[PATCH v7 29/42] target/arm: Use mte_checkN for sve unpredicated stores

[Richard Henderson <richard.henderson@linaro.org>]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/helper-sve.h    |  1 +
 target/arm/sve_helper.c    | 63 ++++++++++++++++++++++++++-
 target/arm/translate-sve.c | 88 ++++++++++++++------------------------
 3 files changed, 94 insertions(+), 58 deletions(-)

diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h
index 550ae7f7e6..197cd72999 100644
--- a/target/arm/helper-sve.h
+++ b/target/arm/helper-sve.h
@@ -1139,6 +1139,7 @@ DEF_HELPER_FLAGS_5(sve_ftmad_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_5(sve_ftmad_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32)
 
 DEF_HELPER_FLAGS_4(sve_ldr, TCG_CALL_NO_WG, void, env, ptr, tl, int)
+DEF_HELPER_FLAGS_4(sve_str, TCG_CALL_NO_WG, void, env, ptr, tl, int)
 
 DEF_HELPER_FLAGS_4(sve_ld1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32)
 DEF_HELPER_FLAGS_4(sve_ld2bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32)
diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c
index 04340cbe0c..507fba37c9 100644
--- a/target/arm/sve_helper.c
+++ b/target/arm/sve_helper.c
@@ -4130,7 +4130,7 @@ static bool sve_probe_page(SVEHostPage *info, bool nofault,
 }
 
 /*
- * Load contiguous data, unpredicated.
+ * Load/store contiguous data, unpredicated.
  *
  * Note that unpredicated load/store of vector/predicate registers
  * are defined as a stream of bytes, which equates to little-endian
@@ -4198,6 +4198,67 @@ void HELPER(sve_ldr)(CPUARMState *env, void *vd, target_ulong addr, int size)
     }
 }
 
+void HELPER(sve_str)(CPUARMState *env, void *vd, target_ulong addr, int size)
+{
+    int mem_idx = cpu_mmu_index(env, false);
+    int in_page = -((int)addr | TARGET_PAGE_MASK);
+    uintptr_t ra = GETPC();
+    uint64_t val;
+    void *host;
+    int i;
+
+    /* Small stores are expanded inline. */
+    tcg_debug_assert(size > 2 * 8);
+
+    if (likely(size <= in_page)) {
+        host = probe_write(env, addr, size, mem_idx, ra);
+        if (likely(host != NULL)) {
+            for (i = 0; i + 8 <= size; i += 8) {
+                stq_le_p(host + i, *(uint64_t *)(vd + i));
+            }
+
+            /* Predicate load length may be any multiple of 2. */
+            if (unlikely(i != size)) {
+                val = *(uint64_t *)(vd + i);
+                if (size & 4) {
+                    stl_le_p(host + i, val);
+                    i += 4;
+                    val >>= 32;
+                }
+                if (size & 2) {
+                    stw_le_p(host + i, val);
+                }
+            }
+            return;
+        }
+    } else {
+        (void)probe_write(env, addr, in_page, mem_idx, ra);
+        (void)probe_write(env, addr + in_page, size - in_page, mem_idx, ra);
+    }
+
+    /*
+     * Note there is no endian-specific target store function, so to handle
+     * aarch64_be-linux-user we need to bswap the big-endian store.
+     */
+    for (i = 0; i + 8 <= size; i += 8) {
+        val = *(uint64_t *)(vd + i);
+        cpu_stq_le_data_ra(env, addr + i, val, ra);
+    }
+
+    /* Predicate load length may be any multiple of 2. */
+    if (unlikely(i != size)) {
+        val = *(uint64_t *)(vd + i);
+        if (size & 4) {
+            cpu_stl_le_data_ra(env, addr + i, val, ra);
+            i += 4;
+            val >>= 32;
+        }
+        if (size & 2) {
+            cpu_stw_le_data_ra(env, addr + i, val, ra);
+        }
+    }
+}
+
 /*
  * Analyse contiguous data, protected by a governing predicate.
  */
diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c
index 35b979fa13..e515646db2 100644
--- a/target/arm/translate-sve.c
+++ b/target/arm/translate-sve.c
@@ -4410,78 +4410,52 @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm)
     int len_remain = len % 8;
     int nparts = len / 8 + ctpop8(len_remain);
     int midx = get_mem_index(s);
-    TCGv_i64 addr, t0;
+    TCGv_i64 dirty_addr, clean_addr, t0;
+    int i;
+
+    dirty_addr = read_cpu_reg_sp(s, rn, true);
+    tcg_gen_addi_i64(dirty_addr, dirty_addr, imm);
+
+    clean_addr = gen_mte_checkN(s, dirty_addr, true, rn != 31, len, MO_8);
+
+    /* Limit tcg code expansion by doing large loads out of line. */
+    if (nparts > 4) {
+        TCGv_ptr t_rd = tcg_temp_new_ptr();
+        TCGv_i32 t_len = tcg_const_i32(len);
+
+        tcg_gen_addi_ptr(t_rd, cpu_env, vofs);
+        gen_helper_sve_str(cpu_env, t_rd, clean_addr, t_len);
+        tcg_temp_free_ptr(t_rd);
+        tcg_temp_free_i32(t_len);
+        return;
+    }
 
