target/arm: Implement MVE VQDMLSDH and VQRDMLSDH

Implement the MVE VQDMLSDH and VQRDMLSDH insns, which are
like VQDMLADH and VQRDMLADH except that products are subtracted
rather than added.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210617121628.20116-38-peter.maydell@linaro.org

target/arm/helper-mve.h

@@ -217,6 +217,22 @@ DEF_HELPER_FLAGS_4(mve_vqrdmladhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
 DEF_HELPER_FLAGS_4(mve_vqrdmladhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
 DEF_HELPER_FLAGS_4(mve_vqrdmladhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
 
+DEF_HELPER_FLAGS_4(mve_vqdmlsdhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmlsdhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmlsdhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
+DEF_HELPER_FLAGS_4(mve_vqdmlsdhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmlsdhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmlsdhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
+DEF_HELPER_FLAGS_4(mve_vqrdmlsdhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmlsdhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmlsdhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
+DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
 DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32)

target/arm/mve.decode

@@ -147,6 +147,11 @@ VQDMLADHX        1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op
 VQRDMLADH        1110 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op
 VQRDMLADHX       1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op
 
+VQDMLSDH         1111 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 0 @2op
+VQDMLSDHX        1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op
+VQRDMLSDH        1111 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op
+VQRDMLSDHX       1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op
+
 # Vector miscellaneous
 
 VCLS             1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op

target/arm/mve_helper.c

@@ -702,6 +702,36 @@ static int32_t do_vqdmladh_w(int32_t a, int32_t b, int32_t c, int32_t d,
     return r >> 32;
 }
 
+static int8_t do_vqdmlsdh_b(int8_t a, int8_t b, int8_t c, int8_t d,
+                            int round, bool *sat)
+{
+    int64_t r = ((int64_t)a * b - (int64_t)c * d) * 2 + (round << 7);
+    return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8;
+}
+
+static int16_t do_vqdmlsdh_h(int16_t a, int16_t b, int16_t c, int16_t d,
+                             int round, bool *sat)
+{
+    int64_t r = ((int64_t)a * b - (int64_t)c * d) * 2 + (round << 15);
+    return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16;
+}
+
+static int32_t do_vqdmlsdh_w(int32_t a, int32_t b, int32_t c, int32_t d,
+                             int round, bool *sat)
+{
+    int64_t m1 = (int64_t)a * b;
+    int64_t m2 = (int64_t)c * d;
+    int64_t r;
+    /* The same ordering issue as in do_vqdmladh_w applies here too */
+    if (ssub64_overflow(m1, m2, &r) ||
+        sadd64_overflow(r, (round << 30), &r) ||
+        sadd64_overflow(r, r, &r)) {
+        *sat = true;
+        return r < 0 ? INT32_MAX : INT32_MIN;
+    }
+    return r >> 32;
+}
+
 DO_VQDMLADH_OP(vqdmladhb, 1, int8_t, 0, 0, do_vqdmladh_b)
 DO_VQDMLADH_OP(vqdmladhh, 2, int16_t, 0, 0, do_vqdmladh_h)
 DO_VQDMLADH_OP(vqdmladhw, 4, int32_t, 0, 0, do_vqdmladh_w)
@@ -716,6 +746,20 @@ DO_VQDMLADH_OP(vqrdmladhxb, 1, int8_t, 1, 1, do_vqdmladh_b)
 DO_VQDMLADH_OP(vqrdmladhxh, 2, int16_t, 1, 1, do_vqdmladh_h)
 DO_VQDMLADH_OP(vqrdmladhxw, 4, int32_t, 1, 1, do_vqdmladh_w)
 
+DO_VQDMLADH_OP(vqdmlsdhb, 1, int8_t, 0, 0, do_vqdmlsdh_b)
+DO_VQDMLADH_OP(vqdmlsdhh, 2, int16_t, 0, 0, do_vqdmlsdh_h)
+DO_VQDMLADH_OP(vqdmlsdhw, 4, int32_t, 0, 0, do_vqdmlsdh_w)
+DO_VQDMLADH_OP(vqdmlsdhxb, 1, int8_t, 1, 0, do_vqdmlsdh_b)
+DO_VQDMLADH_OP(vqdmlsdhxh, 2, int16_t, 1, 0, do_vqdmlsdh_h)
+DO_VQDMLADH_OP(vqdmlsdhxw, 4, int32_t, 1, 0, do_vqdmlsdh_w)
+
+DO_VQDMLADH_OP(vqrdmlsdhb, 1, int8_t, 0, 1, do_vqdmlsdh_b)
+DO_VQDMLADH_OP(vqrdmlsdhh, 2, int16_t, 0, 1, do_vqdmlsdh_h)
+DO_VQDMLADH_OP(vqrdmlsdhw, 4, int32_t, 0, 1, do_vqdmlsdh_w)
+DO_VQDMLADH_OP(vqrdmlsdhxb, 1, int8_t, 1, 1, do_vqdmlsdh_b)
+DO_VQDMLADH_OP(vqrdmlsdhxh, 2, int16_t, 1, 1, do_vqdmlsdh_h)
+DO_VQDMLADH_OP(vqrdmlsdhxw, 4, int32_t, 1, 1, do_vqdmlsdh_w)
+
 #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN)                              \
     void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn,   \
                                 uint32_t rm)                            \

target/arm/translate-mve.c

@@ -414,6 +414,10 @@ DO_2OP(VQDMLADH, vqdmladh)
 DO_2OP(VQDMLADHX, vqdmladhx)
 DO_2OP(VQRDMLADH, vqrdmladh)
 DO_2OP(VQRDMLADHX, vqrdmladhx)
+DO_2OP(VQDMLSDH, vqdmlsdh)
+DO_2OP(VQDMLSDHX, vqdmlsdhx)
+DO_2OP(VQRDMLSDH, vqrdmlsdh)
+DO_2OP(VQRDMLSDHX, vqrdmlsdhx)
 
 static bool do_2op_scalar(DisasContext *s, arg_2scalar *a,
                           MVEGenTwoOpScalarFn fn)