From 5f69e58f3743c12e68691373421c893b367076ef Mon Sep 17 00:00:00 2001
From: Yang Liu <liuyang22@iscas.ac.cn>
Date: Fri, 21 Apr 2023 00:23:25 +0800
Subject: [RV64_DYNAREC] Added more opcodes (#718)

* Added F0 01 LOCK ADD opcode

* Added F2 0F 51 SQRTSD opcode

* Added FE /0,/1 INC/DEC opcodes
---
 src/dynarec/rv64/dynarec_rv64_00_3.c      |  21 ++++++
 src/dynarec/rv64/dynarec_rv64_emit_math.c | 109 ++++++++++++++++++++++++++++++
 src/dynarec/rv64/dynarec_rv64_f0.c        |  22 ++++++
 src/dynarec/rv64/dynarec_rv64_f20f.c      |   7 ++
 src/dynarec/rv64/dynarec_rv64_helper.h    |   4 +-
 5 files changed, 161 insertions(+), 2 deletions(-)

(limited to 'src')

diff --git a/src/dynarec/rv64/dynarec_rv64_00_3.c b/src/dynarec/rv64/dynarec_rv64_00_3.c
index f6c2a4ed..a1e7421e 100644
--- a/src/dynarec/rv64/dynarec_rv64_00_3.c
+++ b/src/dynarec/rv64/dynarec_rv64_00_3.c
@@ -914,6 +914,27 @@ uintptr_t dynarec64_00_3(dynarec_rv64_t* dyn, uintptr_t addr, uintptr_t ip, int
             INST_NAME("CLD");
             ANDI(xFlags, xFlags, ~(1<<F_CF));
             break;
+        case 0xFE:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("INC Eb");
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
+                    GETEB(x1, 0);
+                    emit_inc8(dyn, ninst, ed, x2, x4, x5);
+                    EBBACK(x5, 0);
+                    break;
+                case 1:
+                    INST_NAME("DEC Eb");
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
+                    GETEB(x1, 0);
+                    emit_dec8(dyn, ninst, ed, x2, x4, x5);
+                    EBBACK(x5, 0);
+                    break;
+                default:
+                    DEFAULT;
+            }
+            break;
         case 0xFF:
             nextop = F8;
             switch((nextop>>3)&7) {
diff --git a/src/dynarec/rv64/dynarec_rv64_emit_math.c b/src/dynarec/rv64/dynarec_rv64_emit_math.c
index 2f4138d6..df70eac4 100644
--- a/src/dynarec/rv64/dynarec_rv64_emit_math.c
+++ b/src/dynarec/rv64/dynarec_rv64_emit_math.c
@@ -557,6 +557,115 @@ void emit_sub32c(dynarec_rv64_t* dyn, int ninst, rex_t rex, int s1, int64_t c, i
     }
 }
 
+// emit INC8 instruction, from s1, store result in s1 using s2, s3 and s4 as scratch
+void emit_inc8(dynarec_rv64_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_ALL) {
+        ANDI(xFlags, xFlags, ~((1UL<<F_AF) | (1UL<<F_OF2) | (1UL<<F_ZF) | (1UL<<F_SF) | (1UL<<F_PF)));
+    }
+    IFX(X_PEND) {
+        SB(s1, xEmu, offsetof(x64emu_t, op1));
+        SET_DF(s3, d_inc8);
+    } else IFX(X_ALL) {
+        SET_DFNONE();
+    }
+    IFX(X_AF | X_OF) {
+        ORI(s3, s1, 1);      // s3 = op1 | op2
+        ANDI(s4, s1, 1);      // s5 = op1 & op2
+    }
+
+    ADDIW(s1, s1, 1);
+
+    IFX(X_PEND) {
+        SB(s1, xEmu, offsetof(x64emu_t, res));
+    }
+    IFX(X_AF | X_OF) {
+        NOT(s2, s1);   // s2 = ~res
+        AND(s3, s2, s3);   // s3 = ~res & (op1 | op2)
+        OR(s3, s3, s4);   // cc = (~res & (op1 | op2)) | (op1 & op2)
+        IFX(X_AF) {
+            ANDI(s2, s3, 0x08); // AF: cc & 0x08
+            BEQZ(s2, 8);
+            ORI(xFlags, xFlags, 1 << F_AF);
+        }
+        IFX(X_OF) {
+            SRLI(s3, s3, 6);
+            SRLI(s2, s3, 1);
+            XOR(s3, s3, s2);
+            ANDI(s3, s3, 1); // OF: xor of two MSB's of cc
+            BEQZ(s3, 8);
+            ORI(xFlags, xFlags, 1 << F_OF2);
+        }
+    }
+    IFX(X_SF) {
+        BGE(s1, xZR, 8);
+        ORI(xFlags, xFlags, 1 << F_SF);
+    }
+    ANDI(s1, s1, 0xff);
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s2);
+    }
+    IFX(X_ZF) {
+        BNEZ(s1, 8);
+        ORI(xFlags, xFlags, 1 << F_ZF);
+    }
+}
+
+
+// emit DEC8 instruction, from s1, store result in s1 using s2, s3 and s4 as scratch
+void emit_dec8(dynarec_rv64_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_ALL) {
+        ANDI(xFlags, xFlags, ~((1UL<<F_AF) | (1UL<<F_OF2) | (1UL<<F_ZF) | (1UL<<F_SF) | (1UL<<F_PF)));
+    }
+    IFX(X_PEND) {
+        SB(s1, xEmu, offsetof(x64emu_t, op1));
+        SET_DF(s3, d_dec8);
+    } else IFX(X_ALL) {
+        SET_DFNONE();
+    }
+    IFX(X_AF | X_OF) {
+        ORI(s3, s1, 1);      // s3 = op1 | op2
+        ANDI(s4, s1, 1);      // s4 = op1 & op2
+    }
+
+    ADDIW(s1, s1, -1);
+
+    IFX(X_PEND) {
+        SB(s1, xEmu, offsetof(x64emu_t, res));
+    }
+    IFX(X_AF | X_OF) {
+        NOT(s2, s1);   // s2 = ~res
+        AND(s3, s2, s3);   // s3 = ~res & (op1 | op2)
+        OR(s3, s3, s4);   // cc = (~res & (op1 | op2)) | (op1 & op2)
+        IFX(X_AF) {
+            ANDI(s2, s3, 0x08); // AF: cc & 0x08
+            BEQZ(s2, 8);
+            ORI(xFlags, xFlags, 1 << F_AF);
+        }
+        IFX(X_OF) {
+            SRLI(s3, s3, 6);
+            SRLI(s2, s3, 1);
+            XOR(s3, s3, s2);
+            ANDI(s3, s3, 1); // OF: xor of two MSB's of cc
+            BEQZ(s3, 8);
+            ORI(xFlags, xFlags, 1 << F_OF2);
+        }
+    }
+    IFX(X_SF) {
+        BGE(s1, xZR, 8);
+        ORI(xFlags, xFlags, 1 << F_SF);
+    }
+    ANDI(s1, s1, 0xff);
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s2);
+    }
+    IFX(X_ZF) {
+        BNEZ(s1, 8);
+        ORI(xFlags, xFlags, 1 << F_ZF);
+    }
+}
+
 // emit INC32 instruction, from s1, store result in s1 using s3 and s4 as scratch
 void emit_inc32(dynarec_rv64_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3, int s4, int s5)
 {
diff --git a/src/dynarec/rv64/dynarec_rv64_f0.c b/src/dynarec/rv64/dynarec_rv64_f0.c
index ff15db2b..f5ae3950 100644
--- a/src/dynarec/rv64/dynarec_rv64_f0.c
+++ b/src/dynarec/rv64/dynarec_rv64_f0.c
@@ -57,6 +57,28 @@ uintptr_t dynarec64_F0(dynarec_rv64_t* dyn, uintptr_t addr, uintptr_t ip, int ni
     // TODO: Take care of unligned memory access for all the LOCK ones.
     // https://github.com/ptitSeb/box64/pull/604
     switch(opcode) {
+        case 0x01:
+            INST_NAME("LOCK ADD Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            SMDMB();
+            if((nextop&0xC0)==0xC0) {
+                ed = xRAX+(nextop&7)+(rex.b<<3);
+                emit_add32(dyn, ninst, rex, ed, gd, x3, x4, x5);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, x1, &fixedaddress, rex, LOCK_LOCK, 0, 0);
+                MARKLOCK;
+                LRxw(x1, wback, 1, 1);
+                ADDxw(x4, x1, gd);
+                SCxw(x3, x4, wback, 1, 1);
+                BNEZ_MARKLOCK(x3);
+                IFX(X_ALL|X_PEND) {
+                    emit_add32(dyn, ninst, rex, x1, gd, x3, x4, x5);
+                }
+            }
+            SMDMB();
+            break;
         case 0x09:
             INST_NAME("LOCK OR Ed, Gd");
             SETFLAGS(X_ALL, SF_SET_PENDING);
diff --git a/src/dynarec/rv64/dynarec_rv64_f20f.c b/src/dynarec/rv64/dynarec_rv64_f20f.c
index c8976aff..11b4f506 100644
--- a/src/dynarec/rv64/dynarec_rv64_f20f.c
+++ b/src/dynarec/rv64/dynarec_rv64_f20f.c
@@ -178,6 +178,13 @@ uintptr_t dynarec64_F20F(dynarec_rv64_t* dyn, uintptr_t addr, uintptr_t ip, int
                     DEFAULT;
             }
             break;
+        case 0x51: // TODO: fastnan handling
+            INST_NAME("SQRTSD Gx, Ex");
+            nextop = F8;
+            GETEXSD(d0, 0);
+            GETGXSD_empty(d1);
+            FSQRTD(d1, d0);
+            break;
         case 0x58:
             INST_NAME("ADDSD Gx, Ex");
             nextop = F8;
diff --git a/src/dynarec/rv64/dynarec_rv64_helper.h b/src/dynarec/rv64/dynarec_rv64_helper.h
index 2785ce99..90d2598e 100644
--- a/src/dynarec/rv64/dynarec_rv64_helper.h
+++ b/src/dynarec/rv64/dynarec_rv64_helper.h
@@ -945,10 +945,10 @@ void emit_and16(dynarec_rv64_t* dyn, int ninst, int s1, int s2, int s3, int s4);
 //void emit_and16c(dynarec_rv64_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
 void emit_inc32(dynarec_rv64_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3, int s4, int s5);
 //void emit_inc16(dynarec_rv64_t* dyn, int ninst, int s1, int s3, int s4);
-//void emit_inc8(dynarec_rv64_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_inc8(dynarec_rv64_t* dyn, int ninst, int s1, int s2, int s3, int s4);
 void emit_dec32(dynarec_rv64_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3, int s4, int s5);
 //void emit_dec16(dynarec_rv64_t* dyn, int ninst, int s1, int s3, int s4);
-//void emit_dec8(dynarec_rv64_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_dec8(dynarec_rv64_t* dyn, int ninst, int s1, int s2, int s3, int s4);
 void emit_adc32(dynarec_rv64_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3, int s4, int s5);
 //void emit_adc32c(dynarec_rv64_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
 //void emit_adc8(dynarec_rv64_t* dyn, int ninst, int s1, int s2, int s3, int s4);
-- 
cgit 1.4.1