fix YAML template

.gitignore

@@ -31,3 +31,4 @@ language.settings.xml
 /*.out
 /dump.json
 /src-gen/
+/*.yaml

gen_input/TGC_B.core_desc

@@ -0,0 +1,14 @@
+import "CoreDSL-Instruction-Set-Description/RV32I.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVM.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVC.core_desc"
+
+Core TGC_B provides RV32I {
+	architectural_state {
+        XLEN=32;
+        // definitions for the architecture wrapper
+        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
+        unsigned MISA_VAL = 0b01000000000000000000000100000000;
+        unsigned MARCHID_VAL = 0x80000002;
+	}
+}
+

gen_input/TGC_C.core_desc

@@ -0,0 +1,13 @@
+import "CoreDSL-Instruction-Set-Description/RV32I.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVM.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVC.core_desc"
+
+Core TGC_C provides RV32I, RV32M, RV32IC {
+    architectural_state {
+        XLEN=32;
+        // definitions for the architecture wrapper
+        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
+        unsigned MISA_VAL = 0b01000000000000000001000100000100;
+        unsigned MARCHID_VAL = 0x80000003;
+    }
+}

gen_input/TGC_D.core_desc

@@ -0,0 +1,13 @@
+import "CoreDSL-Instruction-Set-Description/RV32I.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVM.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVC.core_desc"
+
+Core TGC_D provides RV32I, RV32M, RV32IC {
+    architectural_state {
+        XLEN=32;
+        // definitions for the architecture wrapper
+        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
+        unsigned MISA_VAL = 0b01000000000000000001000100000100;
+        unsigned MARCHID_VAL = 0x80000004;
+    }
+}

gen_input/TGC_D_XRB_MAC.core_desc

@@ -0,0 +1,73 @@
+import "CoreDSL-Instruction-Set-Description/RISCVBase.core_desc"
+import "CoreDSL-Instruction-Set-Description/RV32I.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVM.core_desc"
+import "CoreDSL-Instruction-Set-Description/RVC.core_desc"
+
+InstructionSet X_RB_MAC extends RISCVBase {
+  architectural_state {
+    register unsigned<64> ACC;
+  }
+
+  instructions {
+    RESET_ACC { // v-- funct7       v-- funct3
+      encoding: 7'd0 :: 10'b0 :: 3'd0 :: 5'b0 :: 7'b0001011;
+      behavior: ACC = 0;
+    }
+    
+    GET_ACC_LO {
+      encoding: 7'd1 :: 10'b0 :: 3'd0 :: rd[4:0] :: 7'b0001011;
+      behavior: if (rd != 0) X[rd] = ACC[31:0];
+    }
+    
+    GET_ACC_HI {
+      encoding: 7'd2 :: 10'b0 :: 3'd0 :: rd[4:0] :: 7'b0001011;
+      behavior: if (rd != 0) X[rd] = ACC[63:32];
+    }
+    
+    MACU_32 {
+      encoding: 7'd0 :: rs2[4:0] :: rs1[4:0] :: 3'd1 :: 5'b0 :: 7'b0001011;
+      behavior: {
+        unsigned<64> mul = X[rs1]    * X[rs2];
+        unsigned<33> add = mul[31:0] + ACC[31:0];
+        ACC = add[31:0];
+      }
+    }
+    
+    MACS_32 {
+      encoding: 7'd1 :: rs2[4:0] :: rs1[4:0] :: 3'd1 :: 5'b0 :: 7'b0001011;
+      behavior: {
+        signed<64> mul = ((signed) X[rs1])    * ((signed) X[rs2]);
+        signed<33> add = ((signed) mul[31:0]) + ((signed) ACC[31:0]);
+        ACC = add[31:0]; // bit range always yields unsigned type
+      }
+    }
+    
+    MACU_64 {
+      encoding: 7'd0 :: rs2[4:0] :: rs1[4:0] :: 3'd2 :: 5'b0 :: 7'b0001011;
+      behavior: {
+        unsigned<64> mul = X[rs1] * X[rs2];
+        unsigned<65> add = mul    + ACC;
+        ACC = add[63:0];
+      }
+    }
+    
+    MACS_64 {
+      encoding: 7'd1 :: rs2[4:0] :: rs1[4:0] :: 3'd2 :: 5'b0 :: 7'b0001011;
+      behavior: {
+        signed<64> mul = ((signed) X[rs1]) * ((signed) X[rs2]);
+        signed<65> add =           mul     + ((signed) ACC);
+        ACC = add[63:0];
+      }
+    }
+  }
+}
+
+Core TGC_D_XRB_MAC provides RV32I, RV32M, RV32IC, X_RB_MAC {
+    architectural_state {
+        XLEN=32;
+        // definitions for the architecture wrapper
+        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
+        unsigned MISA_VAL = 0b01000000000000000001000100000100;
+        unsigned MARCHID_VAL = 0x80000004;
+    }
+}

