target/arm: Implement SCLAMP, UCLAMP

This is an SVE instruction that operates using the SVE vector
length but that it is present only if SME is implemented.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220708151540.18136-31-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

4 files changed, 149 insertions, 0 deletions

diff --git a/target/arm/helper.h b/target/arm/helper.h
index 3a8ce42..92f36d9 100644
--- a/target/arm/helper.h
+++ b/target/arm/helper.h
@@ -1019,6 +1019,24 @@ DEF_HELPER_FLAGS_6(gvec_bfmlal, TCG_CALL_NO_RWG,
 DEF_HELPER_FLAGS_6(gvec_bfmlal_idx, TCG_CALL_NO_RWG,
                    void, ptr, ptr, ptr, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_5(gvec_sclamp_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_sclamp_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_sclamp_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_sclamp_d, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_5(gvec_uclamp_b, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_uclamp_h, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_uclamp_s, TCG_CALL_NO_RWG,
+                   void, ptr, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_5(gvec_uclamp_d, 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/sve.decode b/target/arm/sve.decode
index a9e48f0..14b3a69 100644
--- a/target/arm/sve.decode
+++ b/target/arm/sve.decode
@@ -1695,3 +1695,8 @@ PSEL            00100101 .. 1 100 .. 01 .... 0 .... 0 ....  \
                 @psel esz=2 imm=%psel_imm_s
 PSEL            00100101 .1 1 000 .. 01 .... 0 .... 0 ....  \
                 @psel esz=3 imm=%psel_imm_d
+
+### SVE clamp
+
+SCLAMP          01000100 .. 0 ..... 110000 ..... .....          @rda_rn_rm
+UCLAMP          01000100 .. 0 ..... 110001 ..... .....          @rda_rn_rm
diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c
index 9ed3b26..41f8b12 100644
--- a/target/arm/translate-sve.c
+++ b/target/arm/translate-sve.c
@@ -7478,3 +7478,105 @@ static bool trans_PSEL(DisasContext *s, arg_psel *a)
     tcg_temp_free_ptr(ptr);
     return true;
 }
+
+static void gen_sclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a)
+{
+    tcg_gen_smax_i32(d, a, n);
+    tcg_gen_smin_i32(d, d, m);
+}
+
+static void gen_sclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a)
+{
+    tcg_gen_smax_i64(d, a, n);
+    tcg_gen_smin_i64(d, d, m);
+}
+
+static void gen_sclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
+                           TCGv_vec m, TCGv_vec a)
+{
+    tcg_gen_smax_vec(vece, d, a, n);
+    tcg_gen_smin_vec(vece, d, d, m);
+}
+
+static void gen_sclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
+                       uint32_t a, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop[] = {
+        INDEX_op_smin_vec, INDEX_op_smax_vec, 0
+    };
+    static const GVecGen4 ops[4] = {
+        { .fniv = gen_sclamp_vec,
+          .fno  = gen_helper_gvec_sclamp_b,
+          .opt_opc = vecop,
+          .vece = MO_8 },
+        { .fniv = gen_sclamp_vec,
+          .fno  = gen_helper_gvec_sclamp_h,
+          .opt_opc = vecop,
+          .vece = MO_16 },
+        { .fni4 = gen_sclamp_i32,
+          .fniv = gen_sclamp_vec,
+          .fno  = gen_helper_gvec_sclamp_s,
+          .opt_opc = vecop,
+          .vece = MO_32 },
+        { .fni8 = gen_sclamp_i64,
+          .fniv = gen_sclamp_vec,
+          .fno  = gen_helper_gvec_sclamp_d,
+          .opt_opc = vecop,
+          .vece = MO_64,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64 }
+    };
+    tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]);
+}
+
+TRANS_FEAT(SCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_sclamp, a)
+
+static void gen_uclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a)
+{
+    tcg_gen_umax_i32(d, a, n);
+    tcg_gen_umin_i32(d, d, m);
+}
+
+static void gen_uclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a)
+{
+    tcg_gen_umax_i64(d, a, n);
+    tcg_gen_umin_i64(d, d, m);
+}
+
+static void gen_uclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
+                           TCGv_vec m, TCGv_vec a)
+{
+    tcg_gen_umax_vec(vece, d, a, n);
+    tcg_gen_umin_vec(vece, d, d, m);
+}
+
+static void gen_uclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
+                       uint32_t a, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop[] = {
+        INDEX_op_umin_vec, INDEX_op_umax_vec, 0
+    };
+    static const GVecGen4 ops[4] = {
+        { .fniv = gen_uclamp_vec,
+          .fno  = gen_helper_gvec_uclamp_b,
+          .opt_opc = vecop,
+          .vece = MO_8 },
+        { .fniv = gen_uclamp_vec,
+          .fno  = gen_helper_gvec_uclamp_h,
+          .opt_opc = vecop,
+          .vece = MO_16 },
+        { .fni4 = gen_uclamp_i32,
+          .fniv = gen_uclamp_vec,
+          .fno  = gen_helper_gvec_uclamp_s,
+          .opt_opc = vecop,
+          .vece = MO_32 },
+        { .fni8 = gen_uclamp_i64,
+          .fniv = gen_uclamp_vec,
+          .fno  = gen_helper_gvec_uclamp_d,
+          .opt_opc = vecop,
+          .vece = MO_64,
+          .prefer_i64 = TCG_TARGET_REG_BITS == 64 }
+    };
+    tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]);
+}
+
+TRANS_FEAT(UCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_uclamp, a)
diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c
index 9a9c034..f59d3b2 100644
--- a/target/arm/vec_helper.c
+++ b/target/arm/vec_helper.c
@@ -2690,3 +2690,27 @@ void HELPER(gvec_bfmlal_idx)(void *vd, void *vn, void *vm,
     }
     clear_tail(d, opr_sz, simd_maxsz(desc));
 }
+
+#define DO_CLAMP(NAME, TYPE) \
+void HELPER(NAME)(void *d, void *n, void *m, void *a, uint32_t desc)    \
+{                                                                       \
+    intptr_t i, opr_sz = simd_oprsz(desc);                              \
+    for (i = 0; i < opr_sz; i += sizeof(TYPE)) {                        \
+        TYPE aa = *(TYPE *)(a + i);                                     \
+        TYPE nn = *(TYPE *)(n + i);                                     \
+        TYPE mm = *(TYPE *)(m + i);                                     \
+        TYPE dd = MIN(MAX(aa, nn), mm);                                 \
+        *(TYPE *)(d + i) = dd;                                          \
+    }                                                                   \
+    clear_tail(d, opr_sz, simd_maxsz(desc));                            \
+}
+
+DO_CLAMP(gvec_sclamp_b, int8_t)
+DO_CLAMP(gvec_sclamp_h, int16_t)
+DO_CLAMP(gvec_sclamp_s, int32_t)
+DO_CLAMP(gvec_sclamp_d, int64_t)
+
+DO_CLAMP(gvec_uclamp_b, uint8_t)
+DO_CLAMP(gvec_uclamp_h, uint16_t)
+DO_CLAMP(gvec_uclamp_s, uint32_t)
+DO_CLAMP(gvec_uclamp_d, uint64_t)