blob: 83118cfb5d19f011974c678b55a9ce5c3069cd77 [file] [log] [blame]
// Copyright 2024 The Dawn & Tint Authors
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
//
// 1. Redistributions of source code must retain the above copyright notice, this
// list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// 3. Neither the name of the copyright holder nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include "src/tint/lang/glsl/writer/raise/texture_polyfill.h"
#include <string>
#include <utility>
#include <vector>
#include "src/tint/lang/core/fluent_types.h" // IWYU pragma: export
#include "src/tint/lang/core/ir/builder.h"
#include "src/tint/lang/core/ir/module.h"
#include "src/tint/lang/core/ir/validator.h"
#include "src/tint/lang/core/type/depth_multisampled_texture.h"
#include "src/tint/lang/core/type/depth_texture.h"
#include "src/tint/lang/core/type/multisampled_texture.h"
#include "src/tint/lang/core/type/sampled_texture.h"
#include "src/tint/lang/core/type/storage_texture.h"
#include "src/tint/lang/glsl/ir/builtin_call.h"
#include "src/tint/lang/glsl/ir/member_builtin_call.h"
namespace tint::glsl::writer::raise {
namespace {
using namespace tint::core::fluent_types; // NOLINT
using namespace tint::core::number_suffixes; // NOLINT
/// PIMPL state for the transform.
struct State {
/// The IR module.
core::ir::Module& ir;
/// The configuration
const TexturePolyfillConfig& cfg;
/// The IR builder.
core::ir::Builder b{ir};
/// The type manager.
core::type::Manager& ty{ir.Types()};
// A map of single texture to the replacement var. Note, this doesn't just re-use the
// `texture_sampler_to_replacment` with the placeholder sampler because we can share an
// individual texture with a texture,sampler pair. So, if we create the texture as `t1,s1` we
// want to use that `t1` individually but if we look it up with `t1,sp` then we won't find it.
// This secondary map exists to allow us access to textures which may have been created
// with a sampler.
Hashmap<core::ir::Var*, std::optional<core::ir::Var*>, 2> texture_to_replacement_{};
// A map of the <texture,sampler> binding pair to the replacement var.
Hashmap<binding::CombinedTextureSamplerPair, core::ir::Var*, 2>
texture_sampler_to_replacement_{};
// The list of textures and samplers that were replaced. There may have been textures which
// existed but were unused. We don't want to delete them, so we only delete replaced values.
Hashset<core::ir::Var*, 4> replaced_textures_and_samplers_{};
/// Process the module.
void Process() {
/// Converts all 1D texture types and accesses to 2D. This is required for GLSL ES, which
/// does not support 1D textures. We do this for desktop GL as well for consistency. This
/// could be relaxed in the future if desired.
UpgradeTexture1DVars();
UpgradeTexture1DParams();
PopulateTextureInformation();
Vector<core::ir::CoreBuiltinCall*, 4> call_worklist;
for (auto* inst : ir.Instructions()) {
if (auto* call = inst->As<core::ir::CoreBuiltinCall>()) {
switch (call->Func()) {
case core::BuiltinFn::kTextureDimensions:
case core::BuiltinFn::kTextureGather:
case core::BuiltinFn::kTextureGatherCompare:
case core::BuiltinFn::kTextureLoad:
case core::BuiltinFn::kTextureNumLayers:
case core::BuiltinFn::kTextureSample:
case core::BuiltinFn::kTextureSampleBias:
case core::BuiltinFn::kTextureSampleCompare:
case core::BuiltinFn::kTextureSampleCompareLevel:
case core::BuiltinFn::kTextureSampleGrad:
case core::BuiltinFn::kTextureSampleLevel:
case core::BuiltinFn::kTextureStore:
call_worklist.Push(call);
break;
default:
break;
}
continue;
}
}
// Replace the builtin calls that we found
for (auto* call : call_worklist) {
switch (call->Func()) {
case core::BuiltinFn::kTextureDimensions:
TextureDimensions(call);
break;
case core::BuiltinFn::kTextureGather:
TextureGather(call);
break;
case core::BuiltinFn::kTextureGatherCompare:
TextureGatherCompare(call);
break;
case core::BuiltinFn::kTextureLoad:
TextureLoad(call);
break;
case core::BuiltinFn::kTextureNumLayers:
TextureNumLayers(call);
break;
case core::BuiltinFn::kTextureSample:
TextureSample(call);
break;
case core::BuiltinFn::kTextureSampleBias:
TextureSampleBias(call);
break;
case core::BuiltinFn::kTextureSampleCompare:
TextureSampleCompare(call);
break;
case core::BuiltinFn::kTextureSampleCompareLevel:
TextureSampleCompareLevel(call);
break;
case core::BuiltinFn::kTextureSampleGrad:
TextureSampleGrad(call);
break;
case core::BuiltinFn::kTextureSampleLevel:
TextureSampleLevel(call);
break;
case core::BuiltinFn::kTextureStore:
TextureStore(call);
break;
default:
TINT_UNREACHABLE() << call->Func();
}
}
// Remove all replaced textures and samplers as they have been replaced by new globals.
