[Qemu-arm] [PATCH v2 27/67] target/arm: Implement SVE Permute - Unpredicated Group

[Richard Henderson]

Signed-off-by: Richard Henderson <address@hidden>
---
 target/arm/helper-sve.h    |  23 +++++++++
 target/arm/translate-a64.h |  14 +++---
 target/arm/sve_helper.c    | 114 +++++++++++++++++++++++++++++++++++++++++++++
 target/arm/translate-sve.c | 113 ++++++++++++++++++++++++++++++++++++++++++++
 target/arm/sve.decode      |  29 +++++++++++-
 5 files changed, 285 insertions(+), 8 deletions(-)

diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h
index 94f4356ce9..0c9aad575e 100644
--- a/target/arm/helper-sve.h
+++ b/target/arm/helper-sve.h
@@ -416,6 +416,29 @@ DEF_HELPER_FLAGS_4(sve_cpy_z_d, TCG_CALL_NO_RWG, void, 
ptr, ptr, i64, i32)
 
 DEF_HELPER_FLAGS_4(sve_ext, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 
+DEF_HELPER_FLAGS_4(sve_insr_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32)
+DEF_HELPER_FLAGS_4(sve_insr_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32)
+DEF_HELPER_FLAGS_4(sve_insr_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32)
+DEF_HELPER_FLAGS_4(sve_insr_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32)
+
+DEF_HELPER_FLAGS_3(sve_rev_b, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_rev_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_rev_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_rev_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_4(sve_tbl_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(sve_tbl_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(sve_tbl_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(sve_tbl_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_3(sve_sunpk_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_sunpk_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_sunpk_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_3(sve_uunpk_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_uunpk_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+DEF_HELPER_FLAGS_3(sve_uunpk_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
+
 DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, 
i32)
 DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, 
i32)
 DEF_HELPER_FLAGS_5(sve_eor_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, 
i32)
diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h
index e519aee314..328aa7fce1 100644
--- a/target/arm/translate-a64.h
+++ b/target/arm/translate-a64.h
@@ -66,18 +66,18 @@ static inline void assert_fp_access_checked(DisasContext *s)
 static inline int vec_reg_offset(DisasContext *s, int regno,
                                  int element, TCGMemOp size)
 {
-    int offs = 0;
+    int element_size = 1 << size;
+    int offs = element * element_size;
 #ifdef HOST_WORDS_BIGENDIAN
     /* This is complicated slightly because vfp.zregs[n].d[0] is
      * still the low half and vfp.zregs[n].d[1] the high half
      * of the 128 bit vector, even on big endian systems.
-     * Calculate the offset assuming a fully bigendian 128 bits,
-     * then XOR to account for the order of the two 64 bit halves.
+     * Calculate the offset assuming a fully little-endian 128 bits,
+     * then XOR to account for the order of the 64 bit units.
      */
-    offs += (16 - ((element + 1) * (1 << size)));
-    offs ^= 8;
-#else
-    offs += element * (1 << size);
+    if (element_size < 8) {
+        offs ^= 8 - element_size;
+    }
 #endif
     offs += offsetof(CPUARMState, vfp.zregs[regno]);
     assert_fp_access_checked(s);
diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c
index fb3f54300b..466a209c1e 100644
--- a/target/arm/sve_helper.c
+++ b/target/arm/sve_helper.c
@@ -1550,3 +1550,117 @@ void HELPER(sve_ext)(void *vd, void *vn, void *vm, 
uint32_t desc)
         memcpy(vd + n_siz, &tmp, n_ofs);
     }
 }
+
+#define DO_INSR(NAME, TYPE, H) \
+void HELPER(NAME)(void *vd, void *vn, uint64_t val, uint32_t desc) \
+{                                                                  \
+    intptr_t opr_sz = simd_oprsz(desc);                            \
+    swap_memmove(vd + sizeof(TYPE), vn, opr_sz - sizeof(TYPE));    \
+    *(TYPE *)(vd + H(0)) = val;                                    \
+}
+
+DO_INSR(sve_insr_b, uint8_t, H1)
