0
0
Fork 0

Fix shader FSWZADD instruction (#4069)

* Fix shader FSWZADD instruction

* Shader cache version bump
This commit is contained in:
gdkchan 2022-12-08 14:08:07 -03:00 committed by GitHub
parent 9a0330f7f8
commit 8428bb6541
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 6 additions and 3 deletions

View file

@ -22,7 +22,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
private const ushort FileFormatVersionMajor = 1; private const ushort FileFormatVersionMajor = 1;
private const ushort FileFormatVersionMinor = 2; private const ushort FileFormatVersionMinor = 2;
private const uint FileFormatVersionPacked = ((uint)FileFormatVersionMajor << 16) | FileFormatVersionMinor; private const uint FileFormatVersionPacked = ((uint)FileFormatVersionMajor << 16) | FileFormatVersionMinor;
private const uint CodeGenVersion = 4067; private const uint CodeGenVersion = 4069;
private const string SharedTocFileName = "shared.toc"; private const string SharedTocFileName = "shared.toc";
private const string SharedDataFileName = "shared.data"; private const string SharedDataFileName = "shared.data";

View file

@ -2,6 +2,6 @@ float Helper_SwizzleAdd(float x, float y, int mask)
{ {
vec4 xLut = vec4(1.0, -1.0, 1.0, 0.0); vec4 xLut = vec4(1.0, -1.0, 1.0, 0.0);
vec4 yLut = vec4(1.0, 1.0, -1.0, 1.0); vec4 yLut = vec4(1.0, 1.0, -1.0, 1.0);
int lutIdx = mask >> int($SUBGROUP_INVOCATION$ & 3u) * 2; int lutIdx = (mask >> (int($SUBGROUP_INVOCATION$ & 3u) * 2)) & 3;
return x * xLut[lutIdx] + y * yLut[lutIdx]; return x * xLut[lutIdx] + y * yLut[lutIdx];
} }

View file

@ -1449,10 +1449,13 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
var xLut = context.ConstantComposite(v4float, one, minusOne, one, zero); var xLut = context.ConstantComposite(v4float, one, minusOne, one, zero);
var yLut = context.ConstantComposite(v4float, one, one, minusOne, one); var yLut = context.ConstantComposite(v4float, one, one, minusOne, one);
var three = context.Constant(context.TypeU32(), 3);
var threadId = context.GetAttribute(AggregateType.U32, AttributeConsts.LaneId, false); var threadId = context.GetAttribute(AggregateType.U32, AttributeConsts.LaneId, false);
var shift = context.BitwiseAnd(context.TypeU32(), threadId, context.Constant(context.TypeU32(), 3)); var shift = context.BitwiseAnd(context.TypeU32(), threadId, three);
shift = context.ShiftLeftLogical(context.TypeU32(), shift, context.Constant(context.TypeU32(), 1)); shift = context.ShiftLeftLogical(context.TypeU32(), shift, context.Constant(context.TypeU32(), 1));
var lutIdx = context.ShiftRightLogical(context.TypeU32(), mask, shift); var lutIdx = context.ShiftRightLogical(context.TypeU32(), mask, shift);
lutIdx = context.BitwiseAnd(context.TypeU32(), lutIdx, three);
var xLutValue = context.VectorExtractDynamic(context.TypeFP32(), xLut, lutIdx); var xLutValue = context.VectorExtractDynamic(context.TypeFP32(), xLut, lutIdx);
var yLutValue = context.VectorExtractDynamic(context.TypeFP32(), yLut, lutIdx); var yLutValue = context.VectorExtractDynamic(context.TypeFP32(), yLut, lutIdx);