[vk, SPIR-V] Various changes in an attempt to fix MH on Android
This commit is contained in:
Ribbit
parent
eb98fcd8ab
commit
d6551b049c
10 changed files with 287 additions and 44 deletions
|
|
@ -2,6 +2,7 @@
|
|||
// SPDX-License-Identifier: GPL-2.0-or-later
|
||||
|
||||
#include <boost/container/static_vector.hpp>
|
||||
#include <cmath>
|
||||
|
||||
#include "shader_recompiler/backend/spirv/emit_spirv.h"
|
||||
#include "shader_recompiler/backend/spirv/emit_spirv_instructions.h"
|
||||
|
|
@ -185,6 +186,84 @@ private:
|
|||
spv::ImageOperandsMask mask{};
|
||||
};
|
||||
|
||||
|
||||
Id SampledVectorType(EmitContext& ctx, TextureComponentType component_type) {
|
||||
switch (component_type) {
|
||||
case TextureComponentType::Float:
|
||||
return ctx.F32[4];
|
||||
case TextureComponentType::Sint:
|
||||
return ctx.S32[4];
|
||||
case TextureComponentType::Uint:
|
||||
return ctx.U32[4];
|
||||
}
|
||||
throw LogicError("Unhandled texture component type {}", static_cast<u32>(component_type));
|
||||
}
|
||||
|
||||
bool ExpectsFloatResult(const IR::Inst* inst) {
|
||||
switch (inst->Type()) {
|
||||
case IR::Type::F32:
|
||||
case IR::Type::F32x2:
|
||||
case IR::Type::F32x3:
|
||||
case IR::Type::F32x4:
|
||||
return true;
|
||||
default:
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
Id MakeFloatVector(EmitContext& ctx, float value) {
|
||||
const Id scalar{ctx.Const(value)};
|
||||
return ctx.ConstantComposite(ctx.F32[4], scalar, scalar, scalar, scalar);
|
||||
}
|
||||
|
||||
Id NormalizeUnsignedSample(EmitContext& ctx, u32 component_bits, Id value) {
|
||||
if (component_bits == 0) {
|
||||
return value;
|
||||
}
|
||||
const double max_value = std::exp2(static_cast<double>(component_bits)) - 1.0;
|
||||
if (!(max_value > 0.0)) {
|
||||
return value;
|
||||
}
|
||||
const float inv_max = static_cast<float>(1.0 / max_value);
|
||||
return ctx.OpFMul(ctx.F32[4], value, MakeFloatVector(ctx, inv_max));
|
||||
}
|
||||
|
||||
Id NormalizeSignedSample(EmitContext& ctx, u32 component_bits, Id value) {
|
||||
if (component_bits == 0) {
|
||||
return value;
|
||||
}
|
||||
const double positive_max = component_bits > 0 ? std::exp2(static_cast<double>(component_bits - 1)) - 1.0 : 0.0;
|
||||
if (!(positive_max > 0.0)) {
|
||||
return ctx.OpFClamp(ctx.F32[4], value, MakeFloatVector(ctx, -1.0f), MakeFloatVector(ctx, 1.0f));
|
||||
}
|
||||
const float inv_pos = static_cast<float>(1.0 / positive_max);
|
||||
const Id scaled{ctx.OpFMul(ctx.F32[4], value, MakeFloatVector(ctx, inv_pos))};
|
||||
return ctx.OpFClamp(ctx.F32[4], scaled, MakeFloatVector(ctx, -1.0f), MakeFloatVector(ctx, 1.0f));
|
||||
}
|
||||
|
||||
Id ConvertSampleToExpectedType(EmitContext& ctx, const IR::Inst* inst,
|
||||
const TextureDefinition* texture_def, Id value) {
|
||||
if (!texture_def || texture_def->component_type == TextureComponentType::Float) {
|
||||
return value;
|
||||
}
|
||||
if (!ExpectsFloatResult(inst)) {
|
||||
return value;
|
||||
}
|
||||
switch (texture_def->component_type) {
|
||||
case TextureComponentType::Sint: {
|
||||
const Id as_float{ctx.OpConvertSToF(ctx.F32[4], value)};
|
||||
return NormalizeSignedSample(ctx, texture_def->component_bit_size, as_float);
|
||||
}
|
||||
case TextureComponentType::Uint: {
|
||||
const Id as_float{ctx.OpConvertUToF(ctx.F32[4], value)};
|
||||
return NormalizeUnsignedSample(ctx, texture_def->component_bit_size, as_float);
|
||||
}
|
||||
case TextureComponentType::Float:
|
||||
break;
|
||||
}
|
||||
return value;
|
||||
}
|
||||
|
||||
Id Texture(EmitContext& ctx, IR::TextureInstInfo info, [[maybe_unused]] const IR::Value& index) {
|
||||
