From 9a49f8aec92f7707037f5d1e677078451d07036b Mon Sep 17 00:00:00 2001
From: riperiperi <rhy3756547@hotmail.com>
Date: Wed, 24 Jun 2020 01:43:44 +0100
Subject: [PATCH] Fix VMVN (immediate), Add VPMIN, VPMAX, VMVN (register)
(#1303)
* Add Vmvn (register), tests for both Vmvn variants.
* Add Vpmin, Vpmax, improve Non-FastFp accuracy for Vpadd
* Rebase on top of PTC.
* Add Nopcode
* Increment PTC version.
* Fix nits.
---
ARMeilleure/Decoders/OpCodeTable.cs | 12 ++-
.../Instructions/InstEmitSimdArithmetic32.cs | 62 ++++++++++++-
.../Instructions/InstEmitSimdMove32.cs | 18 +++-
ARMeilleure/Instructions/InstName.cs | 2 +
ARMeilleure/Translation/PTC/Ptc.cs | 2 +-
Ryujinx.Tests/Cpu/CpuTestSimdMov32.cs | 84 +++++++++++++++++-
Ryujinx.Tests/Cpu/CpuTestSimdReg32.cs | 86 ++++++++++++++++---
7 files changed, 243 insertions(+), 23 deletions(-)
diff --git a/ARMeilleure/Decoders/OpCodeTable.cs b/ARMeilleure/Decoders/OpCodeTable.cs
index ec7b8bd91c..8567e1ce69 100644
--- a/ARMeilleure/Decoders/OpCodeTable.cs
+++ b/ARMeilleure/Decoders/OpCodeTable.cs
@@ -704,6 +704,7 @@ namespace ARMeilleure.Decoders
SetA32("<<<<0011111x0000xxxxxxxxxxxxxxxx", InstName.Mvn, InstEmit32.Mvn, typeof(OpCode32AluImm));
SetA32("<<<<0001111x0000xxxxxxxxxxx0xxxx", InstName.Mvn, InstEmit32.Mvn, typeof(OpCode32AluRsImm));
SetA32("<<<<0001111x0000xxxxxxxx0xx1xxxx", InstName.Mvn, InstEmit32.Mvn, typeof(OpCode32AluRsReg));
+ SetA32("<<<<0011001000001111000000000000", InstName.Nop, InstEmit32.Nop, typeof(OpCode32));
SetA32("<<<<0011100xxxxxxxxxxxxxxxxxxxxx", InstName.Orr, InstEmit32.Orr, typeof(OpCode32AluImm));
SetA32("<<<<0001100xxxxxxxxxxxxxxxx0xxxx", InstName.Orr, InstEmit32.Orr, typeof(OpCode32AluRsImm));
SetA32("<<<<0001100xxxxxxxxxxxxx0xx1xxxx", InstName.Orr, InstEmit32.Orr, typeof(OpCode32AluRsReg));
@@ -878,9 +879,10 @@ namespace ARMeilleure.Decoders
SetA32("1111001x1x<<xxxxxxx01010x1x0xxxx", InstName.Vmull, InstEmit32.Vmull_1, typeof(OpCode32SimdRegElemLong));
SetA32("1111001x1x<<xxxxxxx01100x0x0xxxx", InstName.Vmull, InstEmit32.Vmull_I, typeof(OpCode32SimdRegLong));
SetA32("111100101x00xxxxxxx01110x0x0xxxx", InstName.Vmull, InstEmit32.Vmull_I, typeof(OpCode32SimdRegLong)); // Polynomial
- SetA32("1111001x1x000xxxxxxx0xx00x11xxxx", InstName.Vmvn, InstEmit32.Vmvn_I, typeof(OpCode32SimdImm)); // D/Q vector I32.
- SetA32("1111001x1x000xxxxxxx10x00x11xxxx", InstName.Vmvn, InstEmit32.Vmvn_I, typeof(OpCode32SimdImm));
- SetA32("1111001x1x000xxxxxxx110x0x11xxxx", InstName.Vmvn, InstEmit32.Vmvn_I, typeof(OpCode32SimdImm));
+ SetA32("111100111x110000xxxx01011xx0xxxx", InstName.Vmvn, InstEmit32.Vmvn_I, typeof(OpCode32SimdBinary));
+ SetA32("1111001x1x000xxxxxxx0xx00x11xxxx", InstName.Vmvn, InstEmit32.Vmvn_II, typeof(OpCode32SimdImm)); // D/Q vector I32.
