Fix fast jit int rem_s and const shl issues (#1213)

int rem_s -1 should return 0
int32 lhs << int32 rhs may cause sanitizer check failure
fix codegen I8TOI64, I16TOI64, I64TOI8, I64TOI16
implement codegen neg operations

Author: wenyongh

core/iwasm/fast-jit/cg/x86-64/jit_codegen_x86_64.cpp

@@ -2081,7 +2081,9 @@ convert_r_f64_to_r_u32(x86::Assembler &a, int32 reg_no_dst, int32 reg_no_src)
 static bool
 neg_imm_to_r_i32(x86::Assembler &a, int32 reg_no, int32 data)
 {
-    return false;
+    Imm imm(-data);
+    a.mov(regs_i32[reg_no], imm);
+    return true;
 }
 
 /**
@@ -2096,7 +2098,9 @@ neg_imm_to_r_i32(x86::Assembler &a, int32 reg_no, int32 data)
 static bool
 neg_r_to_r_i32(x86::Assembler &a, int32 reg_no_dst, int32 reg_no_src)
 {
-    return false;
+    mov_r_to_r_i32(a, reg_no_dst, reg_no_src);
+    a.neg(regs_i32[reg_no_dst]);
+    return true;
 }
 
 /**
@@ -2111,7 +2115,9 @@ neg_r_to_r_i32(x86::Assembler &a, int32 reg_no_dst, int32 reg_no_src)
 static bool
 neg_imm_to_r_i64(x86::Assembler &a, int32 reg_no, int64 data)
 {
-    return false;
+    Imm imm(-data);
+    a.mov(regs_i64[reg_no], imm);
+    return true;
 }
 
 /**
@@ -2126,7 +2132,9 @@ neg_imm_to_r_i64(x86::Assembler &a, int32 reg_no, int64 data)
 static bool
 neg_r_to_r_i64(x86::Assembler &a, int32 reg_no_dst, int32 reg_no_src)
 {
-    return false;
+    mov_r_to_r_i64(a, reg_no_dst, reg_no_src);
+    a.neg(regs_i64[reg_no_dst]);
+    return true;
 }
 
 /**
@@ -5923,7 +5931,7 @@ jit_codegen_gen_native(JitCompContext *cc)
 
                 case JIT_OP_I8TOI64:
                     LOAD_2ARGS();
-                    CONVERT_R_R(I64, I64, i64, i8, int8);
+                    CONVERT_R_R(I64, I32, i64, i8, int8);
                     break;
 
                 case JIT_OP_I16TOI32:
@@ -5933,7 +5941,7 @@ jit_codegen_gen_native(JitCompContext *cc)
 
                 case JIT_OP_I16TOI64:
                     LOAD_2ARGS();
-                    CONVERT_R_R(I64, I64, i64, i16, int16);
+                    CONVERT_R_R(I64, I32, i64, i16, int16);
                     break;
 
                 case JIT_OP_I32TOI8:
@@ -5988,12 +5996,12 @@ jit_codegen_gen_native(JitCompContext *cc)
 
                 case JIT_OP_I64TOI8:
                     LOAD_2ARGS();
-                    CONVERT_R_R(I64, I64, i8, i64, int64);
+                    CONVERT_R_R(I32, I64, i8, i64, int64);
                     break;
 
                 case JIT_OP_I64TOI16:
                     LOAD_2ARGS();
-                    CONVERT_R_R(I64, I64, i16, i64, int64);
+                    CONVERT_R_R(I32, I64, i16, i64, int64);
                     break;
 
                 case JIT_OP_I64TOI32:

core/iwasm/fast-jit/fe/jit_emit_conversion.c

@@ -185,7 +185,7 @@ jit_compile_op_i64_extend_i64(JitCompContext *cc, int8 bitwidth)
 
     POP_I64(value);
 
-    tmp = jit_cc_new_reg_I64(cc);
+    tmp = jit_cc_new_reg_I32(cc);
     res = jit_cc_new_reg_I64(cc);
 
     switch (bitwidth) {

core/iwasm/fast-jit/fe/jit_emit_numberic.c

@@ -701,7 +701,7 @@ compile_int_div(JitCompContext *cc, IntArithmetic arith_op, bool is_i32,
 
         switch (arith_op) {
             case INT_DIV_S:
-            case INT_REM_S:
+            {
                 /* Check integer overflow */
                 GEN_INSN(CMP, cc->cmp_reg, left,
                          is_i32 ? NEW_CONST(I32, INT32_MIN)
@@ -723,11 +723,27 @@ compile_int_div(JitCompContext *cc, IntArithmetic arith_op, bool is_i32,
                 /* Build default div and rem */
                 return compile_int_div_no_check(cc, arith_op, is_i32, left,
                                                 right, res);
-                return true;
+            }
+            case INT_REM_S:
+            {
+                GEN_INSN(CMP, cc->cmp_reg, right,
+                         is_i32 ? NEW_CONST(I32, -1) : NEW_CONST(I64, -1LL));
+                if (is_i32)
+                    GEN_INSN(SELECTEQ, left, cc->cmp_reg, NEW_CONST(I32, 0),
+                             left);
+                else
+                    GEN_INSN(SELECTEQ, left, cc->cmp_reg, NEW_CONST(I64, 0),
+                             left);
+                /* Build default div and rem */
+                return compile_int_div_no_check(cc, arith_op, is_i32, left,
+                                                right, res);
+            }
             default:
+            {
                 /* Build default div and rem */
                 return compile_int_div_no_check(cc, arith_op, is_i32, left,
                                                 right, res);
+            }
         }
     }
 
@@ -990,7 +1006,17 @@ DEF_UNI_INT_CONST_OPS(shru)
     return 0;
 }
 
-DEF_BI_INT_CONST_OPS(shl, <<)
+static int32
+do_i32_const_shl(int32 lhs, int32 rhs)
+{
+    return (int32)((uint32)lhs << (uint32)rhs);
+}
+
+static int64
+do_i64_const_shl(int64 lhs, int64 rhs)
+{
+    return (int32)((uint64)lhs << (uint64)rhs);
+}
 
 DEF_BI_INT_CONST_OPS(shrs, >>)
 
@@ -1505,7 +1531,6 @@ compile_op_float_arithmetic(JitCompContext *cc, FloatArithmetic arith_op,
         }
         case FLOAT_DIV:
         {
-            /*TODO: add divided by zero interception */
             GEN_INSN(DIV_S, res, lhs, rhs);
             break;
         }