target/arm: Implement SVE Integer Compare - Scalars Group

Backports commit caf1cefc72be98497e0907d0e07f4327fc641e96 from qemu

qemu/aarch64.h

@@ -3688,6 +3688,7 @@
 #define helper_sve_uzp_h helper_sve_uzp_h_aarch64
 #define helper_sve_uzp_s helper_sve_uzp_s_aarch64
 #define helper_sve_uzp_p helper_sve_uzp_p_aarch64
+#define helper_sve_while helper_sve_while_aarch64
 #define helper_sve_zip_b helper_sve_zip_b_aarch64
 #define helper_sve_zip_d helper_sve_zip_d_aarch64
 #define helper_sve_zip_h helper_sve_zip_h_aarch64

qemu/aarch64eb.h

@@ -3688,6 +3688,7 @@
 #define helper_sve_uzp_h helper_sve_uzp_h_aarch64eb
 #define helper_sve_uzp_s helper_sve_uzp_s_aarch64eb
 #define helper_sve_uzp_p helper_sve_uzp_p_aarch64eb
+#define helper_sve_while helper_sve_while_aarch64eb
 #define helper_sve_zip_b helper_sve_zip_b_aarch64eb
 #define helper_sve_zip_d helper_sve_zip_d_aarch64eb
 #define helper_sve_zip_h helper_sve_zip_h_aarch64eb

qemu/header_gen.py

@@ -3709,6 +3709,7 @@ aarch64_symbols = (
     'helper_sve_uzp_h',
     'helper_sve_uzp_s',
     'helper_sve_uzp_p',
+    'helper_sve_while',
     'helper_sve_zip_b',
     'helper_sve_zip_d',
     'helper_sve_zip_h',

qemu/target/arm/helper-sve.h

@@ -678,3 +678,5 @@ DEF_HELPER_FLAGS_4(sve_brkn, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(sve_brkns, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32)
 
 DEF_HELPER_FLAGS_3(sve_cntp, TCG_CALL_NO_RWG, i64, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_3(sve_while, TCG_CALL_NO_RWG, i32, ptr, i32, i32)

qemu/target/arm/sve.decode

@@ -606,6 +606,14 @@ SINCDECP_r_64   00100101 .. 1010 d:1 u:1 10001 10 .... .....    @incdec_pred
 # SVE saturating inc/dec vector by predicate count
 SINCDECP_z      00100101 .. 1010 d:1 u:1 10000 00 .... .....    @incdec2_pred
 
+### SVE Integer Compare - Scalars Group
+
+# SVE conditionally terminate scalars
+CTERM           00100101 1 sf:1 1 rm:5 001000 rn:5 ne:1 0000
+
+# SVE integer compare scalar count and limit
+WHILE           00100101 esz:2 1 rm:5 000 sf:1 u:1 1 rn:5 eq:1 rd:4
+
 ### SVE Memory - 32-bit Gather and Unsized Contiguous Group
 
 # SVE load predicate register

qemu/target/arm/sve_helper.c

@@ -2737,3 +2737,34 @@ uint64_t HELPER(sve_cntp)(void *vn, void *vg, uint32_t pred_desc)
     }
     return sum;
 }
+
+uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc)
+{
+    uintptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2;
+    intptr_t esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2);
+    uint64_t esz_mask = pred_esz_masks[esz];
+    ARMPredicateReg *d = vd;
+    uint32_t flags;
+    intptr_t i;
+
+    /* Begin with a zero predicate register.  */
+    flags = do_zero(d, oprsz);
+    if (count == 0) {
+        return flags;
+    }
+
+    /* Scale from predicate element count to bits.  */
+    count <<= esz;
+    /* Bound to the bits in the predicate.  */
+    count = MIN(count, oprsz * 8);
+
+    /* Set all of the requested bits.  */
+    for (i = 0; i < count / 64; ++i) {
+        d->p[i] = esz_mask;
+    }
+    if (count & 63) {
+        d->p[i] = MAKE_64BIT_MASK(0, count & 63) & esz_mask;
+    }
+
+    return predtest_ones(d, oprsz, esz_mask);
+}

qemu/target/arm/translate-sve.c

@@ -3205,6 +3205,107 @@ static bool trans_SINCDECP_z(DisasContext *s, arg_incdec2_pred *a,
     return true;
 }
 
