From 91d29459fb55cb0d28006639e7a38134c5a368ec Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Marcin=20Miko=C5=82ajczyk?= <marcinmikolajcz@gmail.com>
Date: Fri, 6 Jun 2025 05:19:05 +0200
Subject: [PATCH] Implement PM4CondExec (#3046)

---
 src/video_core/amdgpu/liverpool.cpp | 13 +++++++++++++
 src/video_core/amdgpu/pm4_cmds.h    | 21 +++++++++++++++++++++
 2 files changed, 34 insertions(+)

diff --git a/src/video_core/amdgpu/liverpool.cpp b/src/video_core/amdgpu/liverpool.cpp
index 118c43cef..e031d0ebc 100644
--- a/src/video_core/amdgpu/liverpool.cpp
+++ b/src/video_core/amdgpu/liverpool.cpp
@@ -765,6 +765,19 @@ Liverpool::Task Liverpool::ProcessGraphics(std::span<const u32> dcb, std::span<c
                 LOG_WARNING(Render_Vulkan, "Unimplemented IT_GET_LOD_STATS");
                 break;
             }
+            case PM4ItOpcode::CondExec: {
+                const auto* cond_exec = reinterpret_cast<const PM4CmdCondExec*>(header);
+                if (cond_exec->command.Value() != 0) {
+                    LOG_WARNING(Render, "IT_COND_EXEC used a reserved command");
+                }
+                const auto skip = *cond_exec->Address() == false;
+                if (skip) {
+                    dcb = NextPacket(dcb,
+                                     header->type3.NumWords() + 1 + cond_exec->exec_count.Value());
+                    continue;
+                }
+                break;
+            }
             default:
                 UNREACHABLE_MSG("Unknown PM4 type 3 opcode {:#x} with count {}",
                                 static_cast<u32>(opcode), count);
diff --git a/src/video_core/amdgpu/pm4_cmds.h b/src/video_core/amdgpu/pm4_cmds.h
index 011e47bf0..23c1b8f21 100644
--- a/src/video_core/amdgpu/pm4_cmds.h
+++ b/src/video_core/amdgpu/pm4_cmds.h
@@ -1159,4 +1159,25 @@ struct PM4CmdMemSemaphore {
     }
 };
 
+struct PM4CmdCondExec {
+    PM4Type3Header header;
+    union {
+        BitField<2, 30, u32> bool_addr_lo; ///< low 32 address bits for the block in memory from
+                                           ///< where the CP will fetch the condition
+    };
+    union {
+        BitField<0, 16, u32> bool_addr_hi; ///< high address bits for the condition
+        BitField<28, 4, u32> command;
+    };
+    union {
+        BitField<0, 14, u32> exec_count; ///< Number of DWords that the CP will skip
+                                         ///< if bool pointed to is zero
+    };
+
+    bool* Address() const {
+        return std::bit_cast<bool*>(u64(bool_addr_hi.Value()) << 32 | u64(bool_addr_lo.Value())
+                                                                          << 2);
+    }
+};
+
 } // namespace AmdGpu