[Qemu-devel] [4691] PPC TCG Fixes

[malc]

Revision: 4691
          http://svn.sv.gnu.org/viewvc/?view=rev&root=qemu&revision=4691
Author:   malc
Date:     2008-06-07 20:31:33 +0000 (Sat, 07 Jun 2008)

Log Message:
-----------
PPC TCG Fixes

* Fix typo in aliased div2
* "Optimize" aliased div2/divu2
* Fix two remaining branch retranslation problems
  (Kudos to Andrzej Zaborowski)
* Rework goto_tb and set_jmp_target1
* Use correct size when flushing icache
* Use correct register selection for ORI
  (Was harmless since in both cases srcreg was equal to dstreg)

Modified Paths:
--------------
    trunk/exec-all.h
    trunk/tcg/ppc/tcg-target.c

Modified: trunk/exec-all.h
===================================================================
--- trunk/exec-all.h    2008-06-07 08:07:37 UTC (rev 4690)
+++ trunk/exec-all.h    2008-06-07 20:31:33 UTC (rev 4691)
@@ -184,32 +184,37 @@
 #if defined(USE_DIRECT_JUMP)
 
 #if defined(__powerpc__)
+static inline void flush_icache_range(unsigned long start, unsigned long stop);
 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long 
addr)
 {
-    uint32_t val, *ptr;
+    /* This must be in concord with INDEX_op_goto_tb inside tcg_out_op */
+    uint32_t *ptr;
     long disp = addr - jmp_addr;
+    unsigned long patch_size;
 
     ptr = (uint32_t *)jmp_addr;
-    val = *ptr;
 
     if ((disp << 6) >> 6 != disp) {
-        uint16_t *p1;
-
-        p1 = (uint16_t *) ptr;
-        *ptr = (val & ~0x03fffffc) | 4;
-        p1[3] = addr >> 16;
-        p1[5] = addr & 0xffff;
+        ptr[0] = 0x3c000000 | (addr >> 16);    /* lis 0,address@hidden */
+        ptr[1] = 0x60000000 | (addr & 0xffff); /* la  0,address@hidden(0) */
+        ptr[2] = 0x7c0903a6;                   /* mtctr 0 */
+        ptr[3] = 0x4e800420;                   /* brctr */
+        patch_size = 16;
     } else {
         /* patch the branch destination */
-        val = (val & ~0x03fffffc) | (disp & 0x03fffffc);
-        *ptr = val;
+        if (disp != 16) {
+            *ptr = 0x48000000 | (disp & 0x03fffffc); /* b disp */
+            patch_size = 4;
+        } else {
+            ptr[0] = 0x60000000; /* nop */
+            ptr[1] = 0x60000000;
+            ptr[2] = 0x60000000;
+            ptr[3] = 0x60000000;
+            patch_size = 16;
+        }
     }
     /* flush icache */
-    asm volatile ("dcbst 0,%0" : : "r"(ptr) : "memory");
-    asm volatile ("sync" : : : "memory");
-    asm volatile ("icbi 0,%0" : : "r"(ptr) : "memory");
-    asm volatile ("sync" : : : "memory");
-    asm volatile ("isync" : : : "memory");
+    flush_icache_range(jmp_addr, jmp_addr + patch_size);
 }
 #elif defined(__i386__) || defined(__x86_64__)
 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long 
addr)

Modified: trunk/tcg/ppc/tcg-target.c
===================================================================
--- trunk/tcg/ppc/tcg-target.c  2008-06-07 08:07:37 UTC (rev 4690)
+++ trunk/tcg/ppc/tcg-target.c  2008-06-07 20:31:33 UTC (rev 4691)
@@ -388,7 +388,7 @@
     else {
         tcg_out32 (s, ADDIS | RT (ret) | RA (0) | ((arg >> 16) & 0xffff));
         if (arg & 0xffff)
-            tcg_out32 (s, ORI | RT (ret) | RA (ret) | (arg & 0xffff));
+            tcg_out32 (s, ORI | RS (ret) | RA (ret) | (arg & 0xffff));
     }
 }
 
@@ -939,18 +939,14 @@
             tcg_out32 (s, op | RA (arg1) | RB (arg2));
     }
 
