target-sparc: Change fpr representation to doubles.

This allows a more efficient representation for 64-bit hosts.
It should be about the same for 32-bit hosts, as we can still
access the individual pieces of the double.

Signed-off-by: Richard Henderson <rth@twiddle.net>

5 files changed, 114 insertions, 140 deletions

diff --git a/target-sparc/cpu.h b/target-sparc/cpu.h
index 4eace33c47..38a707466c 100644
--- a/target-sparc/cpu.h
+++ b/target-sparc/cpu.h
@@ -3,16 +3,17 @@
 
 #include "config.h"
 #include "qemu-common.h"
+#include "bswap.h"
 
 #if !defined(TARGET_SPARC64)
 #define TARGET_LONG_BITS 32
-#define TARGET_FPREGS 32
+#define TARGET_DPREGS 16
 #define TARGET_PAGE_BITS 12 /* 4k */
 #define TARGET_PHYS_ADDR_SPACE_BITS 36
 #define TARGET_VIRT_ADDR_SPACE_BITS 32
 #else
 #define TARGET_LONG_BITS 64
-#define TARGET_FPREGS 64
+#define TARGET_DPREGS 32
 #define TARGET_PAGE_BITS 13 /* 8k */
 #define TARGET_PHYS_ADDR_SPACE_BITS 41
 # ifdef TARGET_ABI32
@@ -395,7 +396,7 @@ typedef struct CPUSPARCState {
 
     uint32_t psr;      /* processor state register */
     target_ulong fsr;      /* FPU state register */
-    float32 fpr[TARGET_FPREGS];  /* floating point registers */
+    CPU_DoubleU fpr[TARGET_DPREGS];  /* floating point registers */
     uint32_t cwp;      /* index of current register window (extracted
                           from PSR) */
 #if !defined(TARGET_SPARC64) || defined(TARGET_ABI32)
diff --git a/target-sparc/cpu_init.c b/target-sparc/cpu_init.c
index 6954800af0..c7269b54a8 100644
--- a/target-sparc/cpu_init.c
+++ b/target-sparc/cpu_init.c
@@ -813,11 +813,11 @@ void cpu_dump_state(CPUState *env, FILE *f, fprintf_function cpu_fprintf,
         }
     }
     cpu_fprintf(f, "\nFloating Point Registers:\n");
-    for (i = 0; i < TARGET_FPREGS; i++) {
+    for (i = 0; i < TARGET_DPREGS; i++) {
         if ((i & 3) == 0) {
-            cpu_fprintf(f, "%%f%02d:", i);
+            cpu_fprintf(f, "%%f%02d:", i * 2);
         }
-        cpu_fprintf(f, " %016f", *(float *)&env->fpr[i]);
+        cpu_fprintf(f, " %016" PRIx64, env->fpr[i].ll);
         if ((i & 3) == 3) {
             cpu_fprintf(f, "\n");
         }
diff --git a/target-sparc/ldst_helper.c b/target-sparc/ldst_helper.c
index 80e540818f..b59707ecd2 100644
--- a/target-sparc/ldst_helper.c
+++ b/target-sparc/ldst_helper.c
@@ -2045,7 +2045,7 @@ void helper_ldda_asi(target_ulong addr, int asi, int rd)
 void helper_ldf_asi(target_ulong addr, int asi, int size, int rd)
 {
     unsigned int i;
-    CPU_DoubleU u;
+    target_ulong val;
 
