[PATCH V10 3/8] target/mips: Add loongson-ext lswc2 group of instructions (Part 1)

[Huacai Chen <zltjiangshi@gmail.com>]

From: Jiaxun Yang <jiaxun.yang@flygoat.com>

LWC2 & SWC2 have been rewritten by Loongson EXT vendor ASE
as "load/store quad word" and "shifted load/store" groups of
instructions.

This patch add implementation of these instructions:
gslq: load 16 bytes to GPR
gssq: store 16 bytes from GPR
gslqc1: load 16 bytes to FPR
gssqc1: store 16 bytes from FPR

Details of Loongson-EXT is here:
https://github.com/FlyGoat/loongson-insn/blob/master/loongson-ext.md

Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
---
 target/mips/translate.c | 87 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 87 insertions(+)

diff --git a/target/mips/translate.c b/target/mips/translate.c
index 398edf7289..cb0adde310 100644
--- a/target/mips/translate.c
+++ b/target/mips/translate.c
@@ -460,6 +460,17 @@ enum {
     R6_OPC_SCD         = 0x27 | OPC_SPECIAL3,
 };
 
+/* Loongson EXT load/store quad word opcodes */
+#define MASK_LOONGSON_GSLSQ(op)           (MASK_OP_MAJOR(op) | (op & 0x8020))
+enum {
+    OPC_GSLQ        = 0x0020 | OPC_LWC2,
+    OPC_GSLQC1      = 0x8020 | OPC_LWC2,
+    OPC_GSSHFL      = OPC_LWC2,
+    OPC_GSSQ        = 0x0020 | OPC_SWC2,
+    OPC_GSSQC1      = 0x8020 | OPC_SWC2,
+    OPC_GSSHFS      = OPC_SWC2,
+};
+
 /* BSHFL opcodes */
 #define MASK_BSHFL(op)              (MASK_SPECIAL3(op) | (op & (0x1F << 6)))
 
@@ -5910,6 +5921,80 @@ no_rd:
     tcg_temp_free_i64(t1);
 }
 
+static void gen_loongson_lswc2(DisasContext *ctx, int rt,
+                                int rs, int rd)
+{
+    TCGv t0, t1, t2;
+    TCGv_i32 fp0;
+#if defined(TARGET_MIPS64)
+    int lsq_rt1 = ctx->opcode & 0x1f;
+    int lsq_offset = ((int)((ctx->opcode >> 6) & 0x1ff) << 23) >> 19;
+#endif
+    int shf_offset = (int8_t)(ctx->opcode >> 6);
+
+    t0 = tcg_temp_new();
+
+    switch (MASK_LOONGSON_GSLSQ(ctx->opcode)) {
+#if defined(TARGET_MIPS64)
+    case OPC_GSLQ:
+        t1 = tcg_temp_new();
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset);
+        tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset + 8);
+        tcg_gen_qemu_ld_tl(t0, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        gen_store_gpr(t1, rt);
+        gen_store_gpr(t0, lsq_rt1);
+        tcg_temp_free(t1);
+        break;
+    case OPC_GSLQC1:
+        check_cp1_enabled(ctx);
+        t1 = tcg_temp_new();
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset);
+        tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset + 8);
+        tcg_gen_qemu_ld_tl(t0, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        gen_store_fpr64(ctx, t1, rt);
+        gen_store_fpr64(ctx, t0, lsq_rt1);
+        tcg_temp_free(t1);
+        break;
+    case OPC_GSSQ:
+        t1 = tcg_temp_new();
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset);
+        gen_load_gpr(t1, rt);
+        tcg_gen_qemu_st_tl(t1, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset + 8);
+        gen_load_gpr(t1, lsq_rt1);
+        tcg_gen_qemu_st_tl(t1, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        tcg_temp_free(t1);
+        break;
+    case OPC_GSSQC1:
+        check_cp1_enabled(ctx);
+        t1 = tcg_temp_new();
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset);
+        gen_load_fpr64(ctx, t1, rt);
+        tcg_gen_qemu_st_tl(t1, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        gen_base_offset_addr(ctx, t0, rs, lsq_offset + 8);
+        gen_load_fpr64(ctx, t1, lsq_rt1);
+        tcg_gen_qemu_st_tl(t1, t0, ctx->mem_idx, MO_TEQ |
+                            ctx->default_tcg_memop_mask);
+        tcg_temp_free(t1);
+        break;
+#endif
+    default:
+        MIPS_INVAL("loongson_gslsq");
+        generate_exception_end(ctx, EXCP_RI);
+        break;
+    }
+    tcg_temp_free(t0);
+}
+
 /* Traps */
 static void gen_trap(DisasContext *ctx, uint32_t opc,
                      int rs, int rt, int16_t imm)
@@ -30774,6 +30859,8 @@ static void decode_opc(CPUMIPSState *env, DisasContext *ctx)
             /* OPC_BC, OPC_BALC */
             gen_compute_compact_branch(ctx, op, 0, 0,
                                        sextract32(ctx->opcode << 2, 0, 28));
+        } else if (ctx->insn_flags & ASE_LEXT) {
+            gen_loongson_lswc2(ctx, rt, rs, rd);
         } else {
             /* OPC_LWC2, OPC_SWC2 */
             /* COP2: Not implemented. */
-- 
2.17.2