for (auto* var : replaced_textures_and_samplers_.Vector()) {
var->Result(0)->ForEachUseUnsorted([](core::ir::Usage use) {
TINT_ASSERT(use.instruction->Is<core::ir::Load>());
use.instruction->Destroy();
});
var->Destroy();
}
}
core::ir::Var* GetReplacement(core::ir::Var* tex,
core::ir::Var* sampler,
const core::type::Pointer* tex_ty) {
// Don't change storage textures
if (tex->Result(0)->Type()->UnwrapPtr()->Is<core::type::StorageTexture>()) {
return tex;
}
if (!sampler) {
auto existing_var = texture_to_replacement_.Get(tex);
if (existing_var) {
return existing_var->value();
}
replaced_textures_and_samplers_.Add(tex);
// If the texture wasn't already in the map this means it was an individual texture we
// hadn't seen yet. Create it and insert into the map for future use.
binding::CombinedTextureSamplerPair key{tex->BindingPoint().value(),
cfg.placeholder_sampler_bind_point};
auto* replacement = MakeVar(key, tex, nullptr, tex_ty);
texture_to_replacement_.Add(tex, replacement);
return replacement;
}
auto tex_bp = tex->BindingPoint();
auto samp_bp = sampler->BindingPoint();
TINT_ASSERT(tex_bp.has_value() && samp_bp.has_value());
replaced_textures_and_samplers_.Add(tex);
replaced_textures_and_samplers_.Add(sampler);
binding::CombinedTextureSamplerPair key{tex_bp.value(), samp_bp.value()};
auto var = texture_sampler_to_replacement_.Get(key);
TINT_ASSERT(var);
return *(var.value);
}
// Get the `var` for a texture/sampler value. This means the value must be the result of a load.
core::ir::Var* VarForValue(core::ir::Value* val) {
if (!val) {
return nullptr;
}
auto* load = LoadForValue(val);
TINT_ASSERT(load);
auto* from = load->From()->As<core::ir::InstructionResult>();
TINT_ASSERT(from);
auto* var = from->Instruction()->As<core::ir::Var>();
TINT_ASSERT(var);
return var;
}
core::ir::Load* LoadForValue(core::ir::Value* val) {
if (!val) {
return nullptr;
}
auto* res = val->As<core::ir::InstructionResult>();
TINT_ASSERT(res);
auto* load = res->Instruction()->As<core::ir::Load>();
TINT_ASSERT(load);
return load;
}
struct SamplerTextureVars {
core::ir::Var* texture;
core::ir::Var* sampler;
};
SamplerTextureVars GetTextureSamplerFor(core::ir::CoreBuiltinCall* call) {
auto args = call->Args();
switch (call->Func()) {
case core::BuiltinFn::kTextureDimensions:
case core::BuiltinFn::kTextureLoad:
case core::BuiltinFn::kTextureNumLayers:
case core::BuiltinFn::kTextureStore:
return {VarForValue(args[0]), nullptr};
case core::BuiltinFn::kTextureGather: {
if (args[0]->Type()->Is<core::type::Texture>()) {
return {VarForValue(args[0]), VarForValue(args[1])};
}
return {VarForValue(args[1]), VarForValue(args[2])};
}
case core::BuiltinFn::kTextureGatherCompare:
case core::BuiltinFn::kTextureSample:
case core::BuiltinFn::kTextureSampleBaseClampToEdge:
case core::BuiltinFn::kTextureSampleBias:
case core::BuiltinFn::kTextureSampleCompare:
case core::BuiltinFn::kTextureSampleCompareLevel:
case core::BuiltinFn::kTextureSampleGrad:
case core::BuiltinFn::kTextureSampleLevel:
return {VarForValue(args[0]), VarForValue(args[1])};
default:
TINT_UNREACHABLE() << "unhandled texture function: " << call->Func();
}
}
core::ir::Var* MakeVar(binding::CombinedTextureSamplerPair& key,
core::ir::Var* tex,
core::ir::Var* sampler,
const core::type::Pointer* tex_ty) {
std::string name;
auto it = (cfg.sampler_texture_to_name.find(key));
if (it != cfg.sampler_texture_to_name.end()) {
name = it->second;
} else {
name = ir.NameOf(tex).Name();
if (name.empty()) {
name = "t";
}
if (sampler) {
auto sampler_name = ir.NameOf(sampler).Name();
if (sampler_name.empty()) {
sampler_name = "s";
}
name += "_" + sampler_name;
}
if (name.empty()) {
name = "v";
}
}
core::ir::Var* var = nullptr;
// We may already be inside an insert block, so make a new insert block instead of
// appending directly to the root block.