     helper_check_align(addr, 3);
     addr = asi_address_mask(env, asi, addr);
@@ -2060,13 +2060,11 @@ void helper_ldf_asi(target_ulong addr, int asi, int size, int rd)
             return;
         }
         helper_check_align(addr, 0x3f);
-        for (i = 0; i < 16; i++) {
-            *(uint32_t *)&env->fpr[rd++] = helper_ld_asi(addr, asi & 0x8f, 4,
-                                                         0);
-            addr += 4;
+        for (i = 0; i < 8; i++, rd += 2, addr += 8) {
+            env->fpr[rd/2].ll = helper_ld_asi(addr, asi & 0x8f, 8, 0);
         }
-
         return;
+
     case 0x16: /* UA2007 Block load primary, user privilege */
     case 0x17: /* UA2007 Block load secondary, user privilege */
     case 0x1e: /* UA2007 Block load primary LE, user privilege */
@@ -2080,13 +2078,11 @@ void helper_ldf_asi(target_ulong addr, int asi, int size, int rd)
             return;
         }
         helper_check_align(addr, 0x3f);
-        for (i = 0; i < 16; i++) {
-            *(uint32_t *)&env->fpr[rd++] = helper_ld_asi(addr, asi & 0x19, 4,
-                                                         0);
-            addr += 4;
+        for (i = 0; i < 8; i++, rd += 2, addr += 4) {
+            env->fpr[rd/2].ll = helper_ld_asi(addr, asi & 0x19, 8, 0);
         }
-
         return;
+
     default:
         break;
     }
@@ -2094,20 +2090,19 @@ void helper_ldf_asi(target_ulong addr, int asi, int size, int rd)
     switch (size) {
     default:
     case 4:
-        *((uint32_t *)&env->fpr[rd]) = helper_ld_asi(addr, asi, size, 0);
+        val = helper_ld_asi(addr, asi, size, 0);
+        if (rd & 1) {
+            env->fpr[rd/2].l.lower = val;
+        } else {
+            env->fpr[rd/2].l.upper = val;
+        }
         break;
     case 8:
-        u.ll = helper_ld_asi(addr, asi, size, 0);
-        *((uint32_t *)&env->fpr[rd++]) = u.l.upper;
-        *((uint32_t *)&env->fpr[rd++]) = u.l.lower;
+        env->fpr[rd/2].ll = helper_ld_asi(addr, asi, size, 0);
         break;
     case 16:
-        u.ll = helper_ld_asi(addr, asi, 8, 0);
-        *((uint32_t *)&env->fpr[rd++]) = u.l.upper;
-        *((uint32_t *)&env->fpr[rd++]) = u.l.lower;
-        u.ll = helper_ld_asi(addr + 8, asi, 8, 0);
-        *((uint32_t *)&env->fpr[rd++]) = u.l.upper;
-        *((uint32_t *)&env->fpr[rd++]) = u.l.lower;
+        env->fpr[rd/2].ll = helper_ld_asi(addr, asi, 8, 0);
+        env->fpr[rd/2 + 1].ll = helper_ld_asi(addr + 8, asi, 8, 0);
         break;
     }
 }
@@ -2115,8 +2110,7 @@ void helper_ldf_asi(target_ulong addr, int asi, int size, int rd)
 void helper_stf_asi(target_ulong addr, int asi, int size, int rd)
 {
     unsigned int i;
-    target_ulong val = 0;
-    CPU_DoubleU u;
+    target_ulong val;
 
     helper_check_align(addr, 3);
     addr = asi_address_mask(env, asi, addr);
@@ -2133,10 +2127,8 @@ void helper_stf_asi(target_ulong addr, int asi, int size, int rd)
             return;
         }
         helper_check_align(addr, 0x3f);
-        for (i = 0; i < 16; i++) {
-            val = *(uint32_t *)&env->fpr[rd++];
-            helper_st_asi(addr, val, asi & 0x8f, 4);
-            addr += 4;
+        for (i = 0; i < 8; i++, rd += 2, addr += 8) {
+            helper_st_asi(addr, env->fpr[rd/2].ll, asi & 0x8f, 8);
         }
 
         return;
@@ -2153,10 +2145,8 @@ void helper_stf_asi(target_ulong addr, int asi, int size, int rd)
             return;
         }
         helper_check_align(addr, 0x3f);
-        for (i = 0; i < 16; i++) {
-            val = *(uint32_t *)&env->fpr[rd++];
-            helper_st_asi(addr, val, asi & 0x19, 4);
-            addr += 4;
+        for (i = 0; i < 8; i++, rd += 2, addr += 8) {
+            helper_st_asi(addr, env->fpr[rd/2].ll, asi & 0x19, 8);
         }
 
