rjx-mirror/Ryujinx.Tests/Cpu/CpuTestSimdShImm32.cs
LDj3SNuD e36e97c64d
CPU: This PR fixes Fpscr, among other things. (#1433)
* CPU: This PR fixes Fpscr, among other things.

* Add Fpscr.Qc = 1 if sat. for Vqrshrn & Vqrshrun.

* Fix Vcmp & Vcmpe opcode table.

* Revert "Fix Vcmp & Vcmpe opcode table."

This reverts commit c117d9410d.

* Address PR feedbacks.
2020-08-08 17:18:51 +02:00

298 lines
11 KiB
C#

#define SimdShImm32
using ARMeilleure.State;
using NUnit.Framework;
namespace Ryujinx.Tests.Cpu
{
[Category("SimdShImm32")]
public sealed class CpuTestSimdShImm32 : CpuTest32
{
#if SimdShImm32
#region "ValueSource (Types)"
private static ulong[] _1D_()
{
return new ulong[] { 0x0000000000000000ul, 0x7FFFFFFFFFFFFFFFul,
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
}
private static ulong[] _2S_()
{
return new ulong[] { 0x0000000000000000ul, 0x7FFFFFFF7FFFFFFFul,
0x8000000080000000ul, 0xFFFFFFFFFFFFFFFFul };
}
private static ulong[] _4H_()
{
return new ulong[] { 0x0000000000000000ul, 0x7FFF7FFF7FFF7FFFul,
0x8000800080008000ul, 0xFFFFFFFFFFFFFFFFul };
}
private static ulong[] _8B_()
{
return new ulong[] { 0x0000000000000000ul, 0x7F7F7F7F7F7F7F7Ful,
0x8080808080808080ul, 0xFFFFFFFFFFFFFFFFul };
}
#endregion
#region "ValueSource (Opcodes)"
private static uint[] _Vshr_Imm_SU8_()
{
return new uint[]
{
0xf2880110u, // VSRA.S8 D0, D0, #8
0xf2880210u, // VRSHR.S8 D0, D0, #8
0xf2880010u // VSHR.S8 D0, D0, #8
};
}
private static uint[] _Vshr_Imm_SU16_()
{
return new uint[]
{
0xf2900110u, // VSRA.S16 D0, D0, #16
0xf2900210u, // VRSHR.S16 D0, D0, #16
0xf2900010u // VSHR.S16 D0, D0, #16
};
}
private static uint[] _Vshr_Imm_SU32_()
{
return new uint[]
{
0xf2a00110u, // VSRA.S32 D0, D0, #32
0xf2a00210u, // VRSHR.S32 D0, D0, #32
0xf2a00010u // VSHR.S32 D0, D0, #32
};
}
private static uint[] _Vshr_Imm_SU64_()
{
return new uint[]
{
0xf2800190u, // VSRA.S64 D0, D0, #64
0xf2800290u, // VRSHR.S64 D0, D0, #64
0xf2800090u // VSHR.S64 D0, D0, #64
};
}
#endregion
private const int RndCnt = 2;
private const int RndCntShiftImm = 2;
[Test, Pairwise]
public void Vshr_Imm_SU8([ValueSource("_Vshr_Imm_SU8_")] uint opcode,
[Range(0u, 3u)] uint rd,
[Range(0u, 3u)] uint rm,
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
[ValueSource("_8B_")] [Random(RndCnt)] ulong b,
[Values(1u, 8u)] [Random(2u, 7u, RndCntShiftImm)] uint shiftImm,
[Values] bool u,
[Values] bool q)
{
uint imm6 = 16 - shiftImm;
Vshr_Imm_SU(opcode, rd, rm, z, b, imm6, u, q);
}
[Test, Pairwise]
public void Vshr_Imm_SU16([ValueSource("_Vshr_Imm_SU16_")] uint opcode,
[Range(0u, 3u)] uint rd,
[Range(0u, 3u)] uint rm,
[ValueSource("_4H_")] [Random(RndCnt)] ulong z,
[ValueSource("_4H_")] [Random(RndCnt)] ulong b,
[Values(1u, 16u)] [Random(2u, 15u, RndCntShiftImm)] uint shiftImm,
[Values] bool u,
[Values] bool q)
{
uint imm6 = 32 - shiftImm;
Vshr_Imm_SU(opcode, rd, rm, z, b, imm6, u, q);
}
[Test, Pairwise]
public void Vshr_Imm_SU32([ValueSource("_Vshr_Imm_SU32_")] uint opcode,
[Range(0u, 3u)] uint rd,
[Range(0u, 3u)] uint rm,
[ValueSource("_2S_")] [Random(RndCnt)] ulong z,
[ValueSource("_2S_")] [Random(RndCnt)] ulong b,
[Values(1u, 32u)] [Random(2u, 31u, RndCntShiftImm)] uint shiftImm,
[Values] bool u,
[Values] bool q)
{
uint imm6 = 64 - shiftImm;
Vshr_Imm_SU(opcode, rd, rm, z, b, imm6, u, q);
}
[Test, Pairwise]
public void Vshr_Imm_SU64([ValueSource("_Vshr_Imm_SU64_")] uint opcode,
[Range(0u, 3u)] uint rd,
[Range(0u, 3u)] uint rm,
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
[ValueSource("_1D_")] [Random(RndCnt)] ulong b,
[Values(1u, 64u)] [Random(2u, 63u, RndCntShiftImm)] uint shiftImm,
[Values] bool u,
[Values] bool q)
{
uint imm6 = 64 - shiftImm;
Vshr_Imm_SU(opcode, rd, rm, z, b, imm6, u, q);
