target/loongarch: Implement vld vst

This patch includes:
- VLD[X], VST[X];
- VLDREPL.{B/H/W/D};
- VSTELM.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-42-gaosong@loongson.cn>

4 files changed, 239 insertions, 0 deletions

diff --git a/target/loongarch/disas.c b/target/loongarch/disas.c
index 0b62bbb..8627908 100644
--- a/target/loongarch/disas.c
+++ b/target/loongarch/disas.c
@@ -21,11 +21,21 @@ static inline int plus_1(DisasContext *ctx, int x)
     return x + 1;
 }
 
+static inline int shl_1(DisasContext *ctx, int x)
+{
+    return x << 1;
+}
+
 static inline int shl_2(DisasContext *ctx, int x)
 {
     return x << 2;
 }
 
+static inline int shl_3(DisasContext *ctx, int x)
+{
+    return x << 3;
+}
+
 #define CSR_NAME(REG) \
     [LOONGARCH_CSR_##REG] = (#REG)
 
@@ -823,6 +833,11 @@ static void output_vr_i(DisasContext *ctx, arg_vr_i *a, const char *mnemonic)
     output(ctx, mnemonic, "v%d, r%d, 0x%x", a->vd, a->rj, a->imm);
 }
 
+static void output_vr_ii(DisasContext *ctx, arg_vr_ii *a, const char *mnemonic)
+{
+    output(ctx, mnemonic, "v%d, r%d, 0x%x, 0x%x", a->vd, a->rj, a->imm, a->imm2);
+}
+
 static void output_rv_i(DisasContext *ctx, arg_rv_i *a, const char *mnemonic)
 {
     output(ctx, mnemonic, "r%d, v%d, 0x%x", a->rd, a->vj,  a->imm);
@@ -838,6 +853,11 @@ static void output_vvr(DisasContext *ctx, arg_vvr *a, const char *mnemonic)
     output(ctx, mnemonic, "v%d, v%d, r%d", a->vd, a->vj, a->rk);
 }
 
