mirror of
https://github.com/Ryujinx/Ryujinx.git
synced 2024-10-01 12:30:00 +02:00
CPU: Produce non-inf results for RSQRTE instruction with subnormal inputs (#6634)
* CPU: Produce non-inf results for RSQRTE instruction with subnormal inputs * PPTC version bump
This commit is contained in:
parent
338ff79e1e
commit
543d75a587
2 changed files with 11 additions and 3 deletions
|
@ -2426,7 +2426,11 @@ namespace ARMeilleure.Instructions
|
||||||
}
|
}
|
||||||
else if (Optimizations.FastFP && Optimizations.UseSse41 && sizeF == 0)
|
else if (Optimizations.FastFP && Optimizations.UseSse41 && sizeF == 0)
|
||||||
{
|
{
|
||||||
Operand res = EmitSse41Round32Exp8OpF(context, context.AddIntrinsic(Intrinsic.X86Rsqrtss, GetVec(op.Rn)), scalar: true);
|
// RSQRTSS handles subnormals as zero, which differs from Arm, so we can't use it here.
|
||||||
|
|
||||||
|
Operand res = context.AddIntrinsic(Intrinsic.X86Sqrtss, GetVec(op.Rn));
|
||||||
|
res = context.AddIntrinsic(Intrinsic.X86Rcpss, res);
|
||||||
|
res = EmitSse41Round32Exp8OpF(context, res, scalar: true);
|
||||||
|
|
||||||
context.Copy(GetVec(op.Rd), context.VectorZeroUpper96(res));
|
context.Copy(GetVec(op.Rd), context.VectorZeroUpper96(res));
|
||||||
}
|
}
|
||||||
|
@ -2451,7 +2455,11 @@ namespace ARMeilleure.Instructions
|
||||||
}
|
}
|
||||||
else if (Optimizations.FastFP && Optimizations.UseSse41 && sizeF == 0)
|
else if (Optimizations.FastFP && Optimizations.UseSse41 && sizeF == 0)
|
||||||
{
|
{
|
||||||
Operand res = EmitSse41Round32Exp8OpF(context, context.AddIntrinsic(Intrinsic.X86Rsqrtps, GetVec(op.Rn)), scalar: false);
|
// RSQRTPS handles subnormals as zero, which differs from Arm, so we can't use it here.
|
||||||
|
|
||||||
|
Operand res = context.AddIntrinsic(Intrinsic.X86Sqrtps, GetVec(op.Rn));
|
||||||
|
res = context.AddIntrinsic(Intrinsic.X86Rcpps, res);
|
||||||
|
res = EmitSse41Round32Exp8OpF(context, res, scalar: false);
|
||||||
|
|
||||||
if (op.RegisterSize == RegisterSize.Simd64)
|
if (op.RegisterSize == RegisterSize.Simd64)
|
||||||
{
|
{
|
||||||
|
|
|
@ -29,7 +29,7 @@ namespace ARMeilleure.Translation.PTC
|
||||||
private const string OuterHeaderMagicString = "PTCohd\0\0";
|
private const string OuterHeaderMagicString = "PTCohd\0\0";
|
||||||
private const string InnerHeaderMagicString = "PTCihd\0\0";
|
private const string InnerHeaderMagicString = "PTCihd\0\0";
|
||||||
|
|
||||||
private const uint InternalVersion = 6613; //! To be incremented manually for each change to the ARMeilleure project.
|
private const uint InternalVersion = 6634; //! To be incremented manually for each change to the ARMeilleure project.
|
||||||
|
|
||||||
private const string ActualDir = "0";
|
private const string ActualDir = "0";
|
||||||
private const string BackupDir = "1";
|
private const string BackupDir = "1";
|
||||||
|
|
Loading…
Reference in a new issue