gen_input/TGFS.core_desc

@@ -1,37 +0,0 @@
-import "CoreDSL-Instruction-Set-Description/RV32I.core_desc"
-import "CoreDSL-Instruction-Set-Description/RVM.core_desc"
-import "CoreDSL-Instruction-Set-Description/RVC.core_desc"
-
-Core TGC_B provides RV32I {
-	architectural_state {
-        unsigned XLEN=32;
-        unsigned PCLEN=32;
-        // definitions for the architecture wrapper
-        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
-        unsigned MISA_VAL = 0b01000000000000000000000100000000;
-        unsigned PGSIZE = 0x1000; //1 << 12;
-        unsigned PGMASK = 0xfff; //PGSIZE-1
-	}
-}
-
-Core TGC_C provides RV32I, RV32M, RV32IC {
-    architectural_state {
-        unsigned XLEN=32;
-        unsigned PCLEN=32;
-        // definitions for the architecture wrapper
-        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
-        unsigned MISA_VAL = 0b01000000000000000001000100000100;
-        unsigned PGSIZE = 0x1000; //1 << 12;
-        unsigned PGMASK = 0xfff; //PGSIZE-1
-    }
-}
-
-Core TGC_D provides RV32I, RV32M, RV32IC {
-    architectural_state {
-        unsigned XLEN=32;
-        unsigned PCLEN=32;
-        // definitions for the architecture wrapper
-        //                    XL    ZYXWVUTSRQPONMLKJIHGFEDCBA
-        unsigned MISA_VAL = 0b01000000000000000001000100000100;
-    }
-}

gen_input/templates/CORENAME_instr.yaml.gtl

@@ -0,0 +1,16 @@
+<% def getInstructionGroups() {
+    def instrGroups = [:]
+    instructions.each {
+        def groupName = it['instruction'].eContainer().name
+        if(!instrGroups.containsKey(groupName)) {
+            instrGroups[groupName]=[]
+        }
+        instrGroups[groupName]+=it;
+    }
+    instrGroups
+}%><%getInstructionGroups().each{name, instrList -> %>
+${name}: <% instrList.findAll{!it.instruction.name.startsWith("__")}.each { %>
+  - ${it.instruction.name}:
+    encoding: ${it.encoding}
+    mask: ${it.mask}<%}}%>
+

gen_input/templates/interp/CORENAME.cpp.gtl

@@ -324,6 +324,7 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
             if (is_jump_to_self_enabled(cond) &&
                     (insn == 0x0000006f || (insn&0xffff)==0xa001)) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
             auto f = decode_inst(insn);
+            auto old_pc = pc.val;
             pc = (this->*f)(pc, insn);
         }
     }

incl/iss/arch/riscv_hart_m_p.h

@@ -307,7 +307,7 @@ riscv_hart_m_p<BASE>::riscv_hart_m_p()
     // reset values
     csr[misa] = traits<BASE>::MISA_VAL;
     csr[mvendorid] = 0x669;
-    csr[marchid] = 0x80000003;
+    csr[marchid] = traits<BASE>::MARCHID_VAL;
     csr[mimpid] = 1;
 