+DO_INSR(sve_insr_h, uint16_t, H1_2)
+DO_INSR(sve_insr_s, uint32_t, H1_4)
+DO_INSR(sve_insr_d, uint64_t, )
+
+#undef DO_INSR
+
+void HELPER(sve_rev_b)(void *vd, void *vn, uint32_t desc)
+{
+    intptr_t i, j, opr_sz = simd_oprsz(desc);
+    for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) {
+        uint64_t f = *(uint64_t *)(vn + i);
+        uint64_t b = *(uint64_t *)(vn + j);
+        *(uint64_t *)(vd + i) = bswap64(b);
+        *(uint64_t *)(vd + j) = bswap64(f);
+    }
+}
+
+static inline uint64_t hswap64(uint64_t h)
+{
+    uint64_t m = 0x0000ffff0000ffffull;
+    h = rol64(h, 32);
+    return ((h & m) << 16) | ((h >> 16) & m);
+}
+
+void HELPER(sve_rev_h)(void *vd, void *vn, uint32_t desc)
+{
+    intptr_t i, j, opr_sz = simd_oprsz(desc);
+    for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) {
+        uint64_t f = *(uint64_t *)(vn + i);
+        uint64_t b = *(uint64_t *)(vn + j);
+        *(uint64_t *)(vd + i) = hswap64(b);
+        *(uint64_t *)(vd + j) = hswap64(f);
+    }
+}
+
+void HELPER(sve_rev_s)(void *vd, void *vn, uint32_t desc)
+{
+    intptr_t i, j, opr_sz = simd_oprsz(desc);
+    for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) {
+        uint64_t f = *(uint64_t *)(vn + i);
+        uint64_t b = *(uint64_t *)(vn + j);
+        *(uint64_t *)(vd + i) = rol64(b, 32);
+        *(uint64_t *)(vd + j) = rol64(f, 32);
+    }
+}
+
+void HELPER(sve_rev_d)(void *vd, void *vn, uint32_t desc)
+{
+    intptr_t i, j, opr_sz = simd_oprsz(desc);
+    for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) {
+        uint64_t f = *(uint64_t *)(vn + i);
+        uint64_t b = *(uint64_t *)(vn + j);
+        *(uint64_t *)(vd + i) = b;
+        *(uint64_t *)(vd + j) = f;
+    }
+}
+
+#define DO_TBL(NAME, TYPE, H) \
+void HELPER(NAME)(void *vd, void *vn, void *vm, uint32_t desc) \
+{                                                              \
+    intptr_t i, opr_sz = simd_oprsz(desc);                     \
+    uintptr_t elem = opr_sz / sizeof(TYPE);                    \
+    TYPE *d = vd, *n = vn, *m = vm;                            \
+    ARMVectorReg tmp;                                          \
+    if (unlikely(vd == vn)) {                                  \
+        n = memcpy(&tmp, vn, opr_sz);                          \
+    }                                                          \
+    for (i = 0; i < elem; i++) {                               \
+        TYPE j = m[H(i)];                                      \
+        d[H(i)] = j < elem ? n[H(j)] : 0;                      \
+    }                                                          \
+}
+
+DO_TBL(sve_tbl_b, uint8_t, H1)
+DO_TBL(sve_tbl_h, uint16_t, H2)
+DO_TBL(sve_tbl_s, uint32_t, H4)
+DO_TBL(sve_tbl_d, uint64_t, )
+
+#undef TBL
+
+#define DO_UNPK(NAME, TYPED, TYPES, HD, HS) \
+void HELPER(NAME)(void *vd, void *vn, uint32_t desc)           \
+{                                                              \
+    intptr_t i, opr_sz = simd_oprsz(desc);                     \
+    TYPED *d = vd;                                             \
+    TYPES *n = vn;                                             \
+    ARMVectorReg tmp;                                          \
+    if (unlikely(vn - vd < opr_sz)) {                          \
+        n = memcpy(&tmp, n, opr_sz / 2);                       \
+    }                                                          \
+    for (i = 0; i < opr_sz / sizeof(TYPED); i++) {             \
+        d[HD(i)] = n[HS(i)];                                   \
+    }                                                          \
+}
+
+DO_UNPK(sve_sunpk_h, int16_t, int8_t, H2, H1)
+DO_UNPK(sve_sunpk_s, int32_t, int16_t, H4, H2)
+DO_UNPK(sve_sunpk_d, int64_t, int32_t, , H4)
+
+DO_UNPK(sve_uunpk_h, uint16_t, uint8_t, H2, H1)
+DO_UNPK(sve_uunpk_s, uint32_t, uint16_t, H4, H2)
+DO_UNPK(sve_uunpk_d, uint64_t, uint32_t, , H4)
+
+#undef DO_UNPK
diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c
index 07a5eac092..3724f6290c 100644
--- a/target/arm/translate-sve.c
+++ b/target/arm/translate-sve.c
@@ -1819,6 +1819,119 @@ static void trans_EXT(DisasContext *s, arg_EXT *a, 
uint32_t insn)
     tcg_gen_gvec_3_ool(d, n, m, vsz, vsz, n_ofs, gen_helper_sve_ext);
 }
 
