mirror of
https://github.com/KytyPS5/KytyPS5.git
synced 2026-10-20 19:07:01 +02:00
shader: support returning 32-bit LDS compare-exchange (DS_CMPST_RTN_B32) (#1578)
* 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:
@@ -115,6 +115,7 @@ constexpr MemoryOpcodeInfo DS_OPCODE_LIST[] = {
|
||||
{0x27u, Opcode::DS_MIN_RTN_U32, 1, 32}, {0x28u, Opcode::DS_MAX_RTN_U32, 1, 32},
|
||||
{0x29u, Opcode::DS_AND_RTN_B32, 1, 32}, {0x2au, Opcode::DS_OR_RTN_B32, 1, 32},
|
||||
{0x2bu, Opcode::DS_XOR_RTN_B32, 1, 32}, {0x2du, Opcode::DS_WRXCHG_RTN_B32, 1, 32},
|
||||
{0x30u, Opcode::DS_CMPST_RTN_B32, 1, 32},
|
||||
{0x35u, Opcode::DS_SWIZZLE_B32, 1, 32}, {0x36u, Opcode::DS_READ_B32, 1, 32},
|
||||
{0x37u, Opcode::DS_READ2_B32, 2, 32}, {0x38u, Opcode::DS_READ2ST64_B32, 2, 32},
|
||||
{0x39u, Opcode::DS_READ_I8, 1, 8, true}, {0x3au, Opcode::DS_READ_U8, 1, 8},
|
||||
@@ -201,6 +202,7 @@ bool IsDsAtomicOpcode(Opcode opcode) {
|
||||
case Opcode::DS_MAX_U32:
|
||||
case Opcode::DS_MAX_RTN_U32:
|
||||
case Opcode::DS_MAX_RTN_U64:
|
||||
case Opcode::DS_CMPST_RTN_B32:
|
||||
case Opcode::DS_CMPST_RTN_B64:
|
||||
case Opcode::DS_AND_B32:
|
||||
case Opcode::DS_AND_RTN_B32:
|
||||
@@ -221,6 +223,7 @@ uint32_t DsSourceCount(Opcode opcode) {
|
||||
case Opcode::DS_WRITE2ST64_B32:
|
||||
case Opcode::DS_WRITE2_B64:
|
||||
case Opcode::DS_WRITE2ST64_B64:
|
||||
case Opcode::DS_CMPST_RTN_B32:
|
||||
case Opcode::DS_CMPST_RTN_B64:
|
||||
case Opcode::DS_MSKOR_B32: return 3u;
|
||||
case Opcode::DS_MIN_F32:
|
||||
@@ -455,6 +458,9 @@ void DecodeDs(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
|
||||
if (inst.opcode == Opcode::DS_SWIZZLE_B32 && inst.offset >= 0xe000u) {
|
||||
SetUnsupported(inst, Family::DS, opcode, "DS swizzle FFT mode is not implemented");
|
||||
}
|
||||
if (inst.gds && inst.opcode == Opcode::DS_CMPST_RTN_B32) {
|
||||
SetUnsupported(inst, Family::DS, opcode, "DS compare-store return B32 GDS variant is not implemented");
|
||||
}
|
||||
if (inst.gds &&
|
||||
(inst.opcode == Opcode::DS_SWIZZLE_B32 || inst.opcode == Opcode::DS_PERMUTE_B32 ||
|
||||
inst.opcode == Opcode::DS_BPERMUTE_B32 ||
|
||||
|
||||
@@ -718,6 +718,7 @@ std::string InstructionToString(const Instruction& inst) {
|
||||
case Opcode::DS_MAX_U32:
|
||||
case Opcode::DS_MAX_RTN_U32:
|
||||
case Opcode::DS_MAX_RTN_U64:
|
||||
case Opcode::DS_CMPST_RTN_B32:
|
||||
case Opcode::DS_CMPST_RTN_B64:
|
||||
case Opcode::DS_AND_B32:
|
||||
case Opcode::DS_AND_RTN_B32:
|
||||
|
||||
@@ -632,6 +632,7 @@ enum class Opcode {
|
||||
DS_MAX_U32,
|
||||
DS_MAX_RTN_U32,
|
||||
DS_MAX_RTN_U64,
|
||||
DS_CMPST_RTN_B32,
|
||||
DS_CMPST_RTN_B64,
|
||||
DS_AND_B32,
|
||||
DS_AND_RTN_B32,
|
||||
|
||||
@@ -1085,6 +1085,8 @@ void Translator::EmitMemory(const Decoder::Instruction& inst) {
|
||||
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicUMax32, true);
|
||||
case Decoder::Opcode::DS_MAX_RTN_U64:
|
||||
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicUMax64, true);
|
||||
case Decoder::Opcode::DS_CMPST_RTN_B32:
|
||||
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicCmpSwap32, true);
|
||||
case Decoder::Opcode::DS_CMPST_RTN_B64:
|
||||
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicCmpSwap64, true);
|
||||
case Decoder::Opcode::DS_AND_B32:
|
||||
|
||||
Reference in New Issue
Block a user