Add !=,<,<=,>,>=; both signed and not

4 files changed, 299 insertions, 8 deletions

diff --git a/executor/executor.h b/executor/executor.h
index 245e64c..70d9637 100644
--- a/executor/executor.h
+++ b/executor/executor.h
@@ -45,7 +45,17 @@ CHECK_RESULT int amal_exec_call(amal_executor *self, u32 code_offset, AmalReg ds
 void amal_exec_call_overwrite(amal_executor *self, u32 call_code_offset, i32 new_target_rel32);
 /*CHECK_RESULT int amal_exec_callr(AmalReg dst_reg, BufferView data);*/
 CHECK_RESULT int amal_exec_calle(amal_executor *self, void *func, AmalReg dst_reg);
-CHECK_RESULT int amal_exec_cmp(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_eq(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_neq(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_ilt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_ile(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_igt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_ige(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_lt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_le(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_gt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_ge(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
+CHECK_RESULT int amal_exec_and(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2);
 CHECK_RESULT int amal_exec_jz(amal_executor *self, AmalReg reg, u16 target_label);
 CHECK_RESULT int amal_exec_jmp(amal_executor *self, u16 target_label);
 CHECK_RESULT int amal_exec_ret(amal_executor *self, AmalReg reg);
diff --git a/executor/x86_64/asm.c b/executor/x86_64/asm.c
index f7bc19c..a400656 100644
--- a/executor/x86_64/asm.c
+++ b/executor/x86_64/asm.c
@@ -410,6 +410,14 @@ void asm_mov_rr(Asm *self, Reg64 dst, Reg64 src) {
     ins_end(self, "mov %s, %s", reg64_to_str(dst), reg64_to_str(src));
 }
 
+void asm_and_mr(Asm *self, AsmPtr *dst, Reg64 src) {
+    ins_start(self);
+    *self->code_it++ = rex_rm(dst, src);
+    *self->code_it++ = 0x21;
+    asm_rm(self, dst, src);
+    ins_end(self, "and %s, %s", asm_ptr_to_string(dst), reg64_to_str(src));
+}
+
 void asm_add_rr(Asm *self, Reg64 dst, Reg64 src) {
     ins_start(self);
     *self->code_it++ = rex_rr(dst, src);
@@ -515,6 +523,87 @@ void asm_sete_r(Asm *self, Reg64 dst) {
     ins_end(self, "sete %s", reg64_to_str(dst));
 }
 
+void asm_setne_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x95;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setne %s", reg64_to_str(dst));
+}
+
+void asm_setb_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x92;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setb %s", reg64_to_str(dst));
+}
+
+void asm_setbe_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x96;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setbe %s", reg64_to_str(dst));
+}
+
+void asm_seta_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x97;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "seta %s", reg64_to_str(dst));
+}
+
+void asm_setae_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x93;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setae %s", reg64_to_str(dst));
+}
+
+void asm_setl_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x9C;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setl %s", reg64_to_str(dst));
+}
+
+void asm_setle_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x9E;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setle %s", reg64_to_str(dst));
+}
+
+void asm_setg_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x9F;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setg %s", reg64_to_str(dst));
+}
+
+void asm_setge_r(Asm *self, Reg64 dst) {
+    assert(dst != RSP && dst != RBP && dst != RSI && dst != RDI);
+    ins_start(self);
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0x9D;
+    asm_rr(self, dst, 0x0); /* the @src bits are not used */
+    ins_end(self, "setge %s", reg64_to_str(dst));
+}
+
 /*
     Note: This is sometimes called with @relative INT32_MAX-(2 or 6) (will print jz 0x7ffffff9), in which case it's most likely a dummy
     jump until the relative position is later changed with @asm_overwrite_jcc_rel32.
diff --git a/executor/x86_64/asm.h b/executor/x86_64/asm.h
index dacc248..7d68bc0 100644
--- a/executor/x86_64/asm.h
+++ b/executor/x86_64/asm.h
@@ -80,6 +80,8 @@ void asm_mov_rm(Asm *self, Reg64 dst, AsmPtr *src);
 void asm_mov_ri(Asm *self, Reg64 dst, i64 immediate);
 void asm_mov_rr(Asm *self, Reg64 dst, Reg64 src);
 
+void asm_and_mr(Asm *self, AsmPtr *dst, Reg64 src);
+
 void asm_add_rr(Asm *self, Reg64 dst, Reg64 src);
 void asm_sub_rr(Asm *self, Reg64 dst, Reg64 src);
 void asm_imul_rr(Asm *self, Reg64 dst, Reg64 src);
@@ -110,6 +112,23 @@ void asm_cmp_rm(Asm *self, Reg64 reg1, AsmPtr *reg2);
 */
 void asm_sete_m(Asm *self, AsmPtr *dst);
 void asm_sete_r(Asm *self, Reg64 dst);
+void asm_setne_r(Asm *self, Reg64 dst);
+/* Unsigned */
+void asm_setb_r(Asm *self, Reg64 dst);
+/* Unsigned */
+void asm_setbe_r(Asm *self, Reg64 dst);
+/* Unsigned */
+void asm_seta_r(Asm *self, Reg64 dst);
+/* Unsigned */
+void asm_setae_r(Asm *self, Reg64 dst);
+/* Signed */
+void asm_setl_r(Asm *self, Reg64 dst);
+/* Signed */
+void asm_setle_r(Asm *self, Reg64 dst);
+/* Signed */
+void asm_setg_r(Asm *self, Reg64 dst);
+/* Signed */
+void asm_setge_r(Asm *self, Reg64 dst);
 /*
     In x86 assembly, the @relative position starts from the next instruction.
     This offset shouldn't be calculated by the caller and is instead managed
diff --git a/executor/x86_64/executor.c b/executor/x86_64/executor.c
index 6f3c1de..7591dcf 100644
--- a/executor/x86_64/executor.c
+++ b/executor/x86_64/executor.c
@@ -389,10 +389,15 @@ int amal_exec_call(amal_executor *self, u32 code_offset, AmalReg dst_reg) {
     /* TODO: This assumes all arguments are isize */
     /* Do the function call */
     isize asm_offset = asm_get_size(&impl->asm);
-    int num_pushed_stack = impl->num_pushed_values + impl->num_saved_params_for_call - (int)NUM_REG_PARAMS;
+    /* TODO: Do not push */
+    int num_pushed_stack = impl->num_pushed_values;/* + impl->num_saved_params_for_call - (int)NUM_REG_PARAMS;*/
     ASM_ENSURE_CAPACITY
 
-    assert((num_pushed_stack <= 0 || num_pushed_stack % 2 == 0) && "TODO: Align stack to 16-bytes before calling functions");
+    /*assert((num_pushed_stack <= 0 || num_pushed_stack % 2 == 0) && "TODO: Align stack to 16-bytes before calling functions");*/
+    if(num_pushed_stack & 1) {
+        ++num_pushed_stack;
+        asm_sub_rm64_imm(&impl->asm, RSP, sizeof(isize));
+    }
     assert(code_offset < asm_offset);
     asm_call_rel32(&impl->asm, (isize)code_offset - asm_offset);
 
@@ -427,10 +432,15 @@ int amal_exec_calle(amal_executor *self, void *func, AmalReg dst_reg) {
     AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
     AsmOperand rax_op = asm_reg_to_operand(RAX);
     amal_executor_impl *impl = (amal_executor_impl*)self;
-    int num_pushed_stack = impl->num_pushed_values + impl->num_saved_params_for_call - (int)NUM_REG_PARAMS;
+    /* TODO: Do not push */
+    int num_pushed_stack = impl->num_pushed_values;/* + impl->num_saved_params_for_call - (int)NUM_REG_PARAMS;*/
     ASM_ENSURE_CAPACITY
 
-    assert((num_pushed_stack <= 0 || num_pushed_stack % 2 == 0) && "TODO: Align stack to 16-bytes before calling functions");
+    /*assert((num_pushed_stack <= 0 || num_pushed_stack % 2 == 0) && "TODO: Align stack to 16-bytes before calling functions");*/
+    if(num_pushed_stack & 1) {
+        ++num_pushed_stack;
+        asm_sub_rm64_imm(&impl->asm, RSP, sizeof(isize));
+    }
 
     /* TODO: Preserve necessary registers before call? */
     /* TODO: This assumes all arguments are isize */
@@ -449,7 +459,7 @@ int amal_exec_callr(AmalReg dst_reg, BufferView data) {
 }
 */
 
-int amal_exec_cmp(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+int amal_exec_eq(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
     AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
     AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
     AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
@@ -458,9 +468,7 @@ int amal_exec_cmp(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalRe
     amal_executor_impl *impl = (amal_executor_impl*)self;
     ASM_ENSURE_CAPACITY
 
-    asm_mov(&impl->asm, &rcx_op, &dst_op);
     asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
-
     asm_mov(&impl->asm, &rax_op, &src_op1);
     asm_cmp(&impl->asm, &rax_op, &src_op2);
     asm_sete_r(&impl->asm, rcx_op.value.reg);
@@ -468,6 +476,171 @@ int amal_exec_cmp(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalRe
     return 0;
 }
 
+int amal_exec_neq(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setne_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_ilt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setb_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_ile(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setbe_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_igt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_seta_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_ige(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setae_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_lt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setl_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_le(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setle_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_gt(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setg_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_ge(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    AsmOperand rax_op = asm_reg_to_operand(RAX);
+    AsmOperand rcx_op = asm_reg_to_operand(RCX);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_xor_rm64(&impl->asm, rcx_op.value.reg, rcx_op.value.reg);
+    asm_mov(&impl->asm, &rax_op, &src_op1);
+    asm_cmp(&impl->asm, &rax_op, &src_op2);
+    asm_setge_r(&impl->asm, rcx_op.value.reg);
+    asm_mov(&impl->asm, &dst_op, &rcx_op);
+    return 0;
+}
+
+int amal_exec_and(amal_executor *self, AmalReg dst_reg, AmalReg src_reg1, AmalReg src_reg2) {
+    AsmOperand dst_op = amal_reg_to_asm_operand(dst_reg);
+    AsmOperand src_op1 = amal_reg_to_asm_operand(src_reg1);
+    AsmOperand src_op2 = amal_reg_to_asm_operand(src_reg2);
+    amal_executor_impl *impl = (amal_executor_impl*)self;
+    ASM_ENSURE_CAPACITY
+
+    asm_and_rm64(&impl->asm, src_op1.value.reg, src_op2.value.reg);
+    asm_mov(&impl->asm, &dst_op, &src_op1);
+    return 0;
+}
+
 int amal_exec_jz(amal_executor *self, AmalReg reg, u16 target_label) {
     AsmOperand op = amal_reg_to_asm_operand(reg);
     AsmOperand rax_op = asm_reg_to_operand(RAX);