diff mbox series

[v2,18/23] target/arm: Convert atomic memory ops to decodetree

Message ID 20230611160032.274823-19-peter.maydell@linaro.org
State Superseded
Headers show
Series target/arm: Convert exception, system, loads and stores to decodetree | expand

Commit Message

Peter Maydell June 11, 2023, 4 p.m. UTC
Convert the insns in the atomic memory operations group to
decodetree.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230602155223.2040685-16-peter.maydell@linaro.org
---
 target/arm/tcg/a64.decode      |  15 ++++
 target/arm/tcg/translate-a64.c | 153 ++++++++++++---------------------
 2 files changed, 70 insertions(+), 98 deletions(-)
diff mbox series

Patch

diff --git a/target/arm/tcg/a64.decode b/target/arm/tcg/a64.decode
index 5c086d6af6d..799c5ecb77a 100644
--- a/target/arm/tcg/a64.decode
+++ b/target/arm/tcg/a64.decode
@@ -442,3 +442,18 @@  STR_v           sz:2 111 1 00 00 1 ..... ... . 10 ..... ..... @ldst sign=0 ext=0
 STR_v           00 111 1 00 10 1 ..... ... . 10 ..... ..... @ldst sign=0 ext=0 sz=4
 LDR_v           sz:2 111 1 00 01 1 ..... ... . 10 ..... ..... @ldst sign=0 ext=0
 LDR_v           00 111 1 00 11 1 ..... ... . 10 ..... ..... @ldst sign=0 ext=0 sz=4
+
+# Atomic memory operations
+&atomic         rs rn rt a r sz
+@atomic         sz:2 ... . .. a:1 r:1 . rs:5 . ... .. rn:5 rt:5 &atomic
+LDADD           .. 111 0 00 . . 1 ..... 0000 00 ..... ..... @atomic
+LDCLR           .. 111 0 00 . . 1 ..... 0001 00 ..... ..... @atomic
+LDEOR           .. 111 0 00 . . 1 ..... 0010 00 ..... ..... @atomic
+LDSET           .. 111 0 00 . . 1 ..... 0011 00 ..... ..... @atomic
+LDSMAX          .. 111 0 00 . . 1 ..... 0100 00 ..... ..... @atomic
+LDSMIN          .. 111 0 00 . . 1 ..... 0101 00 ..... ..... @atomic
+LDUMAX          .. 111 0 00 . . 1 ..... 0110 00 ..... ..... @atomic
+LDUMIN          .. 111 0 00 . . 1 ..... 0111 00 ..... ..... @atomic
+SWP             .. 111 0 00 . . 1 ..... 1000 00 ..... ..... @atomic
+
+LDAPR           sz:2 111 0 00 1 0 1 11111 1100 00 rn:5 rt:5
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 3d161169411..ba072e557e1 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -3244,113 +3244,32 @@  static bool trans_STR_v(DisasContext *s, arg_ldst *a)
     return true;
 }
 
-/* Atomic memory operations
- *
- *  31  30      27  26    24    22  21   16   15    12    10    5     0
- * +------+-------+---+-----+-----+---+----+----+-----+-----+----+-----+
- * | size | 1 1 1 | V | 0 0 | A R | 1 | Rs | o3 | opc | 0 0 | Rn |  Rt |
- * +------+-------+---+-----+-----+--------+----+-----+-----+----+-----+
- *
- * Rt: the result register
- * Rn: base address or SP
- * Rs: the source register for the operation
- * V: vector flag (always 0 as of v8.3)
- * A: acquire flag
- * R: release flag
- */
-static void disas_ldst_atomic(DisasContext *s, uint32_t insn,
-                              int size, int rt, bool is_vector)
+
+static bool do_atomic_ld(DisasContext *s, arg_atomic *a, AtomicThreeOpFn *fn,
+                         int sign, bool invert)
 {
-    int rs = extract32(insn, 16, 5);
-    int rn = extract32(insn, 5, 5);
-    int o3_opc = extract32(insn, 12, 4);
-    bool r = extract32(insn, 22, 1);
-    bool a = extract32(insn, 23, 1);
-    TCGv_i64 tcg_rs, tcg_rt, clean_addr;
-    AtomicThreeOpFn *fn = NULL;
-    MemOp mop = size;
+    MemOp mop = a->sz | sign;
+    TCGv_i64 clean_addr, tcg_rs, tcg_rt;
 
