raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-20 02:47:02 +02:00
Code Issues Packages Projects Releases Wiki Activity

shader: support returning 32-bit LDS compare-exchange (DS_CMPST_RTN_B32) (#1578)

Browse Source 
* shader: support DS_CMPST_RTN_B32

Assassin's Creed Valhalla (PPSA01490, #281) stops on DS opcode 0x30 in a
compute shader. DS compares DATA and stores DATA2, the opposite of
BUFFER_ATOMIC_CMPSWAP, so the 32-bit atomic path swaps the two for this
opcode, as the 64-bit form does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* shader: reject unsupported GDS compare-store variant

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: nmzik <Nmzik@mail.ru>
This commit is contained in:
Nawid3333
2026-10-17 04:37:16 +02:00
committed by GitHub
parent 0fd41ae206
commit e93b4b0378
10 changed files with 97 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/ShaderRecompilerComputeTests.cpp
+58 -1
View File
@@ -21015,6 +21015,7 @@ CoverageClass ClassifyOpcode(ShaderOpcode opcode,
case Opcode::DS_ADD_U32:
case Opcode::DS_ADD_U64:
case Opcode::DS_MAX_RTN_U64:
case Opcode::DS_CMPST_RTN_B32:
case Opcode::DS_CMPST_RTN_B64:
case Opcode::DS_ADD_RTN_U32:
case Opcode::DS_SUB_U32:
@@ -34698,6 +34699,60 @@ TestCase DsAtomicReturn64() {
return test;
}
TestCase DsCompareStoreReturn32() {
using O = ShaderOpcode;
TestCase test;
test.name = "DsCompareStoreReturn32";
auto &code = test.code;
const auto result = [&](u32 reg, u32 value) {
AppendStoreVgpr(&code, reg, static_cast<u32>(test.expected.size()));
test.expected.push_back(value);
};
struct Case { u32 before, compare, replacement; };
constexpr std::array cases{Case{0x12345678u, 0x12345678u, 0x9abcdef0u},
Case{0x12345678u, 0x9abcdef0u, 0x12345678u},
Case{0x80000000u, 0u, 0x7fffffffu},
Case{UINT32_MAX, UINT32_MAX, 0u}};
AppendVMovU32(&code, 7, 0);
for (const auto &item : cases) {
AppendVMovLiteral(&code, 2, item.before);
code.push_back(EncodeDs0(0x0d, 8));
code.push_back(EncodeDs1(0, 2, 7));
AppendVMovLiteral(&code, 5, item.compare);
AppendVMovLiteral(&code, 19, item.replacement);
AppendVMovU32(&code, 4, 8);
// CS04b812daaacd532f pc 0x3e8: DATA compares, DATA2 replaces, VDst aliases ADDR.
code.insert(code.end(), {0xd8c00000u, 0x04130504u});
result(4, item.before);
code.push_back(EncodeDs0(0x36, 8));
code.push_back(EncodeDs1(8, 0, 7));
result(8, item.before == item.compare ? item.replacement : item.before);
}
AppendVMovU32(&code, 5, 0);
AppendVMovU32(&code, 4, 12); // Past the three LDS DWORDs.
code.insert(code.end(), {0xd8c00000u, 0x04130504u});
result(4, 0);
AppendVMovLiteral(&code, 2, 0x12345678u);
code.push_back(EncodeDs0(0x0d, 8));
code.push_back(EncodeDs1(0, 2, 7));
AppendVMovLiteral(&code, 5, 0x12345678u);
AppendVMovU32(&code, 4, 8);
code.push_back(EncodeSMovB32(126, InlineU32(0)));
code.insert(code.end(), {0xd8c00000u, 0x04130504u});
code.push_back(EncodeSMovB32(126, InlineU32(1)));
result(4, 8); // An inactive lane keeps its destination and leaves memory alone.
code.push_back(EncodeDs0(0x36, 8));
code.push_back(EncodeDs1(8, 0, 7));
result(8, 0x12345678u);
AppendEnd(&code);
test.compute_info.lds_size_dwords = 3;
test.opcodes = {O::V_MOV_B32, O::S_MOV_B32, O::DS_WRITE_B32, O::DS_CMPST_RTN_B32,
O::DS_READ_B32, O::BUFFER_STORE_DWORD, O::S_ENDPGM};
test.required_spirv = {"OpAtomicCompareExchange %uint"};
test.forbidden_spirv = {"OpLoopMerge"};
return test;
}
TestCase DsAtomic64Contention(bool add, u32 wave_size) {
using O = ShaderOpcode;
TestCase test;
@@ -42276,6 +42331,7 @@ std::vector<TestCase> MakeCases() {
}
AddCase(DsAtomicReturnVariants);
AddCase(DsAtomicReturn64);
AddCase(DsCompareStoreReturn32);
for (bool add : {false, true}) {
cases.push_back(DsAtomic64Bounds(add));
for (u32 wave_size : {32, 64}) {
@@ -47800,7 +47856,8 @@ int main(int argc, char **argv) {
RunCase(&vulkan, DsWideLdsPartialBounds());
RunCase(&vulkan, DsAtomic64Bounds(false));
RunCase(&vulkan, DsAtomic64Bounds(true));
for (auto test : {DsWideLdsPartialBounds(), DsAtomic64Bounds(false), DsAtomic64Bounds(true), DsAtomicReturn64()}) {
for (auto test : {DsWideLdsPartialBounds(), DsAtomic64Bounds(false), DsAtomic64Bounds(true), DsAtomicReturn64(),
DsCompareStoreReturn32()}) {
test.compute_info.lds_storage = true;
std::erase(test.required_spirv, "WorkgroupMemoryExplicitLayoutKHR");
test.required_spirv.push_back("StorageBuffer");