}
private void Vshr_Imm_SU(uint opcode, uint rd, uint rm, ulong z, ulong b, uint imm6, bool u, bool q)
{
if (u)
{
opcode |= 1 << 24;
}
if (q)
{
opcode |= 1 << 6;
rd >>= 1; rd <<= 1;
rm >>= 1; rm <<= 1;
}
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= (imm6 & 0x3f) << 16;
V128 v0 = MakeVectorE0E1(z, ~z);
V128 v1 = MakeVectorE0E1(b, ~b);
SingleOpcode(opcode, v0: v0, v1: v1);
CompareAgainstUnicorn();
}
[Test, Pairwise, Description("VSHL.<size> {<Vd>}, <Vm>, #<imm>")]
public void Vshl_Imm([Values(0u)] uint rd,
[Values(2u, 0u)] uint rm,
[Values(0u, 1u, 2u, 3u)] uint size,
[Random(RndCntShiftImm)] [Values(0u)] uint shiftImm,
[Random(RndCnt)] ulong z,
[Random(RndCnt)] ulong a,
[Random(RndCnt)] ulong b,
[Values] bool q)
{
uint opcode = 0xf2800510u; // VORR.I32 D0, #0 (immediate value changes it into SHL)
if (q)
{
opcode |= 1 << 6;
rm <<= 1;
rd <<= 1;
}
uint imm = 1u << ((int)size + 3);
imm |= shiftImm & (imm - 1);
opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((imm & 0x3f) << 16) | ((imm & 0x40) << 1);
V128 v0 = MakeVectorE0E1(z, z);
V128 v1 = MakeVectorE0E1(a, z);
V128 v2 = MakeVectorE0E1(b, z);
SingleOpcode(opcode, v0: v0, v1: v1, v2: v2);
CompareAgainstUnicorn();
}
[Test, Pairwise, Description("VSHRN.<size> <Vd>, <Vm>, #<imm>")]
public void Vshrn_Imm([Values(0u, 1u)] uint rd,
[Values(2u, 0u)] uint rm,
[Values(0u, 1u, 2u)] uint size,
[Random(RndCntShiftImm)] [Values(0u)] uint shiftImm,
[Random(RndCnt)] ulong z,
[Random(RndCnt)] ulong a,
[Random(RndCnt)] ulong b)
{
uint opcode = 0xf2800810u; // VMOV.I16 D0, #0 (immediate value changes it into SHRN)
uint imm = 1u << ((int)size + 3);
imm |= shiftImm & (imm - 1);
opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((imm & 0x3f) << 16);
V128 v0 = MakeVectorE0E1(z, z);
V128 v1 = MakeVectorE0E1(a, z);
V128 v2 = MakeVectorE0E1(b, z);
SingleOpcode(opcode, v0: v0, v1: v1, v2: v2);
CompareAgainstUnicorn();
}
[Test, Pairwise, Description("VQRSHRN.<type><size> <Vd>, <Vm>, #<imm>")]
public void Vqrshrn_Imm([Values(0u, 1u)] uint rd,
[Values(2u, 0u)] uint rm,
[Values(0u, 1u, 2u)] uint size,
[Random(RndCntShiftImm)] [Values(0u)] uint shiftImm,
[Random(RndCnt)] ulong z,
[Random(RndCnt)] ulong a,
[Random(RndCnt)] ulong b,
[Values] bool u)
{
uint opcode = 0xf2800950u; // VORR.I16 Q0, #0 (immediate value changes it into QRSHRN)
uint imm = 1u << ((int)size + 3);
imm |= shiftImm & (imm - 1);
opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((imm & 0x3f) << 16);
if (u)
{
opcode |= 1u << 24;
}
V128 v0 = MakeVectorE0E1(z, z);
V128 v1 = MakeVectorE0E1(a, z);
V128 v2 = MakeVectorE0E1(b, z);
int fpscr = (int)TestContext.CurrentContext.Random.NextUInt() & (int)Fpsr.Qc;
SingleOpcode(opcode, v0: v0, v1: v1, v2: v2, fpscr: fpscr);
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
}
[Test, Pairwise, Description("VQRSHRUN.<type><size> <Vd>, <Vm>, #<imm>")]
public void Vqrshrun_Imm([Values(0u, 1u)] uint rd,
[Values(2u, 0u)] uint rm,
[Values(0u, 1u, 2u)] uint size,
[Random(RndCntShiftImm)] [Values(0u)] uint shiftImm,
[Random(RndCnt)] ulong z,
[Random(RndCnt)] ulong a,
[Random(RndCnt)] ulong b)
{
uint opcode = 0xf3800850u; // VMOV.I16 Q0, #0x80 (immediate value changes it into QRSHRUN)
uint imm = 1u << ((int)size + 3);
imm |= shiftImm & (imm - 1);
opcode |= ((rm & 0xf) << 0) | ((rm & 0x10) << 1);
opcode |= ((rd & 0xf) << 12) | ((rd & 0x10) << 18);
opcode |= ((imm & 0x3f) << 16);
V128 v0 = MakeVectorE0E1(z, z);
V128 v1 = MakeVectorE0E1(a, z);
V128 v2 = MakeVectorE0E1(b, z);
int fpscr = (int)TestContext.CurrentContext.Random.NextUInt() & (int)Fpsr.Qc;
SingleOpcode(opcode, v0: v0, v1: v1, v2: v2, fpscr: fpscr);
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
}
#endif
}
}