+/*
+ *** SVE Permute - Unpredicated Group
+ */
+
+static void trans_DUP_s(DisasContext *s, arg_DUP_s *a, uint32_t insn)
+{
+    unsigned vsz = vec_full_reg_size(s);
+    tcg_gen_gvec_dup_i64(a->esz, vec_full_reg_offset(s, a->rd),
+                         vsz, vsz, cpu_reg_sp(s, a->rn));
+}
+
+static void trans_DUP_x(DisasContext *s, arg_DUP_x *a, uint32_t insn)
+{
+    unsigned vsz = vec_full_reg_size(s);
+    unsigned dofs = vec_full_reg_offset(s, a->rd);
+    unsigned esz, index;
+
+    if ((a->imm & 0x1f) == 0) {
+        unallocated_encoding(s);
+        return;
+    }
+    esz = ctz32(a->imm);
+    index = a->imm >> (esz + 1);
+
+    if ((index << esz) < vsz) {
+        unsigned nofs = vec_reg_offset(s, a->rn, index, esz);
+        tcg_gen_gvec_dup_mem(esz, dofs, nofs, vsz, vsz);
+    } else {
+        tcg_gen_gvec_dup64i(dofs, vsz, vsz, 0);
+    }
+}
+
+static void do_insr_i64(DisasContext *s, arg_rrr_esz *a, TCGv_i64 val)
+{
+    typedef void gen_insr(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv_i32);
+    static gen_insr * const fns[4] = {
+        gen_helper_sve_insr_b, gen_helper_sve_insr_h,
+        gen_helper_sve_insr_s, gen_helper_sve_insr_d,
+    };
+    unsigned vsz = vec_full_reg_size(s);
+    TCGv_i32 desc = tcg_const_i32(simd_desc(vsz, vsz, 0));
+    TCGv_ptr t_zd = tcg_temp_new_ptr();
+    TCGv_ptr t_zn = tcg_temp_new_ptr();
+
+    tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, a->rd));
+    tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, a->rn));
+
+    fns[a->esz](t_zd, t_zn, val, desc);
+
+    tcg_temp_free_ptr(t_zd);
+    tcg_temp_free_ptr(t_zn);
+    tcg_temp_free_i32(desc);
+}
+
+static void trans_INSR_f(DisasContext *s, arg_rrr_esz *a, uint32_t insn)
+{
+    TCGv_i64 t = tcg_temp_new_i64();
+    tcg_gen_ld_i64(t, cpu_env, vec_reg_offset(s, a->rm, 0, MO_64));
+    do_insr_i64(s, a, t);
+    tcg_temp_free_i64(t);
+}
+
+static void trans_INSR_r(DisasContext *s, arg_rrr_esz *a, uint32_t insn)
+{
+    do_insr_i64(s, a, cpu_reg(s, a->rm));
+}
+
+static void trans_REV_v(DisasContext *s, arg_rr_esz *a, uint32_t insn)
+{
+    static gen_helper_gvec_2 * const fns[4] = {
+        gen_helper_sve_rev_b, gen_helper_sve_rev_h,
+        gen_helper_sve_rev_s, gen_helper_sve_rev_d
+    };
+    unsigned vsz = vec_full_reg_size(s);
+
+    tcg_gen_gvec_2_ool(vec_full_reg_offset(s, a->rd),
+                       vec_full_reg_offset(s, a->rn),
+                       vsz, vsz, 0, fns[a->esz]);
+}
+
+static void trans_TBL(DisasContext *s, arg_rrr_esz *a, uint32_t insn)
+{
+    static gen_helper_gvec_3 * const fns[4] = {
+        gen_helper_sve_tbl_b, gen_helper_sve_tbl_h,
+        gen_helper_sve_tbl_s, gen_helper_sve_tbl_d
+    };
+    unsigned vsz = vec_full_reg_size(s);
+
+    tcg_gen_gvec_3_ool(vec_full_reg_offset(s, a->rd),
+                       vec_full_reg_offset(s, a->rn),
+                       vec_full_reg_offset(s, a->rm),
+                       vsz, vsz, 0, fns[a->esz]);
+}
+
+static void trans_UNPK(DisasContext *s, arg_UNPK *a, uint32_t insn)
+{
+    static gen_helper_gvec_2 * const fns[4][2] = {
+        { NULL, NULL },
+        { gen_helper_sve_sunpk_h, gen_helper_sve_uunpk_h },
+        { gen_helper_sve_sunpk_s, gen_helper_sve_uunpk_s },
+        { gen_helper_sve_sunpk_d, gen_helper_sve_uunpk_d },
+    };
+    unsigned vsz = vec_full_reg_size(s);
+
+    if (a->esz == 0) {
+        unallocated_encoding(s);
+        return;
+    }
+    tcg_gen_gvec_2_ool(vec_full_reg_offset(s, a->rd),
+                       vec_full_reg_offset(s, a->rn) + (a->h ? vsz / 2 : 0),
+                       vsz, vsz, 0, fns[a->esz][a->u]);
+}
+
 /*
  *** SVE Memory - 32-bit Gather and Unsized Contiguous Group
  */
diff --git a/target/arm/sve.decode b/target/arm/sve.decode
index 5e3a9839d4..8af47ad27b 100644
--- a/target/arm/sve.decode
+++ b/target/arm/sve.decode
@@ -24,6 +24,7 @@
 
 %imm4_16_p1    16:4 !function=plus1
 %imm6_22_5     22:1 5:5
+%imm7_22_16    22:2 16:5
 %imm8_16_10    16:5 10:3
 %imm9_16_10    16:s6 10:3
 %preg4_5       5:4
@@ -85,7 +86,9 @@
 @pd_pg_pn_pm_s ........ . s:1 .. rm:4 .. pg:4 . rn:4 . rd:4    &rprr_s
 
 # Three operand, vector element size
address@hidden  ........ esz:2 . rm:5  ... ...  rn:5 rd:5       &rrr_esz
address@hidden  ........ esz:2 . rm:5 ... ... rn:5 rd:5         &rrr_esz
address@hidden          ........ esz:2 ...... ...... rm:5 rd:5 \
+               &rrr_esz rn=%reg_movprfx
 
 # Three operand with "memory" size, aka immediate left shift
 @rd_rn_msz_rm  ........ ... rm:5 .... imm:2 rn:5 rd:5          &rrri
@@ -370,6 +373,30 @@ CPY_z_i            00000101 .. 01 .... 00 . ........ ..... 
  @rdn_pg4 imm=%sh8_i8s
 EXT            00000101 001 ..... 000 ... rm:5 rd:5 \
                &rrri rn=%reg_movprfx imm=%imm8_16_10
 
+### SVE Permute - Unpredicated Group
+
+# SVE broadcast general register
+DUP_s          00000101 .. 1 00000 001110 ..... .....          @rd_rn
+
+# SVE broadcast indexed element
+DUP_x          00000101 .. 1 ..... 001000 rn:5 rd:5 \
+               &rri imm=%imm7_22_16
+
+# SVE insert SIMD&FP scalar register
+INSR_f         00000101 .. 1 10100 001110 ..... .....          @rdn_rm
+
+# SVE insert general register
+INSR_r         00000101 .. 1 00100 001110 ..... .....          @rdn_rm
+
+# SVE reverse vector elements
+REV_v          00000101 .. 1 11000 001110 ..... .....          @rd_rn
+
+# SVE vector table lookup
+TBL            00000101 .. 1 ..... 001100 ..... .....          @rd_rn_rm
+
+# SVE unpack vector elements
+UNPK           00000101 esz:2 1100 u:1 h:1 001110 rn:5 rd:5
+
 ### SVE Predicate Logical Operations Group
 
 # SVE predicate logical operations
-- 
2.14.3