         return;
@@ -2167,20 +2157,19 @@ void helper_stf_asi(target_ulong addr, int asi, int size, int rd)
     switch (size) {
     default:
     case 4:
-        helper_st_asi(addr, *(uint32_t *)&env->fpr[rd], asi, size);
+        if (rd & 1) {
+            val = env->fpr[rd/2].l.lower;
+        } else {
+            val = env->fpr[rd/2].l.upper;
+        }
+        helper_st_asi(addr, val, asi, size);
         break;
     case 8:
-        u.l.upper = *(uint32_t *)&env->fpr[rd++];
-        u.l.lower = *(uint32_t *)&env->fpr[rd++];
-        helper_st_asi(addr, u.ll, asi, size);
+        helper_st_asi(addr, env->fpr[rd/2].ll, asi, size);
         break;
     case 16:
-        u.l.upper = *(uint32_t *)&env->fpr[rd++];
-        u.l.lower = *(uint32_t *)&env->fpr[rd++];
-        helper_st_asi(addr, u.ll, asi, 8);
-        u.l.upper = *(uint32_t *)&env->fpr[rd++];
-        u.l.lower = *(uint32_t *)&env->fpr[rd++];
-        helper_st_asi(addr + 8, u.ll, asi, 8);
+        helper_st_asi(addr, env->fpr[rd/2].ll, asi, 8);
+        helper_st_asi(addr + 8, env->fpr[rd/2 + 1].ll, asi, 8);
         break;
     }
 }
diff --git a/target-sparc/machine.c b/target-sparc/machine.c
index 56ae0412cd..235b088a45 100644
--- a/target-sparc/machine.c
+++ b/target-sparc/machine.c
@@ -21,13 +21,9 @@ void cpu_save(QEMUFile *f, void *opaque)
         qemu_put_betls(f, &env->regbase[i]);
 
     /* FPU */
-    for(i = 0; i < TARGET_FPREGS; i++) {
-        union {
-            float32 f;
-            uint32_t i;
-        } u;
-        u.f = env->fpr[i];
-        qemu_put_be32(f, u.i);
+    for (i = 0; i < TARGET_DPREGS; i++) {
+        qemu_put_be32(f, env->fpr[i].l.upper);
+        qemu_put_be32(f, env->fpr[i].l.lower);
     }
 
     qemu_put_betls(f, &env->pc);
@@ -128,13 +124,9 @@ int cpu_load(QEMUFile *f, void *opaque, int version_id)
         qemu_get_betls(f, &env->regbase[i]);
 
     /* FPU */
-    for(i = 0; i < TARGET_FPREGS; i++) {
-        union {
-            float32 f;
-            uint32_t i;
-        } u;
-        u.i = qemu_get_be32(f);
-        env->fpr[i] = u.f;
+    for (i = 0; i < TARGET_DPREGS; i++) {
+        env->fpr[i].l.upper = qemu_get_be32(f);
+        env->fpr[i].l.lower = qemu_get_be32(f);
     }
 
     qemu_get_betls(f, &env->pc);
diff --git a/target-sparc/translate.c b/target-sparc/translate.c
index 0b95b64ca4..2c123b1e09 100644
--- a/target-sparc/translate.c
+++ b/target-sparc/translate.c
@@ -63,7 +63,7 @@ static TCGv cpu_tmp0;
 static TCGv_i32 cpu_tmp32;
 static TCGv_i64 cpu_tmp64;
 /* Floating point registers */
-static TCGv_i32 cpu_fpr[TARGET_FPREGS];
+static TCGv_i64 cpu_fpr[TARGET_DPREGS];
 
 static target_ulong gen_opc_npc[OPC_BUF_SIZE];
 static target_ulong gen_opc_jump_pc[2];
@@ -82,8 +82,8 @@ typedef struct DisasContext {
     uint32_t cc_op;  /* current CC operation */
     struct TranslationBlock *tb;
     sparc_def_t *def;
-    TCGv_i64 t64[3];
-    int n_t64;
+    TCGv_i32 t32[3];
+    int n_t32;
 } DisasContext;
 