const TextureDefinition& def{ctx.textures.at(info.descriptor_index)};
|
||||
if (def.count > 1) {
|
||||
|
|
@ -449,31 +528,39 @@ Id EmitBoundImageWrite(EmitContext&) {
|
|||
Id EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
|
||||
Id bias_lc, const IR::Value& offset) {
|
||||
const auto info{inst->Flags<IR::TextureInstInfo>()};
|
||||
const TextureDefinition* texture_def =
|
||||
info.type == TextureType::Buffer ? nullptr : &ctx.textures.at(info.descriptor_index);
|
||||
const Id result_type =
|
||||
texture_def ? SampledVectorType(ctx, texture_def->component_type) : ctx.F32[4];
|
||||
Id sample{};
|
||||
if (ctx.stage == Stage::Fragment) {
|
||||
const ImageOperands operands(ctx, info.has_bias != 0, false, info.has_lod_clamp != 0,
|
||||
bias_lc, offset);
|
||||
return Emit(&EmitContext::OpImageSparseSampleImplicitLod,
|
||||
&EmitContext::OpImageSampleImplicitLod, ctx, inst, ctx.F32[4],
|
||||
Texture(ctx, info, index), coords, operands.MaskOptional(), operands.Span());
|
||||
sample = Emit(&EmitContext::OpImageSparseSampleImplicitLod,
|
||||
&EmitContext::OpImageSampleImplicitLod, ctx, inst, result_type,
|
||||
Texture(ctx, info, index), coords, operands.MaskOptional(), operands.Span());
|
||||
} else {
|
||||
// We can't use implicit lods on non-fragment stages on SPIR-V. Maxwell hardware behaves as
|
||||
// if the lod was explicitly zero. This may change on Turing with implicit compute
|
||||
// derivatives
|
||||
const Id lod{ctx.Const(0.0f)};
|
||||
const ImageOperands operands(ctx, false, true, info.has_lod_clamp != 0, lod, offset);
|
||||
return Emit(&EmitContext::OpImageSparseSampleExplicitLod,
|
||||
&EmitContext::OpImageSampleExplicitLod, ctx, inst, ctx.F32[4],
|
||||
Texture(ctx, info, index), coords, operands.Mask(), operands.Span());
|
||||
sample = Emit(&EmitContext::OpImageSparseSampleExplicitLod,
|
||||
&EmitContext::OpImageSampleExplicitLod, ctx, inst, result_type,
|
||||
Texture(ctx, info, index), coords, operands.Mask(), operands.Span());
|
||||
}
|
||||
return ConvertSampleToExpectedType(ctx, inst, texture_def, sample);
|
||||
}
|
||||
|
||||
Id EmitImageSampleExplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
|
||||
Id lod, const IR::Value& offset) {
|
||||
const auto info{inst->Flags<IR::TextureInstInfo>()};
|
||||
const TextureDefinition* texture_def =
|
||||
info.type == TextureType::Buffer ? nullptr : &ctx.textures.at(info.descriptor_index);
|
||||
const Id result_type =
|
||||
texture_def ? SampledVectorType(ctx, texture_def->component_type) : ctx.F32[4];
|
||||
const ImageOperands operands(ctx, false, true, false, lod, offset);
|
||||
return Emit(&EmitContext::OpImageSparseSampleExplicitLod,
|
||||
&EmitContext::OpImageSampleExplicitLod, ctx, inst, ctx.F32[4],
|
||||
Texture(ctx, info, index), coords, operands.Mask(), operands.Span());
|
||||
const Id sample = Emit(&EmitContext::OpImageSparseSampleExplicitLod,
|
||||
&EmitContext::OpImageSampleExplicitLod, ctx, inst, result_type,
|
||||
Texture(ctx, info, index), coords, operands.Mask(), operands.Span());
|
||||
return ConvertSampleToExpectedType(ctx, inst, texture_def, sample);
|
||||
}
|
||||
|
||||
Id EmitImageSampleDrefImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index,
|
||||
|
|
@ -509,13 +596,19 @@ Id EmitImageSampleDrefExplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Va
|
|||
Id EmitImageGather(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
|
||||
const IR::Value& offset, const IR::Value& offset2) {
|
||||
const auto info{inst->Flags<IR::TextureInstInfo>()};
|
||||
const TextureDefinition* texture_def =
|
||||
info.type == TextureType::Buffer ? nullptr : &ctx.textures.at(info.descriptor_index);
