b/target-xtensa/translate.c
@@ -44,6 +44,7 @@ typedef struct DisasContext {
     const XtensaConfig *config;
     TranslationBlock *tb;
     uint32_t pc;
+    uint32_t lend;
     int is_jmp;
     int singlestep_enabled;
 } DisasContext;
@@ -57,6 +58,9 @@ static TCGv_i32 cpu_UR[256];
 #include "gen-icount.h"
 
 static const char * const sregnames[256] = {
+    [LBEG] = "LBEG",
+    [LEND] = "LEND",
+    [LCOUNT] = "LCOUNT",
     [SAR] = "SAR",
     [SCOMPARE1] = "SCOMPARE1",
     [WINDOW_BASE] = "WINDOW_BASE",
@@ -126,6 +130,11 @@ static void gen_rsr(TCGv_i32 d, int sr)
     }
 }
 
+static void gen_wsr_lend(DisasContext *dc, uint32_t sr, TCGv_i32 v)
+{
+    gen_helper_wsr_lend(v);
+}
+
 static void gen_wsr_windowbase(DisasContext *dc, uint32_t sr, TCGv_i32 v)
 {
     gen_helper_wsr_windowbase(v);
@@ -135,6 +144,7 @@ static void gen_wsr(DisasContext *dc, uint32_t sr, TCGv_i32 s)
 {
     static void (* const wsr_handler[256])(DisasContext *dc,
             uint32_t sr, TCGv_i32 v) = {
+        [LEND] = gen_wsr_lend,
         [WINDOW_BASE] = gen_wsr_windowbase,
     };
 
@@ -212,10 +222,36 @@ static void gen_callw(DisasContext *dc, int _callinc, TCGv_i32 target)
     gen_jump(dc, target);
 }
 
+static void gen_check_loop_end(DisasContext *dc)
+{
+    if (option_enabled(dc, XTENSA_OPTION_LOOP) &&
+            dc->pc == dc->lend) {
+        TCGv_i32 tmp = tcg_temp_new_i32();
+        int label = gen_new_label();
+
+        tcg_gen_andi_i32(tmp, cpu_SR[PS], PS_EXCM);
+        tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, PS_EXCM, label);
+        tcg_gen_brcondi_i32(TCG_COND_NE, cpu_SR[LEND], dc->pc, label);
+        tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_SR[LCOUNT], 0, label);
+        tcg_gen_subi_i32(cpu_SR[LCOUNT], cpu_SR[LCOUNT], 1);
+        gen_jump(dc, cpu_SR[LBEG]);
+        gen_set_label(label);
+        gen_jumpi(dc, dc->pc);
+        tcg_temp_free(tmp);
+    }
+}
+
+static void gen_jumpi_check_loop_end(DisasContext *dc, uint32_t dest)
+{
+    dc->pc = dest;
+    gen_check_loop_end(dc);
+    gen_jumpi(dc, dest);
+}
+
 static void disas_xtensa_insn(DisasContext *dc)
 {
 #define HAS_OPTION(opt) do { \
-        if (!(dc->config->options & (((uint64_t)1) << (opt)))) { \
+        if (!option_enabled(dc, opt)) { \
             goto invalid_opcode; \
         } \
     } while (0)
@@ -1383,7 +1419,7 @@ static void disas_xtensa_insn(DisasContext *dc)
                 }
                 gen_jumpi(dc, dc->pc + 4 + BRI12_IMM12_SE);
                 gen_set_label(label);
-                gen_jumpi(dc, dc->pc + 3);
+                gen_jumpi_check_loop_end(dc, dc->pc + 3);
             }
             break;
 
@@ -1405,7 +1441,7 @@ static void disas_xtensa_insn(DisasContext *dc)
                 }
                 gen_jumpi(dc, dc->pc + 4 + BRI8_IMM8_SE);
                 gen_set_label(label);
-                gen_jumpi(dc, dc->pc + 3);
+                gen_jumpi_check_loop_end(dc, dc->pc + 3);
             }
             break;
 
@@ -1437,15 +1473,29 @@ static void disas_xtensa_insn(DisasContext *dc)
                     break;
 
                 case 8: /*LOOP*/
-                    TBD();
-                    break;
-
                 case 9: /*LOOPNEZ*/
-                    TBD();
-                    break;
-
                 case 10: /*LOOPGTZ*/
-                    TBD();
+                    HAS_OPTION(XTENSA_OPTION_LOOP);
+                    {
+                        uint32_t lend = dc->pc + RRI8_IMM8 + 4;
+                        TCGv_i32 tmp = tcg_const_i32(lend);
+
+                        tcg_gen_subi_i32(cpu_SR[LCOUNT], cpu_R[RRI8_S], 1);
+                        tcg_gen_movi_i32(cpu_SR[LBEG], dc->pc + 3);
+                        gen_wsr_lend(dc, LEND, tmp);
+                        tcg_temp_free(tmp);
+
+                        if (BRI8_R > 8) {
+                            int label = gen_new_label();
+                            tcg_gen_brcondi_i32(
+                                    BRI8_R == 9 ? TCG_COND_NE : TCG_COND_GT,
+                                    cpu_R[RRI8_S], 0, label);
+                            gen_jumpi(dc, lend);
+                            gen_set_label(label);
+                        }
+
+                        gen_jumpi(dc, dc->pc + 3);
+                    }
                     break;
 
                 default: /*reserved*/
@@ -1466,7 +1516,7 @@ static void disas_xtensa_insn(DisasContext *dc)
 
                     gen_jumpi(dc, dc->pc + 4 + BRI8_IMM8_SE);
                     gen_set_label(label);
-                    gen_jumpi(dc, dc->pc + 3);
+                    gen_jumpi_check_loop_end(dc, dc->pc + 3);
                 }
                 break;
             }
@@ -1545,7 +1595,7 @@ static void disas_xtensa_insn(DisasContext *dc)
             }
             gen_jumpi(dc, dc->pc + 4 + RRI8_IMM8_SE);
             gen_set_label(label);
-            gen_jumpi(dc, dc->pc + 3);
+            gen_jumpi_check_loop_end(dc, dc->pc + 3);
         }
         break;
 
@@ -1586,7 +1636,7 @@ static void disas_xtensa_insn(DisasContext *dc)
                     cpu_R[RRRN_S], 0, label);
             gen_jumpi(dc, dc->pc + 4 + (RRRN_R | ((RRRN_T & 3) << 4)));
             gen_set_label(label);
-            gen_jumpi(dc, dc->pc + 2);
+            gen_jumpi_check_loop_end(dc, dc->pc + 2);
         }
         break;
 
@@ -1646,6 +1696,9 @@ static void disas_xtensa_insn(DisasContext *dc)
     } else {
         dc->pc += 3;
     }
+
+    gen_check_loop_end(dc);
+
     return;
 
 invalid_opcode:
@@ -1686,6 +1739,7 @@ static void gen_intermediate_code_internal(
     dc.singlestep_enabled = env->singlestep_enabled;
     dc.tb = tb;
     dc.pc = env->pc;
+    dc.lend = env->sregs[LEND];
     dc.is_jmp = DISAS_NEXT;
 
     gen_icount_start();
-- 
1.7.3.4