+ SetA32("1111001x1x000xxxxxxx10x00x11xxxx", InstName.Vmvn, InstEmit32.Vmvn_II, typeof(OpCode32SimdImm));
+ SetA32("1111001x1x000xxxxxxx110x0x11xxxx", InstName.Vmvn, InstEmit32.Vmvn_II, typeof(OpCode32SimdImm));
SetA32("<<<<11101x110001xxxx101x01x0xxxx", InstName.Vneg, InstEmit32.Vneg_S, typeof(OpCode32SimdS));
SetA32("111100111x11xx01xxxx0x111xx0xxxx", InstName.Vneg, InstEmit32.Vneg_V, typeof(OpCode32Simd));
SetA32("<<<<11100x01xxxxxxxx101xx1x0xxxx", InstName.Vnmla, InstEmit32.Vnmla_S, typeof(OpCode32SimdRegS));
@@ -890,6 +892,10 @@ namespace ARMeilleure.Decoders
SetA32("1111001x1x000xxxxxxx0xx10x01xxxx", InstName.Vorr, InstEmit32.Vorr_II, typeof(OpCode32SimdImm));
SetA32("111100100x<<xxxxxxxx1011x0x1xxxx", InstName.Vpadd, InstEmit32.Vpadd_I, typeof(OpCode32SimdReg));
SetA32("111100110x00xxxxxxxx1101x0x0xxxx", InstName.Vpadd, InstEmit32.Vpadd_V, typeof(OpCode32SimdReg));
+ SetA32("1111001x0x<<xxxxxxxx1010x0x0xxxx", InstName.Vpmax, InstEmit32.Vpmax_I, typeof(OpCode32SimdReg));
+ SetA32("111100110x00xxxxxxxx1111x0x0xxxx", InstName.Vpmax, InstEmit32.Vpmax_V, typeof(OpCode32SimdReg));
+ SetA32("1111001x0x<<xxxxxxxx1010x0x1xxxx", InstName.Vpmin, InstEmit32.Vpmin_I, typeof(OpCode32SimdReg));
+ SetA32("111100110x10xxxxxxxx1111x0x0xxxx", InstName.Vpmin, InstEmit32.Vpmin_V, typeof(OpCode32SimdReg));
SetA32("1111001x1x>>>xxxxxxx100101x1xxx0", InstName.Vqrshrn, InstEmit32.Vqrshrn, typeof(OpCode32SimdShImmNarrow));
SetA32("111100111x>>>xxxxxxx100001x1xxx0", InstName.Vqrshrun, InstEmit32.Vqrshrun, typeof(OpCode32SimdShImmNarrow));
SetA32("111100111x111011xxxx010x0xx0xxxx", InstName.Vrecpe, InstEmit32.Vrecpe, typeof(OpCode32SimdSqrte));
diff --git a/ARMeilleure/Instructions/InstEmitSimdArithmetic32.cs b/ARMeilleure/Instructions/InstEmitSimdArithmetic32.cs
index fdc1bb469f..82f57d63e1 100644
--- a/ARMeilleure/Instructions/InstEmitSimdArithmetic32.cs
+++ b/ARMeilleure/Instructions/InstEmitSimdArithmetic32.cs
@@ -817,7 +817,7 @@ namespace ARMeilleure.Instructions
}
else
{
- EmitVectorPairwiseOpF32(context, (op1, op2) => context.Add(op1, op2));
+ EmitVectorPairwiseOpF32(context, (op1, op2) => EmitSoftFloatCallDefaultFpscr(context, nameof(SoftFloat32.FPAddFpscr), op1, op2));
}
}
@@ -835,6 +835,66 @@ namespace ARMeilleure.Instructions
}
}
+ public static void Vpmax_V(ArmEmitterContext context)
+ {
+ if (Optimizations.FastFP && Optimizations.UseSse2)
+ {
+ EmitSse2VectorPairwiseOpF32(context, Intrinsic.X86Maxps);
+ }
+ else
+ {
+ EmitVectorPairwiseOpF32(context, (op1, op2) => EmitSoftFloatCallDefaultFpscr(context, nameof(SoftFloat64.FPMaxFpscr), op1, op2));
+ }
+ }
+
+ public static void Vpmax_I(ArmEmitterContext context)
+ {
+ OpCode32SimdReg op = (OpCode32SimdReg)context.CurrOp;
+
+ if (Optimizations.UseSsse3)
+ {
+ EmitSsse3VectorPairwiseOp32(context, op.U ? X86PmaxuInstruction : X86PmaxsInstruction);
+ }
+ else
+ {
+ EmitVectorPairwiseOpI32(context, (op1, op2) =>