     uart_buf.str("");
@@ -362,8 +362,8 @@ riscv_hart_m_p<BASE>::riscv_hart_m_p()
     csr_rd_cb[mie] = &this_class::read_ie;
     csr_wr_cb[mie] = &this_class::write_ie;
     csr_rd_cb[mhartid] = &this_class::read_hartid;
-    csr_rd_cb[mcounteren] = &this_class::read_null;
-    csr_wr_cb[mcounteren] = &this_class::write_null;
+//    csr_rd_cb[mcounteren] = &this_class::read_null;
+//    csr_wr_cb[mcounteren] = &this_class::write_null;
     csr_wr_cb[misa] = &this_class::write_null;
     csr_wr_cb[mvendorid] = &this_class::write_null;
     csr_wr_cb[marchid] = &this_class::write_null;
@@ -920,7 +920,16 @@ template <typename BASE> uint64_t riscv_hart_m_p<BASE>::enter_trap(uint64_t flag
     if (trap_id == 0) { // exception
         // store ret addr in xepc register
         csr[mepc] = static_cast<reg_t>(addr) & get_pc_mask(); // store actual address instruction of exception
-        csr[mtval] = cause==2?((instr & 0x3)==3?instr:instr&0xffff):fault_data;
+        switch(cause){
+        case 0:
+            csr[mtval] = addr;
+            break;
+        case 2:
+            csr[mtval] = (instr & 0x3)==3?instr:instr&0xffff;
+            break;
+        default:
+            csr[mtval] = fault_data;
+        }
         fault_data = 0;
     } else {
         csr[mepc] = this->reg.NEXT_PC & get_pc_mask(); // store next address if interrupt

incl/iss/arch/riscv_hart_msu_vp.h

@@ -145,7 +145,7 @@ public:
 
         mstatus_t mstatus;
 
-        static const reg_t mstatus_reset_val = 0;
+        static const reg_t mstatus_reset_val = 0x1800;
 
         void write_mstatus(T val, unsigned priv_lvl) {
             auto mask = get_mask(priv_lvl);
@@ -398,7 +398,7 @@ private:
     iss::status read_ip(unsigned addr, reg_t &val);
     iss::status write_ip(unsigned addr, reg_t val);
     iss::status read_hartid(unsigned addr, reg_t &val);
-    iss::status write_mepc(unsigned addr, reg_t val);
+    iss::status write_epc(unsigned addr, reg_t val);
     iss::status read_satp(unsigned addr, reg_t &val);
     iss::status write_satp(unsigned addr, reg_t val);
     iss::status read_fcsr(unsigned addr, reg_t &val);
@@ -419,7 +419,7 @@ riscv_hart_msu_vp<BASE>::riscv_hart_msu_vp()
     // reset values
     csr[misa] = traits<BASE>::MISA_VAL;
     csr[mvendorid] = 0x669;
-    csr[marchid] = 0x80000003;
+    csr[marchid] = traits<BASE>::MARCHID_VAL;
     csr[mimpid] = 1;
 
     uart_buf.str("");
@@ -954,8 +954,8 @@ template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_ip(unsigned
     return iss::Ok;
 }
 
-template <typename BASE> iss::status riscv_hart_m_p<BASE>::write_epc(unsigned addr, reg_t val) {
-     csr[addr] = val & get_pc_mask();
+template <typename BASE> iss::status riscv_hart_msu_vp<BASE>::write_epc(unsigned addr, reg_t val) {
+    csr[addr] = val & get_pc_mask();
     return iss::Ok;
 }
 

incl/iss/arch/riscv_hart_mu_p.h

@@ -146,7 +146,7 @@ public:
 
         mstatus_t mstatus;
 