b.Append(ir.root_block, [&] { var = b.Var(name, tex_ty); });
return var;
}
// This function builds up replacement combined textures. It creates a global mapping, one for
// all the texture,sampler pairs and one with individual textures. The individual textures will
// attempt to populate with the first texture from a texture,sampler pair but if we didn't see
// the texture in any pair we'll create it on the fly when getting the replacemnt later.
void PopulateTextureInformation() {
for (auto* inst : ir.Instructions()) {
auto* call = inst->As<core::ir::CoreBuiltinCall>();
if (!call || (!core::IsTexture(call->Func()) && !core::IsImageQuery(call->Func()))) {
continue;
}
auto tex_sampler = GetTextureSamplerFor(call);
auto* tex = tex_sampler.texture;
auto* sampler = tex_sampler.sampler;
// No sampler, then we aren't going to be creating a combined sampler.
if (!sampler) {
continue;
}
BindingPoint tex_bp = tex->BindingPoint().value();
BindingPoint samp_bp = sampler->BindingPoint().value();
binding::CombinedTextureSamplerPair key{tex_bp, samp_bp};
auto* replacement = texture_sampler_to_replacement_.GetOrAdd(key, [&] {
return MakeVar(key, tex, sampler,
tex->Result(0)->Type()->As<core::type::Pointer>());
});
// Don't add depth textures here because the unsampled depth texture will need to be
// created as a sampled texture, instead of a depth texture.
if (!tex->Result(0)->Type()->UnwrapPtr()->Is<core::type::DepthTexture>()) {
texture_to_replacement_.Add(tex, replacement);
}
}
}
std::optional<const core::type::Type*> UpgradeTexture1D(core::ir::Value* value) {
bool is_1d = false;
const core::type::Type* new_type = nullptr;
tint::Switch(
value->Type()->UnwrapPtr(),
[&](const core::type::SampledTexture* s) {
is_1d = s->Dim() == core::type::TextureDimension::k1d;
new_type = ty.Get<core::type::SampledTexture>(core::type::TextureDimension::k2d,
s->Type());
},
[&](const core::type::StorageTexture* s) {
is_1d = s->Dim() == core::type::TextureDimension::k1d;
new_type = ty.Get<core::type::StorageTexture>(
core::type::TextureDimension::k2d, s->TexelFormat(), s->Access(), s->Type());
});
if (!is_1d) {
return std::nullopt;
}
if (auto* ptr = value->Type()->As<core::type::Pointer>()) {
new_type = ty.ptr(ptr->AddressSpace(), new_type, ptr->Access());
}
// For each 1d texture usage we have to make sure return values and arguments are modified
// to fit the 2d texture.
for (auto usage : value->UsagesUnsorted()) {
if (auto* call = usage->instruction->As<core::ir::CoreBuiltinCall>()) {
switch (call->Func()) {
case core::BuiltinFn::kTextureDimensions: {
// Upgrade result to a vec2 and swizzle out the `x` component.
auto* res = call->DetachResult();
call->SetResults(b.InstructionResult(ty.vec2<u32>()));
b.InsertAfter(call, [&] {
auto* s = b.Swizzle(res->Type(), call, Vector<uint32_t, 1>{0});
res->ReplaceAllUsesWith(s->Result(0));
});
break;
}
case core::BuiltinFn::kTextureLoad:
case core::BuiltinFn::kTextureStore: {
// Add a new coord item so it's a vec2.
auto arg = call->Args()[1];
b.InsertBefore(call, [&] {
call->SetArg(1,
b.Construct(ty.vec2(arg->Type()), arg, b.Zero(arg->Type()))
->Result(0));
});
break;
}
case core::BuiltinFn::kTextureSample: {
// Add a new coord item so it's a vec2.
auto arg = call->Args()[2];
b.InsertBefore(call, [&] {
call->SetArg(2,
b.Construct(ty.vec2(arg->Type()), arg, 0.5_f)->Result(0));
});
break;
}
default:
TINT_UNREACHABLE() << "unknown usage instruction for texture";
}
}
}
return {new_type};
}
void UpgradeTexture1DVars() {
for (auto* inst : *ir.root_block) {
auto* var = inst->As<core::ir::Var>();
if (!var) {
continue;
}
auto new_type = UpgradeTexture1D(var->Result(0));
if (!new_type.has_value()) {
continue;
}
var->Result(0)->SetType(new_type.value());
// All of the usages of the textures should involve loading them as the `var`
// declarations will be pointers and the function usages require non-pointer textures.
for (auto usage : var->Result(0)->UsagesUnsorted()) {
UpgradeLoadOf1DTexture(usage->instruction);
}
}
}
void UpgradeTexture1DParams() {
for (auto func : ir.functions) {
for (auto* param : func->Params()) {
auto new_type = UpgradeTexture1D(param);
if (!new_type.has_value()) {
continue;
}
param->SetType(new_type.value());
}
}
}
void UpgradeLoadOf1DTexture(core::ir::Instruction* inst) {
auto* ld = inst->As<core::ir::Load>();
TINT_ASSERT(ld);
auto new_type = UpgradeTexture1D(ld->Result(0));
if (!new_type.has_value()) {
return;
}
ld->Result(0)->SetType(new_type.value());
}
// Must be called inside an insertion block
core::ir::Value* GetNewTexture(core::ir::Value* tex, core::ir::Value* sampler = nullptr) {
auto* t = VarForValue(tex);
auto* s = VarForValue(sampler);
auto* tex_ty = t->Result(0)->Type()->As<core::type::Pointer>();
TINT_ASSERT(tex_ty);
// A depth texture gets turned into a SampledTexture of type `f32` when there is no
// sampler.
if (!sampler) {
if (tex_ty->StoreType()->Is<core::type::DepthTexture>()) {
tex_ty =
ty.ptr(tex_ty->AddressSpace(),
ty.Get<core::type::SampledTexture>(
tex_ty->UnwrapPtr()->As<core::type::Texture>()->Dim(), ty.f32()),
tex_ty->Access());
}
}
auto* replacement = GetReplacement(t, s, tex_ty);
TINT_ASSERT(replacement);
// In the storage case, we'll return the original texture. Nothing else to do in that
// case.
if (replacement == t) {
return tex;
}
return b.Load(replacement)->Result(0);
}
// `textureDimensions` returns an unsigned scalar / vector in WGSL. `textureSize` and
// `imageSize` return a signed scalar / vector in GLSL. So, we need to cast the result
// to the needed WGSL type.
void TextureDimensions(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
auto func = glsl::BuiltinFn::kTextureSize;
uint32_t idx = 0;
Vector<core::ir::Value*, 2> new_args;
auto* tex = GetNewTexture(args[idx++]);
auto* tex_ty = tex->Type()->As<core::type::Texture>();
new_args.Push(tex);
if (tex_ty->Is<core::type::StorageTexture>()) {
func = glsl::BuiltinFn::kImageSize;
}
if (!(tex_ty->Is<core::type::StorageTexture>() ||
tex_ty->Is<core::type::MultisampledTexture>() ||
tex_ty->Is<core::type::DepthMultisampledTexture>())) {
// Add a LOD to any texture other then storage, and multi-sampled textures
// which does not already have an LOD.
if (args.Length() == 1) {
new_args.Push(b.Constant(0_i));
} else {
// Make sure the LOD is a i32
new_args.Push(b.Bitcast(ty.i32(), args[idx++])->Result(0));
}
}
auto ret_type = call->Result(0)->Type();
// In GLSL the array dimensions return a 3rd parameter.