+ {
+ Operand greater = op.U ? context.ICompareGreaterUI(op1, op2) : context.ICompareGreater(op1, op2);
+ return context.ConditionalSelect(greater, op1, op2);
+ }, !op.U);
+ }
+ }
+
+ public static void Vpmin_V(ArmEmitterContext context)
+ {
+ if (Optimizations.FastFP && Optimizations.UseSse2)
+ {
+ EmitSse2VectorPairwiseOpF32(context, Intrinsic.X86Minps);
+ }
+ else
+ {
+ EmitVectorPairwiseOpF32(context, (op1, op2) => EmitSoftFloatCallDefaultFpscr(context, nameof(SoftFloat32.FPMinFpscr), op1, op2));
+ }
+ }
+
+ public static void Vpmin_I(ArmEmitterContext context)
+ {
+ OpCode32SimdReg op = (OpCode32SimdReg)context.CurrOp;
+
+ if (Optimizations.UseSsse3)
+ {
+ EmitSsse3VectorPairwiseOp32(context, op.U ? X86PminuInstruction : X86PminsInstruction);
+ }
+ else
+ {
+ EmitVectorPairwiseOpI32(context, (op1, op2) =>
+ {
+ Operand greater = op.U ? context.ICompareLessUI(op1, op2) : context.ICompareLess(op1, op2);
+ return context.ConditionalSelect(greater, op1, op2);
+ }, !op.U);
+ }
+ }
+
public static void Vrev(ArmEmitterContext context)
{
OpCode32SimdRev op = (OpCode32SimdRev)context.CurrOp;
diff --git a/ARMeilleure/Instructions/InstEmitSimdMove32.cs b/ARMeilleure/Instructions/InstEmitSimdMove32.cs
index f11f9cc593..b484381fe8 100644
--- a/ARMeilleure/Instructions/InstEmitSimdMove32.cs
+++ b/ARMeilleure/Instructions/InstEmitSimdMove32.cs
@@ -34,7 +34,23 @@ namespace ARMeilleure.Instructions
public static void Vmvn_I(ArmEmitterContext context)
{
- EmitVectorImmUnaryOp32(context, (op1) => context.BitwiseExclusiveOr(op1, op1));
+ if (Optimizations.UseSse2)
+ {
+ EmitVectorUnaryOpSimd32(context, (op1) =>
+ {
+ Operand mask = X86GetAllElements(context, -1L);
+ return context.AddIntrinsic(Intrinsic.X86Pandn, op1, mask);
+ });
+ }
+ else
+ {
+ EmitVectorUnaryOpZx32(context, (op1) => context.BitwiseNot(op1));
+ }
+ }
+
+ public static void Vmvn_II(ArmEmitterContext context)
+ {
+ EmitVectorImmUnaryOp32(context, (op1) => context.BitwiseNot(op1));
}
public static void Vmov_GS(ArmEmitterContext context)
diff --git a/ARMeilleure/Instructions/InstName.cs b/ARMeilleure/Instructions/InstName.cs
index 9bf319aaf5..e4d084560d 100644
--- a/ARMeilleure/Instructions/InstName.cs
+++ b/ARMeilleure/Instructions/InstName.cs
@@ -582,6 +582,8 @@ namespace ARMeilleure.Instructions
Vnmls,
Vorr,
Vpadd,
+ Vpmax,
+ Vpmin,
Vqrshrn,
Vqrshrun,
Vrev,
diff --git a/ARMeilleure/Translation/PTC/Ptc.cs b/ARMeilleure/Translation/PTC/Ptc.cs
index 135a45f035..2b4059ec29 100644
--- a/ARMeilleure/Translation/PTC/Ptc.cs
+++ b/ARMeilleure/Translation/PTC/Ptc.cs
@@ -20,7 +20,7 @@ namespace ARMeilleure.Translation.PTC
{
private const string HeaderMagic = "PTChd";
- private const int InternalVersion = 1; //! To be incremented manually for each change to the ARMeilleure project.
+ private const int InternalVersion = 2; //! To be incremented manually for each change to the ARMeilleure project.