-        static const reg_t mstatus_reset_val = 0;
+        static const reg_t mstatus_reset_val = 0x1800; // MPP set to 1
 
         void write_mstatus(T val, unsigned priv_lvl) {
             auto mask = get_mask(priv_lvl);
@@ -349,7 +349,7 @@ riscv_hart_mu_p<BASE, FEAT>::riscv_hart_mu_p()
     // reset values
     csr[misa] = traits<BASE>::MISA_VAL;
     csr[mvendorid] = 0x669;
-    csr[marchid] = 0x80000004;
+    csr[marchid] = traits<BASE>::MARCHID_VAL;
     csr[mimpid] = 1;
     csr[mclicbase] = 0xc0000000; // TODO: should be taken from YAML file
 
@@ -543,25 +543,27 @@ template <typename BASE, features_e FEAT> bool riscv_hart_mu_p<BASE, FEAT>::pmp_
     constexpr auto PMP_NA4 =0x2U;
     constexpr auto PMP_NAPOT =0x3U;
     reg_t base = 0;
+    auto any_active = false;
     for (size_t i = 0; i < 16; i++) {
         reg_t tor = csr[pmpaddr0+i] << PMP_SHIFT;
         uint8_t cfg = csr[pmpcfg0+(i/4)]>>(i%4);
         if (cfg & PMP_A) {
+            any_active=true;
             auto pmp_a = (cfg & PMP_A) >> 3;
-            bool is_tor = pmp_a == PMP_TOR;
-            bool is_na4 = pmp_a == PMP_NA4;
+            auto is_tor = pmp_a == PMP_TOR;
+            auto is_na4 = pmp_a == PMP_NA4;
 
             reg_t mask = (csr[pmpaddr0+i] << 1) | (!is_na4);
             mask = ~(mask & ~(mask + 1)) << PMP_SHIFT;
 
             // Check each 4-byte sector of the access
-            bool any_match = false;
-            bool all_match = true;
+            auto any_match = false;
+            auto all_match = true;
             for (reg_t offset = 0; offset < len; offset += 1 << PMP_SHIFT) {
                 reg_t cur_addr = addr + offset;
-                bool napot_match = ((cur_addr ^ tor) & mask) == 0;
-                bool tor_match = base <= cur_addr && cur_addr < tor;
-                bool match = is_tor ? tor_match : napot_match;
+                auto napot_match = ((cur_addr ^ tor) & mask) == 0;
+                auto tor_match = base <= cur_addr && cur_addr < tor;
+                auto match = is_tor ? tor_match : napot_match;
                 any_match |= match;
                 all_match &= match;
             }
@@ -577,7 +579,7 @@ template <typename BASE, features_e FEAT> bool riscv_hart_mu_p<BASE, FEAT>::pmp_
         }
         base = tor;
     }
-    return this->reg.PRIV == PRIV_M;
+    return !any_active || this->reg.PRIV == PRIV_M;
 }
 
 
@@ -926,8 +928,6 @@ template <typename BASE, features_e FEAT> iss::status riscv_hart_mu_p<BASE, FEAT
 
 template <typename BASE, features_e FEAT> iss::status riscv_hart_mu_p<BASE, FEAT>::write_ie(unsigned addr, reg_t val) {
     auto mask = get_irq_wrmask((addr >> 8) & 0x3);
-    if(this->reg.PRIV==0)
-        mask&= ~(0xff<<4); // STIE and UTIE are read only in user and supervisor mode
     csr[mie] = (csr[mie] & ~mask) | (val & mask);
     check_interrupt();
     return iss::Ok;
@@ -1256,27 +1256,33 @@ template <typename BASE, features_e FEAT> uint64_t riscv_hart_mu_p<BASE, FEAT>::
 template <typename BASE, features_e FEAT> uint64_t riscv_hart_mu_p<BASE, FEAT>::leave_trap(uint64_t flags) {
     auto cur_priv = this->reg.PRIV;
     auto inst_priv = (flags & 0x3)? 3:0;
-    auto status = state.mstatus;
-    // pop the relevant lower-privilege interrupt enable and privilege mode stack
-    // clear respective yIE
-    switch (inst_priv) {
-    case PRIV_M:
-        this->reg.PRIV = state.mstatus.MPP;
-        state.mstatus.MPP = 0; // clear mpp to U mode
-        state.mstatus.MIE = state.mstatus.MPIE;
-        state.mstatus.MPIE = 1;
-        break;
-    case PRIV_U:
-        this->reg.PRIV = 0;
-        state.mstatus.UIE = state.mstatus.UPIE;
-        state.mstatus.UPIE = 1;
-        break;
+    if(inst_priv>cur_priv){
+        auto trap_val =  0x80ULL << 24 | (2 << 16); // illegal instruction
+        this->reg.trap_state = trap_val;
+        this->reg.NEXT_PC = std::numeric_limits<uint32_t>::max();
+    } else {
+        auto status = state.mstatus;
+        // pop the relevant lower-privilege interrupt enable and privilege mode stack
+        // clear respective yIE
+        switch (inst_priv) {
+        case PRIV_M:
+            this->reg.PRIV = state.mstatus.MPP;
+            state.mstatus.MPP = 0; // clear mpp to U mode
+            state.mstatus.MIE = state.mstatus.MPIE;
+            state.mstatus.MPIE = 1;
+            break;
+        case PRIV_U:
+            this->reg.PRIV = 0;
+            state.mstatus.UIE = state.mstatus.UPIE;
+            state.mstatus.UPIE = 1;
+            break;
+        }
+        // sets the pc to the value stored in the x epc register.
+        this->reg.NEXT_PC = csr[uepc | inst_priv << 8];
+        CLOG(INFO, disass) << "Executing xRET , changing privilege level from " << lvl[cur_priv] << " to "
+                           << lvl[this->reg.PRIV];
+        check_interrupt();
     }
-    // sets the pc to the value stored in the x epc register.
-    this->reg.NEXT_PC = csr[uepc | inst_priv << 8];
-    CLOG(INFO, disass) << "Executing xRET , changing privilege level from " << lvl[cur_priv] << " to "
-                       << lvl[this->reg.PRIV];
-    check_interrupt();
     return this->reg.NEXT_PC;
 }
 

incl/iss/arch/tgc_c.h

@@ -51,9 +51,9 @@ template <> struct traits<tgc_c> {
         {"X0", "X1", "X2", "X3", "X4", "X5", "X6", "X7", "X8", "X9", "X10", "X11", "X12", "X13", "X14", "X15", "X16", "X17", "X18", "X19", "X20", "X21", "X22", "X23", "X24", "X25", "X26", "X27", "X28", "X29", "X30", "X31", "PC", "NEXT_PC", "PRIV"}};
  
     static constexpr std::array<const char*, 35> reg_aliases{
-        {"X0", "X1", "X2", "X3", "X4", "X5", "X6", "X7", "X8", "X9", "X10", "X11", "X12", "X13", "X14", "X15", "X16", "X17", "X18", "X19", "X20", "X21", "X22", "X23", "X24", "X25", "X26", "X27", "X28", "X29", "X30", "X31", "PC", "NEXT_PC", "PRIV"}};
+        {"ZERO", "RA", "SP", "GP", "TP", "T0", "T1", "T2", "S0", "S1", "A0", "A1", "A2", "A3", "A4", "A5", "A6", "A7", "S2", "S3", "S4", "S5", "S6", "S7", "S8", "S9", "S10", "S11", "T3", "T4", "T5", "T6", "PC", "NEXT_PC", "PRIV"}};
 
