[PULL 61/85] target/hppa: Implement HADD

[Richard Henderson]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/hppa/helper.h     |  3 +++
 target/hppa/insns.decode |  8 +++++++-
 target/hppa/op_helper.c  | 32 ++++++++++++++++++++++++++++++++
 target/hppa/translate.c  | 37 +++++++++++++++++++++++++++++++++++++
 4 files changed, 79 insertions(+), 1 deletion(-)

diff --git a/target/hppa/helper.h b/target/hppa/helper.h
index 57ea5447b6..db662f0c33 100644
--- a/target/hppa/helper.h
+++ b/target/hppa/helper.h
@@ -14,6 +14,9 @@ DEF_HELPER_FLAGS_3(stdby_e_parallel, TCG_CALL_NO_WG, void, 
env, tl, tl)
 
 DEF_HELPER_FLAGS_1(ldc_check, TCG_CALL_NO_RWG, void, tl)
 
+DEF_HELPER_FLAGS_2(hadd_ss, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(hadd_us, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+
 DEF_HELPER_FLAGS_4(probe, TCG_CALL_NO_WG, tl, env, tl, i32, i32)
 
 DEF_HELPER_FLAGS_1(loaded_fr0, TCG_CALL_NO_RWG, void, env)
diff --git a/target/hppa/insns.decode b/target/hppa/insns.decode
index 820049b0c5..4bcfc94b1c 100644
--- a/target/hppa/insns.decode
+++ b/target/hppa/insns.decode
@@ -65,6 +65,7 @@
 &ldst           t b x disp sp m scale size
 
 &rr_cf_d        t r cf d
+&rrr            t r1 r2
 &rrr_cf         t r1 r2 cf
 &rrr_cf_d       t r1 r2 cf d
 &rrr_cf_d_sh    t r1 r2 cf d sh
@@ -81,6 +82,7 @@
 ####
 
 @rr_cf_d        ...... r:5 ..... cf:4 ...... d:1 t:5    &rr_cf_d
+@rrr            ...... r2:5 r1:5 .... ....... t:5       &rrr
 @rrr_cf         ...... r2:5 r1:5 cf:4 ....... t:5       &rrr_cf
 @rrr_cf_d       ...... r2:5 r1:5 cf:4 ...... d:1 t:5    &rrr_cf_d
 @rrr_cf_d_sh    ...... r2:5 r1:5 cf:4 .... sh:2 d:1 t:5 &rrr_cf_d_sh
@@ -208,6 +210,10 @@ subi_tsv        100101 ..... ..... .... 1 ...........   
@rri_cf
 
 cmpiclr         100100 ..... ..... .... . ...........   @rri_cf_d
 
+hadd            000010 ..... ..... 00000011 11 0 .....  @rrr
+hadd_ss         000010 ..... ..... 00000011 01 0 .....  @rrr
+hadd_us         000010 ..... ..... 00000011 00 0 .....  @rrr
+
 ####
 # Index Mem
 ####
@@ -429,7 +435,7 @@ fmpyfadd_d      101110 rm1:5 rm2:5 ... 0 1 ..0 0 0 neg:1 
t:5    ra3=%rc32
 
 @f0e_f_3        ...... ..... ..... ... .0 110 ..0 .....    \
                 &fclass3 r1=%ra64 r2=%rb64 t=%rt64
-@f0e_d_3        ...... r1:5  r2:5  ... 01 110 000 t:5
+@f0e_d_3        ...... r1:5  r2:5  ... 01 110 000 t:5      &fclass3
 
 # Floating point class 0
 
diff --git a/target/hppa/op_helper.c b/target/hppa/op_helper.c
index a25e6df7e4..d93801ca47 100644
--- a/target/hppa/op_helper.c
+++ b/target/hppa/op_helper.c
@@ -377,3 +377,35 @@ target_ulong HELPER(read_interval_timer)(void)
     return qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) >> 2;
 #endif
 }
+
+uint64_t HELPER(hadd_ss)(uint64_t r1, uint64_t r2)
+{
+    uint64_t ret = 0;
+
+    for (int i = 0; i < 64; i += 16) {
+        int f1 = sextract64(r1, i, 16);
+        int f2 = sextract64(r2, i, 16);
+        int fr = f1 + f2;
+
+        fr = MIN(fr, INT16_MAX);
+        fr = MAX(fr, INT16_MIN);
+        ret = deposit64(ret, i, 16, fr);
+    }
+    return ret;
+}
+
+uint64_t HELPER(hadd_us)(uint64_t r1, uint64_t r2)
+{
+    uint64_t ret = 0;
+
+    for (int i = 0; i < 64; i += 16) {
+        int f1 = extract64(r1, i, 16);
+        int f2 = sextract64(r2, i, 16);
+        int fr = f1 + f2;
+
+        fr = MIN(fr, UINT16_MAX);
+        fr = MAX(fr, 0);
+        ret = deposit64(ret, i, 16, fr);
+    }
+    return ret;
+}
diff --git a/target/hppa/translate.c b/target/hppa/translate.c
index 1d749e2a78..dc8e37f99c 100644
--- a/target/hppa/translate.c
+++ b/target/hppa/translate.c
@@ -23,6 +23,7 @@
 #include "qemu/host-utils.h"
 #include "exec/exec-all.h"
 #include "tcg/tcg-op.h"
+#include "tcg/tcg-op-gvec.h"
 #include "exec/helper-proto.h"
 #include "exec/helper-gen.h"
 #include "exec/translator.h"
@@ -2767,6 +2768,42 @@ static bool trans_cmpiclr(DisasContext *ctx, 
arg_rri_cf_d *a)
     return nullify_end(ctx);
 }
 
+static bool do_multimedia(DisasContext *ctx, arg_rrr *a,
+                          void (*fn)(TCGv_i64, TCGv_i64, TCGv_i64))
+{
+    TCGv_i64 r1, r2, dest;
+
+    if (!ctx->is_pa20) {
+        return false;
+    }
+
+    nullify_over(ctx);
+
+    r1 = load_gpr(ctx, a->r1);
+    r2 = load_gpr(ctx, a->r2);
+    dest = dest_gpr(ctx, a->t);
+
+    fn(dest, r1, r2);
+    save_gpr(ctx, a->t, dest);
+
+    return nullify_end(ctx);
+}
+
+static bool trans_hadd(DisasContext *ctx, arg_rrr *a)
+{
+    return do_multimedia(ctx, a, tcg_gen_vec_add16_i64);
+}
+
+static bool trans_hadd_ss(DisasContext *ctx, arg_rrr *a)
+{
+    return do_multimedia(ctx, a, gen_helper_hadd_ss);
+}
+
+static bool trans_hadd_us(DisasContext *ctx, arg_rrr *a)
+{
+    return do_multimedia(ctx, a, gen_helper_hadd_us);
+}
+
 static bool trans_ld(DisasContext *ctx, arg_ldst *a)
 {
     if (!ctx->is_pa20 && a->size > MO_32) {
-- 
2.34.1