if (tex_ty->Dim() == core::type::TextureDimension::k2dArray ||
tex_ty->Dim() == core::type::TextureDimension::kCubeArray) {
ret_type = ty.vec(ty.i32(), 3);
} else {
ret_type = ty.MatchWidth(ty.i32(), call->Result(0)->Type());
}
core::ir::Value* result =
b.Call<glsl::ir::BuiltinCall>(ret_type, func, new_args)->Result(0);
// `textureSize` on array samplers returns the array size in the final
// component, WGSL requires a 2 component response, so drop the array size
if (tex_ty->Dim() == core::type::TextureDimension::k2dArray ||
tex_ty->Dim() == core::type::TextureDimension::kCubeArray) {
ret_type = ty.MatchWidth(ty.i32(), call->Result(0)->Type());
result = b.Swizzle(ret_type, result, {0, 1})->Result(0);
}
b.BitcastWithResult(call->DetachResult(), result)->Result(0);
});
call->Destroy();
}
// `textureNumLayers` returns an unsigned scalar in WGSL. `textureSize` and `imageSize`
// return a signed scalar / vector in GLSL.
//
// For the `textureSize` and `imageSize` calls the valid WGSL values always produce a
// `vec3` in GLSL so we extract the `z` component for the number of layers.
void TextureNumLayers(core::ir::BuiltinCall* call) {
b.InsertBefore(call, [&] {
auto args = call->Args();
auto* tex = GetNewTexture(args[0]);
auto* tex_ty = tex->Type()->As<core::type::Texture>();
auto func = glsl::BuiltinFn::kTextureSize;
if (tex_ty->Is<core::type::StorageTexture>()) {
func = glsl::BuiltinFn::kImageSize;
}
Vector<core::ir::Value*, 2> new_args;
new_args.Push(tex);
// Non-storage textures require a LOD
if (!tex_ty->Is<core::type::StorageTexture>()) {
new_args.Push(b.Constant(0_i));
}
auto* new_call = b.Call<glsl::ir::BuiltinCall>(ty.vec(ty.i32(), 3), func, new_args);
auto* swizzle = b.Swizzle(ty.i32(), new_call, {2});
b.BitcastWithResult(call->DetachResult(), swizzle->Result(0));
});
call->Destroy();
}
void TextureLoad(core::ir::CoreBuiltinCall* call) {
b.InsertBefore(call, [&] {
uint32_t idx = 0;
auto args = call->Args();
auto* source_tex = args[idx++];
bool source_was_depth =
source_tex->Type()
->IsAnyOf<core::type::DepthTexture, core::type::DepthMultisampledTexture>();
auto* tex = GetNewTexture(source_tex);
// No loading from a depth texture in GLSL, so we should never have gotten here.
TINT_ASSERT(!tex->Type()->Is<core::type::DepthTexture>());
auto* tex_type = tex->Type()->As<core::type::Texture>();
glsl::BuiltinFn func = glsl::BuiltinFn::kNone;
if (tex_type->Is<core::type::StorageTexture>()) {
func = glsl::BuiltinFn::kImageLoad;
} else {
func = glsl::BuiltinFn::kTexelFetch;
}
bool is_ms = tex_type->Is<core::type::MultisampledTexture>();
bool is_storage = tex_type->Is<core::type::StorageTexture>();
Vector<core::ir::Value*, 3> call_args{tex};
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d: {
call_args.Push(b.Convert(ty.vec2<i32>(), args[idx++])->Result(0));
if (is_ms) {
call_args.Push(b.Convert(ty.i32(), args[idx++])->Result(0));
} else {
if (!is_storage) {
call_args.Push(b.Convert(ty.i32(), args[idx++])->Result(0));
}
}
break;
}
case core::type::TextureDimension::k2dArray: {
auto* coord = b.Convert(ty.vec2<i32>(), args[idx++]);
auto* ary_idx = b.Convert(ty.i32(), args[idx++]);
call_args.Push(b.Construct(ty.vec3<i32>(), coord, ary_idx)->Result(0));
if (!is_storage) {
call_args.Push(b.Convert(ty.i32(), args[idx++])->Result(0));
}
break;
}
case core::type::TextureDimension::k3d: {
call_args.Push(b.Convert(ty.vec3<i32>(), args[idx++])->Result(0));
if (!is_storage) {
call_args.Push(b.Convert(ty.i32(), args[idx++])->Result(0));
}
break;
}
default:
TINT_UNREACHABLE();
}
// If we had a depth texture source, then that means we've swapped the depth type for a
// sampled 2d texture. Sampled 2d returns a `vec4<f32>` from the texel fetch but the
// depth texture is expecting an `f32`. So, swap the types to the call and swizzle out
// the `x` component if needed.