-    if (l->has_value) {
-        tcg_target_long disp;
-
-        disp = (tcg_target_long) s->code_ptr - l->u.value;
-        if (disp != (int16_t) disp)
-            tcg_abort ();
-
+    if (l->has_value)
         tcg_out32 (s, tcg_to_bc[cond] | reloc_pc14_val (s->code_ptr,
                                                         l->u.value));
-    }
     else {
-        tcg_out32 (s, tcg_to_bc[cond]);
+        uint16_t val = *(uint16_t *) &s->code_ptr[2];
+
+        /* Thanks to Andrzej Zaborowski */
+        tcg_out32 (s, tcg_to_bc[cond] | (val & 0xfffc));
         tcg_out_reloc (s, s->code_ptr - 4, R_PPC_REL14, label_index, 0);
     }
 }
@@ -1029,24 +1025,9 @@
     case INDEX_op_goto_tb:
         if (s->tb_jmp_offset) {
             /* direct jump method */
-            uint32_t val;
-            uint16_t *p;
 
             s->tb_jmp_offset[args[0]] = s->code_ptr - s->code_buf;
-            /* Thanks to Andrzej Zaborowski for this */
-            val = *(uint32_t *) s->code_ptr & 0x3fffffc;
-
-            tcg_out32 (s, B | val);
-
-            /* For branches outside of LL range
-               This must be in concord with tb_set_jmp_target1 */
-            p = (uint16_t *) s->code_ptr;
-            p[0] = (ADDIS | RT (0) | RA (0)) >> 16;
-            p[2] = (ORI | RT (0) | RA (0)) >> 16;
-            s->code_ptr += 8;
-
-            tcg_out32 (s, MTSPR | RS (0) | CTR);
-            tcg_out32 (s, BCCTR | BO_ALWAYS);
+            s->code_ptr += 16;
         }
         else {
             tcg_abort ();
@@ -1061,7 +1042,10 @@
                 tcg_out_b (s, 0, l->u.value);
             }
             else {
-                tcg_out32 (s, B);
+                uint32_t val = *(uint32_t *) s->code_ptr;
+
+                /* Thanks to Andrzej Zaborowski */
+                tcg_out32 (s, B | (val & 0x3fffffc));
                 tcg_out_reloc (s, s->code_ptr - 4, R_PPC_REL24, args[0], 0);
             }
         }
@@ -1222,10 +1206,10 @@
     case INDEX_op_div2_i32:
         if (args[0] == args[2] || args[0] == args[3]) {
             tcg_out32 (s, DIVW | TAB (0, args[2], args[3]));
+            tcg_out32 (s, MTSPR | RS (0) | CTR);
             tcg_out32 (s, MULLW | TAB (0, 0, args[3]));
-            tcg_out32 (s, SUBF | TAB (0, 0, args[2]));
-            tcg_out32 (s, DIVW | TAB (args[0], args[2], args[3]));
-            tcg_out_mov (s, args[1], 0);
+            tcg_out32 (s, SUBF | TAB (args[1], 0, args[2]));
+            tcg_out32 (s, MFSPR | RT (args[0]) | CTR);
         }
         else {
             tcg_out32 (s, DIVW | TAB (args[0], args[2], args[3]));
@@ -1236,10 +1220,10 @@
     case INDEX_op_divu2_i32:
         if (args[0] == args[2] || args[0] == args[3]) {
             tcg_out32 (s, DIVWU | TAB (0, args[2], args[3]));
+            tcg_out32 (s, MTSPR | RS (0) | CTR);
             tcg_out32 (s, MULLW | TAB (0, 0, args[3]));
-            tcg_out32 (s, SUBF | TAB (0, 0, args[2]));
-            tcg_out32 (s, DIVWU | TAB (args[0], args[2], args[3]));
-            tcg_out_mov (s, args[1], 0);
+            tcg_out32 (s, SUBF | TAB (args[1], 0, args[2]));
+            tcg_out32 (s, MFSPR | RT (args[0]) | CTR);
         }
         else {
             tcg_out32 (s, DIVWU | TAB (args[0], args[2], args[3]));