 // This function uses non-native bit order
@@ -126,12 +126,44 @@ static inline void gen_update_fprs_dirty(int rd)
 /* floating point registers moves */
 static TCGv_i32 gen_load_fpr_F(DisasContext *dc, unsigned int src)
 {
-    return cpu_fpr[src];
+#if TCG_TARGET_REG_BITS == 32
+    if (src & 1) {
+        return TCGV_LOW(cpu_fpr[src / 2]);
+    } else {
+        return TCGV_HIGH(cpu_fpr[src / 2]);
+    }
+#else
+    if (src & 1) {
+        return MAKE_TCGV_I32(GET_TCGV_I64(cpu_fpr[src / 2]));
+    } else {
+        TCGv_i32 ret = tcg_temp_local_new_i32();
+        TCGv_i64 t = tcg_temp_new_i64();
+
+        tcg_gen_shri_i64(t, cpu_fpr[src / 2], 32);
+        tcg_gen_trunc_i64_i32(ret, t);
+        tcg_temp_free_i64(t);
+
+        dc->t32[dc->n_t32++] = ret;
+        assert(dc->n_t32 <= ARRAY_SIZE(dc->t32));
+
+        return ret;
+    }
+#endif
 }
 
 static void gen_store_fpr_F(DisasContext *dc, unsigned int dst, TCGv_i32 v)
 {
-    tcg_gen_mov_i32(cpu_fpr[dst], v);
+#if TCG_TARGET_REG_BITS == 32
+    if (dst & 1) {
+        tcg_gen_mov_i32(TCGV_LOW(cpu_fpr[dst / 2]), v);
+    } else {
+        tcg_gen_mov_i32(TCGV_HIGH(cpu_fpr[dst / 2]), v);
+    }
+#else
+    TCGv_i64 t = MAKE_TCGV_I64(GET_TCGV_I32(v));
+    tcg_gen_deposit_i64(cpu_fpr[dst / 2], cpu_fpr[dst / 2], t,
+                        (dst & 1 ? 0 : 32), 32);
+#endif
     gen_update_fprs_dirty(dst);
 }
 
@@ -142,42 +174,14 @@ static TCGv_i32 gen_dest_fpr_F(void)
 
 static TCGv_i64 gen_load_fpr_D(DisasContext *dc, unsigned int src)
 {
-    TCGv_i64 ret = tcg_temp_new_i64();
     src = DFPREG(src);
-
-#if TCG_TARGET_REG_BITS == 32
-    tcg_gen_mov_i32(TCGV_HIGH(ret), cpu_fpr[src]);
-    tcg_gen_mov_i32(TCGV_LOW(ret), cpu_fpr[src + 1]);
-#else
-    {
-        TCGv_i64 t = tcg_temp_new_i64();
-        tcg_gen_extu_i32_i64(ret, cpu_fpr[src]);
-        tcg_gen_extu_i32_i64(t, cpu_fpr[src + 1]);
-        tcg_gen_shli_i64(ret, ret, 32);
-        tcg_gen_or_i64(ret, ret, t);
-        tcg_temp_free_i64(t);
-    }
-#endif
-
-    dc->t64[dc->n_t64++] = ret;
-    assert(dc->n_t64 <= ARRAY_SIZE(dc->t64));
-
-    return ret;
+    return cpu_fpr[src / 2];
 }
 
 static void gen_store_fpr_D(DisasContext *dc, unsigned int dst, TCGv_i64 v)
 {
     dst = DFPREG(dst);
-
-#if TCG_TARGET_REG_BITS == 32
-    tcg_gen_mov_i32(cpu__fpu[dst], TCGV_HIGH(v));
-    tcg_gen_mov_i32(cpu__fpu[dst + 1], TCGV_LOW(v));
-#else
-    tcg_gen_trunc_i64_i32(cpu_fpr[dst + 1], v);
-    tcg_gen_shri_i64(v, v, 32);
-    tcg_gen_trunc_i64_i32(cpu_fpr[dst], v);
-#endif
-
+    tcg_gen_mov_i64(cpu_fpr[dst / 2], v);
     gen_update_fprs_dirty(dst);
 }
 