const core::type::Type* fetch_ty = call->Result(0)->Type();
if (source_was_depth) {
fetch_ty = ty.vec4<f32>();
}
core::ir::Instruction* new_call =
b.Call<glsl::ir::BuiltinCall>(fetch_ty, func, std::move(call_args));
if (source_was_depth) {
new_call = b.Swizzle(ty.f32(), new_call, {0});
}
call->Result(0)->ReplaceAllUsesWith(new_call->Result(0));
});
call->Destroy();
}
void TextureStore(core::ir::BuiltinCall* call) {
b.InsertBefore(call, [&] {
uint32_t idx = 0;
auto args = call->Args();
auto* tex = GetNewTexture(args[idx++]);
auto* tex_type = tex->Type()->As<core::type::StorageTexture>();
TINT_ASSERT(tex_type);
Vector<core::ir::Value*, 3> new_args;
new_args.Push(tex);
if (tex_type->Dim() == core::type::TextureDimension::k2dArray) {
auto* coords = args[idx++];
if (!coords->Type()->DeepestElement()->Is<core::type::I32>()) {
coords = b.Convert(ty.vec2<i32>(), coords)->Result(0);
}
auto* array = b.Convert(ty.i32(), args[idx++]);
auto* coords_ty = coords->Type()->As<core::type::Vector>();
TINT_ASSERT(coords_ty);
auto* new_coords = b.Construct(ty.vec3<i32>(), coords, array);
new_args.Push(new_coords->Result(0));
new_args.Push(args[idx++]);
} else {
auto* coords = args[idx++];
if (!coords->Type()->DeepestElement()->Is<core::type::I32>()) {
coords = b.Convert(ty.MatchWidth(ty.i32(), coords->Type()), coords)->Result(0);
}
new_args.Push(coords);
new_args.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(
call->DetachResult(), glsl::BuiltinFn::kImageStore, std::move(new_args));
});
call->Destroy();
}
void TextureGather(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
core::ir::Value* coords = nullptr;
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
core::ir::Value* component = nullptr;
if (!args[idx]->Type()->Is<core::type::Texture>()) {
component = args[idx++];
TINT_ASSERT(component);
}
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
bool is_depth = tex_type->Is<core::type::DepthTexture>();
params.Push(tex);
coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray:
params.Push(b.Construct(ty.vec3<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
break;
case core::type::TextureDimension::kCube:
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
break;
default:
TINT_UNREACHABLE();
}
// Depth gather requires a `refz` param in GLSL.
if (is_depth) {
params.Push(b.Constant(0.0_f));
}
auto fn = glsl::BuiltinFn::kTextureGather;
if (idx < args.Length()) {
fn = glsl::BuiltinFn::kTextureGatherOffset;
params.Push(args[idx++]);
}
// Push the component onto the end of the list if needed.
if (component != nullptr) {
params.Push(b.Convert(ty.i32(), component)->Result(0));
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureGatherCompare(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
Vector<core::ir::Value*, 4> params;
params.Push(tex);
auto* coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray:
params.Push(b.Construct(ty.vec3<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
break;
case core::type::TextureDimension::kCube:
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
break;
default:
TINT_UNREACHABLE();
}
params.Push(args[idx++]);
auto fn = glsl::BuiltinFn::kTextureGather;
if (idx < args.Length()) {
fn = glsl::BuiltinFn::kTextureGatherOffset;
params.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureSample(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
params.Push(tex);
bool is_depth = tex_type->Is<core::type::DepthTexture>();
bool is_array = false;
auto depth_ref = 0_f;
core::ir::Value* coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
if (is_depth) {
coords = b.Construct(ty.vec3<f32>(), coords, depth_ref)->Result(0);
}
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray: {
is_array = true;
Vector<core::ir::Value*, 3> new_coords;
new_coords.Push(coords);
new_coords.Push(b.Convert<f32>(args[idx++])->Result(0));
uint32_t vec_width = 3;
if (is_depth) {
new_coords.Push(b.Value(depth_ref));
++vec_width;
}
params.Push(b.Construct(ty.vec(ty.f32(), vec_width), new_coords)->Result(0));
break;
}
case core::type::TextureDimension::k3d:
case core::type::TextureDimension::kCube:
if (is_depth) {
coords = b.Construct(ty.vec4<f32>(), coords, depth_ref)->Result(0);
}
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
is_array = true;
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
if (is_depth) {
params.Push(b.Value(depth_ref));
}
break;
default:
TINT_UNREACHABLE();
}
auto fn = glsl::BuiltinFn::kTexture;
if (idx < args.Length()) {
// In GLSL ES `textureOffset` does not support depth 2d array textures. In order to
// support this texture we polyfill with a `textureGradOffset` and explicitly
// calculate the `dPdx` and `dPdy` values.
