From 4ca33c54a270be5da35eab829f7f0768e3c41b1a Mon Sep 17 00:00:00 2001
From: Peter Maydell <peter.maydell@linaro.org>
Date: Thu, 25 Feb 2021 12:20:56 -0500
Subject: [PATCH] target/arm: Convert Neon 2-reg-misc VSHLL to decodetree

Convert the VSHLL insn in the 2-reg-misc Neon group to decodetree.

Backports commit 749e2be36d75f11d5fa8f8277e2a0569bd2a1c97 from qemu
---
 qemu/target/arm/neon-dp.decode       |  2 ++
 qemu/target/arm/translate-neon.inc.c | 53 ++++++++++++++++++++++++++++
 qemu/target/arm/translate.c          | 36 +------------------
 3 files changed, 56 insertions(+), 35 deletions(-)

diff --git a/qemu/target/arm/neon-dp.decode b/qemu/target/arm/neon-dp.decode
index 2277b4c7..0102aa72 100644
--- a/qemu/target/arm/neon-dp.decode
+++ b/qemu/target/arm/neon-dp.decode
@@ -459,6 +459,8 @@ Vimm_1r          1111 001 . 1 . 000 ... .... cmode:4 0 . op:1 1 .... @1reg_imm
     # VQMOVN: signed result, source may be signed (_S) or unsigned (_U)
     VQMOVN_S     1111 001 11 . 11 .. 10 .... 0 0101 0 . 0 .... @2misc_q0
     VQMOVN_U     1111 001 11 . 11 .. 10 .... 0 0101 1 . 0 .... @2misc_q0
+
+    VSHLL        1111 001 11 . 11 .. 10 .... 0 0110 0 . 0 .... @2misc_q0
   ]
 
   # Subgroup for size != 0b11
diff --git a/qemu/target/arm/translate-neon.inc.c b/qemu/target/arm/translate-neon.inc.c
index 6de6c66b..f13ce1f9 100644
--- a/qemu/target/arm/translate-neon.inc.c
+++ b/qemu/target/arm/translate-neon.inc.c
@@ -3343,3 +3343,56 @@ DO_VMOVN(VMOVN, gen_neon_narrow_u)
 DO_VMOVN(VQMOVUN, gen_helper_neon_unarrow_sat)
 DO_VMOVN(VQMOVN_S, gen_helper_neon_narrow_sat_s)
 DO_VMOVN(VQMOVN_U, gen_helper_neon_narrow_sat_u)
+
+static bool trans_VSHLL(DisasContext *s, arg_2misc *a)
+{
+    TCGContext *tcg_ctx = s->uc->tcg_ctx;
+    TCGv_i32 rm0, rm1;
+    TCGv_i64 rd;
+    static NeonGenWidenFn * const widenfns[] = {
+        gen_helper_neon_widen_u8,
+        gen_helper_neon_widen_u16,
+        tcg_gen_extu_i32_i64,
+        NULL,
+    };
+    NeonGenWidenFn *widenfn = widenfns[a->size];
+
+    if (!arm_dc_feature(s, ARM_FEATURE_NEON)) {
+        return false;
+    }
+
+    /* UNDEF accesses to D16-D31 if they don't exist. */
+    if (!dc_isar_feature(aa32_simd_r32, s) &&
+        ((a->vd | a->vm) & 0x10)) {
+        return false;
+    }
+
+    if (a->vd & 1) {
+        return false;
+    }
+
+    if (!widenfn) {
+        return false;
+    }
+
+    if (!vfp_access_check(s)) {
+        return true;
+    }
+
+    rd = tcg_temp_new_i64(tcg_ctx);
+
+    rm0 = neon_load_reg(s, a->vm, 0);
+    rm1 = neon_load_reg(s, a->vm, 1);
+
+    widenfn(tcg_ctx, rd, rm0);
+    tcg_gen_shli_i64(tcg_ctx, rd, rd, 8 << a->size);
+    neon_store_reg64(s, rd, a->vd);
+    widenfn(tcg_ctx, rd, rm1);
+    tcg_gen_shli_i64(tcg_ctx, rd, rd, 8 << a->size);
+    neon_store_reg64(s, rd, a->vd + 1);
+
+    tcg_temp_free_i64(tcg_ctx, rd);
+    tcg_temp_free_i32(tcg_ctx, rm0);
+    tcg_temp_free_i32(tcg_ctx, rm1);
+    return true;
+}
diff --git a/qemu/target/arm/translate.c b/qemu/target/arm/translate.c
index c56985a6..0518c574 100644
--- a/qemu/target/arm/translate.c
+++ b/qemu/target/arm/translate.c
@@ -3081,27 +3081,6 @@ static void gen_neon_trn_u16(DisasContext *s, TCGv_i32 t0, TCGv_i32 t1)
     tcg_temp_free_i32(tcg_ctx, rd);
 }
 
-static inline void gen_neon_widen(DisasContext *s, TCGv_i64 dest, TCGv_i32 src, int size, int u)
-{
-    TCGContext *tcg_ctx = s->uc->tcg_ctx;
-    if (u) {
-        switch (size) {
-        case 0: gen_helper_neon_widen_u8(tcg_ctx, dest, src); break;
-        case 1: gen_helper_neon_widen_u16(tcg_ctx, dest, src); break;
-        case 2: tcg_gen_extu_i32_i64(tcg_ctx, dest, src); break;
-        default: abort();
-        }
-    } else {
-        switch (size) {
-        case 0: gen_helper_neon_widen_s8(tcg_ctx, dest, src); break;
-        case 1: gen_helper_neon_widen_s16(tcg_ctx, dest, src); break;
-        case 2: tcg_gen_ext_i32_i64(tcg_ctx, dest, src); break;
-        default: abort();
-        }
-    }
-    tcg_temp_free_i32(tcg_ctx, src);
-}
-
 /* Symbolic constants for op fields for Neon 2-register miscellaneous.
  * The values correspond to bits [17:16,10:7]; see the ARM ARM DDI0406B
  * table A7-13.
@@ -5052,6 +5031,7 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn)
                 case NEON_2RM_VUZP:
                 case NEON_2RM_VZIP:
                 case NEON_2RM_VMOVN: case NEON_2RM_VQMOVN:
+                case NEON_2RM_VSHLL:
                     /* handled by decodetree */
                     return 1;
                 case NEON_2RM_VTRN:
@@ -5067,20 +5047,6 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn)
                         goto elementwise;
                     }
                     break;
-                case NEON_2RM_VSHLL:
-                    if (q || (rd & 1)) {
-                        return 1;
-                    }
-                    tmp = neon_load_reg(s, rm, 0);
-                    tmp2 = neon_load_reg(s, rm, 1);
-                    for (pass = 0; pass < 2; pass++) {
-                        if (pass == 1)
-                            tmp = tmp2;
-                        gen_neon_widen(s, s->V0, tmp, size, 1);
-                        tcg_gen_shli_i64(tcg_ctx, s->V0, s->V0, 8 << size);
-                        neon_store_reg64(s, s->V0, rd + pass);
-                    }
-                    break;
                 case NEON_2RM_VCVT_F16_F32:
                 {
                     TCGv_ptr fpst;