[Qemu-devel] [PATCH 26/52] target-m68k: Inline shifts

[Laurent Vivier <laurent@vivier.eu>]

From: Richard Henderson <rth@twiddle.net>

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

fix arithmetical/logical switch

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
---
 target-m68k/helper.c    | 52 ---------------------------
 target-m68k/helper.h    |  3 --
 target-m68k/translate.c | 94 +++++++++++++++++++++++++++++++++++++------------
 3 files changed, 72 insertions(+), 77 deletions(-)

diff --git a/target-m68k/helper.c b/target-m68k/helper.c
index 8758016..0ca8d3f 100644
--- a/target-m68k/helper.c
+++ b/target-m68k/helper.c
@@ -323,58 +323,6 @@ void HELPER(set_sr)(CPUM68KState *env, uint32_t val)
     m68k_switch_sp(env);
 }
 
-uint32_t HELPER(shl_cc)(CPUM68KState *env, uint32_t val, uint32_t shift)
-{
-    uint64_t result;
-
-    shift &= 63;
-    result = (uint64_t)val << shift;
-
-    env->cc_c = (result >> 32) & 1;
-    env->cc_n = result;
-    env->cc_z = result;
-    env->cc_v = 0;
-    env->cc_x = shift ? env->cc_c : env->cc_x;
-
-    return result;
-}
-
-uint32_t HELPER(shr_cc)(CPUM68KState *env, uint32_t val, uint32_t shift)
-{
-    uint64_t temp;
-    uint32_t result;
-
-    shift &= 63;
-    temp = (uint64_t)val << 32 >> shift;
-    result = temp >> 32;
-
-    env->cc_c = (temp >> 31) & 1;
-    env->cc_n = result;
-    env->cc_z = result;
-    env->cc_v = 0;
-    env->cc_x = shift ? env->cc_c : env->cc_x;
-
-    return result;
-}
-
-uint32_t HELPER(sar_cc)(CPUM68KState *env, uint32_t val, uint32_t shift)
-{
-    uint64_t temp;
-    uint32_t result;
-
-    shift &= 63;
-    temp = (int64_t)val << 32 >> shift;
-    result = temp >> 32;
-
-    env->cc_c = (temp >> 31) & 1;
-    env->cc_n = result;
-    env->cc_z = result;
-    env->cc_v = result ^ val;
-    env->cc_x = shift ? env->cc_c : env->cc_x;
-
-    return result;
-}
-
 /* FPU helpers.  */
 uint32_t HELPER(f64_to_i32)(CPUM68KState *env, float64 val)
 {
diff --git a/target-m68k/helper.h b/target-m68k/helper.h
index c868148..9985f9b 100644
--- a/target-m68k/helper.h
+++ b/target-m68k/helper.h
@@ -5,9 +5,6 @@ DEF_HELPER_2(divu, void, env, i32)
 DEF_HELPER_2(divs, void, env, i32)
 DEF_HELPER_3(addx_cc, i32, env, i32, i32)
 DEF_HELPER_3(subx_cc, i32, env, i32, i32)
-DEF_HELPER_3(shl_cc, i32, env, i32, i32)
-DEF_HELPER_3(shr_cc, i32, env, i32, i32)
-DEF_HELPER_3(sar_cc, i32, env, i32, i32)
 DEF_HELPER_2(set_sr, void, env, i32)
 DEF_HELPER_3(movec, void, env, i32, i32)
 
diff --git a/target-m68k/translate.c b/target-m68k/translate.c
index 58cc9d9..1e8cb37 100644
--- a/target-m68k/translate.c
+++ b/target-m68k/translate.c
@@ -2038,48 +2038,98 @@ DISAS_INSN(addx)
     gen_helper_addx_cc(reg, cpu_env, reg, src);
 }
 
-/* TODO: This could be implemented without helper functions.  */
 DISAS_INSN(shift_im)
 {
-    TCGv reg;
-    int tmp;
-    TCGv shift;
+    TCGv reg = DREG(insn, 0);
+    int count = (insn >> 9) & 7;
+    int logical = insn & 8;
 
-    set_cc_op(s, CC_OP_FLAGS);
+    if (count == 0) {
+        count = 8;
+    }
 