@@ -188,50 +192,36 @@ static TCGv_i64 gen_dest_fpr_D(void)
 
 static void gen_op_load_fpr_QT0(unsigned int src)
 {
-    tcg_gen_st_i32(cpu_fpr[src], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.upmost));
-    tcg_gen_st_i32(cpu_fpr[src + 1], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.upper));
-    tcg_gen_st_i32(cpu_fpr[src + 2], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.lower));
-    tcg_gen_st_i32(cpu_fpr[src + 3], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.lowest));
+    tcg_gen_st_i64(cpu_fpr[src / 2], cpu_env, offsetof(CPUSPARCState, qt0) +
+                   offsetof(CPU_QuadU, ll.upper));
+    tcg_gen_st_i64(cpu_fpr[src/2 + 1], cpu_env, offsetof(CPUSPARCState, qt0) +
+                   offsetof(CPU_QuadU, ll.lower));
 }
 
 static void gen_op_load_fpr_QT1(unsigned int src)
 {
-    tcg_gen_st_i32(cpu_fpr[src], cpu_env, offsetof(CPUSPARCState, qt1) +
-                   offsetof(CPU_QuadU, l.upmost));
-    tcg_gen_st_i32(cpu_fpr[src + 1], cpu_env, offsetof(CPUSPARCState, qt1) +
-                   offsetof(CPU_QuadU, l.upper));
-    tcg_gen_st_i32(cpu_fpr[src + 2], cpu_env, offsetof(CPUSPARCState, qt1) +
-                   offsetof(CPU_QuadU, l.lower));
-    tcg_gen_st_i32(cpu_fpr[src + 3], cpu_env, offsetof(CPUSPARCState, qt1) +
-                   offsetof(CPU_QuadU, l.lowest));
+    tcg_gen_st_i64(cpu_fpr[src / 2], cpu_env, offsetof(CPUSPARCState, qt1) +
+                   offsetof(CPU_QuadU, ll.upper));
+    tcg_gen_st_i64(cpu_fpr[src/2 + 1], cpu_env, offsetof(CPUSPARCState, qt1) +
+                   offsetof(CPU_QuadU, ll.lower));
 }
 
 static void gen_op_store_QT0_fpr(unsigned int dst)
 {
-    tcg_gen_ld_i32(cpu_fpr[dst], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.upmost));
-    tcg_gen_ld_i32(cpu_fpr[dst + 1], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.upper));
-    tcg_gen_ld_i32(cpu_fpr[dst + 2], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.lower));
-    tcg_gen_ld_i32(cpu_fpr[dst + 3], cpu_env, offsetof(CPUSPARCState, qt0) +
-                   offsetof(CPU_QuadU, l.lowest));
+    tcg_gen_ld_i64(cpu_fpr[dst / 2], cpu_env, offsetof(CPUSPARCState, qt0) +
+                   offsetof(CPU_QuadU, ll.upper));
+    tcg_gen_ld_i64(cpu_fpr[dst/2 + 1], cpu_env, offsetof(CPUSPARCState, qt0) +
+                   offsetof(CPU_QuadU, ll.lower));
 }
 
 #ifdef TARGET_SPARC64
-static void gen_move_Q(int rd, int rs)
+static void gen_move_Q(unsigned int rd, unsigned int rs)
 {
     rd = QFPREG(rd);
     rs = QFPREG(rs);
 
