summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorFernando Sahmkow <fsahmkow27@gmail.com>2020-04-20 08:16:56 +0200
committerFernando Sahmkow <fsahmkow27@gmail.com>2020-04-23 14:52:55 +0200
commit3fedcc2f6e001f0ed1fd791de4f9692570359eef (patch)
tree49109516beab33d825cc653d4e885107304da332
parentMerge pull request #3730 from lioncash/time (diff)
downloadyuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.tar
yuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.tar.gz
yuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.tar.bz2
yuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.tar.lz
yuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.tar.xz
yuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.tar.zst
yuzu-3fedcc2f6e001f0ed1fd791de4f9692570359eef.zip
-rw-r--r--src/video_core/dma_pusher.cpp30
-rw-r--r--src/video_core/dma_pusher.h1
-rw-r--r--src/video_core/engines/fermi_2d.cpp6
-rw-r--r--src/video_core/engines/fermi_2d.h3
-rw-r--r--src/video_core/engines/kepler_compute.cpp6
-rw-r--r--src/video_core/engines/kepler_compute.h3
-rw-r--r--src/video_core/engines/kepler_memory.cpp6
-rw-r--r--src/video_core/engines/kepler_memory.h3
-rw-r--r--src/video_core/engines/maxwell_3d.cpp52
-rw-r--r--src/video_core/engines/maxwell_3d.h4
-rw-r--r--src/video_core/engines/maxwell_dma.cpp6
-rw-r--r--src/video_core/engines/maxwell_dma.h3
-rw-r--r--src/video_core/gpu.cpp47
-rw-r--r--src/video_core/gpu.h8
14 files changed, 164 insertions, 14 deletions
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index 324dafdcd..16311f05e 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -71,16 +71,22 @@ bool DmaPusher::Step() {
gpu.MemoryManager().ReadBlockUnsafe(dma_get, command_headers.data(),
command_list_header.size * sizeof(u32));
- for (const CommandHeader& command_header : command_headers) {
-
- // now, see if we're in the middle of a command
- if (dma_state.length_pending) {
- // Second word of long non-inc methods command - method count
- dma_state.length_pending = 0;
- dma_state.method_count = command_header.method_count_;
- } else if (dma_state.method_count) {
+ for (std::size_t index = 0; index < command_headers.size();) {
+ const CommandHeader& command_header = command_headers[index];
+
+ if (dma_state.method_count) {
// Data word of methods command
- CallMethod(command_header.argument);
+ if (dma_state.non_incrementing) {
+ const u32 max_write = static_cast<u32>(
+ std::min<std::size_t>(index + dma_state.method_count, command_headers.size()) -
+ index);
+ CallMultiMethod(&command_header.argument, max_write);
+ dma_state.method_count -= max_write;
+ index += max_write;
+ continue;
+ } else {
+ CallMethod(command_header.argument);
+ }
if (!dma_state.non_incrementing) {
dma_state.method++;
@@ -120,6 +126,7 @@ bool DmaPusher::Step() {
break;
}
}
+ index++;
}
if (!non_main) {
@@ -140,4 +147,9 @@ void DmaPusher::CallMethod(u32 argument) const {
gpu.CallMethod({dma_state.method, argument, dma_state.subchannel, dma_state.method_count});
}
+void DmaPusher::CallMultiMethod(const u32* base_start, u32 num_methods) const {
+ gpu.CallMultiMethod(dma_state.method, dma_state.subchannel, base_start, num_methods,
+ dma_state.method_count);
+}
+
} // namespace Tegra
diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h
index d6188614a..6cef71306 100644
--- a/src/video_core/dma_pusher.h
+++ b/src/video_core/dma_pusher.h
@@ -75,6 +75,7 @@ private:
void SetState(const CommandHeader& command_header);
void CallMethod(u32 argument) const;
+ void CallMultiMethod(const u32* base_start, u32 num_methods) const;
std::vector<CommandHeader> command_headers; ///< Buffer for list of commands fetched at once
diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp
index bace6affb..8a47614d2 100644
--- a/src/video_core/engines/fermi_2d.cpp
+++ b/src/video_core/engines/fermi_2d.cpp
@@ -28,6 +28,12 @@ void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
}
}
+void Fermi2D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+ for (std::size_t i = 0; i < amount; i++) {
+ CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+ }
+}
+
static std::pair<u32, u32> DelimitLine(u32 src_1, u32 src_2, u32 dst_1, u32 dst_2, u32 src_line) {
const u32 line_a = src_2 - src_1;
const u32 line_b = dst_2 - dst_1;
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index dba342c70..939a5966d 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -39,6 +39,9 @@ public:
/// Write the value to the register identified by method.
void CallMethod(const GPU::MethodCall& method_call);
+ /// Write multiple values to the register identified by method.
+ void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
enum class Origin : u32 {
Center = 0,
Corner = 1,
diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp
index 368c75a66..894300f57 100644
--- a/src/video_core/engines/kepler_compute.cpp
+++ b/src/video_core/engines/kepler_compute.cpp
@@ -51,6 +51,12 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
}
}
+void KeplerCompute::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+ for (std::size_t i = 0; i < amount; i++) {
+ CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+ }
+}
+
Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) const {
const std::bitset<8> cbuf_mask = launch_description.const_buffer_enable_mask.Value();
ASSERT(cbuf_mask[regs.tex_cb_index]);
diff --git a/src/video_core/engines/kepler_compute.h b/src/video_core/engines/kepler_compute.h
index eeb79c56f..fe55fdfd0 100644
--- a/src/video_core/engines/kepler_compute.h
+++ b/src/video_core/engines/kepler_compute.h
@@ -202,6 +202,9 @@ public:
/// Write the value to the register identified by method.
void CallMethod(const GPU::MethodCall& method_call);
+ /// Write multiple values to the register identified by method.
+ void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
Texture::FullTextureInfo GetTexture(std::size_t offset) const;
/// Given a texture handle, returns the TSC and TIC entries.
diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp
index 597872e43..e906a1124 100644
--- a/src/video_core/engines/kepler_memory.cpp
+++ b/src/video_core/engines/kepler_memory.cpp
@@ -41,4 +41,10 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) {
}
}
+void KeplerMemory::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+ for (std::size_t i = 0; i < amount; i++) {
+ CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+ }
+}
+
} // namespace Tegra::Engines
diff --git a/src/video_core/engines/kepler_memory.h b/src/video_core/engines/kepler_memory.h
index 396fb6e86..bb26fb030 100644
--- a/src/video_core/engines/kepler_memory.h
+++ b/src/video_core/engines/kepler_memory.h
@@ -40,6 +40,9 @@ public:
/// Write the value to the register identified by method.
void CallMethod(const GPU::MethodCall& method_call);
+ /// Write multiple values to the register identified by method.
+ void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
struct Regs {
static constexpr size_t NUM_REGS = 0x7F;
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 2824ed707..879ce542a 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -280,6 +280,36 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
}
}
+void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+ switch (method) {
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[3]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[4]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[5]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[6]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[7]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[8]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[9]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[10]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[11]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[12]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[13]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[14]):
+ case MAXWELL3D_REG_INDEX(const_buffer.cb_data[15]): {
+ ProcessCBMultiData(method, base_start, amount);
+ break;
+ }
+ default: {
+ for (std::size_t i = 0; i < amount; i++) {
+ CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+ }
+ }
+ }
+
+}
+
void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
if (mme_draw.current_mode == MMEDrawMode::Undefined) {
if (mme_draw.gl_begin_consume) {
@@ -570,6 +600,28 @@ void Maxwell3D::StartCBData(u32 method) {
ProcessCBData(regs.const_buffer.cb_data[cb_data_state.id]);
}
+void Maxwell3D::ProcessCBMultiData(u32 method, const u32* start_base, u32 amount) {
+ if (cb_data_state.current != method) {
+ if (cb_data_state.current != null_cb_data) {
+ FinishCBData();
+ }
+ constexpr u32 first_cb_data = MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]);
+ cb_data_state.start_pos = regs.const_buffer.cb_pos;
+ cb_data_state.id = method - first_cb_data;
+ cb_data_state.current = method;
+ cb_data_state.counter = 0;
+ }
+ const std::size_t id = cb_data_state.id;
+ const std::size_t size = amount;
+ std::size_t i = 0;
+ for (; i < size; i++) {
+ cb_data_state.buffer[id][cb_data_state.counter] = start_base[i];
+ cb_data_state.counter++;
+ }
+ // Increment the current buffer position.
+ regs.const_buffer.cb_pos = regs.const_buffer.cb_pos + 4 * amount;
+}
+
void Maxwell3D::FinishCBData() {
// Write the input value to the current const buffer at the current position.
const GPUVAddr buffer_address = regs.const_buffer.BufferAddress();
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 59d5752d2..cfcda4f53 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -1358,6 +1358,9 @@ public:
/// Write the value to the register identified by method.
void CallMethod(const GPU::MethodCall& method_call);
+ /// Write multiple values to the register identified by method.
+ void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
/// Write the value to the register identified by method.
void CallMethodFromMME(const GPU::MethodCall& method_call);
@@ -1511,6 +1514,7 @@ private:
/// Handles a write to the CB_DATA[i] register.
void StartCBData(u32 method);
void ProcessCBData(u32 value);
+ void ProcessCBMultiData(u32 method, const u32* start_base, u32 amount);
void FinishCBData();
/// Handles a write to the CB_BIND register.
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index 3bfed6ab8..51e606a10 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -36,6 +36,12 @@ void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
#undef MAXWELLDMA_REG_INDEX
}
+void MaxwellDMA::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+ for (std::size_t i = 0; i < amount; i++) {
+ CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+ }
+}
+
void MaxwellDMA::HandleCopy() {
LOG_TRACE(HW_GPU, "Requested a DMA copy");
diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h
index 4f40d1d1f..c43ed8194 100644
--- a/src/video_core/engines/maxwell_dma.h
+++ b/src/video_core/engines/maxwell_dma.h
@@ -35,6 +35,9 @@ public:
/// Write the value to the register identified by method.
void CallMethod(const GPU::MethodCall& method_call);
+ /// Write multiple values to the register identified by method.
+ void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
struct Regs {
static constexpr std::size_t NUM_REGS = 0x1D6;
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 3b7572d61..54e7876a6 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -209,16 +209,31 @@ void GPU::CallMethod(const MethodCall& method_call) {
ASSERT(method_call.subchannel < bound_engines.size());
- if (ExecuteMethodOnEngine(method_call)) {
+ if (ExecuteMethodOnEngine(method_call.method)) {
CallEngineMethod(method_call);
} else {
CallPullerMethod(method_call);
}
}
-bool GPU::ExecuteMethodOnEngine(const MethodCall& method_call) {
- const auto method = static_cast<BufferMethods>(method_call.method);
- return method >= BufferMethods::NonPullerMethods;
+void GPU::CallMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount, u32 methods_pending) {
+ LOG_TRACE(HW_GPU, "Processing method {:08X} on subchannel {}", method,
+ subchannel);
+
+ ASSERT(subchannel < bound_engines.size());
+
+ if (ExecuteMethodOnEngine(method)) {
+ CallEngineMultiMethod(method, subchannel, base_start, amount, methods_pending);
+ } else {
+ for (std::size_t i = 0; i < amount; i++) {
+ CallPullerMethod({method, base_start[i], subchannel, methods_pending - static_cast<u32>(i)});
+ }
+ }
+}
+
+bool GPU::ExecuteMethodOnEngine(u32 method) {
+ const auto buffer_method = static_cast<BufferMethods>(method);
+ return buffer_method >= BufferMethods::NonPullerMethods;
}
void GPU::CallPullerMethod(const MethodCall& method_call) {
@@ -298,6 +313,30 @@ void GPU::CallEngineMethod(const MethodCall& method_call) {
}
}
+void GPU::CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount, u32 methods_pending) {
+ const EngineID engine = bound_engines[subchannel];
+
+ switch (engine) {
+ case EngineID::FERMI_TWOD_A:
+ fermi_2d->CallMultiMethod(method, base_start, amount, methods_pending);
+ break;
+ case EngineID::MAXWELL_B:
+ maxwell_3d->CallMultiMethod(method, base_start, amount, methods_pending);
+ break;
+ case EngineID::KEPLER_COMPUTE_B:
+ kepler_compute->CallMultiMethod(method, base_start, amount, methods_pending);
+ break;
+ case EngineID::MAXWELL_DMA_COPY_A:
+ maxwell_dma->CallMultiMethod(method, base_start, amount, methods_pending);
+ break;
+ case EngineID::KEPLER_INLINE_TO_MEMORY_B:
+ kepler_memory->CallMultiMethod(method, base_start, amount, methods_pending);
+ break;
+ default:
+ UNIMPLEMENTED_MSG("Unimplemented engine");
+ }
+}
+
void GPU::ProcessBindMethod(const MethodCall& method_call) {
// Bind the current subchannel to the desired engine id.
LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel,
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 5e3eb94e9..4d7e2651c 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -155,6 +155,9 @@ public:
/// Calls a GPU method.
void CallMethod(const MethodCall& method_call);
+ /// Calls a GPU multivalue method.
+ void CallMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount, u32 methods_pending);
+
/// Flush all current written commands into the host GPU for execution.
void FlushCommands();
/// Synchronizes CPU writes with Host GPU memory.
@@ -309,8 +312,11 @@ private:
/// Calls a GPU engine method.
void CallEngineMethod(const MethodCall& method_call);
+ /// Calls a GPU engine multivalue method.
+ void CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount, u32 methods_pending);
+
/// Determines where the method should be executed.
- bool ExecuteMethodOnEngine(const MethodCall& method_call);
+ bool ExecuteMethodOnEngine(u32 method);
protected:
std::unique_ptr<Tegra::DmaPusher> dma_pusher;