summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp25
-rw-r--r--src/video_core/macro/macro_jit_x64.cpp62
-rw-r--r--src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp84
-rw-r--r--src/video_core/renderer_vulkan/vk_staging_buffer_pool.h3
-rw-r--r--src/web_service/web_backend.cpp3
5 files changed, 97 insertions, 80 deletions
diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp b/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp
index b5c08d611..7e8f37563 100644
--- a/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp
+++ b/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp
@@ -13,9 +13,6 @@ namespace Shader::Backend::GLASM {
namespace {
void GetCbuf(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, ScalarU32 offset,
std::string_view size) {
- if (!binding.IsImmediate()) {
- throw NotImplementedException("Indirect constant buffer loading");
- }
const Register ret{ctx.reg_alloc.Define(inst)};
if (offset.type == Type::U32) {
// Avoid reading arrays out of bounds, matching hardware's behavior
@@ -24,7 +21,27 @@ void GetCbuf(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, ScalarU
return;
}
}
- ctx.Add("LDC.{} {},c{}[{}];", size, ret, binding.U32(), offset);
+
+ if (binding.IsImmediate()) {
+ ctx.Add("LDC.{} {},c{}[{}];", size, ret, binding.U32(), offset);
+ return;
+ }
+
+ const ScalarU32 idx{ctx.reg_alloc.Consume(binding)};
+ for (u32 i = 0; i < Info::MAX_INDIRECT_CBUFS; i++) {
+ ctx.Add("SEQ.S.CC RC.x,{},{};"
+ "IF NE.x;"
+ "LDC.{} {},c{}[{}];",
+ idx, i, size, ret, i, offset);
+
+ if (i != Info::MAX_INDIRECT_CBUFS - 1) {
+ ctx.Add("ELSE;");
+ }
+ }
+
+ for (u32 i = 0; i < Info::MAX_INDIRECT_CBUFS; i++) {
+ ctx.Add("ENDIF;");
+ }
}
bool IsInputArray(Stage stage) {
diff --git a/src/video_core/macro/macro_jit_x64.cpp b/src/video_core/macro/macro_jit_x64.cpp
index aca25d902..a302a9603 100644
--- a/src/video_core/macro/macro_jit_x64.cpp
+++ b/src/video_core/macro/macro_jit_x64.cpp
@@ -279,28 +279,13 @@ void MacroJITx64Impl::Compile_ExtractInsert(Macro::Opcode opcode) {
auto dst = Compile_GetRegister(opcode.src_a, RESULT);
auto src = Compile_GetRegister(opcode.src_b, eax);
- if (opcode.bf_src_bit != 0 && opcode.bf_src_bit != 31) {
- shr(src, opcode.bf_src_bit);
- } else if (opcode.bf_src_bit == 31) {
- xor_(src, src);
- }
- // Don't bother masking the whole register since we're using a 32 bit register
- if (opcode.bf_size != 31 && opcode.bf_size != 0) {
- and_(src, opcode.GetBitfieldMask());
- } else if (opcode.bf_size == 0) {
- xor_(src, src);
- }
- if (opcode.bf_dst_bit != 31 && opcode.bf_dst_bit != 0) {
- shl(src, opcode.bf_dst_bit);
- } else if (opcode.bf_dst_bit == 31) {
- xor_(src, src);
- }
-
const u32 mask = ~(opcode.GetBitfieldMask() << opcode.bf_dst_bit);
- if (mask != 0xffffffff) {
- and_(dst, mask);
- }
+ and_(dst, mask);
+ shr(src, opcode.bf_src_bit);
+ and_(src, opcode.GetBitfieldMask());
+ shl(src, opcode.bf_dst_bit);
or_(dst, src);
+
Compile_ProcessResult(opcode.result_operation, opcode.dst);
}
@@ -309,17 +294,9 @@ void MacroJITx64Impl::Compile_ExtractShiftLeftImmediate(Macro::Opcode opcode) {
const auto src = Compile_GetRegister(opcode.src_b, RESULT);
shr(src, dst.cvt8());
- if (opcode.bf_size != 0 && opcode.bf_size != 31) {
- and_(src, opcode.GetBitfieldMask());
- } else if (opcode.bf_size == 0) {
- xor_(src, src);
- }
+ and_(src, opcode.GetBitfieldMask());
+ shl(src, opcode.bf_dst_bit);
- if (opcode.bf_dst_bit != 0 && opcode.bf_dst_bit != 31) {
- shl(src, opcode.bf_dst_bit);
- } else if (opcode.bf_dst_bit == 31) {
- xor_(src, src);
- }
Compile_ProcessResult(opcode.result_operation, opcode.dst);
}
@@ -327,13 +304,8 @@ void MacroJITx64Impl::Compile_ExtractShiftLeftRegister(Macro::Opcode opcode) {
const auto dst = Compile_GetRegister(opcode.src_a, ecx);
const auto src = Compile_GetRegister(opcode.src_b, RESULT);
- if (opcode.bf_src_bit != 0) {
- shr(src, opcode.bf_src_bit);
- }
-
- if (opcode.bf_size != 31) {
- and_(src, opcode.GetBitfieldMask());
- }
+ shr(src, opcode.bf_src_bit);
+ and_(src, opcode.GetBitfieldMask());
shl(src, dst.cvt8());
Compile_ProcessResult(opcode.result_operation, opcode.dst);
@@ -429,17 +401,11 @@ void MacroJITx64Impl::Compile_Branch(Macro::Opcode opcode) {