-    enum constants {XLEN=32, PCLEN=32, MISA_VAL=0b01000000000000000001000100000100, PGSIZE=0x1000, PGMASK=0b111111111111, CSR_SIZE=4096, fence=0, fencei=1, fencevmal=2, fencevmau=3, MUL_LEN=64};
+    enum constants {MISA_VAL=0b01000000000000000001000100000100, MARCHID_VAL=0x80000003, XLEN=32, CSR_SIZE=4096, INSTR_ALIGNMENT=2, fence=0, fencei=1, fencevmal=2, fencevmau=3, MUL_LEN=64};
 
     constexpr static unsigned FP_REGS_SIZE = 0;
 

src/main.cpp

@@ -49,6 +49,11 @@ using tgc_b_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc_b>;
 #include "iss/arch/tgc_d.h"
 using tgc_d_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_d, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
 #endif
+#ifdef CORE_TGC_D_XRB_MAC
+#include "iss/arch/riscv_hart_mu_p.h"
+#include "iss/arch/tgc_d_xrb_mac.h"
+using tgc_d_xrb_mac_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_d_xrb_mac, (iss::arch::features_e)(iss::arch::FEAT_PMP | iss::arch::FEAT_CLIC | iss::arch::FEAT_EXT_N)>;
+#endif
 #ifdef WITH_LLVM
 #include <iss/llvm/jit_helper.h>
 #endif
@@ -138,6 +143,12 @@ int main(int argc, char *argv[]) {
             std::tie(cpu, vm) =
                 iss::create_cpu<tgc_d_plat_type>(clim["backend"].as<std::string>(), clim["gdb-port"].as<unsigned>());
         } else
+#endif
+#ifdef CORE_TGC_D_XRB_MAC
+        if (isa_opt == "tgc_d_xrb_mac") {
+            std::tie(cpu, vm) =
+                iss::create_cpu<tgc_d_xrb_mac_plat_type>(clim["backend"].as<std::string>(), clim["gdb-port"].as<unsigned>());
+        } else
 #endif
         {
             LOG(ERROR) << "Illegal argument value for '--isa': " << clim["isa"].as<std::string>() << std::endl;

src/sysc/core_complex.cpp

@@ -44,6 +44,11 @@ using tgc_c_plat_type = iss::arch::riscv_hart_m_p<iss::arch::tgc_c>;
 #include "iss/arch/tgc_d.h"
 using tgc_d_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_d, iss::arch::FEAT_PMP>;
 #endif
+#ifdef CORE_TGC_D_XRB_MAC
+#include "iss/arch/riscv_hart_mu_p.h"
+#include "iss/arch/tgc_d_xrb_mac.h"
+using tgc_d_xrb_mac_plat_type = iss::arch::riscv_hart_mu_p<iss::arch::tgc_d_xrb_mac, iss::arch::FEAT_PMP>;
+#endif
 #include "iss/debugger/encoderdecoder.h"
 #include "iss/debugger/gdb_session.h"
 #include "iss/debugger/server.h"
@@ -285,6 +290,9 @@ public:
 #endif
 #ifdef CORE_TGC_D
         CREATE_CORE(tgc_d)
+#endif
+#ifdef CORE_TGC_D_XRB_MACD
+        CREATE_CORE(tgc_d_xrb_mac)
 #endif
         {
             LOG(ERROR) << "Illegal argument value for core type: " << type << std::endl;

