zdoom/qzdoom
0
0
Fork 0
mirror of https://github.com/ZDoom/qzdoom.git synced 2024-11-14 08:30:49 +00:00
Code Issues Projects Releases Packages Wiki Activity

Merge pull request #538 from Gutawer/asmjit

Browse source 
Added all integer comparison opcodes
This commit is contained in:
Magnus Norddahl 2018-08-14 16:47:18 +02:00 committed by GitHub
commit f862b96763
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 177 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

190
src/scripting/vm/jit.cpp
View file

@ -8,6 +8,7 @@
#include <asmjit/asmjit.h>
#include <asmjit/x86.h>
#include <functional>
class AsmJitException : public std::exception
{
@ -221,6 +222,29 @@ static bool CanJit(VMScriptFunction *sfunc)
return true;
}
template <typename Func>
void emitComparisonOpcode(asmjit::X86Compiler& cc, const TArray<asmjit::Label>& labels, const VMOP* pc, int i, Func compFunc) {
using namespace asmjit;
Label elseLabel = cc.newLabel();
auto tmp0 = cc.newInt32();
auto tmp1 = cc.newInt32();
compFunc(tmp0);
cc.mov(tmp1, A);
cc.and_(tmp1, CMP_CHECK);
cc.cmp(tmp0, tmp1);
cc.jne(elseLabel);
cc.jmp(labels[i + 2 + JMPOFS(pc + 1)]);
cc.bind(elseLabel);
cc.jmp(labels[i + 2]);
}
JitFuncPtr JitCompile(VMScriptFunction *sfunc)
{
#if 0 // For debugging
@ -309,6 +333,11 @@ JitFuncPtr JitCompile(VMScriptFunction *sfunc)
*/
int size = sfunc->CodeSize;
TArray<Label> labels(size, true);
for (int i = 0; i < size; i++) labels[i] = cc.newLabel();
for (int i = 0; i < size; i++)
{
const VMOP *pc = sfunc->Code + i;
@ -316,6 +345,8 @@ JitFuncPtr JitCompile(VMScriptFunction *sfunc)
int a = pc->a;
int b;// , c;
cc.bind(labels[i]);
switch (op)
{
default:
@ -905,20 +936,153 @@ JitFuncPtr JitCompile(VMScriptFunction *sfunc)
cc.not_(regD[a]);
break;
case OP_EQ_R: // if ((dB == dkC) != A) then pc++
case OP_EQ_K:
case OP_LT_RR: // if ((dkB < dkC) != A) then pc++
case OP_LT_RK:
case OP_LT_KR:
case OP_LE_RR: // if ((dkB <= dkC) != A) then pc++
case OP_LE_RK:
case OP_LE_KR:
case OP_LTU_RR: // if ((dkB < dkC) != A) then pc++ -- unsigned
case OP_LTU_RK:
case OP_LTU_KR:
case OP_LEU_RR: // if ((dkB <= dkC) != A) then pc++ -- unsigned
case OP_LEU_RK:
case OP_LEU_KR:
{
auto compLambda = [&] (X86Gp& result) {
cc.cmp(regD[B], regD[C]);
cc.sete(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_EQ_K:
{
auto compLambda = [&] (X86Gp& result) {
cc.cmp(regD[B], konstd[C]);
cc.sete(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LT_RR: // if ((dkB < dkC) != A) then pc++
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], regD[C]);
cc.setl(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LT_RK:
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], konstd[C]);
cc.setl(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LT_KR:
{
auto compLambda = [&](X86Gp& result) {
auto tmp = cc.newIntPtr();
cc.mov(tmp, reinterpret_cast<ptrdiff_t>(&(konstd[C])));
cc.cmp(x86::ptr(tmp), regD[B]);
cc.setl(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LE_RR: // if ((dkB <= dkC) != A) then pc++
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], regD[C]);
cc.setle(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LE_RK:
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], konstd[C]);
cc.setle(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LE_KR:
{
auto compLambda = [&](X86Gp& result) {
auto tmp = cc.newIntPtr();
cc.mov(tmp, reinterpret_cast<ptrdiff_t>(&(konstd[C])));
cc.cmp(x86::ptr(tmp), regD[B]);
cc.setle(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LTU_RR: // if ((dkB < dkC) != A) then pc++ -- unsigned
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], regD[C]);
cc.setb(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LTU_RK:
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], konstd[C]);
cc.setb(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LTU_KR:
{
auto compLambda = [&](X86Gp& result) {
auto tmp = cc.newIntPtr();
cc.mov(tmp, reinterpret_cast<ptrdiff_t>(&(konstd[C])));
cc.cmp(x86::ptr(tmp), regD[B]);
cc.setb(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LEU_RR: // if ((dkB <= dkC) != A) then pc++ -- unsigned
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], regD[C]);
cc.setbe(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LEU_RK:
{
auto compLambda = [&](X86Gp& result) {
cc.cmp(regD[B], konstd[C]);
cc.setbe(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
case OP_LEU_KR:
{
auto compLambda = [&](X86Gp& result) {
auto tmp = cc.newIntPtr();
cc.mov(tmp, reinterpret_cast<ptrdiff_t>(&(konstd[C])));
cc.cmp(x86::ptr(tmp), regD[B]);
cc.setbe(result);
};
emitComparisonOpcode(cc, labels, pc, i, compLambda);
break;
}
// Double-precision floating point math.
case OP_ADDF_RR: // fA = fB + fkC