-    reg = DREG(insn, 0);
-    tmp = (insn >> 9) & 7;
-    if (tmp == 0)
-        tmp = 8;
-    shift = tcg_const_i32(tmp);
-    /* No need to flush flags becuse we know we will set C flag.  */
     if (insn & 0x100) {
-        gen_helper_shl_cc(reg, cpu_env, reg, shift);
+        tcg_gen_shri_i32(QREG_CC_C, reg, 31 - count);
+        tcg_gen_shli_i32(QREG_CC_N, reg, count);
     } else {
-        if (insn & 8) {
-            gen_helper_shr_cc(reg, cpu_env, reg, shift);
+        tcg_gen_shri_i32(QREG_CC_C, reg, count - 1);
+        if (logical) {
+            tcg_gen_shri_i32(QREG_CC_N, reg, count);
         } else {
-            gen_helper_sar_cc(reg, cpu_env, reg, shift);
+            tcg_gen_sari_i32(QREG_CC_N, reg, count);
         }
     }
+    tcg_gen_andi_i32(QREG_CC_C, QREG_CC_C, 1);
+    tcg_gen_mov_i32(QREG_CC_Z, QREG_CC_N);
+    tcg_gen_mov_i32(QREG_CC_X, QREG_CC_C);
+
+    /* Note that ColdFire always clears V, while M68000 sets it for
+       a change in the sign bit.  */
+    if (!logical && m68k_feature(s->env, M68K_FEATURE_M68000)) {
+        tcg_gen_xor_i32(QREG_CC_V, QREG_CC_N, reg);
+    } else {
+        tcg_gen_movi_i32(QREG_CC_V, 0);
+    }
+
+    tcg_gen_mov_i32(reg, QREG_CC_N);
+    set_cc_op(s, CC_OP_FLAGS);
 }
 
 DISAS_INSN(shift_reg)
 {
-    TCGv reg;
-    TCGv shift;
+    TCGv reg, s32;
+    TCGv_i64 t64, s64;
+    int logical = insn & 8;
 
     reg = DREG(insn, 0);
-    shift = DREG(insn, 9);
+    t64 = tcg_temp_new_i64();
+    s64 = tcg_temp_new_i64();
+    s32 = tcg_temp_new();
+
+    /* Note that m68k truncates the shift count modulo 64, not 32.
+       In addition, a 64-bit shift makes it easy to find "the last
+       bit shifted out", for the carry flag.  */
+    tcg_gen_andi_i32(s32, DREG(insn, 9), 63);
+    tcg_gen_extu_i32_i64(s64, s32);
+
+    /* Non-arithmetic shift clears V.  Use it as a source zero here.  */
+    tcg_gen_movi_i32(QREG_CC_V, 0);
+
     if (insn & 0x100) {
-        gen_helper_shl_cc(reg, cpu_env, reg, shift);
+        tcg_gen_extu_i32_i64(t64, reg);
+        tcg_gen_shl_i64(t64, t64, s64);
+        tcg_temp_free_i64(s64);
+        tcg_gen_extr_i64_i32(QREG_CC_N, QREG_CC_C, t64);
+        tcg_temp_free_i64(t64);
+        tcg_gen_andi_i32(QREG_CC_C, QREG_CC_C, 1);
     } else {
-        if (insn & 8) {
-            gen_helper_shr_cc(reg, cpu_env, reg, shift);
+        tcg_gen_extu_i32_i64(t64, reg);
+        tcg_gen_shli_i64(t64, t64, 32);
+        if (logical) {
+            tcg_gen_shr_i64(t64, t64, s64);
         } else {
-            gen_helper_sar_cc(reg, cpu_env, reg, shift);
+            tcg_gen_sar_i64(t64, t64, s64);
         }
+        tcg_temp_free_i64(s64);
+        tcg_gen_extr_i64_i32(QREG_CC_C, QREG_CC_N, t64);
+        tcg_temp_free_i64(t64);
+        tcg_gen_shri_i32(QREG_CC_C, QREG_CC_C, 31);
     }
+    tcg_gen_mov_i32(QREG_CC_Z, QREG_CC_N);
+
+    /* Note that X = C, but only if the shift count was non-zero.  */
+    tcg_gen_movcond_i32(TCG_COND_NE, QREG_CC_X, s32, QREG_CC_V,
+                        QREG_CC_C, QREG_CC_X);
+    tcg_temp_free(s32);
+
+    /* Note that ColdFire always clears V (which we have done above),
+       while M68000 sets it for a change in the sign bit.  */
+    if (!logical && m68k_feature(s->env, M68K_FEATURE_M68000)) {
+        tcg_gen_xor_i32(QREG_CC_V, QREG_CC_N, reg);
+    }
+
+    /* Write back the result.  */
+    tcg_gen_mov_i32(reg, QREG_CC_N);
     set_cc_op(s, CC_OP_FLAGS);
 }
 
-- 
2.5.5