private const string BaseDir = "Ryujinx";
diff --git a/Ryujinx.Tests/Cpu/CpuTestSimdMov32.cs b/Ryujinx.Tests/Cpu/CpuTestSimdMov32.cs
index 8c9627ce0f..aba0f02dbf 100644
--- a/Ryujinx.Tests/Cpu/CpuTestSimdMov32.cs
+++ b/Ryujinx.Tests/Cpu/CpuTestSimdMov32.cs
@@ -39,8 +39,8 @@ namespace Ryujinx.Tests.Cpu
0b1110_1
};
-
uint opcode = 0xf2800010u; // VMOV.I32 D0, #0
+
uint cmodeOp = variants[variant];
if (q)
@@ -49,11 +49,11 @@ namespace Ryujinx.Tests.Cpu
}
opcode |= ((cmodeOp & 1) << 5) | ((cmodeOp & 0x1e) << 7);
- opcode |= ((q ? 1u : 0u) << 6);
+ opcode |= (q ? 1u : 0u) << 6;
opcode |= (imm & 0xf) | ((imm & 0x70) << 12) | ((imm & 0x80) << 16);
- opcode |= ((vd & 0x10) << 18);
- opcode |= ((vd & 0xf) << 12);
+ opcode |= (vd & 0x10) << 18;
+ opcode |= (vd & 0xf) << 12;
SingleOpcode(opcode);
@@ -258,6 +258,82 @@ namespace Ryujinx.Tests.Cpu
CompareAgainstUnicorn();
}
+ [Test, Pairwise, Description("VMVN.<size> <Vt>, <Vm>")]
+ public void Vmvn([Range(0u, 1u, 2u)] uint size,
+ [Values(0u, 1u, 2u, 3u)] uint vd,
+ [Values(0u, 2u, 4u, 8u)] uint vm,
+ [Values] bool q)
+ {
+ uint opcode = 0xf3b00580u; // VMVN D0, D0
+
+ if (q)
+ {
+ opcode |= 1 << 6;
+ vm <<= 1;
+ vd <<= 1;
+ }
+
+ opcode |= (size & 0x3) << 18;
+ opcode |= (vm & 0x10) << 1;
+ opcode |= (vm & 0xf) << 0;
+
+ opcode |= (vd & 0x10) << 18;
+ opcode |= (vd & 0xf) << 12;
+
+ V128 v0 = new V128(TestContext.CurrentContext.Random.NextULong(), TestContext.CurrentContext.Random.NextULong());
+ V128 v1 = new V128(TestContext.CurrentContext.Random.NextULong(), TestContext.CurrentContext.Random.NextULong());
+ V128 v2 = new V128(TestContext.CurrentContext.Random.NextULong(), TestContext.CurrentContext.Random.NextULong());
+ V128 v3 = new V128(TestContext.CurrentContext.Random.NextULong(), TestContext.CurrentContext.Random.NextULong());
+
+ SingleOpcode(opcode, v0: v0, v1: v1, v2: v2, v3: v3);
+
+ CompareAgainstUnicorn();
+ }
+
+ [Test, Pairwise, Description("VMVN.I<size> <Dd/Qd>, #<imm>")]
+ public void Mvni_V([Range(0u, 7u)] uint variant,
+ [Values(0u, 1u, 2u, 3u)] uint vd,
+ [Values(0x0u)] [Random(1u, 0xffu, RndCntImm)] uint imm,
+ [Values] bool q)
+ {
+ uint[] variants =
+ {
+ // I32
+ 0b0000,
+ 0b0010,
+ 0b0100,
+ 0b0110,
+
+ // I16
+ 0b1000,
+ 0b1010,
+
+ // I32
+ 0b1100,
+ 0b1101,
+ };
+
+ uint opcode = 0xf2800030u; // VMVN.I32 D0, #0
+
+ uint cmodeOp = variants[variant];
+
+ if (q)
+ {
+ vd <<= 1;
+ }
+
+ opcode |= (cmodeOp & 0xf) << 8;
+ opcode |= (q ? 1u : 0u) << 6;
+ opcode |= (imm & 0xf) | ((imm & 0x70) << 12) | ((imm & 0x80) << 16);
+
+ opcode |= (vd & 0x10) << 18;
+ opcode |= (vd & 0xf) << 12;
+
+ SingleOpcode(opcode);
+
+ CompareAgainstUnicorn();
+ }
+
[Test, Pairwise, Description("VTRN.<size> <Vd>, <Vm>")]
public void Vtrn([Values(0u, 1u, 2u, 3u)] uint vm,
[Values(0u, 1u, 2u, 3u)] uint vd,
diff --git a/Ryujinx.Tests/Cpu/CpuTestSimdReg32.cs b/Ryujinx.Tests/Cpu/CpuTestSimdReg32.cs
index 66db63bc49..1581e85044 100644
--- a/Ryujinx.Tests/Cpu/CpuTestSimdReg32.cs
+++ b/Ryujinx.Tests/Cpu/CpuTestSimdReg32.cs
@@ -12,6 +12,31 @@ namespace Ryujinx.Tests.Cpu
{
#if SimdReg32
+#region "ValueSource (Opcodes)"
+ private static uint[] _Vp_Add_Max_Min_F_()
+ {
+ return new uint[]
+ {
+ 0xf3000d00u, // VPADD.F32 D0, D0, D0
+ 0xf3000f00u, // VPMAX.F32 D0, D0, D0
+ 0xf3200f00u // VPMIN.F32 D0, D0, D0
+ };
+ }
+
+ // VPADD does not have an unsigned flag, so we check the opcode before setting it.