-    if (is_vector || !dc_isar_feature(aa64_atomics, s)) {
-        unallocated_encoding(s);
-        return;
-    }
-    switch (o3_opc) {
-    case 000: /* LDADD */
-        fn = tcg_gen_atomic_fetch_add_i64;
-        break;
-    case 001: /* LDCLR */
-        fn = tcg_gen_atomic_fetch_and_i64;
-        break;
-    case 002: /* LDEOR */
-        fn = tcg_gen_atomic_fetch_xor_i64;
-        break;
-    case 003: /* LDSET */
-        fn = tcg_gen_atomic_fetch_or_i64;
-        break;
-    case 004: /* LDSMAX */
-        fn = tcg_gen_atomic_fetch_smax_i64;
-        mop |= MO_SIGN;
-        break;
-    case 005: /* LDSMIN */
-        fn = tcg_gen_atomic_fetch_smin_i64;
-        mop |= MO_SIGN;
-        break;
-    case 006: /* LDUMAX */
-        fn = tcg_gen_atomic_fetch_umax_i64;
-        break;
-    case 007: /* LDUMIN */
-        fn = tcg_gen_atomic_fetch_umin_i64;
-        break;
-    case 010: /* SWP */
-        fn = tcg_gen_atomic_xchg_i64;
-        break;
-    case 014: /* LDAPR, LDAPRH, LDAPRB */
-        if (!dc_isar_feature(aa64_rcpc_8_3, s) ||
-            rs != 31 || a != 1 || r != 0) {
-            unallocated_encoding(s);
-            return;
-        }
-        break;
-    default:
-        unallocated_encoding(s);
-        return;
-    }
-
-    if (rn == 31) {
+    if (a->rn == 31) {
         gen_check_sp_alignment(s);
     }
-
-    mop = check_atomic_align(s, rn, mop);
-    clean_addr = gen_mte_check1(s, cpu_reg_sp(s, rn), false, rn != 31, mop);
-
-    if (o3_opc == 014) {
-        /*
-         * LDAPR* are a special case because they are a simple load, not a
-         * fetch-and-do-something op.
-         * The architectural consistency requirements here are weaker than
-         * full load-acquire (we only need "load-acquire processor consistent"),
-         * but we choose to implement them as full LDAQ.
-         */
-        do_gpr_ld(s, cpu_reg(s, rt), clean_addr, mop, false,
-                  true, rt, disas_ldst_compute_iss_sf(size, false, 0), true);
-        tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
-        return;
-    }
-
-    tcg_rs = read_cpu_reg(s, rs, true);
-    tcg_rt = cpu_reg(s, rt);
-
-    if (o3_opc == 1) { /* LDCLR */
+    mop = check_atomic_align(s, a->rn, mop);
+    clean_addr = gen_mte_check1(s, cpu_reg_sp(s, a->rn), false,
+                                a->rn != 31, mop);
+    tcg_rs = read_cpu_reg(s, a->rs, true);
+    tcg_rt = cpu_reg(s, a->rt);
+    if (invert) {
         tcg_gen_not_i64(tcg_rs, tcg_rs);
     }
-
-    /* The tcg atomic primitives are all full barriers.  Therefore we
+    /*
+     * The tcg atomic primitives are all full barriers.  Therefore we
      * can ignore the Acquire and Release bits of this instruction.
      */
     fn(tcg_rt, clean_addr, tcg_rs, get_mem_index(s), mop);
 
     if (mop & MO_SIGN) {
-        switch (size) {
+        switch (a->sz) {
         case MO_8:
             tcg_gen_ext8u_i64(tcg_rt, tcg_rt);
             break;
@@ -3366,6 +3285,46 @@  static void disas_ldst_atomic(DisasContext *s, uint32_t insn,
             g_assert_not_reached();
         }
     }
+    return true;
+}
+
+TRANS_FEAT(LDADD, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_add_i64, 0, false)
+TRANS_FEAT(LDCLR, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_and_i64, 0, true)
+TRANS_FEAT(LDEOR, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_xor_i64, 0, false)
+TRANS_FEAT(LDSET, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_or_i64, 0, false)
+TRANS_FEAT(LDSMAX, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_smax_i64, MO_SIGN, false)
+TRANS_FEAT(LDSMIN, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_smin_i64, MO_SIGN, false)
+TRANS_FEAT(LDUMAX, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_umax_i64, 0, false)
+TRANS_FEAT(LDUMIN, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_fetch_umin_i64, 0, false)
+TRANS_FEAT(SWP, aa64_atomics, do_atomic_ld, a, tcg_gen_atomic_xchg_i64, 0, false)
+
+static bool trans_LDAPR(DisasContext *s, arg_LDAPR *a)
+{
+    bool iss_sf = ldst_iss_sf(a->sz, false, false);
+    TCGv_i64 clean_addr;
+    MemOp mop;
+
+    if (!dc_isar_feature(aa64_atomics, s) ||
+        !dc_isar_feature(aa64_rcpc_8_3, s)) {
+        return false;
+    }
+    if (a->rn == 31) {
+        gen_check_sp_alignment(s);
+    }
+    mop = check_atomic_align(s, a->rn, a->sz);
+    clean_addr = gen_mte_check1(s, cpu_reg_sp(s, a->rn), false,
+                                a->rn != 31, mop);
+    /*
+     * LDAPR* are a special case because they are a simple load, not a
+     * fetch-and-do-something op.
+     * The architectural consistency requirements here are weaker than
+     * full load-acquire (we only need "load-acquire processor consistent"),
+     * but we choose to implement them as full LDAQ.
+     */
+    do_gpr_ld(s, cpu_reg(s, a->rt), clean_addr, mop, false,
+              true, a->rt, iss_sf, true);
+    tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ);
+    return true;
 }
 
 /*
@@ -3532,8 +3491,6 @@  static void disas_ldst_reg(DisasContext *s, uint32_t insn)
         }
         switch (extract32(insn, 10, 2)) {
         case 0:
-            disas_ldst_atomic(s, insn, size, rt, is_vector);
-            return;
         case 2:
             break;
         default: