From 06095918a943f3d7bc8036525a8f4932e17d3a29 Mon Sep 17 00:00:00 2001
From: Billy Laws <blaws05@gmail.com>
Date: Sat, 12 Nov 2022 13:58:15 +0000
Subject: [PATCH] Introduce per-channel sequence number for invalidation
 tracking

For cases like shaders, which may be uploaded through I2M (which no longer causes an execution) we need a way to cause an invalidation on all writes
---
 app/src/main/cpp/skyline/soc/gm20b/channel.cpp              | 6 +++++-
 app/src/main/cpp/skyline/soc/gm20b/channel.h                | 1 +
 app/src/main/cpp/skyline/soc/gm20b/engines/gpfifo.cpp       | 3 +++
 .../main/cpp/skyline/soc/gm20b/engines/inline2memory.cpp    | 1 +
 4 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/app/src/main/cpp/skyline/soc/gm20b/channel.cpp b/app/src/main/cpp/skyline/soc/gm20b/channel.cpp
index 848c10d5..16d30428 100644
--- a/app/src/main/cpp/skyline/soc/gm20b/channel.cpp
+++ b/app/src/main/cpp/skyline/soc/gm20b/channel.cpp
@@ -13,5 +13,9 @@ namespace skyline::soc::gm20b {
           keplerCompute{state, *this},
           inline2Memory{state, *this},
           gpfifo{state, *this, numEntries},
-          globalChannelLock{state.gpu->channelLock} {}
+          globalChannelLock{state.gpu->channelLock} {
+        executor.AddFlushCallback([this] {
+            channelSequenceNumber++;
+        });
+    }
 }
diff --git a/app/src/main/cpp/skyline/soc/gm20b/channel.h b/app/src/main/cpp/skyline/soc/gm20b/channel.h
index b79a60b1..75f73cbd 100644
--- a/app/src/main/cpp/skyline/soc/gm20b/channel.h
+++ b/app/src/main/cpp/skyline/soc/gm20b/channel.h
@@ -31,6 +31,7 @@ namespace skyline::soc::gm20b {
         engine::Inline2Memory inline2Memory;
         ChannelGpfifo gpfifo;
         std::mutex &globalChannelLock;
+        size_t channelSequenceNumber{};
 
         ChannelContext(const DeviceState &state, std::shared_ptr<AddressSpaceContext> asCtx, size_t numEntries);
 
diff --git a/app/src/main/cpp/skyline/soc/gm20b/engines/gpfifo.cpp b/app/src/main/cpp/skyline/soc/gm20b/engines/gpfifo.cpp
index 1b760ef2..33477212 100644
--- a/app/src/main/cpp/skyline/soc/gm20b/engines/gpfifo.cpp
+++ b/app/src/main/cpp/skyline/soc/gm20b/engines/gpfifo.cpp
@@ -26,6 +26,7 @@ namespace skyline::soc::gm20b::engine {
 
                     // Wait forever for another channel to increment
 
+                    channelCtx.executor.Submit();
                     channelCtx.Unlock();
                     syncpoints.at(action.index).Wait(registers.syncpoint->payload, std::chrono::steady_clock::duration::max());
                     channelCtx.Lock();
@@ -44,6 +45,7 @@ namespace skyline::soc::gm20b::engine {
                 switch (action.operation) {
                     case Registers::Semaphore::Operation::Acquire:
                         Logger::Debug("Acquire semaphore: 0x{:X} payload: {}", address, registers.semaphore->payload);
+                        channelCtx.executor.Submit();
                         channelCtx.Unlock();
 
                         while (channelCtx.asCtx->gmmu.Read<u32>(address) != registers.semaphore->payload)
@@ -57,6 +59,7 @@ namespace skyline::soc::gm20b::engine {
                         break;
                     case Registers::Semaphore::Operation::AcqGeq    :
                         Logger::Debug("Acquire semaphore: 0x{:X} payload: {}", address, registers.semaphore->payload);
+                        channelCtx.executor.Submit();
                         channelCtx.Unlock();
 
                         while (channelCtx.asCtx->gmmu.Read<u32>(address) < registers.semaphore->payload)
diff --git a/app/src/main/cpp/skyline/soc/gm20b/engines/inline2memory.cpp b/app/src/main/cpp/skyline/soc/gm20b/engines/inline2memory.cpp
index 130f9744..22aa5e8a 100644
--- a/app/src/main/cpp/skyline/soc/gm20b/engines/inline2memory.cpp
+++ b/app/src/main/cpp/skyline/soc/gm20b/engines/inline2memory.cpp
@@ -21,6 +21,7 @@ namespace skyline::soc::gm20b::engine {
 
         if (state.launchDma.layout == RegisterState::DmaDstMemoryLayout::Pitch && state.lineCount == 1) {
             Logger::Debug("range: 0x{:X} -> 0x{:X}", u64{state.offsetOut}, u64{state.offsetOut} + buffer.size() * 0x4);
+            channelCtx.channelSequenceNumber++;
             interconnect.Upload(u64{state.offsetOut}, span{buffer});
         } else {
             channelCtx.executor.Submit();