-    addr = tcg_temp_new_i64();
     t0 = tcg_temp_new_i64();
-
-    /* Note that unpredicated load/store of vector/predicate registers
-     * are defined as a stream of bytes, which equates to little-endian
-     * operations on larger quantities.  There is no nice way to force
-     * a little-endian store for aarch64_be-linux-user out of line.
-     *
-     * Attempt to keep code expansion to a minimum by limiting the
-     * amount of unrolling done.
-     */
-    if (nparts <= 4) {
-        int i;
-
-        for (i = 0; i < len_align; i += 8) {
-            tcg_gen_ld_i64(t0, cpu_env, vofs + i);
-            tcg_gen_addi_i64(addr, cpu_reg_sp(s, rn), imm + i);
-            tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEQ);
-        }
-    } else {
-        TCGLabel *loop = gen_new_label();
-        TCGv_ptr t2, i = tcg_const_local_ptr(0);
-
-        gen_set_label(loop);
-
-        t2 = tcg_temp_new_ptr();
-        tcg_gen_add_ptr(t2, cpu_env, i);
-        tcg_gen_ld_i64(t0, t2, vofs);
-
-        /* Minimize the number of local temps that must be re-read from
-         * the stack each iteration.  Instead, re-compute values other
-         * than the loop counter.
-         */
-        tcg_gen_addi_ptr(t2, i, imm);
-        tcg_gen_extu_ptr_i64(addr, t2);
-        tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, rn));
-        tcg_temp_free_ptr(t2);
-
-        tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEQ);
-
-        tcg_gen_addi_ptr(i, i, 8);
-
-        tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop);
-        tcg_temp_free_ptr(i);
+    for (i = 0; i < len_align; i += 8) {
+        tcg_gen_ld_i64(t0, cpu_env, vofs + i);
+        tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEQ);
+        tcg_gen_addi_i64(clean_addr, clean_addr, 8);
     }
 
     /* Predicate register stores can be any multiple of 2.  */
     if (len_remain) {
         tcg_gen_ld_i64(t0, cpu_env, vofs + len_align);
-        tcg_gen_addi_i64(addr, cpu_reg_sp(s, rn), imm + len_align);
-
         switch (len_remain) {
-        case 2:
-        case 4:
-        case 8:
-            tcg_gen_qemu_st_i64(t0, addr, midx, MO_LE | ctz32(len_remain));
-            break;
-
         case 6:
-            tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEUL);
-            tcg_gen_addi_i64(addr, addr, 4);
+            tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUL);
+            tcg_gen_addi_i64(clean_addr, clean_addr, 4);
             tcg_gen_shri_i64(t0, t0, 32);
-            tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEUW);
+            /* fall through */
+        case 2:
+            tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUW);
+            break;
+        case 4:
+            tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUL);
             break;
-
         default:
             g_assert_not_reached();
         }
     }
-    tcg_temp_free_i64(addr);
     tcg_temp_free_i64(t0);
 }
 
-- 
2.25.1