summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/buffer_cache/buffer_cache.h11
-rw-r--r--src/video_core/dma_pusher.cpp19
-rw-r--r--src/video_core/dma_pusher.h8
-rw-r--r--src/video_core/engines/engine_upload.cpp4
-rw-r--r--src/video_core/engines/engine_upload.h7
-rw-r--r--src/video_core/engines/maxwell_dma.cpp34
-rw-r--r--src/video_core/engines/maxwell_dma.h8
-rw-r--r--src/video_core/host1x/vic.cpp6
-rw-r--r--src/video_core/host1x/vic.h7
9 files changed, 48 insertions, 56 deletions
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h
index 502b4d90a..6c8d98946 100644
--- a/src/video_core/buffer_cache/buffer_cache.h
+++ b/src/video_core/buffer_cache/buffer_cache.h
@@ -20,6 +20,7 @@
#include "common/lru_cache.h"
#include "common/microprofile.h"
#include "common/polyfill_ranges.h"
+#include "common/scratch_buffer.h"
#include "common/settings.h"
#include "core/memory.h"
#include "video_core/buffer_cache/buffer_base.h"
@@ -422,8 +423,7 @@ private:
IntervalSet common_ranges;
std::deque<IntervalSet> committed_ranges;
- size_t immediate_buffer_capacity = 0;
- std::unique_ptr<u8[]> immediate_buffer_alloc;
+ Common::ScratchBuffer<u8> immediate_buffer_alloc;
struct LRUItemParams {
using ObjectType = BufferId;
@@ -1926,11 +1926,8 @@ std::span<const u8> BufferCache<P>::ImmediateBufferWithData(VAddr cpu_addr, size
template <class P>
std::span<u8> BufferCache<P>::ImmediateBuffer(size_t wanted_capacity) {
- if (wanted_capacity > immediate_buffer_capacity) {
- immediate_buffer_capacity = wanted_capacity;
- immediate_buffer_alloc = std::make_unique<u8[]>(wanted_capacity);
- }
- return std::span<u8>(immediate_buffer_alloc.get(), wanted_capacity);
+ immediate_buffer_alloc.resize_destructive(wanted_capacity);
+ return std::span<u8>(immediate_buffer_alloc.data(), wanted_capacity);
}
template <class P>
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index 9835e3ac1..322de2606 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -56,7 +56,7 @@ bool DmaPusher::Step() {
if (command_list.prefetch_command_list.size()) {
// Prefetched command list from nvdrv, used for things like synchronization
- command_headers = std::move(command_list.prefetch_command_list);
+ ProcessCommands(command_list.prefetch_command_list);
dma_pushbuffer.pop();
} else {
const CommandListHeader command_list_header{
@@ -74,7 +74,7 @@ bool DmaPusher::Step() {
}
// Push buffer non-empty, read a word
- command_headers.resize(command_list_header.size);
+ command_headers.resize_destructive(command_list_header.size);
if (Settings::IsGPULevelHigh()) {
memory_manager.ReadBlock(dma_get, command_headers.data(),
command_list_header.size * sizeof(u32));
@@ -82,16 +82,21 @@ bool DmaPusher::Step() {
memory_manager.ReadBlockUnsafe(dma_get, command_headers.data(),
command_list_header.size * sizeof(u32));
}
+ ProcessCommands(command_headers);
}
- for (std::size_t index = 0; index < command_headers.size();) {
- const CommandHeader& command_header = command_headers[index];
+
+ return true;
+}
+
+void DmaPusher::ProcessCommands(std::span<const CommandHeader> commands) {
+ for (std::size_t index = 0; index < commands.size();) {
+ const CommandHeader& command_header = commands[index];
if (dma_state.method_count) {
// Data word of methods command
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);
+ std::min<std::size_t>(index + dma_state.method_count, commands.size()) - index);
CallMultiMethod(&command_header.argument, max_write);
dma_state.method_count -= max_write;
dma_state.is_last_call = true;
@@ -142,8 +147,6 @@ bool DmaPusher::Step() {
}
index++;
}
-
- return true;
}
void DmaPusher::SetState(const CommandHeader& command_header) {
diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h
index 938f0f11c..6f00de937 100644
--- a/src/video_core/dma_pusher.h
+++ b/src/video_core/dma_pusher.h
@@ -4,11 +4,13 @@