+/*
+ *** SVE Integer Compare Scalars Group
+ */
+
+static bool trans_CTERM(DisasContext *s, arg_CTERM *a, uint32_t insn)
+{
+    if (!sve_access_check(s)) {
+        return true;
+    }
+
+    TCGContext *tcg_ctx = s->uc->tcg_ctx;
+    TCGCond cond = (a->ne ? TCG_COND_NE : TCG_COND_EQ);
+    TCGv_i64 rn = read_cpu_reg(s, a->rn, a->sf);
+    TCGv_i64 rm = read_cpu_reg(s, a->rm, a->sf);
+    TCGv_i64 cmp = tcg_temp_new_i64(tcg_ctx);
+
+    tcg_gen_setcond_i64(tcg_ctx, cond, cmp, rn, rm);
+    tcg_gen_extrl_i64_i32(tcg_ctx, tcg_ctx->cpu_NF, cmp);
+    tcg_temp_free_i64(tcg_ctx, cmp);
+
+    /* VF = !NF & !CF.  */
+    tcg_gen_xori_i32(tcg_ctx, tcg_ctx->cpu_VF, tcg_ctx->cpu_NF, 1);
+    tcg_gen_andc_i32(tcg_ctx, tcg_ctx->cpu_VF, tcg_ctx->cpu_VF, tcg_ctx->cpu_CF);
+
+    /* Both NF and VF actually look at bit 31.  */
+    tcg_gen_neg_i32(tcg_ctx, tcg_ctx->cpu_NF, tcg_ctx->cpu_NF);
+    tcg_gen_neg_i32(tcg_ctx, tcg_ctx->cpu_VF, tcg_ctx->cpu_VF);
+    return true;
+}
+
+static bool trans_WHILE(DisasContext *s, arg_WHILE *a, uint32_t insn)
+{
+    if (!sve_access_check(s)) {
+        return true;
+    }
+
+    TCGContext *tcg_ctx = s->uc->tcg_ctx;
+    TCGv_i64 op0 = read_cpu_reg(s, a->rn, 1);
+    TCGv_i64 op1 = read_cpu_reg(s, a->rm, 1);
+    TCGv_i64 t0 = tcg_temp_new_i64(tcg_ctx);
+    TCGv_i64 t1 = tcg_temp_new_i64(tcg_ctx);
+    TCGv_i32 t2, t3;
+    TCGv_ptr ptr;
+    unsigned desc, vsz = vec_full_reg_size(s);
+    TCGCond cond;
+
+    if (!a->sf) {
+        if (a->u) {
+            tcg_gen_ext32u_i64(tcg_ctx, op0, op0);
+            tcg_gen_ext32u_i64(tcg_ctx, op1, op1);
+        } else {
+            tcg_gen_ext32s_i64(tcg_ctx, op0, op0);
+            tcg_gen_ext32s_i64(tcg_ctx, op1, op1);
+        }
+    }
+
+    /* For the helper, compress the different conditions into a computation
+     * of how many iterations for which the condition is true.
+     *
+     * This is slightly complicated by 0 <= UINT64_MAX, which is nominally
+     * 2**64 iterations, overflowing to 0.  Of course, predicate registers
+     * aren't that large, so any value >= predicate size is sufficient.
+     */
+    tcg_gen_sub_i64(tcg_ctx, t0, op1, op0);
+
+    /* t0 = MIN(op1 - op0, vsz).  */
+    tcg_gen_movi_i64(tcg_ctx, t1, vsz);
+    tcg_gen_umin_i64(tcg_ctx, t0, t0, t1);
+    if (a->eq) {
+        /* Equality means one more iteration.  */
+        tcg_gen_addi_i64(tcg_ctx, t0, t0, 1);
+    }
+
+    /* t0 = (condition true ? t0 : 0).  */
+    cond = (a->u
+            ? (a->eq ? TCG_COND_LEU : TCG_COND_LTU)
+            : (a->eq ? TCG_COND_LE : TCG_COND_LT));
+    tcg_gen_movi_i64(tcg_ctx, t1, 0);
+    tcg_gen_movcond_i64(tcg_ctx, cond, t0, op0, op1, t0, t1);
+
+    t2 = tcg_temp_new_i32(tcg_ctx);
+    tcg_gen_extrl_i64_i32(tcg_ctx, t2, t0);
+    tcg_temp_free_i64(tcg_ctx, t0);
+    tcg_temp_free_i64(tcg_ctx, t1);
+
+    desc = (vsz / 8) - 2;
+    desc = deposit32(desc, SIMD_DATA_SHIFT, 2, a->esz);
+    t3 = tcg_const_i32(tcg_ctx, desc);
+
+    ptr = tcg_temp_new_ptr(tcg_ctx);
+    tcg_gen_addi_ptr(tcg_ctx, ptr, tcg_ctx->cpu_env, pred_full_reg_offset(s, a->rd));
+
+    gen_helper_sve_while(tcg_ctx, t2, ptr, t2, t3);
+    do_pred_flags(s, t2);
+
+    tcg_temp_free_ptr(tcg_ctx, ptr);
+    tcg_temp_free_i32(tcg_ctx, t2);
+    tcg_temp_free_i32(tcg_ctx, t3);
+    return true;
+}
+
 /*
  *** SVE Memory - 32-bit Gather and Unsized Contiguous Group
  */