+ private static uint VpaddI8 = 0xf2000b10u; // VPADD.I8 D0, D0, D0
+
+ private static uint[] _Vp_Add_Max_Min_I_()
+ {
+ return new uint[]
+ {
+ VpaddI8,
+ 0xf2000a00u, // VPMAX.S8 D0, D0, D0
+ 0xf2000a10u // VPMIN.S8 D0, D0, D0
+ };
+ }
+#endregion
+
#region "ValueSource (Types)"
private static ulong[] _1B1H1S1D_()
{
@@ -296,7 +321,7 @@ namespace Ryujinx.Tests.Cpu
{
uint opcode = 0xf2800a00u; // VMLSL.S8 Q0, D0, D0
- opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
+ opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((rn & 0xf) << 16) | ((rn & 0x10) << 3);
@@ -329,7 +354,7 @@ namespace Ryujinx.Tests.Cpu
{
uint opcode = 0xf2800c00u; // VMULL.S8 Q0, D0, D0
- opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
+ opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((rn & 0xf) << 16) | ((rn & 0x10) << 3);
@@ -381,7 +406,7 @@ namespace Ryujinx.Tests.Cpu
opcode |= 1 << 24;
}
- opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
+ opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((rn & 0xf) << 16) | ((rn & 0x10) << 3);
@@ -397,22 +422,57 @@ namespace Ryujinx.Tests.Cpu
}
[Explicit]
- [Test, Pairwise, Description("VPADD.f32 V0, V0, V0")]
- public void Vpadd_f32([Values(0u)] uint rd,
- [Range(0u, 7u)] uint rn,
- [Range(0u, 7u)] uint rm)
+ [Test, Pairwise]
+ public void Vp_Add_Max_Min_F([ValueSource("_Vp_Add_Max_Min_F_")] uint opcode,
+ [Values(0u)] uint rd,
+ [Range(0u, 7u)] uint rn,
+ [Range(0u, 7u)] uint rm,
+ [ValueSource("_2S_F_")] ulong z0,
+ [ValueSource("_2S_F_")] ulong z1,
+ [ValueSource("_2S_F_")] ulong a0,
+ [ValueSource("_2S_F_")] ulong a1,
+ [ValueSource("_2S_F_")] ulong b0,
+ [ValueSource("_2S_F_")] ulong b1)
{
- // not currently a slow path test - just a sanity check for pairwise
- uint opcode = 0xf3000d00u; // VPADD.F32 D0, D0, D0
-
opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((rn & 0xf) << 16) | ((rn & 0x10) << 3);
var rnd = TestContext.CurrentContext.Random;
- V128 v0 = new V128(rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue));
- V128 v1 = new V128(rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue));
- V128 v2 = new V128(rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue), rnd.NextFloat(int.MinValue, int.MaxValue));
+ V128 v0 = MakeVectorE0E1(z0, z1);
+ V128 v1 = MakeVectorE0E1(a0, a1);
+ V128 v2 = MakeVectorE0E1(b0, b1);
+
+ SingleOpcode(opcode, v0: v0, v1: v1, v2: v2);
+
+ CompareAgainstUnicorn();
+ }
+
+ [Test, Pairwise]
+ public void Vp_Add_Max_Min_I([ValueSource("_Vp_Add_Max_Min_I_")] uint opcode,
+ [Values(0u)] uint rd,
+ [Range(0u, 5u)] uint rn,
+ [Range(0u, 5u)] uint rm,
+ [Values(0u, 1u, 2u)] uint size,
+ [Random(RndCnt)] ulong z,
+ [Random(RndCnt)] ulong a,
+ [Random(RndCnt)] ulong b,
+ [Values] bool u)
+ {
+ if (u && opcode != VpaddI8)
+ {
+ opcode |= 1 << 24;
+ }
+
+ opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
+ opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
+ opcode |= ((rn & 0xf) << 16) | ((rn & 0x10) << 3);
+
+ opcode |= size << 20;
+
+ V128 v0 = MakeVectorE0E1(z, z);
+ V128 v1 = MakeVectorE0E1(a, z);
+ V128 v2 = MakeVectorE0E1(b, z);
SingleOpcode(opcode, v0: v0, v1: v1, v2: v2);