target-arm: Fix VQMOVUN Neon instruction.

VQMOVUN does a signed-to-unsigned saturating conversion. This is
different from both the signed-to-signed and unsigned-to-unsigned
conversions already implemented, so we need a new set of helper
functions (neon_unarrow_sat*).

Signed-off-by: Juha Riihimäki <juha.riihimaki@nokia.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-arm/helpers.h

@@ -299,10 +299,13 @@ DEF_HELPER_3(neon_qrdmulh_s32, i32, env, i32, i32)
 
 DEF_HELPER_1(neon_narrow_u8, i32, i64)
 DEF_HELPER_1(neon_narrow_u16, i32, i64)
+DEF_HELPER_2(neon_unarrow_sat8, i32, env, i64)
 DEF_HELPER_2(neon_narrow_sat_u8, i32, env, i64)
 DEF_HELPER_2(neon_narrow_sat_s8, i32, env, i64)
+DEF_HELPER_2(neon_unarrow_sat16, i32, env, i64)
 DEF_HELPER_2(neon_narrow_sat_u16, i32, env, i64)
 DEF_HELPER_2(neon_narrow_sat_s16, i32, env, i64)
+DEF_HELPER_2(neon_unarrow_sat32, i32, env, i64)
 DEF_HELPER_2(neon_narrow_sat_u32, i32, env, i64)
 DEF_HELPER_2(neon_narrow_sat_s32, i32, env, i64)
 DEF_HELPER_1(neon_narrow_high_u8, i32, i64)

target-arm/neon_helper.c

@@ -1053,6 +1053,33 @@ uint32_t HELPER(neon_narrow_round_high_u16)(uint64_t x)
     return ((x >> 16) & 0xffff) | ((x >> 32) & 0xffff0000);
 }
 
+uint32_t HELPER(neon_unarrow_sat8)(CPUState *env, uint64_t x)
+{
+    uint16_t s;
+    uint8_t d;
+    uint32_t res = 0;
+#define SAT8(n) \
+    s = x >> n; \
+    if (s & 0x8000) { \
+        SET_QC(); \
+    } else { \
+        if (s > 0xff) { \
+            d = 0xff; \
+            SET_QC(); \
+        } else  { \
+            d = s; \
+        } \
+        res |= (uint32_t)d << (n / 2); \
+    }
+
+    SAT8(0);
+    SAT8(16);
+    SAT8(32);
+    SAT8(48);
+#undef SAT8
+    return res;
+}
+
 uint32_t HELPER(neon_narrow_sat_u8)(CPUState *env, uint64_t x)
 {
     uint16_t s;
@@ -1099,6 +1126,29 @@ uint32_t HELPER(neon_narrow_sat_s8)(CPUState *env, uint64_t x)
     return res;
 }
 
+uint32_t HELPER(neon_unarrow_sat16)(CPUState *env, uint64_t x)
+{
+    uint32_t high;
+    uint32_t low;
+    low = x;
+    if (low & 0x80000000) {
+        low = 0;
+        SET_QC();
+    } else if (low > 0xffff) {
+        low = 0xffff;
+        SET_QC();
+    }
+    high = x >> 32;
+    if (high & 0x80000000) {
+        high = 0;
+        SET_QC();
+    } else if (high > 0xffff) {
+        high = 0xffff;
+        SET_QC();
+    }
+    return low | (high << 16);
+}
+
 uint32_t HELPER(neon_narrow_sat_u16)(CPUState *env, uint64_t x)
 {
     uint32_t high;
@@ -1133,6 +1183,19 @@ uint32_t HELPER(neon_narrow_sat_s16)(CPUState *env, uint64_t x)
     return (uint16_t)low | (high << 16);
 }
 
+uint32_t HELPER(neon_unarrow_sat32)(CPUState *env, uint64_t x)
+{
+    if (x & 0x8000000000000000ull) {
+        SET_QC();
+        return 0;
+    }
+    if (x > 0xffffffffu) {
+        SET_QC();
+        return 0xffffffffu;
+    }
+    return x;
+}
+
 uint32_t HELPER(neon_narrow_sat_u32)(CPUState *env, uint64_t x)
 {
     if (x > 0xffffffffu) {

target-arm/translate.c

@@ -4065,6 +4065,16 @@ static inline void gen_neon_narrow_satu(int size, TCGv dest, TCGv_i64 src)
     }
 }
 
+static inline void gen_neon_unarrow_sats(int size, TCGv dest, TCGv_i64 src)
+{
+    switch (size) {
+    case 0: gen_helper_neon_unarrow_sat8(dest, cpu_env, src); break;
+    case 1: gen_helper_neon_unarrow_sat16(dest, cpu_env, src); break;
+    case 2: gen_helper_neon_unarrow_sat32(dest, cpu_env, src); break;
+    default: abort();
+    }
+}
+
 static inline void gen_neon_shift_narrow(int size, TCGv var, TCGv shift,
                                          int q, int u)
 {
@@ -5461,12 +5471,18 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     for (pass = 0; pass < 2; pass++) {
                         neon_load_reg64(cpu_V0, rm + pass);
                         tmp = new_tmp();
-                        if (op == 36 && q == 0) {
+                        if (op == 36) {
-                            gen_neon_narrow(size, tmp, cpu_V0);
+                            if (q) { /* VQMOVUN */
-                        } else if (q) {
+                                gen_neon_unarrow_sats(size, tmp, cpu_V0);
-                            gen_neon_narrow_satu(size, tmp, cpu_V0);
+                            } else { /* VMOVN */
-                        } else {
+                                gen_neon_narrow(size, tmp, cpu_V0);
-                            gen_neon_narrow_sats(size, tmp, cpu_V0);
+                            }
+                        } else { /* VQMOVN */
+                            if (q) {
+                                gen_neon_narrow_satu(size, tmp, cpu_V0);
+                            } else {
+                                gen_neon_narrow_sats(size, tmp, cpu_V0);
+                            }
                         }
                         if (pass == 0) {
                             tmp2 = tmp;