[PATCH 20/20] target/arm: Convert sq{, r}dmulh to gvec for aa64 advsimd

[Richard Henderson <richard.henderson@linaro.org>]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/helper.h        | 10 ++++++++
 target/arm/translate-a64.c | 33 ++++++++++++++++++--------
 target/arm/vec_helper.c    | 48 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 81 insertions(+), 10 deletions(-)

diff --git a/target/arm/helper.h b/target/arm/helper.h
index 378bb1898b..3ca73a1764 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -776,6 +776,16 @@ DEF_HELPER_FLAGS_5(gvec_mls_idx_s, TCG_CALL_NO_RWG,
 DEF_HELPER_FLAGS_5(gvec_mls_idx_d, TCG_CALL_NO_RWG,
                    void, ptr, ptr, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_5(neon_sqdmulh_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(neon_sqdmulh_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_5(neon_sqrdmulh_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(neon_sqrdmulh_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+
 #ifdef TARGET_AARCH64
 #include "helper-a64.h"
 #include "helper-sve.h"
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index c74c6e854c..d4da12268c 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -697,6 +697,20 @@ static void gen_gvec_op3_fpst(DisasContext *s, bool is_q, int rd, int rn,
     tcg_temp_free_ptr(fpst);
 }
 
+/* Expand a 3-operand + qc + operation using an out-of-line helper.  */
+static void gen_gvec_op3_qc(DisasContext *s, bool is_q, int rd, int rn,
+                            int rm, gen_helper_gvec_3_ptr *fn)
+{
+    TCGv_ptr qc_ptr = tcg_temp_new_ptr();
+
+    tcg_gen_addi_ptr(qc_ptr, cpu_env, offsetof(CPUARMState, vfp.qc));
+    tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd),
+                       vec_full_reg_offset(s, rn),
+                       vec_full_reg_offset(s, rm), qc_ptr,
+                       is_q ? 16 : 8, vec_full_reg_size(s), 0, fn);
+    tcg_temp_free_ptr(qc_ptr);
+}
+
 /* Set ZF and NF based on a 64 bit result. This is alas fiddlier
  * than the 32 bit equivalent.
  */
@@ -11753,6 +11767,15 @@ static void disas_simd_3same_int(DisasContext *s, uint32_t insn)
             gen_gvec_fn3(s, is_q, rd, rn, rm, gen_gvec_mla, size);
         }
         return;
+    case 0x16: /* SQDMULH, SQRDMULH */
+        {
+            static gen_helper_gvec_3_ptr * const fns[2][2] = {
+                { gen_helper_neon_sqdmulh_h, gen_helper_neon_sqrdmulh_h },
+                { gen_helper_neon_sqdmulh_s, gen_helper_neon_sqrdmulh_s },
+            };
+            gen_gvec_op3_qc(s, is_q, rd, rn, rm, fns[size - 1][u]);
+        }
+        return;
     case 0x11:
         if (!u) { /* CMTST */
             gen_gvec_fn3(s, is_q, rd, rn, rm, gen_gvec_cmtst, size);
@@ -11864,16 +11887,6 @@ static void disas_simd_3same_int(DisasContext *s, uint32_t insn)
                 genenvfn = fns[size][u];
                 break;
             }
-            case 0x16: /* SQDMULH, SQRDMULH */
-            {
-                static NeonGenTwoOpEnvFn * const fns[2][2] = {
-                    { gen_helper_neon_qdmulh_s16, gen_helper_neon_qrdmulh_s16 },
-                    { gen_helper_neon_qdmulh_s32, gen_helper_neon_qrdmulh_s32 },
-                };
-                assert(size == 1 || size == 2);
-                genenvfn = fns[size - 1][u];
-                break;
-            }
             default:
                 g_assert_not_reached();
             }
diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c
index fb53684ce3..73d62c4e4f 100644
--- a/target/arm/vec_helper.c
+++ b/target/arm/vec_helper.c
@@ -109,6 +109,30 @@ void HELPER(gvec_qrdmlsh_s16)(void *vd, void *vn, void *vm,
     clear_tail(d, opr_sz, simd_maxsz(desc));
 }
 
+void HELPER(neon_sqdmulh_h)(void *vd, void *vn, void *vm,
+                            void *vq, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    int16_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz / 2; ++i) {
+        d[i] = do_sqrdmlah_h(n[i], m[i], 0, false, false, vq);
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
+void HELPER(neon_sqrdmulh_h)(void *vd, void *vn, void *vm,
+                             void *vq, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    int16_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz / 2; ++i) {
+        d[i] = do_sqrdmlah_h(n[i], m[i], 0, false, true, vq);
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
 /* Signed saturating rounding doubling multiply-accumulate high half, 32-bit */
 static int32_t do_sqrdmlah_s(int32_t src1, int32_t src2, int32_t src3,
                              bool neg, bool round, uint32_t *sat)
@@ -172,6 +196,30 @@ void HELPER(gvec_qrdmlsh_s32)(void *vd, void *vn, void *vm,
     clear_tail(d, opr_sz, simd_maxsz(desc));
 }
 
+void HELPER(neon_sqdmulh_s)(void *vd, void *vn, void *vm,
+                            void *vq, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    int32_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz / 4; ++i) {
+        d[i] = do_sqrdmlah_s(n[i], m[i], 0, false, false, vq);
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
+void HELPER(neon_sqrdmulh_s)(void *vd, void *vn, void *vm,
+                             void *vq, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    int32_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz / 4; ++i) {
+        d[i] = do_sqrdmlah_s(n[i], m[i], 0, false, true, vq);
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
 /* Integer 8 and 16-bit dot-product.
  *
  * Note that for the loops herein, host endianness does not matter
-- 
2.25.1