Merge pull request #3828 from CouchRiv/fix-shader-combined-pknorm-f16
fix(shader): accept abs, neg and op_sel source modifiers on v_cvt_pknorm_i16_f16 and v_cvt_pknorm_u16_f16
This commit is contained in:
Adrià
GitHub
commit
e53c30bcdf
|
|
@ -0,0 +1,174 @@
|
|||
#include "prx/libSceAgcDriver/Execution/include/VulkanDevice.hpp"
|
||||
#include "prx/libSceAgcDriver/Graphics/include/Draw.hpp"
|
||||
#include "Recompiler.hpp"
|
||||
#include "VulkanTestDevice.hpp"
|
||||
#include <algorithm>
|
||||
#include <array>
|
||||
#include <cstdint>
|
||||
#include <cstdio>
|
||||
#include <iostream>
|
||||
#include <stdexcept>
|
||||
#include <string>
|
||||
#include <vector>
|
||||
|
||||
namespace {
|
||||
|
||||
using AgcDriver::Graphics::Require;
|
||||
using ShaderRecompiler::ShaderStage;
|
||||
|
||||
constexpr std::uint32_t Threads = 32;
|
||||
constexpr std::uint32_t Inputs = 4;
|
||||
constexpr std::uint32_t Results = 2;
|
||||
constexpr std::uint32_t RowCount = 8;
|
||||
constexpr std::uint32_t VariantCount = 8;
|
||||
alignas(256) std::array<std::uint32_t, Threads * Inputs> Input{};
|
||||
alignas(256) std::array<std::uint32_t, Threads * Results> Output{};
|
||||
|
||||
alignas(256) std::array<std::uint32_t, 14> Code{
|
||||
0x34020084u, 0x34060083u, 0xe0381000u, 0x80000401u, 0xbf8c3f70u, 0xd712000au, 0x00020b04u,
|
||||
0xd713000bu, 0x00020b04u, 0xe0701000u, 0x80010a03u, 0xe0701004u, 0x80010b03u, 0xbf810000u,
|
||||
};
|
||||
|
||||
struct Variant {
|
||||
const char* operands;
|
||||
std::uint32_t abs;
|
||||
std::uint32_t neg;
|
||||
std::uint32_t opSel;
|
||||
};
|
||||
|
||||
constexpr Variant Variants[VariantCount] = {
|
||||
{"|v4|, v5", 1, 0, 0},
|
||||
{"v4, |v5|", 2, 0, 0},
|
||||
{"-v4, v5", 0, 1, 0},
|
||||
{"v4, -v5", 0, 2, 0},
|
||||
{"-|v4|, v5", 1, 1, 0},
|
||||
{"v4, -|v5|", 2, 2, 0},
|
||||
{"v4, v5 op_sel:[1,0,0]", 0, 0, 1},
|
||||
{"v4, v5 op_sel:[0,1,0]", 0, 0, 2},
|
||||
};
|
||||
|
||||
constexpr std::uint32_t Rows[RowCount][2] = {
|
||||
{0x00000400u, 0x00000400u},
|
||||
{0x000083ffu, 0x000083ffu},
|
||||
{0x00003555u, 0x0000d640u},
|
||||
{0x0000bc00u, 0x00003c00u},
|
||||
{0x00007bffu, 0x00000400u},
|
||||
{0xbeef0000u, 0xbeef3c00u},
|
||||
{0xbeef8000u, 0xbeef0000u},
|
||||
{0xbeef0000u, 0xbeeffc00u},
|
||||
};
|
||||
|
||||
constexpr std::uint32_t Expected[VariantCount][Results][RowCount] = {
|
||||
{
|
||||
{0x00020002u, 0xfffe0002u, 0x80012aa8u, 0x7fff7fffu, 0x00027fffu, 0x7fff0000u, 0x00000000u, 0x80010000u},
|
||||
{0x00040004u, 0x00000004u, 0x00005550u, 0xffffffffu, 0x0004ffffu, 0xffff0000u, 0x00000000u, 0x00000000u},
|
||||
},
|
||||
{
|
||||
{0x00020002u, 0x0002fffeu, 0x7fff2aa8u, 0x7fff8001u, 0x00027fffu, 0x7fff0000u, 0x00000000u, 0x7fff0000u},
|
||||
{0x00040004u, 0x00040000u, 0xffff5550u, 0xffff0000u, 0x0004ffffu, 0xffff0000u, 0x00000000u, 0xffff0000u},
|
||||
},
|
||||
{
|
||||
{0x0002fffeu, 0xfffe0002u, 0x8001d558u, 0x7fff7fffu, 0x00028001u, 0x7fff0000u, 0x00000000u, 0x80010000u},
|
||||
{0x00040000u, 0x00000004u, 0x00000000u, 0xffffffffu, 0x00040000u, 0xffff0000u, 0x00000000u, 0x00000000u},
|
||||
},
|
||||
{
|
||||
{0xfffe0002u, 0x0002fffeu, 0x7fff2aa8u, 0x80018001u, 0xfffe7fffu, 0x80010000u, 0x00000000u, 0x7fff0000u},
|
||||
{0x00000004u, 0x00040000u, 0xffff5550u, 0x00000000u, 0x0000ffffu, 0x00000000u, 0x00000000u, 0xffff0000u},
|
||||
},
|
||||
{
|
||||
{0x0002fffeu, 0xfffefffeu, 0x8001d558u, 0x7fff8001u, 0x00028001u, 0x7fff0000u, 0x00000000u, 0x80010000u},
|
||||
{0x00040000u, 0x00000000u, 0x00000000u, 0xffff0000u, 0x00040000u, 0xffff0000u, 0x00000000u, 0x00000000u},
|
||||
},
|
||||
{
|
||||
{0xfffe0002u, 0xfffefffeu, 0x80012aa8u, 0x80018001u, 0xfffe7fffu, 0x80010000u, 0x00000000u, 0x80010000u},
|
||||
{0x00000004u, 0x00000000u, 0x00005550u, 0x00000000u, 0x0000ffffu, 0x00000000u, 0x00000000u, 0x00000000u},
|
||||
},
|
||||
{
|
||||
{0x00020000u, 0xfffe0000u, 0x80010000u, 0x7fff0000u, 0x00020000u, 0x7fff8001u, 0x00008001u, 0x80018001u},
|
||||
{0x00040000u, 0x00000000u, 0x00000000u, 0xffff0000u, 0x00040000u, 0xffff0000u, 0x00000000u, 0x00000000u},
|
||||
},
|
||||
{
|
||||
{0x00000002u, 0x0000fffeu, 0x00002aa8u, 0x00008001u, 0x00007fffu, 0x80010000u, 0x80010000u, 0x80010000u},
|
||||
{0x00000004u, 0x00000000u, 0x00005550u, 0x00000000u, 0x0000ffffu, 0x00000000u, 0x00000000u, 0x00000000u},
|
||||
},
|
||||
};
|
||||
|
||||
constexpr const char* Opcodes[Results] = {"v_cvt_pknorm_i16_f16_e64 v10, ", "v_cvt_pknorm_u16_f16_e64 v11, "};
|
||||
|
||||
std::array<std::uint32_t, 4> BufferDescriptor(const void* data, std::uint32_t bytes) {
|
||||
const auto address = reinterpret_cast<std::uintptr_t>(data);
|
||||
return {static_cast<std::uint32_t>(address), static_cast<std::uint32_t>((address >> 32u) & 0xffffu), bytes, 0x31016facu};
|
||||
}
|
||||
|
||||
std::string Hex(std::uint32_t value) {
|
||||
char text[16];
|
||||
std::snprintf(text, sizeof(text), "0x%08x", value);
|
||||
return text;
|
||||
}
|
||||
|
||||
void Run(AgcDriver::VulkanDevice& device, ShaderRecompiler::ShaderFloatMode mode, const Variant& variant) {
|
||||
const std::uint32_t first = 0xd7120000u | variant.opSel << 11u | variant.abs << 8u | 10u;
|
||||
const std::uint32_t second = 0xd7130000u | variant.opSel << 11u | variant.abs << 8u | 11u;
|
||||
const std::uint32_t sources = 0x00020b04u | variant.neg << 29u;
|
||||
Code[5] = first;
|
||||
Code[6] = sources;
|
||||
Code[7] = second;
|
||||
Code[8] = sources;
|
||||
for (std::uint32_t index = 0; index < RowCount; ++index) {
|
||||
std::copy(std::begin(Rows[index]), std::end(Rows[index]), &Input[index * Inputs]);
|
||||
}
|
||||
for (std::uint32_t index = RowCount * Inputs; index < Input.size(); ++index) Input[index] = 0u;
|
||||
Output.fill(0xdeadbeefu);
|
||||
std::vector<std::uint32_t> userData(8, 0u);
|
||||
const auto input = BufferDescriptor(Input.data(), static_cast<std::uint32_t>(Input.size() * 4u));
|
||||
const auto output = BufferDescriptor(Output.data(), static_cast<std::uint32_t>(Output.size() * 4u));
|
||||
std::copy(input.begin(), input.end(), userData.begin());
|
||||
std::copy(output.begin(), output.end(), userData.begin() + 4);
|
||||
const std::span<const std::uint32_t> code(Code);
|
||||
const std::array<ShaderRecompiler::MemoryRegion, 1> memory{{{reinterpret_cast<std::uintptr_t>(code.data()), std::as_bytes(code)}}};
|
||||
const ShaderRecompiler::ShaderComputeStageInfo compute{{Threads, 1, 1}, 0u, {false, false, false}, false, 1};
|
||||
ShaderRecompiler::RecompileRequest request{
|
||||
{ShaderStage::Compute, reinterpret_cast<std::uintptr_t>(code.data()), code, 0, {}},
|
||||
{32, 0, userData, compute, std::nullopt, std::nullopt, memory},
|
||||
device.Target(),
|
||||
{0, 0, 0, 128}
|
||||
};
|
||||
request.useCache = false;
|
||||
request.context.floatMode = mode;
|
||||
const auto result = ShaderRecompiler::Recompile(request);
|
||||
device.Dispatch(result, 1, 1, 1, {}, reinterpret_cast<std::uintptr_t>(code.data()));
|
||||
device.WaitIdle();
|
||||
}
|
||||
|
||||
void Check(const char* modeName, std::uint32_t variantIndex) {
|
||||
for (std::uint32_t slot = 0; slot < Results; ++slot) {
|
||||
for (std::uint32_t index = 0; index < RowCount; ++index) {
|
||||
const std::uint32_t actual = Output[index * Results + slot];
|
||||
const std::uint32_t expected = Expected[variantIndex][slot][index];
|
||||
Require(actual == expected,
|
||||
std::string("v_cvt_pknorm f16 source modifiers: ") + modeName + " row " + std::to_string(index) + " " +
|
||||
Opcodes[slot] + Variants[variantIndex].operands + " of " + Hex(Rows[index][0]) + " and " +
|
||||
Hex(Rows[index][1]) + " is " + Hex(actual) + ", expected " + Hex(expected));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
int main() {
|
||||
try {
|
||||
const auto device = OpenVulkanTestDevice();
|
||||
if (!device) return VulkanTestSkipped;
|
||||
for (std::uint32_t index = 0; index < VariantCount; ++index) {
|
||||
Run(*device, ShaderRecompiler::ShaderFloatMode{0xf0u, true, true, false}, Variants[index]);
|
||||
Check("FLOAT_MODE 0xf0", index);
|
||||
Run(*device, ShaderRecompiler::ShaderFloatMode{0xc0u, true, false, false}, Variants[index]);
|
||||
Check("FLOAT_MODE 0xc0", index);
|
||||
}
|
||||
std::puts("v_cvt_pknorm f16 source modifiers tests passed");
|
||||
return 0;
|
||||
} catch (const std::exception& error) {
|
||||
std::cerr << error.what() << '\n';
|
||||
return 1;
|
||||
}
|
||||
}
|
||||
|
|
@ -740,6 +740,10 @@ bool isNativeVop3F16TernaryOpcode(RdnaOpcode opcode) {
|
|||
opcode == RdnaOpcode::VFmaF16;
|
||||
}
|
||||
|
||||
bool isNativeVop3F16PackOpcode(RdnaOpcode opcode) {
|
||||
return opcode == RdnaOpcode::VPackB32F16 || opcode == RdnaOpcode::VCvtPknormI16F16 || opcode == RdnaOpcode::VCvtPknormU16F16;
|
||||
}
|
||||
|
||||
bool isNativeVop3I16TernaryOpcode(RdnaOpcode opcode) {
|
||||
return opcode == RdnaOpcode::VMed3I16 || opcode == RdnaOpcode::VMed3U16 || opcode == RdnaOpcode::VMax3I16 ||
|
||||
opcode == RdnaOpcode::VMax3U16 || opcode == RdnaOpcode::VMin3I16 || opcode == RdnaOpcode::VMin3U16 ||
|
||||
|
|
@ -1864,7 +1868,7 @@ void applyNativeVop3B16BinaryModifiers(RdnaInstruction& instruction, std::uint32
|
|||
instruction.destination.sdwaSel = (opSel & 0x8u) != 0u ? 5u : 4u;
|
||||
}
|
||||
|
||||
void applyNativeVop3PackB32F16Modifiers(RdnaInstruction& instruction, std::uint32_t opSel, std::uint32_t abs, std::uint32_t neg) {
|
||||
void applyNativeVop3F16PackModifiers(RdnaInstruction& instruction, std::uint32_t opSel, std::uint32_t abs, std::uint32_t neg) {
|
||||
instruction.source0.opSel = (opSel & 0x1u) != 0u;
|
||||
instruction.source1.opSel = (opSel & 0x2u) != 0u;
|
||||
instruction.source0.negate = (neg & 0x1u) != 0u;
|
||||
|
|
@ -2053,7 +2057,7 @@ void checkNativeVop3Modifiers(RdnaOpcode opcode, bool permlane, bool carryInOut,
|
|||
}
|
||||
return;
|
||||
}
|
||||
if (opcode == RdnaOpcode::VPackB32F16) {
|
||||
if (isNativeVop3F16PackOpcode(opcode)) {
|
||||
if ((abs & ~0x3u) != 0u || (opSel & ~0x3u) != 0u || clamp != 0u || omod != 0u || (neg & ~0x3u) != 0u) {
|
||||
throw std::invalid_argument("VOP3 source modifiers are not implemented");
|
||||
}
|
||||
|
|
@ -2430,8 +2434,8 @@ RdnaInstruction DecodeRdnaVop3(std::uint32_t programCounter, std::span<const std
|
|||
applyNativeVop3AlignSelectors(instruction, opSel);
|
||||
} else if (isNativeVop3B16BinaryOpcode(instruction.op)) {
|
||||
applyNativeVop3B16BinaryModifiers(instruction, opSel);
|
||||
} else if (instruction.op == RdnaOpcode::VPackB32F16) {
|
||||
applyNativeVop3PackB32F16Modifiers(instruction, opSel, abs, neg);
|
||||
} else if (isNativeVop3F16PackOpcode(instruction.op)) {
|
||||
applyNativeVop3F16PackModifiers(instruction, opSel, abs, neg);
|
||||
} else if (nativeSourceModifiers) {
|
||||
applyNativeVop3SourceModifiers(instruction, abs, neg);
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in New Issue