summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp13
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvdec.h3
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h3
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_vic.cpp10
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_vic.h3
-rw-r--r--src/video_core/gpu.cpp44
-rw-r--r--src/video_core/gpu.h4
8 files changed, 45 insertions, 40 deletions
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
index 0d7d4ad03..8314d1ec2 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
@@ -20,8 +20,12 @@ NvResult nvhost_nvdec::Ioctl1(DeviceFD fd, Ioctl command, const std::vector<u8>&
switch (command.group) {
case 0x0:
switch (command.cmd) {
- case 0x1:
- return Submit(input, output);
+ case 0x1: {
+ if (!fd_to_id.contains(fd)) {
+ fd_to_id[fd] = next_id++;
+ }
+ return Submit(fd, input, output);
+ }
case 0x2:
return GetSyncpoint(input, output);
case 0x3:
@@ -66,7 +70,10 @@ void nvhost_nvdec::OnOpen(DeviceFD fd) {}
void nvhost_nvdec::OnClose(DeviceFD fd) {
LOG_INFO(Service_NVDRV, "NVDEC video stream ended");
- system.GPU().ClearCdmaInstance();
+ const auto iter = fd_to_id.find(fd);
+ if (iter != fd_to_id.end()) {
+ system.GPU().ClearCdmaInstance(iter->second);
+ }
}
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
index 523d96e3a..a507c4d0a 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
@@ -24,6 +24,9 @@ public:
void OnOpen(DeviceFD fd) override;
void OnClose(DeviceFD fd) override;
+
+private:
+ u32 next_id{};
};
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
index e61261f98..8a05f0668 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
@@ -59,7 +59,8 @@ NvResult nvhost_nvdec_common::SetNVMAPfd(const std::vector<u8>& input) {
return NvResult::Success;
}
-NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
+NvResult nvhost_nvdec_common::Submit(DeviceFD fd, const std::vector<u8>& input,
+ std::vector<u8>& output) {
IoctlSubmit params{};
std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
LOG_DEBUG(Service_NVDRV, "called NVDEC Submit, cmd_buffer_count={}", params.cmd_buffer_count);
@@ -93,7 +94,7 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
system.Memory().ReadBlock(object->addr + cmd_buffer.offset, cmdlist.data(),
cmdlist.size() * sizeof(u32));
- gpu.PushCommandBuffer(cmdlist);
+ gpu.PushCommandBuffer(fd_to_id[fd], cmdlist);
}
std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
// Some games expect command_buffers to be written back
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
index 351625c17..e28c54df6 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
@@ -104,13 +104,14 @@ protected:
/// Ioctl command implementations
NvResult SetNVMAPfd(const std::vector<u8>& input);
- NvResult Submit(const std::vector<u8>& input, std::vector<u8>& output);
+ NvResult Submit(DeviceFD fd, const std::vector<u8>& input, std::vector<u8>& output);
NvResult GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
NvResult GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
NvResult MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
NvResult UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
NvResult SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output);
+ std::unordered_map<DeviceFD, u32> fd_to_id{};
s32_le nvmap_fd{};
u32_le submit_timeout{};
std::shared_ptr<nvmap> nvmap_dev;
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
index eac4dd530..76b39806f 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
@@ -21,7 +21,10 @@ NvResult nvhost_vic::Ioctl1(DeviceFD fd, Ioctl command, const std::vector<u8>& i
case 0x0:
switch (command.cmd) {
case 0x1:
- return Submit(input, output);
+ if (!fd_to_id.contains(fd)) {
+ fd_to_id[fd] = next_id++;
+ }
+ return Submit(fd, input, output);
case 0x2:
return GetSyncpoint(input, output);
case 0x3:
@@ -65,7 +68,10 @@ NvResult nvhost_vic::Ioctl3(DeviceFD fd, Ioctl command, const std::vector<u8>& i
void nvhost_vic::OnOpen(DeviceFD fd) {}
void nvhost_vic::OnClose(DeviceFD fd) {
- system.GPU().ClearCdmaInstance();
+ const auto iter = fd_to_id.find(fd);
+ if (iter != fd_to_id.end()) {
+ system.GPU().ClearCdmaInstance(iter->second);
+ }
}
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.h b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
index 6d7fda9d1..c9732c037 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
@@ -23,5 +23,8 @@ public:
void OnOpen(DeviceFD fd) override;
void OnClose(DeviceFD fd) override;
+
+private:
+ u32 next_id{};
};
} // namespace Service::Nvidia::Devices
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index ab7c21a49..8788f5148 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -185,16 +185,6 @@ struct GPU::Impl {
return *dma_pusher;
}
- /// Returns a reference to the GPU CDMA pusher.
- [[nodiscard]] Tegra::CDmaPusher& CDmaPusher() {
- return *cdma_pusher;
- }
-
- /// Returns a const reference to the GPU CDMA pusher.
- [[nodiscard]] const Tegra::CDmaPusher& CDmaPusher() const {
- return *cdma_pusher;
- }
-
/// Returns a reference to the underlying renderer.
[[nodiscard]] VideoCore::RendererBase& Renderer() {
return *renderer;
@@ -338,25 +328,27 @@ struct GPU::Impl {
}
/// Push GPU command buffer entries to be processed
- void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
+ void PushCommandBuffer(u32 id, Tegra::ChCommandHeaderList& entries) {
if (!use_nvdec) {
return;
}
- if (!cdma_pusher) {
- cdma_pusher = std::make_unique<Tegra::CDmaPusher>(gpu);
+ if (!cdma_pushers.contains(id)) {
+ cdma_pushers.insert_or_assign(id, std::make_unique<Tegra::CDmaPusher>(gpu));
}
// SubmitCommandBuffer would make the nvdec operations async, this is not currently working
// TODO(ameerj): RE proper async nvdec operation
// gpu_thread.SubmitCommandBuffer(std::move(entries));
-
- cdma_pusher->ProcessEntries(std::move(entries));
+ cdma_pushers[id]->ProcessEntries(std::move(entries));
}
/// Frees the CDMAPusher instance to free up resources
- void ClearCdmaInstance() {
- cdma_pusher.reset();
+ void ClearCdmaInstance(u32 id) {
+ const auto iter = cdma_pushers.find(id);
+ if (iter != cdma_pushers.end()) {
+ cdma_pushers.erase(iter);
+ }
}
/// Swap buffers (render frame)
@@ -659,7 +651,7 @@ struct GPU::Impl {
Core::System& system;
std::unique_ptr<Tegra::MemoryManager> memory_manager;
std::unique_ptr<Tegra::DmaPusher> dma_pusher;
- std::unique_ptr<Tegra::CDmaPusher> cdma_pusher;
+ std::map<u32, std::unique_ptr<Tegra::CDmaPusher>> cdma_pushers;
std::unique_ptr<VideoCore::RendererBase> renderer;
VideoCore::RasterizerInterface* rasterizer = nullptr;
const bool use_nvdec;
@@ -811,14 +803,6 @@ const Tegra::DmaPusher& GPU::DmaPusher() const {
return impl->DmaPusher();
}
-Tegra::CDmaPusher& GPU::CDmaPusher() {
- return impl->CDmaPusher();
-}
-
-const Tegra::CDmaPusher& GPU::CDmaPusher() const {
- return impl->CDmaPusher();
-}
-
VideoCore::RendererBase& GPU::Renderer() {
return impl->Renderer();
}
@@ -887,12 +871,12 @@ void GPU::PushGPUEntries(Tegra::CommandList&& entries) {
impl->PushGPUEntries(std::move(entries));
}
-void GPU::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
- impl->PushCommandBuffer(entries);
+void GPU::PushCommandBuffer(u32 id, Tegra::ChCommandHeaderList& entries) {
+ impl->PushCommandBuffer(id, entries);
}
-void GPU::ClearCdmaInstance() {
- impl->ClearCdmaInstance();
+void GPU::ClearCdmaInstance(u32 id) {
+ impl->ClearCdmaInstance(id);
}
void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index c89a5d693..500411176 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -242,10 +242,10 @@ public:
void PushGPUEntries(Tegra::CommandList&& entries);
/// Push GPU command buffer entries to be processed
- void PushCommandBuffer(Tegra::ChCommandHeaderList& entries);
+ void PushCommandBuffer(u32 id, Tegra::ChCommandHeaderList& entries);
/// Frees the CDMAPusher instance to free up resources
- void ClearCdmaInstance();
+ void ClearCdmaInstance(u32 id);
/// Swap buffers (render frame)
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);