if (is_depth && is_array) {
fn = glsl::BuiltinFn::kTextureGradOffset;
auto* dpdx = b.Call(coords->Type(), core::BuiltinFn::kDpdx, coords);
auto* dpdy = b.Call(coords->Type(), core::BuiltinFn::kDpdy, coords);
params.Push(dpdx->Result(0));
params.Push(dpdy->Result(0));
} else {
fn = glsl::BuiltinFn::kTextureOffset;
}
params.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureSampleBias(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
params.Push(tex);
core::ir::Value* coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray: {
Vector<core::ir::Value*, 3> new_coords;
new_coords.Push(coords);
new_coords.Push(b.Convert<f32>(args[idx++])->Result(0));
params.Push(b.Construct(ty.vec3<f32>(), new_coords)->Result(0));
break;
}
case core::type::TextureDimension::k3d:
case core::type::TextureDimension::kCube:
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
break;
default:
TINT_UNREACHABLE();
}
// Bias comes before offset, so pull it out before handling the offset.
auto bias = args[idx++];
auto fn = glsl::BuiltinFn::kTexture;
if (idx < args.Length()) {
fn = glsl::BuiltinFn::kTextureOffset;
params.Push(args[idx++]);
}
params.Push(bias);
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureSampleLevel(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
params.Push(tex);
bool is_depth = tex_type->Is<core::type::DepthTexture>();
bool needs_ext = false;
auto depth_ref = 0_f;
core::ir::Value* coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
if (is_depth) {
coords = b.Construct(ty.vec3<f32>(), coords, depth_ref)->Result(0);
}
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray: {
Vector<core::ir::Value*, 3> new_coords;
new_coords.Push(coords);
new_coords.Push(b.Convert<f32>(args[idx++])->Result(0));
uint32_t vec_width = 3;
if (is_depth) {
needs_ext = true;
new_coords.Push(b.Value(depth_ref));
++vec_width;
}
params.Push(b.Construct(ty.vec(ty.f32(), vec_width), new_coords)->Result(0));
break;
}
case core::type::TextureDimension::k3d:
case core::type::TextureDimension::kCube:
if (is_depth) {
needs_ext = tex_type->Dim() == core::type::TextureDimension::kCube;
coords = b.Construct(ty.vec4<f32>(), coords, depth_ref)->Result(0);
}
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
if (is_depth) {
needs_ext = true;
params.Push(b.Value(depth_ref));
}
break;
default:
TINT_UNREACHABLE();
}
params.Push(b.Convert(ty.f32(), args[idx++])->Result(0));
auto fn = needs_ext ? glsl::BuiltinFn::kExtTextureLod : glsl::BuiltinFn::kTextureLod;
if (idx < args.Length()) {
fn = needs_ext ? glsl::BuiltinFn::kExtTextureLodOffset
: glsl::BuiltinFn::kTextureLodOffset;
params.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureSampleGrad(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
params.Push(tex);
core::ir::Value* coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray: {
Vector<core::ir::Value*, 3> new_coords;
new_coords.Push(coords);
new_coords.Push(b.Convert<f32>(args[idx++])->Result(0));
params.Push(b.Construct(ty.vec3<f32>(), new_coords)->Result(0));
break;
}
case core::type::TextureDimension::k3d:
case core::type::TextureDimension::kCube:
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
break;
default:
TINT_UNREACHABLE();
}
params.Push(args[idx++]); // dPdx
params.Push(args[idx++]); // dPdy
auto fn = glsl::BuiltinFn::kTextureGrad;
if (idx < args.Length()) {
fn = glsl::BuiltinFn::kTextureGradOffset;
params.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureSampleCompare(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
params.Push(tex);
bool is_array = false;
core::ir::Value* coords = args[idx++];
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
coords = b.Construct(ty.vec3<f32>(), coords, args[idx++])->Result(0);
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray: {
is_array = true;
Vector<core::ir::Value*, 3> new_coords;
new_coords.Push(coords);
new_coords.Push(b.Convert<f32>(args[idx++])->Result(0));
new_coords.Push(b.Value(args[idx++]));
params.Push(b.Construct(ty.vec4<f32>(), new_coords)->Result(0));
break;
}
case core::type::TextureDimension::kCube:
coords = b.Construct(ty.vec4<f32>(), coords, args[idx++])->Result(0);
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
is_array = true;
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
params.Push(b.Value(args[idx++]));
break;
default:
TINT_UNREACHABLE();
}
auto fn = glsl::BuiltinFn::kTexture;
if (idx < args.Length()) {
// In GLSL ES `textureOffset` does not support depth 2d array textures. In order to
// support this texture we polyfill with a `textureGradOffset` and explicitly
// calculate the `dPdx` and `dPdy` values.