+static void output_vrr(DisasContext *ctx, arg_vrr *a, const char *mnemonic)
+{
+    output(ctx, mnemonic, "v%d, r%d, r%d", a->vd, a->rj, a->rk);
+}
+
 INSN_LSX(vadd_b,           vvv)
 INSN_LSX(vadd_h,           vvv)
 INSN_LSX(vadd_w,           vvv)
@@ -1654,3 +1674,17 @@ INSN_LSX(vextrins_d,       vv_i)
 INSN_LSX(vextrins_w,       vv_i)
 INSN_LSX(vextrins_h,       vv_i)
 INSN_LSX(vextrins_b,       vv_i)
+
+INSN_LSX(vld,              vr_i)
+INSN_LSX(vst,              vr_i)
+INSN_LSX(vldx,             vrr)
+INSN_LSX(vstx,             vrr)
+
+INSN_LSX(vldrepl_d,        vr_i)
+INSN_LSX(vldrepl_w,        vr_i)
+INSN_LSX(vldrepl_h,        vr_i)
+INSN_LSX(vldrepl_b,        vr_i)
+INSN_LSX(vstelm_d,         vr_ii)
+INSN_LSX(vstelm_w,         vr_ii)
+INSN_LSX(vstelm_h,         vr_ii)
+INSN_LSX(vstelm_b,         vr_ii)
diff --git a/target/loongarch/insn_trans/trans_lsx.c.inc b/target/loongarch/insn_trans/trans_lsx.c.inc
index e1eee6b..86dfd2b 100644
--- a/target/loongarch/insn_trans/trans_lsx.c.inc
+++ b/target/loongarch/insn_trans/trans_lsx.c.inc
@@ -4102,3 +4102,162 @@ TRANS(vextrins_b, gen_vv_i, gen_helper_vextrins_b)
 TRANS(vextrins_h, gen_vv_i, gen_helper_vextrins_h)
 TRANS(vextrins_w, gen_vv_i, gen_helper_vextrins_w)
 TRANS(vextrins_d, gen_vv_i, gen_helper_vextrins_d)
+
+static bool trans_vld(DisasContext *ctx, arg_vr_i *a)
+{
+    TCGv addr, temp;
+    TCGv_i64 rl, rh;
+    TCGv_i128 val;
+
+    CHECK_SXE;
+
+    addr = gpr_src(ctx, a->rj, EXT_NONE);
+    val = tcg_temp_new_i128();
+    rl = tcg_temp_new_i64();
+    rh = tcg_temp_new_i64();
+
+    if (a->imm) {
+        temp = tcg_temp_new();
+        tcg_gen_addi_tl(temp, addr, a->imm);
+        addr = temp;
+    }
+
+    tcg_gen_qemu_ld_i128(val, addr, ctx->mem_idx, MO_128 | MO_TE);
+    tcg_gen_extr_i128_i64(rl, rh, val);
+    set_vreg64(rh, a->vd, 1);
+    set_vreg64(rl, a->vd, 0);
+
+    return true;
+}
+
+static bool trans_vst(DisasContext *ctx, arg_vr_i *a)
+{
+    TCGv addr, temp;
+    TCGv_i128 val;
+    TCGv_i64 ah, al;
+
+    CHECK_SXE;
+
+    addr = gpr_src(ctx, a->rj, EXT_NONE);
+    val = tcg_temp_new_i128();
+    ah = tcg_temp_new_i64();
+    al = tcg_temp_new_i64();
+
+    if (a->imm) {
+        temp = tcg_temp_new();
+        tcg_gen_addi_tl(temp, addr, a->imm);
+        addr = temp;
+    }
+
+    get_vreg64(ah, a->vd, 1);
+    get_vreg64(al, a->vd, 0);
+    tcg_gen_concat_i64_i128(val, al, ah);
+    tcg_gen_qemu_st_i128(val, addr, ctx->mem_idx, MO_128 | MO_TE);
+
+    return true;
+}
+
+static bool trans_vldx(DisasContext *ctx, arg_vrr *a)
+{
+    TCGv addr, src1, src2;
+    TCGv_i64 rl, rh;
+    TCGv_i128 val;
+
+    CHECK_SXE;
+
+    addr = tcg_temp_new();
+    src1 = gpr_src(ctx, a->rj, EXT_NONE);
+    src2 = gpr_src(ctx, a->rk, EXT_NONE);
+    val = tcg_temp_new_i128();
+    rl = tcg_temp_new_i64();
+    rh = tcg_temp_new_i64();
+
+    tcg_gen_add_tl(addr, src1, src2);
+    tcg_gen_qemu_ld_i128(val, addr, ctx->mem_idx, MO_128 | MO_TE);
+    tcg_gen_extr_i128_i64(rl, rh, val);
+    set_vreg64(rh, a->vd, 1);
+    set_vreg64(rl, a->vd, 0);
+
+    return true;
+}
+
+static bool trans_vstx(DisasContext *ctx, arg_vrr *a)
+{
+    TCGv addr, src1, src2;
+    TCGv_i64 ah, al;
+    TCGv_i128 val;
+
+    CHECK_SXE;
+
+    addr = tcg_temp_new();
+    src1 = gpr_src(ctx, a->rj, EXT_NONE);
+    src2 = gpr_src(ctx, a->rk, EXT_NONE);
+    val = tcg_temp_new_i128();
+    ah = tcg_temp_new_i64();
+    al = tcg_temp_new_i64();
+
+    tcg_gen_add_tl(addr, src1, src2);
+    get_vreg64(ah, a->vd, 1);
+    get_vreg64(al, a->vd, 0);
+    tcg_gen_concat_i64_i128(val, al, ah);
+    tcg_gen_qemu_st_i128(val, addr, ctx->mem_idx, MO_128 | MO_TE);
+
+    return true;
+}
+
+#define VLDREPL(NAME, MO)                                                 \
+static bool trans_## NAME (DisasContext *ctx, arg_vr_i *a)                \
+{                                                                         \
+    TCGv addr, temp;                                                      \
+    TCGv_i64 val;                                                         \
+                                                                          \
+    CHECK_SXE;                                                            \
+                                                                          \
+    addr = gpr_src(ctx, a->rj, EXT_NONE);                                 \
+    val = tcg_temp_new_i64();                                             \
+                                                                          \
+    if (a->imm) {                                                         \
+        temp = tcg_temp_new();                                            \
+        tcg_gen_addi_tl(temp, addr, a->imm);                              \
+        addr = temp;                                                      \
+    }                                                                     \
+                                                                          \
+    tcg_gen_qemu_ld_i64(val, addr, ctx->mem_idx, MO);                     \
+    tcg_gen_gvec_dup_i64(MO, vec_full_offset(a->vd), 16, ctx->vl/8, val); \
+                                                                          \
+    return true;                                                          \
+}
+
+VLDREPL(vldrepl_b, MO_8)
+VLDREPL(vldrepl_h, MO_16)
+VLDREPL(vldrepl_w, MO_32)
+VLDREPL(vldrepl_d, MO_64)
+
+#define VSTELM(NAME, MO, E)                                                  \
+static bool trans_## NAME (DisasContext *ctx, arg_vr_ii *a)                  \
+{                                                                            \
+    TCGv addr, temp;                                                         \
+    TCGv_i64 val;                                                            \
+                                                                             \
+    CHECK_SXE;                                                               \
+                                                                             \
+    addr = gpr_src(ctx, a->rj, EXT_NONE);                                    \
+    val = tcg_temp_new_i64();                                                \
+                                                                             \
+    if (a->imm) {                                                            \
+        temp = tcg_temp_new();                                               \
+        tcg_gen_addi_tl(temp, addr, a->imm);                                 \
+        addr = temp;                                                         \
+    }                                                                        \
+                                                                             \
+    tcg_gen_ld_i64(val, cpu_env,                                             \
+                   offsetof(CPULoongArchState, fpr[a->vd].vreg.E(a->imm2))); \
+    tcg_gen_qemu_st_i64(val, addr, ctx->mem_idx, MO);                        \
+                                                                             \
+    return true;                                                             \
+}
+
+VSTELM(vstelm_b, MO_8, B)
+VSTELM(vstelm_h, MO_16, H)
+VSTELM(vstelm_w, MO_32, W)
+VSTELM(vstelm_d, MO_64, D)
diff --git a/target/loongarch/insns.decode b/target/loongarch/insns.decode
index 0263bce..ea6eedb 100644
--- a/target/loongarch/insns.decode
+++ b/target/loongarch/insns.decode
@@ -487,6 +487,17 @@ idle             0000 01100100 10001 ...............      @i15
 dbcl             0000 00000010 10101 ...............      @i15
 
 #
+# LSX Fields
+#
+
+%i9s3     10:s9       !function=shl_3
+%i10s2    10:s10      !function=shl_2
+%i11s1    10:s11      !function=shl_1
+%i8s3     10:s8       !function=shl_3
+%i8s2     10:s8       !function=shl_2
+%i8s1     10:s8       !function=shl_1
+
+#
 # LSX Argument sets
 #
 
@@ -500,6 +511,8 @@ dbcl             0000 00000010 10101 ...............      @i15
 &rv_i         rd vj imm
 &vr           vd rj
 &vvr          vd vj rk
+&vrr          vd rj rk
+&vr_ii        vd rj imm imm2
 
 #
 # LSX Formats
@@ -528,6 +541,15 @@ dbcl             0000 00000010 10101 ...............      @i15
 @rv_ui1      .... ........ ..... .... imm:1 vj:5 rd:5    &rv_i
 @vr               .... ........ ..... ..... rj:5 vd:5    &vr
 @vvr               .... ........ ..... rk:5 vj:5 vd:5    &vvr
+@vr_i9            .... ........ . ......... rj:5 vd:5    &vr_i imm=%i9s3
+@vr_i10            .... ........ .......... rj:5 vd:5    &vr_i imm=%i10s2
+@vr_i11            .... ....... ........... rj:5 vd:5    &vr_i imm=%i11s1
+@vr_i12                 .... ...... imm:s12 rj:5 vd:5    &vr_i
+@vr_i8i1    .... ........ . imm2:1 ........ rj:5 vd:5    &vr_ii imm=%i8s3
+@vr_i8i2      .... ........ imm2:2 ........ rj:5 vd:5    &vr_ii imm=%i8s2
+@vr_i8i3       .... ....... imm2:3 ........ rj:5 vd:5    &vr_ii imm=%i8s1
+@vr_i8i4          .... ...... imm2:4 imm:s8 rj:5 vd:5    &vr_ii
+@vrr               .... ........ ..... rk:5 rj:5 vd:5    &vrr
 
 vadd_b           0111 00000000 10100 ..... ..... .....    @vvv
 vadd_h           0111 00000000 10101 ..... ..... .....    @vvv
@@ -1256,3 +1278,17 @@ vextrins_d       0111 00111000 00 ........ ..... .....    @vv_ui8
 vextrins_w       0111 00111000 01 ........ ..... .....    @vv_ui8
 vextrins_h       0111 00111000 10 ........ ..... .....    @vv_ui8
 vextrins_b       0111 00111000 11 ........ ..... .....    @vv_ui8
+
+vld              0010 110000 ............ ..... .....     @vr_i12
+vst              0010 110001 ............ ..... .....     @vr_i12
+vldx             0011 10000100 00000 ..... ..... .....    @vrr
+vstx             0011 10000100 01000 ..... ..... .....    @vrr
+
+vldrepl_d        0011 00000001 0 ......... ..... .....    @vr_i9
+vldrepl_w        0011 00000010 .......... ..... .....     @vr_i10
+vldrepl_h        0011 0000010 ........... ..... .....     @vr_i11
+vldrepl_b        0011 000010 ............ ..... .....     @vr_i12
+vstelm_d         0011 00010001 0 . ........ ..... .....   @vr_i8i1
+vstelm_w         0011 00010010 .. ........ ..... .....    @vr_i8i2
+vstelm_h         0011 0001010 ... ........ ..... .....    @vr_i8i3
+vstelm_b         0011 000110 .... ........ ..... .....    @vr_i8i4
diff --git a/target/loongarch/translate.c b/target/loongarch/translate.c
index 862847a..c04ed75 100644
--- a/target/loongarch/translate.c
+++ b/target/loongarch/translate.c
@@ -53,11 +53,21 @@ static inline int plus_1(DisasContext *ctx, int x)
     return x + 1;
 }
 
+static inline int shl_1(DisasContext *ctx, int x)
+{
+    return x << 1;
+}
+
 static inline int shl_2(DisasContext *ctx, int x)
 {
     return x << 2;
 }
 
+static inline int shl_3(DisasContext *ctx, int x)
+{
+    return x << 3;
+}
+
 /*
  * LoongArch the upper 32 bits are undefined ("can be any value").
  * QEMU chooses to nanbox, because it is most likely to show guest bugs early.