|
||||
const Id result_type =
|
||||
texture_def ? SampledVectorType(ctx, texture_def->component_type) : ctx.F32[4];
|
||||
const ImageOperands operands(ctx, offset, offset2);
|
||||
if (ctx.profile.need_gather_subpixel_offset) {
|
||||
coords = ImageGatherSubpixelOffset(ctx, info, TextureImage(ctx, info, index), coords);
|
||||
}
|
||||
return Emit(&EmitContext::OpImageSparseGather, &EmitContext::OpImageGather, ctx, inst,
|
||||
ctx.F32[4], Texture(ctx, info, index), coords, ctx.Const(info.gather_component),
|
||||
operands.MaskOptional(), operands.Span());
|
||||
const Id sample = Emit(&EmitContext::OpImageSparseGather, &EmitContext::OpImageGather, ctx, inst,
|
||||
result_type, Texture(ctx, info, index), coords,
|
||||
ctx.Const(info.gather_component), operands.MaskOptional(),
|
||||
operands.Span());
|
||||
return ConvertSampleToExpectedType(ctx, inst, texture_def, sample);
|
||||
}
|
||||
|
||||
Id EmitImageGatherDref(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
|
||||
|
|
@ -538,12 +631,17 @@ Id EmitImageFetch(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id c
|
|||
lod = Id{};
|
||||
}
|
||||
if (Sirit::ValidId(ms)) {
|
||||
// This image is multisampled, lod must be implicit
|
||||
lod = Id{};
|
||||
}
|
||||
const TextureDefinition* texture_def =
|
||||
info.type == TextureType::Buffer ? nullptr : &ctx.textures.at(info.descriptor_index);
|
||||
const Id result_type =
|
||||
texture_def ? SampledVectorType(ctx, texture_def->component_type) : ctx.F32[4];
|
||||
const ImageOperands operands(lod, ms);
|
||||
return Emit(&EmitContext::OpImageSparseFetch, &EmitContext::OpImageFetch, ctx, inst, ctx.F32[4],
|
||||
TextureImage(ctx, info, index), coords, operands.MaskOptional(), operands.Span());
|
||||
const Id sample = Emit(&EmitContext::OpImageSparseFetch, &EmitContext::OpImageFetch, ctx, inst,
|
||||
result_type, TextureImage(ctx, info, index), coords,
|
||||
operands.MaskOptional(), operands.Span());
|
||||
return ConvertSampleToExpectedType(ctx, inst, texture_def, sample);
|
||||
}
|
||||
|
||||
Id EmitImageQueryDimensions(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id lod,
|
||||
|
|
@ -588,14 +686,19 @@ Id EmitImageQueryLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, I
|
|||
Id EmitImageGradient(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords,
|
||||
Id derivatives, const IR::Value& offset, Id lod_clamp) {
|
||||
const auto info{inst->Flags<IR::TextureInstInfo>()};
|
||||
const TextureDefinition* texture_def =
|
||||
info.type == TextureType::Buffer ? nullptr : &ctx.textures.at(info.descriptor_index);
|
||||
const Id result_type =
|
||||
texture_def ? SampledVectorType(ctx, texture_def->component_type) : ctx.F32[4];
|
||||
const auto operands = info.num_derivatives == 3
|
||||
? ImageOperands(ctx, info.has_lod_clamp != 0, derivatives,
|
||||
ctx.Def(offset), {}, lod_clamp)
|
||||
: ImageOperands(ctx, info.has_lod_clamp != 0, derivatives,
|
||||
info.num_derivatives, offset, lod_clamp);
|
||||
return Emit(&EmitContext::OpImageSparseSampleExplicitLod,
|
||||
&EmitContext::OpImageSampleExplicitLod, ctx, inst, ctx.F32[4],
|
||||
Texture(ctx, info, index), coords, operands.Mask(), operands.Span());
|
||||
const Id sample = Emit(&EmitContext::OpImageSparseSampleExplicitLod,
|
||||
&EmitContext::OpImageSampleExplicitLod, ctx, inst, result_type,
|
||||
Texture(ctx, info, index), coords, operands.Mask(), operands.Span());
|
||||
return ConvertSampleToExpectedType(ctx, inst, texture_def, sample);
|
||||
}
|
||||
|
||||
Id EmitImageRead(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords) {
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue