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:
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
@@ -178,6 +178,7 @@ SharedAccess SharedAccessOf(ValueOpcode opcode) {
case ValueOpcode::SharedAtomicSMax32:
case ValueOpcode::SharedAtomicUMax32:
case ValueOpcode::SharedAtomicUMax64:
case ValueOpcode::SharedAtomicCmpSwap32:
case ValueOpcode::SharedAtomicCmpSwap64:
case ValueOpcode::SharedAtomicAnd32:
case ValueOpcode::SharedAtomicOr32:
@@ -307,6 +307,7 @@ VALUE_OPCODE(SharedAtomicUMin32, U32, U32, U32, U1)
VALUE_OPCODE(SharedAtomicSMax32, U32, U32, U32, U1)
VALUE_OPCODE(SharedAtomicUMax32, U32, U32, U32, U1)
VALUE_OPCODE(SharedAtomicUMax64, U64, U32, U64, U1)
VALUE_OPCODE(SharedAtomicCmpSwap32, U32, U32, U32, U32, U1)
VALUE_OPCODE(SharedAtomicCmpSwap64, U64, U32, U64, U64, U1)
VALUE_OPCODE(SharedAtomicAnd32, U32, U32, U32, U1)
VALUE_OPCODE(SharedAtomicOr32, U32, U32, U32, U1)