shader: took fixes from PR 427

- apply c8f34f9's ResourceTracking portability fix
- complete V_CMP_LT_U64 and S_FLBIT_I32_B32 support
- add the sceAmprCommandBufferWaitOnAddress NID from f625a47
- add decoder and compute tests for both shader operations

Co-authored-by: Mac <me@itsmemac.com>

src/graphics/shader/recompiler/frontend/decode/ScalarAluOps.cpp

@@ -45,6 +45,7 @@ constexpr OpcodeMap SOP1_OPCODE_LIST[] = {
     {0x10u, Opcode::S_BCNT1_I32_B64},
     {0x13u, Opcode::S_FF1_I32_B32},
     {0x14u, Opcode::S_FF1_I32_B64},
+    {0x15u, Opcode::S_FLBIT_I32_B32},
     {0x16u, Opcode::S_FLBIT_I32_B64},
     {0x1bu, Opcode::S_BITSET0_B32},
     {0x1du, Opcode::S_BITSET1_B32},

src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.cpp

@@ -476,6 +476,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::S_ABS_I32:
 		case Opcode::S_BREV_B32:
 		case Opcode::S_BCNT1_I32_B32:
+		case Opcode::S_FLBIT_I32_B32:
 		case Opcode::S_FF1_I32_B32:
 		case Opcode::S_FF1_I32_B64:
 		case Opcode::S_NOT_B64:

src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.h

@@ -45,6 +45,7 @@ enum class Opcode {
 	S_BCNT1_I32_B64,
 	S_FF1_I32_B32,
 	S_FF1_I32_B64,
+	S_FLBIT_I32_B32,
 	S_FLBIT_I32_B64,
 	S_BITREPLICATE_B64_B32,
 	S_QUADMASK_B64,
@@ -395,6 +396,7 @@ enum class Opcode {
 	V_CMP_GE_U32,
 	V_CMP_T_U32,
 	V_CMP_EQ_I64,
+	V_CMP_LT_U64,
 	V_CMP_EQ_U64,
 	V_CMP_GT_U64,
 	V_CMP_NE_U64,

src/graphics/shader/recompiler/frontend/decode/VectorAluOps.cpp

@@ -240,8 +240,9 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
     {0xb5u, Opcode::V_CMPX_NE_I64, false}, {0xd1u, Opcode::V_CMPX_LT_U32},
     {0xd2u, Opcode::V_CMPX_EQ_U32},        {0xd3u, Opcode::V_CMPX_LE_U32},
     {0xd4u, Opcode::V_CMPX_GT_U32},        {0xd5u, Opcode::V_CMPX_NE_U32},
-    {0xd6u, Opcode::V_CMPX_GE_U32},        {0xe2u, Opcode::V_CMP_EQ_U64, false},
-    {0xe4u, Opcode::V_CMP_GT_U64, false},  {0xe5u, Opcode::V_CMP_NE_U64, false},
+    {0xd6u, Opcode::V_CMPX_GE_U32},        {0xe1u, Opcode::V_CMP_LT_U64, false},
+    {0xe2u, Opcode::V_CMP_EQ_U64, false},  {0xe4u, Opcode::V_CMP_GT_U64, false},
+    {0xe5u, Opcode::V_CMP_NE_U64, false},
     {0xf5u, Opcode::V_CMPX_NE_U64, false}, {0xc9u, Opcode::V_CMP_LT_F16},
     {0xcau, Opcode::V_CMP_EQ_F16},         {0xcbu, Opcode::V_CMP_LE_F16},
     {0xccu, Opcode::V_CMP_GT_F16},         {0xcdu, Opcode::V_CMP_LG_F16},

src/graphics/shader/recompiler/frontend/translate/Scalar.cpp

@@ -181,6 +181,7 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst) {
 		case O::S_XNOR_B32:
 			return ComposedIntegerBinary(inst, IR::ValueOpcode::BitwiseXor32, false, true, true);
 		case O::S_FF1_I32_B64: return S_FF1_I32_B64(inst);
+		case O::S_FLBIT_I32_B32: return V_FFBH_32(inst, false);
 		case O::S_FLBIT_I32_B64: return S_FLBIT_I32_B64(inst);
 
 		case O::S_BITSET0_B32: return S_BITSET_B32(inst, false);

src/graphics/shader/recompiler/frontend/translate/Vector.cpp

@@ -99,6 +99,9 @@ bool Translator::EmitVector(const Decoder::Instruction& inst) {
 		case O::V_CMP_EQ_U64:
 			EmitIntegerCompare(inst, IR::ValueOpcode::IEqual64, IR::Type::U64, false, false);
 			return true;
+		case O::V_CMP_LT_U64:
+			EmitIntegerCompare(inst, IR::ValueOpcode::ULessThan64, IR::Type::U64, false, false);
+			return true;
 		case O::V_CMP_GT_U64:
 			EmitIntegerCompare(inst, IR::ValueOpcode::UGreaterThan64, IR::Type::U64, false, false);
 			return true;

src/graphics/shader/recompiler/ir/passes/ResourceTracking.cpp

@@ -139,7 +139,8 @@ public:
 		}
 		std::erase_if(m_program.dynamic_reads, [&](Value value) {
 			const auto* inst = value.Resolve().TryInstruction();
-			return std::ranges::any_of(m_indirect_images, [&](const IndirectImagePlan& plan) {
+			return std::any_of(m_indirect_images.begin(), m_indirect_images.end(),
+			                   [&](const IndirectImagePlan& plan) {
 				return std::ranges::find(plan.reads, inst) != plan.reads.end();
 			});
 		});
@@ -432,14 +433,16 @@ private:
 
 	const IndirectImagePlan* FindIndirectImage(const Inst& handle) const {
 		const auto found =
-		    std::ranges::find_if(m_indirect_images, [&](const IndirectImagePlan& plan) {
+		    std::find_if(m_indirect_images.begin(), m_indirect_images.end(),
+		                 [&](const IndirectImagePlan& plan) {
 			    return plan.handle == &handle;
 		    });
 		return found == m_indirect_images.end() ? nullptr : &*found;
 	}
 
 	bool IsIndirectPlanningMemory(uint32_t index) const {
-		return std::ranges::any_of(m_indirect_images, [&](const IndirectImagePlan& plan) {
+		return std::any_of(m_indirect_images.begin(), m_indirect_images.end(),
+		                   [&](const IndirectImagePlan& plan) {
 			return std::ranges::find(plan.memory, index) != plan.memory.end();
 		});
 	}

src/libs/libAmpr.cpp

@@ -2665,6 +2665,7 @@ LIB_DEFINE(InitAmpr_1_Ampr) {
 	LIB_FUNC("dXPaz65HNmk", Ampr::CommandBufferPushMarker);
 	LIB_FUNC("mv0O8Zg0woU", Ampr::CommandBufferPopMarker);
 	LIB_FUNC("DLfoNxTFNVk", Ampr::CommandBufferWaitOnAddress);
+	LIB_FUNC("V7GQTEeUfhw", Ampr::CommandBufferWaitOnAddress);
 	LIB_FUNC("cQb8Zr8Q0Y0", Ampr::CommandBufferWaitOnCounter);
 	LIB_FUNC("j0+3uJMxYJY", Ampr::CommandBufferWriteAddress);
 	LIB_FUNC("sJXyWHjP-F8", Ampr::CommandBufferWriteAddressOnCompletion);