target/mips: Add loongson-ext lswc2 group of instructions (Part 2)

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:

  gslwlc1: similar to lwl but RT is FPR instead of GPR
  gslwrc1: similar to lwr but RT is FPR instead of GPR
  gsldlc1: similar to ldl but RT is FPR instead of GPR
  gsldrc1: similar to ldr but RT is FPR instead of GPR
  gsswlc1: similar to swl but RT is FPR instead of GPR
  gsswrc1: similar to swr but RT is FPR instead of GPR
  gssdlc1: similar to sdl but RT is FPR instead of GPR
  gssdrc1: similar to sdr but RT is FPR instead of GPR

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

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Message-Id: <1602831120-3377-4-git-send-email-chenhc@lemote.com>
[PMD: Reuse t1 on MIPS32, reintroduce t2/fp0]
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>

1 files changed, 180 insertions, 2 deletions

diff --git a/target/mips/translate.c b/target/mips/translate.c
index e83954d..b335645 100644
--- a/target/mips/translate.c
+++ b/target/mips/translate.c
@@ -471,6 +471,19 @@ enum {
     OPC_GSSHFS      = OPC_SWC2,
 };
 
+/* Loongson EXT shifted load/store opcodes */
+#define MASK_LOONGSON_GSSHFLS(op)         (MASK_OP_MAJOR(op) | (op & 0xc03f))
+enum {
+    OPC_GSLWLC1     = 0x4 | OPC_GSSHFL,
+    OPC_GSLWRC1     = 0x5 | OPC_GSSHFL,
+    OPC_GSLDLC1     = 0x6 | OPC_GSSHFL,
+    OPC_GSLDRC1     = 0x7 | OPC_GSSHFL,
+    OPC_GSSWLC1     = 0x4 | OPC_GSSHFS,
+    OPC_GSSWRC1     = 0x5 | OPC_GSSHFS,
+    OPC_GSSDLC1     = 0x6 | OPC_GSSHFS,
+    OPC_GSSDRC1     = 0x7 | OPC_GSSHFS,
+};
+
 /* BSHFL opcodes */
 #define MASK_BSHFL(op)              (MASK_SPECIAL3(op) | (op & (0x1F << 6)))
 
@@ -5924,12 +5937,13 @@ no_rd:
 static void gen_loongson_lswc2(DisasContext *ctx, int rt,
                                int rs, int rd)
 {
-    TCGv t0;
+    TCGv t0, t1, t2;
+    TCGv_i32 fp0;
 #if defined(TARGET_MIPS64)
-    TCGv t1;
     int lsq_rt1 = ctx->opcode & 0x1f;
     int lsq_offset = sextract32(ctx->opcode, 6, 9) << 4;
 #endif
+    int shf_offset = sextract32(ctx->opcode, 6, 8);
 
     t0 = tcg_temp_new();
 
