piwalker/dolphin
1
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2024-11-15 05:47:56 -07:00
Code Issues Actions Packages Projects Releases Wiki Activity

JitArm64: Track single precision in ps_maddXX.

Browse Source
This commit is contained in:
degasus 2016-02-11 11:51:48 +01:00
parent 7fd68c8761
commit c8948ff8c7
1 changed files with 30 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

52
Source/Core/Core/PowerPC/JitArm64/JitArm64_Paired.cpp
View File

@ -105,41 +105,49 @@ void JitArm64::ps_maddXX(UGeckoInstruction inst)
u32 a = inst.FA, b = inst.FB, c = inst.FC, d = inst.FD; u32 a = inst.FA, b = inst.FB, c = inst.FC, d = inst.FD;
u32 op5 = inst.SUBOP5; u32 op5 = inst.SUBOP5;
ARM64Reg VA = fpr.R(a, REG_REG); bool singles = fpr.IsSingle(a) && fpr.IsSingle(b) && fpr.IsSingle(c);
ARM64Reg VB = fpr.R(b, REG_REG); RegType type = singles ? REG_REG_SINGLE : REG_REG;
ARM64Reg VC = fpr.R(c, REG_REG); u8 size = singles ? 32 : 64;
ARM64Reg VD = fpr.RW(d, REG_REG); ARM64Reg (*reg_encoder)(ARM64Reg) = singles ? EncodeRegToDouble : EncodeRegToQuad;
ARM64Reg V0 = fpr.GetReg();
ARM64Reg VA = reg_encoder(fpr.R(a, type));
ARM64Reg VB = reg_encoder(fpr.R(b, type));
ARM64Reg VC = reg_encoder(fpr.R(c, type));
ARM64Reg VD = reg_encoder(fpr.RW(d, type));
ARM64Reg V0Q = fpr.GetReg();
ARM64Reg V0 = reg_encoder(V0Q);
// TODO: Do FMUL and FADD/FSUB in *one* host call to save accuracy.
switch (op5) switch (op5)
{ {
case 14: // ps_madds0 case 14: // ps_madds0
m_float_emit.DUP(64, V0, VC, 0); m_float_emit.DUP(size, V0, VC, 0);
m_float_emit.FMUL(64, V0, V0, VA); m_float_emit.FMUL(size, V0, V0, VA);
m_float_emit.FADD(64, VD, V0, VB); m_float_emit.FADD(size, VD, V0, VB);
break; break;
case 15: // ps_madds1 case 15: // ps_madds1
m_float_emit.DUP(64, V0, VC, 1); m_float_emit.DUP(size, V0, VC, 1);
m_float_emit.FMUL(64, V0, V0, VA); m_float_emit.FMUL(size, V0, V0, VA);
m_float_emit.FADD(64, VD, V0, VB); m_float_emit.FADD(size, VD, V0, VB);
break; break;
case 28: // ps_msub case 28: // ps_msub
m_float_emit.FMUL(64, V0, VA, VC); m_float_emit.FMUL(size, V0, VA, VC);
m_float_emit.FSUB(64, VD, V0, VB); m_float_emit.FSUB(size, VD, V0, VB);
break; break;
case 29: // ps_madd case 29: // ps_madd
m_float_emit.FMUL(64, V0, VA, VC); m_float_emit.FMUL(size, V0, VA, VC);
m_float_emit.FADD(64, VD, V0, VB); m_float_emit.FADD(size, VD, V0, VB);
break; break;
case 30: // ps_nmsub case 30: // ps_nmsub
m_float_emit.FMUL(64, V0, VA, VC); m_float_emit.FMUL(size, V0, VA, VC);
m_float_emit.FSUB(64, VD, V0, VB); m_float_emit.FSUB(size, VD, V0, VB);
m_float_emit.FNEG(64, VD, VD); m_float_emit.FNEG(size, VD, VD);
break; break;
case 31: // ps_nmadd case 31: // ps_nmadd
m_float_emit.FMUL(64, V0, VA, VC); m_float_emit.FMUL(size, V0, VA, VC);
m_float_emit.FADD(64, VD, V0, VB); m_float_emit.FADD(size, VD, V0, VB);
m_float_emit.FNEG(64, VD, VD); m_float_emit.FNEG(size, VD, VD);
break; break;
default: default:
_assert_msg_(DYNA_REC, 0, "ps_madd - invalid op"); _assert_msg_(DYNA_REC, 0, "ps_madd - invalid op");
@ -147,7 +155,7 @@ void JitArm64::ps_maddXX(UGeckoInstruction inst)
} }
fpr.FixSinglePrecision(d); fpr.FixSinglePrecision(d);
fpr.Unlock(V0); fpr.Unlock(V0Q);
} }
void JitArm64::ps_res(UGeckoInstruction inst) void JitArm64::ps_res(UGeckoInstruction inst)