target/arm: Convert Neon VSWP to decodetree

Convert the Neon VSWP insn to decodetree. Since the new implementation
doesn't have to share a pass-loop with the other 2-reg-misc operations
we can implement the swap with 64-bit accesses rather than 32-bits
(which brings us into line with the pseudocode and is more efficient).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200616170844.13318-20-peter.maydell@linaro.org

diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode
index 5507c3e462325e73b2f1b929f66b308834f31065..2f64841de5291a71b172d5c8c7e1fd2619ed5387 100644 (file)
--- a/target/arm/neon-dp.decode
+++ b/target/arm/neon-dp.decode
@@ -488,6 +488,8 @@ Vimm_1r          1111 001 . 1 . 000 ... .... cmode:4 0 . op:1 1 .... @1reg_imm
     VABS_F       1111 001 11 . 11 .. 01 .... 0 1110 . . 0 .... @2misc
     VNEG_F       1111 001 11 . 11 .. 01 .... 0 1111 . . 0 .... @2misc
 
+    VSWP         1111 001 11 . 11 .. 10 .... 0 0000 . . 0 .... @2misc
+
     VUZP         1111 001 11 . 11 .. 10 .... 0 0010 . . 0 .... @2misc
     VZIP         1111 001 11 . 11 .. 10 .... 0 0011 . . 0 .... @2misc
 

diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c
index 29bc161f36aa3671d248ab88262d1b3c77a7419a..01da7fad462883382b93e8cc4051bcd1cda75673 100644 (file)
--- a/target/arm/translate-neon.inc.c
+++ b/target/arm/translate-neon.inc.c
@@ -3927,3 +3927,44 @@ DO_VCVT(VCVTPU, FPROUNDING_POSINF, false)
 DO_VCVT(VCVTPS, FPROUNDING_POSINF, true)
 DO_VCVT(VCVTMU, FPROUNDING_NEGINF, false)
 DO_VCVT(VCVTMS, FPROUNDING_NEGINF, true)
+
+static bool trans_VSWP(DisasContext *s, arg_2misc *a)
+{
+    TCGv_i64 rm, rd;
+    int pass;
+
+    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->size != 0) {
+        return false;
+    }
+
+    if ((a->vd | a->vm) & a->q) {
+        return false;
+    }
+
+    if (!vfp_access_check(s)) {
+        return true;
+    }
+
+    rm = tcg_temp_new_i64();
+    rd = tcg_temp_new_i64();
+    for (pass = 0; pass < (a->q ? 2 : 1); pass++) {
+        neon_load_reg64(rm, a->vm + pass);
+        neon_load_reg64(rd, a->vd + pass);
+        neon_store_reg64(rm, a->vd + pass);
+        neon_store_reg64(rd, a->vm + pass);
+    }
+    tcg_temp_free_i64(rm);
+    tcg_temp_free_i64(rd);
+
+    return true;
+}

diff --git a/target/arm/translate.c b/target/arm/translate.c
index b0181062020387b579469c70d3d03c55d237c1b1..e8cd4a9c61f11bea0bae1bd5f69a3a6f8108f06d 100644 (file)
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -4944,6 +4944,7 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn)
                 case NEON_2RM_VCVTPS:
                 case NEON_2RM_VCVTMU:
                 case NEON_2RM_VCVTMS:
+                case NEON_2RM_VSWP:
                     /* handled by decodetree */
                     return 1;
                 case NEON_2RM_VTRN:
@@ -4965,10 +4966,6 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn)
                     for (pass = 0; pass < (q ? 4 : 2); pass++) {
                         tmp = neon_load_reg(rm, pass);
                         switch (op) {
-                        case NEON_2RM_VSWP:
-                            tmp2 = neon_load_reg(rd, pass);
-                            neon_store_reg(rm, pass, tmp2);
-                            break;
                         case NEON_2RM_VTRN:
                             tmp2 = neon_load_reg(rd, pass);
                             switch (size) {