Xbyak::Label handle_post_exit{};
Xbyak::Label skip{};
jmp(skip, T_NEAR);
- if (opcode.is_exit) {
- L(handle_post_exit);
- // Execute 1 instruction
- mov(BRANCH_HOLDER, end_of_code);
- // Jump to next instruction to skip delay slot check
- jmp(labels[jump_address], T_NEAR);
- } else {
- L(handle_post_exit);
- xor_(BRANCH_HOLDER, BRANCH_HOLDER);
- jmp(labels[jump_address], T_NEAR);
- }
+
+ L(handle_post_exit);
+ xor_(BRANCH_HOLDER, BRANCH_HOLDER);
+ jmp(labels[jump_address], T_NEAR);
+
L(skip);
mov(BRANCH_HOLDER, handle_post_exit);
jmp(delay_skip[pc], T_NEAR);
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
index 06f68d09a..7fb256953 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -26,20 +26,39 @@ using namespace Common::Literals;
constexpr VkDeviceSize MAX_ALIGNMENT = 256;
// Maximum size to put elements in the stream buffer
constexpr VkDeviceSize MAX_STREAM_BUFFER_REQUEST_SIZE = 8_MiB;
-// Stream buffer size in bytes
-constexpr VkDeviceSize STREAM_BUFFER_SIZE = 128_MiB;
-constexpr VkDeviceSize REGION_SIZE = STREAM_BUFFER_SIZE / StagingBufferPool::NUM_SYNCS;
constexpr VkMemoryPropertyFlags HOST_FLAGS =
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
constexpr VkMemoryPropertyFlags STREAM_FLAGS = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT | HOST_FLAGS;
-bool IsStreamHeap(VkMemoryHeap heap) noexcept {
- return STREAM_BUFFER_SIZE < (heap.size * 2) / 3;
+static bool IsStreamHeap(VkMemoryHeap heap, size_t staging_buffer_size) noexcept {
+ return staging_buffer_size < (heap.size * 2) / 3;
+}
+
+static bool HasLargeDeviceLocalHostVisibleMemory(const VkPhysicalDeviceMemoryProperties& props) {
+ const auto flags{VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT | VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT};
+
+ for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
+ const auto& memory_type{props.memoryTypes[type_index]};
+
+ if ((memory_type.propertyFlags & flags) != flags) {
+ // Memory must be device local and host visible
+ continue;
+ }
+
+ const auto& heap{props.memoryHeaps[memory_type.heapIndex]};
+ if (heap.size >= 7168_MiB) {
+ // This is the right type of memory
+ return true;
+ }
+ }
+
+ return false;
}
std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_mask,
- VkMemoryPropertyFlags flags) noexcept {
+ VkMemoryPropertyFlags flags,
+ size_t staging_buffer_size) noexcept {
for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
if (((type_mask >> type_index) & 1) == 0) {
// Memory type is incompatible
@@ -50,7 +69,7 @@ std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& p
// Memory type doesn't have the flags we want
continue;
}
- if (!IsStreamHeap(props.memoryHeaps[memory_type.heapIndex])) {
+ if (!IsStreamHeap(props.memoryHeaps[memory_type.heapIndex], staging_buffer_size)) {
// Memory heap is not suitable for streaming
continue;
}
@@ -61,17 +80,17 @@ std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& p
}
u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_mask,
- bool try_device_local) {
+ bool try_device_local, size_t staging_buffer_size) {
std::optional<u32> type;
if (try_device_local) {
// Try to find a DEVICE_LOCAL_BIT type, Nvidia and AMD have a dedicated heap for this
- type = FindMemoryTypeIndex(props, type_mask, STREAM_FLAGS);
+ type = FindMemoryTypeIndex(props, type_mask, STREAM_FLAGS, staging_buffer_size);
if (type) {
return *type;
}
}
// Otherwise try without the DEVICE_LOCAL_BIT
- type = FindMemoryTypeIndex(props, type_mask, HOST_FLAGS);
+ type = FindMemoryTypeIndex(props, type_mask, HOST_FLAGS, staging_buffer_size);
if (type) {
return *type;
}
@@ -79,20 +98,32 @@ u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_
throw vk::Exception(VK_ERROR_OUT_OF_DEVICE_MEMORY);
}
-size_t Region(size_t iterator) noexcept {
- return iterator / REGION_SIZE;
+size_t Region(size_t iterator, size_t region_size) noexcept {
+ return iterator / region_size;
}
} // Anonymous namespace
StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
Scheduler& scheduler_)
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
+
+ const auto memory_properties{device.GetPhysical().GetMemoryProperties().memoryProperties};
+ if (HasLargeDeviceLocalHostVisibleMemory(memory_properties)) {