-    tcg_gen_mov_i32(cpu_fpr[rd], cpu_fpr[rs]);
-    tcg_gen_mov_i32(cpu_fpr[rd + 1], cpu_fpr[rs + 1]);
-    tcg_gen_mov_i32(cpu_fpr[rd + 2], cpu_fpr[rs + 2]);
-    tcg_gen_mov_i32(cpu_fpr[rd + 3], cpu_fpr[rs + 3]);
+    tcg_gen_mov_i64(cpu_fpr[rd / 2], cpu_fpr[rs / 2]);
+    tcg_gen_mov_i64(cpu_fpr[rd / 2 + 1], cpu_fpr[rs / 2 + 1]);
     gen_update_fprs_dirty(rd);
 }
 #endif
@@ -5001,6 +4991,13 @@ static void disas_sparc_insn(DisasContext * dc)
  egress:
     tcg_temp_free(cpu_tmp1);
     tcg_temp_free(cpu_tmp2);
+    if (dc->n_t32 != 0) {
+        int i;
+        for (i = dc->n_t32 - 1; i >= 0; --i) {
+            tcg_temp_free_i32(dc->t32[i]);
+        }
+        dc->n_t32 = 0;
+    }
 }
 
 static inline void gen_intermediate_code_internal(TranslationBlock * tb,
@@ -5100,9 +5097,6 @@ static inline void gen_intermediate_code_internal(TranslationBlock * tb,
     tcg_temp_free_i64(cpu_tmp64);
     tcg_temp_free_i32(cpu_tmp32);
     tcg_temp_free(cpu_tmp0);
-    for (j = dc->n_t64 - 1; j >= 0; --j) {
-        tcg_temp_free_i64(dc->t64[j]);
-    }
 
     if (tb->cflags & CF_LAST_IO)
         gen_io_end();
@@ -5168,15 +5162,11 @@ void gen_intermediate_code_init(CPUSPARCState *env)
         "g6",
         "g7",
     };
-    static const char * const fregnames[64] = {
-        "f0", "f1", "f2", "f3", "f4", "f5", "f6", "f7",
-        "f8", "f9", "f10", "f11", "f12", "f13", "f14", "f15",
-        "f16", "f17", "f18", "f19", "f20", "f21", "f22", "f23",
-        "f24", "f25", "f26", "f27", "f28", "f29", "f30", "f31",
-        "f32", "f33", "f34", "f35", "f36", "f37", "f38", "f39",
-        "f40", "f41", "f42", "f43", "f44", "f45", "f46", "f47",
-        "f48", "f49", "f50", "f51", "f52", "f53", "f54", "f55",
-        "f56", "f57", "f58", "f59", "f60", "f61", "f62", "f63",
+    static const char * const fregnames[32] = {
+        "f0", "f2", "f4", "f6", "f8", "f10", "f12", "f14",
+        "f16", "f18", "f20", "f22", "f24", "f26", "f28", "f30",
+        "f32", "f34", "f36", "f38", "f40", "f42", "f44", "f46",
+        "f48", "f50", "f52", "f54", "f56", "f58", "f60", "f62",
     };
 
     /* init various static tables */
@@ -5246,14 +5236,16 @@ void gen_intermediate_code_init(CPUSPARCState *env)
         cpu_tbr = tcg_global_mem_new(TCG_AREG0, offsetof(CPUState, tbr),
                                      "tbr");
 #endif
-        for (i = 1; i < 8; i++)
+        for (i = 1; i < 8; i++) {
             cpu_gregs[i] = tcg_global_mem_new(TCG_AREG0,
                                               offsetof(CPUState, gregs[i]),
                                               gregnames[i]);
-        for (i = 0; i < TARGET_FPREGS; i++)
-            cpu_fpr[i] = tcg_global_mem_new_i32(TCG_AREG0,
+        }
+        for (i = 0; i < TARGET_DPREGS; i++) {
+            cpu_fpr[i] = tcg_global_mem_new_i64(TCG_AREG0,
                                                 offsetof(CPUState, fpr[i]),
                                                 fregnames[i]);
+        }
 
         /* register helpers */