if (is_array) {
fn = glsl::BuiltinFn::kTextureGradOffset;
auto* dpdx = b.Call(coords->Type(), core::BuiltinFn::kDpdx, coords);
auto* dpdy = b.Call(coords->Type(), core::BuiltinFn::kDpdy, coords);
params.Push(dpdx->Result(0));
params.Push(dpdy->Result(0));
} else {
fn = glsl::BuiltinFn::kTextureOffset;
}
params.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
void TextureSampleCompareLevel(core::ir::BuiltinCall* call) {
auto args = call->Args();
b.InsertBefore(call, [&] {
Vector<core::ir::Value*, 4> params;
uint32_t idx = 0;
uint32_t tex_arg = idx++;
uint32_t sampler_arg = idx++;
auto* tex = GetNewTexture(args[tex_arg], args[sampler_arg]);
auto* tex_type = tex->Type()->As<core::type::Texture>();
TINT_ASSERT(tex_type);
params.Push(tex);
core::ir::Value* coords = args[idx++];
bool is_array = false;
bool is_depth = tex_type->Is<core::type::DepthTexture>();
switch (tex_type->Dim()) {
case core::type::TextureDimension::k2d:
coords = b.Construct(ty.vec3<f32>(), coords, args[idx++])->Result(0);
params.Push(coords);
break;
case core::type::TextureDimension::k2dArray: {
is_array = true;
Vector<core::ir::Value*, 3> new_coords;
new_coords.Push(coords);
new_coords.Push(b.Convert<f32>(args[idx++])->Result(0));
new_coords.Push(b.Value(args[idx++]));
params.Push(b.Construct(ty.vec4<f32>(), new_coords)->Result(0));
break;
}
case core::type::TextureDimension::kCube:
coords = b.Construct(ty.vec4<f32>(), coords, args[idx++])->Result(0);
params.Push(coords);
break;
case core::type::TextureDimension::kCubeArray:
is_array = true;
params.Push(b.Construct(ty.vec4<f32>(), coords, b.Convert<f32>(args[idx++]))
->Result(0));
params.Push(b.Value(args[idx++]));
break;
default:
TINT_UNREACHABLE();
}
auto fn = glsl::BuiltinFn::kTexture;
if (idx < args.Length()) {
// In GLSL ES `textureOffset` does not support depth 2d array textures. In order to
// support this texture we polyfill with a `textureGradOffset` and pass zero for
// the `dPdx` and `dPdy` values.
if (is_depth && is_array) {
fn = glsl::BuiltinFn::kTextureGradOffset;
params.Push(b.Zero(ty.vec2<f32>()));
params.Push(b.Zero(ty.vec2<f32>()));
} else {
fn = glsl::BuiltinFn::kTextureOffset;
}
params.Push(args[idx++]);
}
b.CallWithResult<glsl::ir::BuiltinCall>(call->DetachResult(), fn, params);
});
call->Destroy();
}
};
} // namespace
Result<SuccessType> TexturePolyfill(core::ir::Module& ir, const TexturePolyfillConfig& cfg) {
auto result = ValidateAndDumpIfNeeded(ir, "glsl.TexturePolyfill");
if (result != Success) {
return result.Failure();
}
State{ir, cfg}.Process();
return Success;
}
} // namespace tint::glsl::writer::raise