#pragma once
#include <array>
+#include <span>
#include <vector>
#include <queue>
#include "common/bit_field.h"
#include "common/common_types.h"
+#include "common/scratch_buffer.h"
#include "video_core/engines/engine_interface.h"
#include "video_core/engines/puller.h"
@@ -136,13 +138,15 @@ private:
static constexpr u32 non_puller_methods = 0x40;
static constexpr u32 max_subchannels = 8;
bool Step();
+ void ProcessCommands(std::span<const CommandHeader> commands);
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
+ Common::ScratchBuffer<CommandHeader>
+ command_headers; ///< Buffer for list of commands fetched at once
std::queue<CommandList> dma_pushbuffer; ///< Queue of command lists to be processed
std::size_t dma_pushbuffer_subindex{}; ///< Index within a command list within the pushbuffer
@@ -159,7 +163,7 @@ private:
DmaState dma_state{};
bool dma_increment_once{};
- bool ib_enable{true}; ///< IB mode enabled
+ const bool ib_enable{true}; ///< IB mode enabled
std::array<Engines::EngineInterface*, max_subchannels> subchannels{};
diff --git a/src/video_core/engines/engine_upload.cpp b/src/video_core/engines/engine_upload.cpp
index e4f8331ab..cea1dd8b0 100644
--- a/src/video_core/engines/engine_upload.cpp
+++ b/src/video_core/engines/engine_upload.cpp
@@ -24,7 +24,7 @@ void State::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
void State::ProcessExec(const bool is_linear_) {
write_offset = 0;
copy_size = regs.line_length_in * regs.line_count;
- inner_buffer.resize(copy_size);
+ inner_buffer.resize_destructive(copy_size);
is_linear = is_linear_;
}
@@ -70,7 +70,7 @@ void State::ProcessData(std::span<const u8> read_buffer) {
const std::size_t dst_size = Tegra::Texture::CalculateSize(
true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth,
regs.dest.BlockHeight(), regs.dest.BlockDepth());
- tmp_buffer.resize(dst_size);
+ tmp_buffer.resize_destructive(dst_size);
memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
Tegra::Texture::SwizzleSubrect(tmp_buffer, read_buffer, bytes_per_pixel, width,
regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
diff --git a/src/video_core/engines/engine_upload.h b/src/video_core/engines/engine_upload.h
index 94fafd9dc..7242d2529 100644
--- a/src/video_core/engines/engine_upload.h
+++ b/src/video_core/engines/engine_upload.h
@@ -4,9 +4,10 @@
#pragma once
#include <span>
-#include <vector>
+
#include "common/bit_field.h"
#include "common/common_types.h"
+#include "common/scratch_buffer.h"
namespace Tegra {
class MemoryManager;
@@ -73,8 +74,8 @@ private:
u32 write_offset = 0;
u32 copy_size = 0;
- std::vector<u8> inner_buffer;
- std::vector<u8> tmp_buffer;
+ Common::ScratchBuffer<u8> inner_buffer;
+ Common::ScratchBuffer<u8> tmp_buffer;
bool is_linear = false;
Registers& regs;
MemoryManager& memory_manager;
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index a189e60ae..f73d7bf0f 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -184,12 +184,8 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
const size_t src_size =
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
- if (read_buffer.size() < src_size) {
- read_buffer.resize(src_size);
- }
- if (write_buffer.size() < dst_size) {
- write_buffer.resize(dst_size);
- }
+ read_buffer.resize_destructive(src_size);
+ write_buffer.resize_destructive(dst_size);
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
@@ -235,12 +231,8 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
const size_t src_size = static_cast<size_t>(regs.pitch_in) * regs.line_count;
- if (read_buffer.size() < src_size) {
- read_buffer.resize(src_size);
- }
- if (write_buffer.size() < dst_size) {
- write_buffer.resize(dst_size);
- }
+ read_buffer.resize_destructive(src_size);
+ write_buffer.resize_destructive(dst_size);
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
if (Settings::IsGPULevelExtreme()) {
@@ -269,12 +261,8 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() {
pos_x = pos_x % x_in_gob;
pos_y = pos_y % 8;
- if (read_buffer.size() < src_size) {
- read_buffer.resize(src_size);
- }
- if (write_buffer.size() < dst_size) {
- write_buffer.resize(dst_size);
- }
+ read_buffer.resize_destructive(src_size);
+ write_buffer.resize_destructive(dst_size);
if (Settings::IsGPULevelExtreme()) {
memory_manager.ReadBlock(regs.offset_in + offset, read_buffer.data(), src_size);
@@ -333,14 +321,10 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
const u32 pitch = x_elements * bytes_per_pixel;
const size_t mid_buffer_size = pitch * regs.line_count;
- if (read_buffer.size() < src_size) {
- read_buffer.resize(src_size);
- }
- if (write_buffer.size() < dst_size) {
- write_buffer.resize(dst_size);
- }
+ read_buffer.resize_destructive(src_size);
+ write_buffer.resize_destructive(dst_size);
- intermediate_buffer.resize(mid_buffer_size);
+ intermediate_buffer.resize_destructive(mid_buffer_size);
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h
index d40d3d302..c88191a61 100644
--- a/src/video_core/engines/maxwell_dma.h
+++ b/src/video_core/engines/maxwell_dma.h
@@ -6,8 +6,10 @@
#include <array>
#include <cstddef>
#include <vector>
+
#include "common/bit_field.h"
#include "common/common_types.h"
+#include "common/scratch_buffer.h"
#include "video_core/engines/engine_interface.h"
namespace Core {
@@ -234,9 +236,9 @@ private:
MemoryManager& memory_manager;
VideoCore::RasterizerInterface* rasterizer = nullptr;
- std::vector<u8> read_buffer;
- std::vector<u8> write_buffer;
- std::vector<u8> intermediate_buffer;
+ Common::ScratchBuffer<u8> read_buffer;
+ Common::ScratchBuffer<u8> write_buffer;
+ Common::ScratchBuffer<u8> intermediate_buffer;
static constexpr std::size_t NUM_REGS = 0x800;
struct Regs {
diff --git a/src/video_core/host1x/vic.cpp b/src/video_core/host1x/vic.cpp
index ac0b7d20e..36a04e4e0 100644
--- a/src/video_core/host1x/vic.cpp
+++ b/src/video_core/host1x/vic.cpp
@@ -155,7 +155,7 @@ void Vic::WriteRGBFrame(const AVFrame* frame, const VicConfig& config) {
// swizzle pitch linear to block linear
const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
const auto size = Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
- luma_buffer.resize(size);
+ luma_buffer.resize_destructive(size);
std::span<const u8> frame_buff(converted_frame_buf_addr, 4 * width * height);
Texture::SwizzleSubrect(luma_buffer, frame_buff, 4, width, height, 1, 0, 0, width, height,
block_height, 0, width * 4);
@@ -181,8 +181,8 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) {
const auto stride = static_cast<size_t>(frame->linesize[0]);
- luma_buffer.resize(aligned_width * surface_height);
- chroma_buffer.resize(aligned_width * surface_height / 2);
+ luma_buffer.resize_destructive(aligned_width * surface_height);
+ chroma_buffer.resize_destructive(aligned_width * surface_height / 2);
// Populate luma buffer
const u8* luma_src = frame->data[0];
diff --git a/src/video_core/host1x/vic.h b/src/video_core/host1x/vic.h
index 2b78786e8..3d9753047 100644
--- a/src/video_core/host1x/vic.h
+++ b/src/video_core/host1x/vic.h
@@ -4,8 +4,9 @@
#pragma once
#include <memory>
-#include <vector>
+
#include "common/common_types.h"
+#include "common/scratch_buffer.h"
struct SwsContext;
@@ -49,8 +50,8 @@ private:
/// size does not change during a stream
using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>;
AVMallocPtr converted_frame_buffer;
- std::vector<u8> luma_buffer;
- std::vector<u8> chroma_buffer;
+ Common::ScratchBuffer<u8> luma_buffer;
+ Common::ScratchBuffer<u8> chroma_buffer;
GPUVAddr config_struct_address{};
GPUVAddr output_surface_luma_address{};
generated by cgit v1.2.3 (git 2.25.1) at 2024-06-26 13:32:59 +0200