+ // Possible on many integrated and newer discrete cards
+ staging_buffer_size = 1_GiB;
+ } else {
+ // Well-supported default size used by most Vulkan PC games
+ staging_buffer_size = 256_MiB;
+ }
+
+ region_size = staging_buffer_size / StagingBufferPool::NUM_SYNCS;
+
const vk::Device& dev = device.GetLogical();
stream_buffer = dev.CreateBuffer(VkBufferCreateInfo{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .size = STREAM_BUFFER_SIZE,
+ .size = staging_buffer_size,
.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
@@ -117,19 +148,18 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
.image = nullptr,
.buffer = *stream_buffer,
};
- const auto memory_properties = device.GetPhysical().GetMemoryProperties().memoryProperties;
VkMemoryAllocateInfo stream_memory_info{
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = make_dedicated ? &dedicated_info : nullptr,
.allocationSize = requirements.size,
- .memoryTypeIndex =
- FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, true),
+ .memoryTypeIndex = FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, true,
+ staging_buffer_size),
};
stream_memory = dev.TryAllocateMemory(stream_memory_info);
if (!stream_memory) {
LOG_INFO(Render_Vulkan, "Dynamic memory allocation failed, trying with system memory");
- stream_memory_info.memoryTypeIndex =
- FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, false);
+ stream_memory_info.memoryTypeIndex = FindMemoryTypeIndex(
+ memory_properties, requirements.memoryTypeBits, false, staging_buffer_size);
stream_memory = dev.AllocateMemory(stream_memory_info);
}
@@ -137,7 +167,7 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
stream_memory.SetObjectNameEXT("Stream Buffer Memory");
}
stream_buffer.BindMemory(*stream_memory, 0);
- stream_pointer = stream_memory.Map(0, STREAM_BUFFER_SIZE);
+ stream_pointer = stream_memory.Map(0, staging_buffer_size);
}
StagingBufferPool::~StagingBufferPool() = default;
@@ -158,25 +188,25 @@ void StagingBufferPool::TickFrame() {
}
StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
- if (AreRegionsActive(Region(free_iterator) + 1,
- std::min(Region(iterator + size) + 1, NUM_SYNCS))) {
+ if (AreRegionsActive(Region(free_iterator, region_size) + 1,
+ std::min(Region(iterator + size, region_size) + 1, NUM_SYNCS))) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}
const u64 current_tick = scheduler.CurrentTick();
- std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + Region(iterator),
- current_tick);
+ std::fill(sync_ticks.begin() + Region(used_iterator, region_size),
+ sync_ticks.begin() + Region(iterator, region_size), current_tick);
used_iterator = iterator;
free_iterator = std::max(free_iterator, iterator + size);
- if (iterator + size >= STREAM_BUFFER_SIZE) {
- std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + NUM_SYNCS,
- current_tick);
+ if (iterator + size >= staging_buffer_size) {
+ std::fill(sync_ticks.begin() + Region(used_iterator, region_size),
+ sync_ticks.begin() + NUM_SYNCS, current_tick);
used_iterator = 0;
iterator = 0;
free_iterator = size;
- if (AreRegionsActive(0, Region(size) + 1)) {
+ if (AreRegionsActive(0, Region(size, region_size) + 1)) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
index 91dc84da8..90c67177f 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
@@ -93,6 +93,9 @@ private:
size_t free_iterator = 0;
std::array<u64, NUM_SYNCS> sync_ticks{};
+ size_t staging_buffer_size = 0;
+ size_t region_size = 0;
+
StagingBuffersCache device_local_cache;
StagingBuffersCache upload_cache;
StagingBuffersCache download_cache;
diff --git a/src/web_service/web_backend.cpp b/src/web_service/web_backend.cpp
index 378804c08..12a7e4922 100644
--- a/src/web_service/web_backend.cpp
+++ b/src/web_service/web_backend.cpp
@@ -111,7 +111,8 @@ struct Client::Impl {
httplib::Error error;
if (!cli->send(request, response, error)) {
- LOG_ERROR(WebService, "{} to {} returned null", method, host + path);
+ LOG_ERROR(WebService, "{} to {} returned null (httplib Error: {})", method, host + path,
+ httplib::to_string(error));
return WebResult{WebResult::Code::LibError, "Null response", ""};
}
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-21 13:31:16 +0100