[Qemu-arm] [PATCH 13/23] target/arm: Implement SVE bitwise shift by wide elements (predicated)

[Richard Henderson]

Signed-off-by: Richard Henderson <address@hidden>
---
 target/arm/helper-sve.h    | 21 +++++++++++++++++++++
 target/arm/sve_helper.c    | 36 ++++++++++++++++++++++++++++++++++++
 target/arm/translate-sve.c | 20 ++++++++++++++++++++
 target/arm/sve.def         |  6 ++++++
 4 files changed, 83 insertions(+)

diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h
index 61b1287269..a2db3e2fd9 100644
--- a/target/arm/helper-sve.h
+++ b/target/arm/helper-sve.h
@@ -189,6 +189,27 @@ DEF_HELPER_FLAGS_5(sve_lsl_zpzz_s, TCG_CALL_NO_RWG,
 DEF_HELPER_FLAGS_5(sve_lsl_zpzz_d, TCG_CALL_NO_RWG,
                    void, ptr, ptr, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_5(sve_asr_zpzw_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(sve_asr_zpzw_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(sve_asr_zpzw_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_5(sve_lsr_zpzw_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(sve_lsr_zpzw_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(sve_lsr_zpzw_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_5(sve_lsl_zpzw_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(sve_lsl_zpzw_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(sve_lsl_zpzw_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+
 DEF_HELPER_FLAGS_3(sve_orv_b, TCG_CALL_NO_RWG, i64, ptr, ptr, i32)
 DEF_HELPER_FLAGS_3(sve_orv_h, TCG_CALL_NO_RWG, i64, ptr, ptr, i32)
 DEF_HELPER_FLAGS_3(sve_orv_s, TCG_CALL_NO_RWG, i64, ptr, ptr, i32)
diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c
index 20f1e60fda..3be6d1ae05 100644
--- a/target/arm/sve_helper.c
+++ b/target/arm/sve_helper.c
@@ -517,6 +517,42 @@ DO_ZPZZ_D(sve_lsl_zpzz_d, uint64_t, DO_LSL)
 #undef DO_ZPZZ
 #undef DO_ZPZZ_D
 
+/* Three-operand expander, controlled by a predicate, in which the
+ * third operand is "wide".  That is, for D = N op M, the same 64-bit
+ * value of M is used with all of the narrower values of N.
+ */
+#define DO_ZPZW(NAME, TYPE, TYPEW, H, OP)                               \
+void HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) \
+{                                                                       \
+    intptr_t iv = 0, ib = 0, opr_sz = simd_oprsz(desc);                 \
+    for (iv = ib = 0; iv < opr_sz; iv += 16, ib += 2) {                 \
+        uint16_t pg = *(uint16_t *)(vg + H2(ib));                       \
+        TYPEW mm = *(TYPEW *)(vm + iv);                                 \
+        intptr_t i = 0;                                                 \
+        do {                                                            \
+            if (pg & 1) {                                               \
+                TYPE nn = *(TYPE *)(vn + iv + H(i));                    \
+                *(TYPE *)(vd + iv + H(i)) = OP(nn, mm);                 \
+            }                                                           \
+            i += sizeof(TYPE), pg >>= sizeof(TYPE);                     \
+        } while (pg);                                                   \
+    }                                                                   \
+}
+
+DO_ZPZW(sve_asr_zpzw_b, int8_t, uint64_t, H1, DO_ASR)
+DO_ZPZW(sve_lsr_zpzw_b, uint8_t, uint64_t, H1, DO_LSR)
+DO_ZPZW(sve_lsl_zpzw_b, uint8_t, uint64_t, H1, DO_LSL)
+
+DO_ZPZW(sve_asr_zpzw_h, int16_t, uint64_t, H1_2, DO_ASR)
+DO_ZPZW(sve_lsr_zpzw_h, uint16_t, uint64_t, H1_2, DO_LSR)
+DO_ZPZW(sve_lsl_zpzw_h, uint16_t, uint64_t, H1_2, DO_LSL)
+
+DO_ZPZW(sve_asr_zpzw_s, int32_t, uint64_t, H1_4, DO_ASR)
+DO_ZPZW(sve_lsr_zpzw_s, uint32_t, uint64_t, H1_4, DO_LSR)
+DO_ZPZW(sve_lsl_zpzw_s, uint32_t, uint64_t, H1_4, DO_LSL)
+
+#undef DO_ZPZW
+
 /* Two-operand reduction expander, controlled by a predicate.
  * The difference between TYPERED and TYPERET has to do with
  * sign-extension.  E.g. for SMAX, TYPERED must be signed,
diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c
index 685a3ba249..91f07d57e3 100644
--- a/target/arm/translate-sve.c
+++ b/target/arm/translate-sve.c
@@ -230,6 +230,26 @@ void trans_UDIV_zpzz(DisasContext *s, arg_rprr_esz *a, 
uint32_t insn)
 
 #undef DO_ZPZZ
 
+#define DO_ZPZW(NAME, name) \
+void trans_##NAME##_zpzw(DisasContext *s, arg_rprr_esz *a, uint32_t insn) \
+{                                                                         \
+    static gen_helper_gvec_4 * const fns[3] = {                           \
+        gen_helper_sve_##name##_zpzw_b, gen_helper_sve_##name##_zpzw_h,   \
+        gen_helper_sve_##name##_zpzw_s,                                   \
+    };                                                                    \
+    if ((unsigned)a->esz < 3) {                                           \
+        do_zpzz_ool(s, a, fns[a->esz]);                                   \
+    } else {                                                              \
+        unallocated_encoding(s);                                          \
+    }                                                                     \
+}
+
+DO_ZPZW(ASR, asr)
+DO_ZPZW(LSR, lsr)
+DO_ZPZW(LSL, lsl)
+
+#undef DO_ZPZW
+
 typedef void gen_helper_gvec_reduc(TCGv_i64, TCGv_ptr, TCGv_ptr, TCGv_i32);
 static void do_vpz_ool(DisasContext *s, arg_rpr_esz *a,
                        gen_helper_gvec_reduc *fn)
diff --git a/target/arm/sve.def b/target/arm/sve.def
index 9f9c0803a0..66be950ca5 100644
--- a/target/arm/sve.def
+++ b/target/arm/sve.def
@@ -141,6 +141,12 @@ ASR_zpzz           00000100 .. 010 100 100 ... ..... ..... 
        @rdm_pg_rn_esz # ASRR
 LSR_zpzz               00000100 .. 010 101 100 ... ..... .....         
@rdm_pg_rn_esz # LSRR
 LSL_zpzz               00000100 .. 010 111 100 ... ..... .....         
@rdm_pg_rn_esz # LSLR
 
+# SVE bitwise shift by wide elements (predicated)
+# Note these require size != 3.
+ASR_zpzw               00000100 .. 011 000 100 ... ..... .....         
@rdn_pg_rm_esz
+LSR_zpzw               00000100 .. 011 001 100 ... ..... .....         
@rdn_pg_rm_esz
+LSL_zpzw               00000100 .. 011 011 100 ... ..... .....         
@rdn_pg_rm_esz
+
 ### SVE Logical - Unpredicated Group
 
 # SVE bitwise logical operations (unpredicated)
-- 
2.14.3