src/vm/interp/vm_tgc_c.cpp

@@ -463,8 +463,13 @@ private:
         // execute instruction
         try {
         {
-            if(rd != 0) *(X+rd) = *PC + 4; 
-            pc_assign(*NEXT_PC) = *PC + (int32_t)sext<21>(imm);
+            if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else {
+                if(rd != 0) *(X+rd) = *PC + 4; 
+                pc_assign(*NEXT_PC) = *PC + (int32_t)sext<21>(imm);
+            }
         }
         } catch(...){}
         // post execution stuff
@@ -507,9 +512,14 @@ private:
         // execute instruction
         try {
         {
-            int32_t new_pc = *(X+rs1) + (int16_t)sext<12>(imm);
-            if(rd != 0) *(X+rd) = *PC + 4; 
-            pc_assign(*NEXT_PC) = new_pc & ~ 0x1;
+            int32_t new_pc = (*(X+rs1) + (int16_t)sext<12>(imm)) & ~ 1;
+            if(new_pc % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else {
+                if(rd != 0) *(X+rd) = *PC + 4; 
+                pc_assign(*NEXT_PC) = new_pc & ~ 0x1;
+            }
         }
         } catch(...){}
         // post execution stuff
@@ -551,7 +561,12 @@ private:
         *NEXT_PC = *PC + 4;
         // execute instruction
         try {
-        if(*(X+rs1) == *(X+rs2)) pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm); 
+        {
+            if(*(X+rs1) == *(X+rs2)) if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm);
+        }
         } catch(...){}
         // post execution stuff
         if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, 4);
@@ -592,7 +607,12 @@ private:
         *NEXT_PC = *PC + 4;
         // execute instruction
         try {
-        if(*(X+rs1) != *(X+rs2)) pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm); 
+        {
+            if(*(X+rs1) != *(X+rs2)) if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm);
+        }
         } catch(...){}
         // post execution stuff
         if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, 5);
@@ -633,7 +653,12 @@ private:
         *NEXT_PC = *PC + 4;
         // execute instruction
         try {
-        if((int32_t)*(X+rs1) < (int32_t)*(X+rs2)) pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm); 
+        {
+            if((int32_t)*(X+rs1) < (int32_t)*(X+rs2)) if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm);
+        }
         } catch(...){}
         // post execution stuff
         if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, 6);
@@ -674,7 +699,12 @@ private:
         *NEXT_PC = *PC + 4;
         // execute instruction
         try {
-        if((int32_t)*(X+rs1) >= (int32_t)*(X+rs2)) pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm); 
+        {
+            if((int32_t)*(X+rs1) >= (int32_t)*(X+rs2)) if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm);
+        }
         } catch(...){}
         // post execution stuff
         if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, 7);
@@ -715,7 +745,12 @@ private:
         *NEXT_PC = *PC + 4;
         // execute instruction
         try {
-        if(*(X+rs1) < *(X+rs2)) pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm); 
+        {
+            if(*(X+rs1) < *(X+rs2)) if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm);
+        }
         } catch(...){}
         // post execution stuff
         if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, 8);
@@ -756,7 +791,12 @@ private:
         *NEXT_PC = *PC + 4;
         // execute instruction
         try {
-        if(*(X+rs1) >= *(X+rs2)) pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm); 
+        {
+            if(*(X+rs1) >= *(X+rs2)) if(imm % traits::INSTR_ALIGNMENT) {
+                raise(0,  0);
+            }
+            else pc_assign(*NEXT_PC) = *PC + (int16_t)sext<13>(imm);
+        }
         } catch(...){}
         // post execution stuff
         if(this->sync_exec && POST_SYNC) this->do_sync(POST_SYNC, 9);
@@ -4138,9 +4178,10 @@ typename vm_base<ARCH>::virt_addr_t vm_impl<ARCH>::execute_inst(finish_cond_e co
             this->do_sync(POST_SYNC, std::numeric_limits<unsigned>::max());
             pc.val = super::core.enter_trap(std::numeric_limits<uint64_t>::max(), pc.val, 0);
         } else {
-            if (is_jump_to_self_enabled(cond) && (insn == 0x0000006f || (insn&0xffff)==0xa001))
-                throw simulation_stopped(0); // 'J 0' or 'C.J 0'
+            if (is_jump_to_self_enabled(cond) &&
+                    (insn == 0x0000006f || (insn&0xffff)==0xa001)) throw simulation_stopped(0); // 'J 0' or 'C.J 0'
             auto f = decode_inst(insn);
+            auto old_pc = pc.val;
             pc = (this->*f)(pc, insn);
         }
     }