@@ -5986,6 +6000,170 @@ static void gen_loongson_lswc2(DisasContext *ctx, int rt,
         tcg_temp_free(t1);
         break;
 #endif
+    case OPC_GSSHFL:
+        switch (MASK_LOONGSON_GSSHFLS(ctx->opcode)) {
+        case OPC_GSLWLC1:
+            check_cp1_enabled(ctx);
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            t1 = tcg_temp_new();
+            tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
+            tcg_gen_andi_tl(t1, t0, 3);
+#ifndef TARGET_WORDS_BIGENDIAN
+            tcg_gen_xori_tl(t1, t1, 3);
+#endif
+            tcg_gen_shli_tl(t1, t1, 3);
+            tcg_gen_andi_tl(t0, t0, ~3);
+            tcg_gen_qemu_ld_tl(t0, t0, ctx->mem_idx, MO_TEUL);
+            tcg_gen_shl_tl(t0, t0, t1);
+            t2 = tcg_const_tl(-1);
+            tcg_gen_shl_tl(t2, t2, t1);
+            fp0 = tcg_temp_new_i32();
+            gen_load_fpr32(ctx, fp0, rt);
+            tcg_gen_ext_i32_tl(t1, fp0);
+            tcg_gen_andc_tl(t1, t1, t2);
+            tcg_temp_free(t2);
+            tcg_gen_or_tl(t0, t0, t1);
+            tcg_temp_free(t1);
+#if defined(TARGET_MIPS64)
+            tcg_gen_extrl_i64_i32(fp0, t0);
+#else
+            tcg_gen_ext32s_tl(fp0, t0);
+#endif
+            gen_store_fpr32(ctx, fp0, rt);
+            tcg_temp_free_i32(fp0);
+            break;
+        case OPC_GSLWRC1:
+            check_cp1_enabled(ctx);
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            t1 = tcg_temp_new();
+            tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
+            tcg_gen_andi_tl(t1, t0, 3);
+#ifdef TARGET_WORDS_BIGENDIAN
+            tcg_gen_xori_tl(t1, t1, 3);
+#endif
+            tcg_gen_shli_tl(t1, t1, 3);
+            tcg_gen_andi_tl(t0, t0, ~3);
+            tcg_gen_qemu_ld_tl(t0, t0, ctx->mem_idx, MO_TEUL);
+            tcg_gen_shr_tl(t0, t0, t1);
+            tcg_gen_xori_tl(t1, t1, 31);
+            t2 = tcg_const_tl(0xfffffffeull);
+            tcg_gen_shl_tl(t2, t2, t1);
+            fp0 = tcg_temp_new_i32();
+            gen_load_fpr32(ctx, fp0, rt);
+            tcg_gen_ext_i32_tl(t1, fp0);
+            tcg_gen_and_tl(t1, t1, t2);
+            tcg_temp_free(t2);
+            tcg_gen_or_tl(t0, t0, t1);
+            tcg_temp_free(t1);
+#if defined(TARGET_MIPS64)
+            tcg_gen_extrl_i64_i32(fp0, t0);
+#else
+            tcg_gen_ext32s_tl(fp0, t0);
+#endif
+            gen_store_fpr32(ctx, fp0, rt);
+            tcg_temp_free_i32(fp0);
+            break;
+#if defined(TARGET_MIPS64)
+        case OPC_GSLDLC1:
+            check_cp1_enabled(ctx);
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            t1 = tcg_temp_new();
+            tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
+            tcg_gen_andi_tl(t1, t0, 7);
+#ifndef TARGET_WORDS_BIGENDIAN
+            tcg_gen_xori_tl(t1, t1, 7);
+#endif
+            tcg_gen_shli_tl(t1, t1, 3);
+            tcg_gen_andi_tl(t0, t0, ~7);
+            tcg_gen_qemu_ld_tl(t0, t0, ctx->mem_idx, MO_TEQ);
+            tcg_gen_shl_tl(t0, t0, t1);
+            t2 = tcg_const_tl(-1);
+            tcg_gen_shl_tl(t2, t2, t1);
+            gen_load_fpr64(ctx, t1, rt);
+            tcg_gen_andc_tl(t1, t1, t2);
+            tcg_temp_free(t2);
+            tcg_gen_or_tl(t0, t0, t1);
+            tcg_temp_free(t1);
+            gen_store_fpr64(ctx, t0, rt);
+            break;
+        case OPC_GSLDRC1:
+            check_cp1_enabled(ctx);
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            t1 = tcg_temp_new();
+            tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
+            tcg_gen_andi_tl(t1, t0, 7);
+#ifdef TARGET_WORDS_BIGENDIAN
+            tcg_gen_xori_tl(t1, t1, 7);
+#endif
+            tcg_gen_shli_tl(t1, t1, 3);
+            tcg_gen_andi_tl(t0, t0, ~7);
+            tcg_gen_qemu_ld_tl(t0, t0, ctx->mem_idx, MO_TEQ);
+            tcg_gen_shr_tl(t0, t0, t1);
+            tcg_gen_xori_tl(t1, t1, 63);
+            t2 = tcg_const_tl(0xfffffffffffffffeull);
+            tcg_gen_shl_tl(t2, t2, t1);
+            gen_load_fpr64(ctx, t1, rt);
+            tcg_gen_and_tl(t1, t1, t2);
+            tcg_temp_free(t2);
+            tcg_gen_or_tl(t0, t0, t1);
+            tcg_temp_free(t1);
+            gen_store_fpr64(ctx, t0, rt);
+            break;
+#endif
+        default:
+            MIPS_INVAL("loongson_gsshfl");
+            generate_exception_end(ctx, EXCP_RI);
+            break;
+        }
+        break;
+    case OPC_GSSHFS:
+        switch (MASK_LOONGSON_GSSHFLS(ctx->opcode)) {
+        case OPC_GSSWLC1:
+            check_cp1_enabled(ctx);
+            t1 = tcg_temp_new();
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            fp0 = tcg_temp_new_i32();
+            gen_load_fpr32(ctx, fp0, rt);
+            tcg_gen_ext_i32_tl(t1, fp0);
+            gen_helper_0e2i(swl, t1, t0, ctx->mem_idx);
+            tcg_temp_free_i32(fp0);
+            tcg_temp_free(t1);
+            break;
+        case OPC_GSSWRC1:
+            check_cp1_enabled(ctx);
+            t1 = tcg_temp_new();
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            fp0 = tcg_temp_new_i32();
+            gen_load_fpr32(ctx, fp0, rt);
+            tcg_gen_ext_i32_tl(t1, fp0);
+            gen_helper_0e2i(swr, t1, t0, ctx->mem_idx);
+            tcg_temp_free_i32(fp0);
+            tcg_temp_free(t1);
+            break;
+#if defined(TARGET_MIPS64)
+        case OPC_GSSDLC1:
+            check_cp1_enabled(ctx);
+            t1 = tcg_temp_new();
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            gen_load_fpr64(ctx, t1, rt);
+            gen_helper_0e2i(sdl, t1, t0, ctx->mem_idx);
+            tcg_temp_free(t1);
+            break;
+        case OPC_GSSDRC1:
+            check_cp1_enabled(ctx);
+            t1 = tcg_temp_new();
+            gen_base_offset_addr(ctx, t0, rs, shf_offset);
+            gen_load_fpr64(ctx, t1, rt);
+            gen_helper_0e2i(sdr, t1, t0, ctx->mem_idx);
+            tcg_temp_free(t1);
+            break;
+#endif
+        default:
+            MIPS_INVAL("loongson_gsshfs");
+            generate_exception_end(ctx, EXCP_RI);
+            break;
+        }
+        break;
     default:
         MIPS_INVAL("loongson_gslsq");
         generate_exception_end(ctx, EXCP_RI);