dxvk/src/dxvk/dxvk_resource.h

139 lines
3.7 KiB
C++

#pragma once
#include "dxvk_include.h"
namespace dxvk {
enum class DxvkAccess : uint32_t {
Read = 0,
Write = 1,
None = 2,
};
using DxvkAccessFlags = Flags<DxvkAccess>;
/**
* \brief DXVK resource
*
* Keeps track of whether the resource is currently in use
* by the GPU. As soon as a command that uses the resource
* is recorded, it will be marked as 'in use'.
*/
class DxvkResource {
static constexpr uint64_t RdAccessShift = 24;
static constexpr uint64_t WrAccessShift = 44;
static constexpr uint64_t RefcountMask = (1ull << RdAccessShift) - 1;
static constexpr uint64_t RdAccessMask = ((1ull << (WrAccessShift - RdAccessShift)) - 1) << RdAccessShift;
static constexpr uint64_t WrAccessMask = ((1ull << (64 - WrAccessShift)) - 1) << WrAccessShift;
static constexpr uint64_t RefcountInc = 1ull;
static constexpr uint64_t RdAccessInc = 1ull << RdAccessShift;
static constexpr uint64_t WrAccessInc = 1ull << WrAccessShift;
public:
DxvkResource();
virtual ~DxvkResource();
/**
* \brief Unique object identifier
*
* Can be used to identify an object even when
* the lifetime of the object is unknown, and
* without referencing the actual object.
* \returns Unique identifier
*/
uint64_t cookie() const {
return m_cookie;
}
/**
* \brief Increments reference count
* \returns New reference count
*/
uint32_t incRef() {
return acquire(DxvkAccess::None);
}
/**
* \brief Decrements reference count
* \returns New reference count
*/
uint32_t decRef() {
return release(DxvkAccess::None);
}
/**
* \brief Acquires resource with given access
*
* Atomically increments both the reference count
* as well as the use count for the given access.
* \returns New reference count
*/
uint32_t acquire(DxvkAccess access) {
return uint32_t((m_useCount += getIncrement(access)) & RefcountMask);
}
/**
* \brief Releases resource with given access
*
* Atomically decrements both the reference count
* as well as the use count for the given access.
* \returns New reference count
*/
uint32_t release(DxvkAccess access) {
return uint32_t((m_useCount -= getIncrement(access)) & RefcountMask);
}
/**
* \brief Checks whether resource is in use
*
* Returns \c true if there are pending accesses to
* the resource by the GPU matching the given access
* type. Note that checking for reads will also return
* \c true if the resource is being written to.
* \param [in] access Access type to check for
* \returns \c true if the resource is in use
*/
bool isInUse(DxvkAccess access = DxvkAccess::Read) const {
uint64_t mask = WrAccessMask;
if (access == DxvkAccess::Read)
mask |= RdAccessMask;
return bool(m_useCount.load() & mask);
}
/**
* \brief Waits for resource to become unused
*
* Blocks calling thread until the GPU finishes
* using the resource with the given access type.
* \param [in] access Access type to check for
*/
void waitIdle(DxvkAccess access = DxvkAccess::Read) const {
sync::spin(50000, [this, access] {
return !isInUse(access);
});
}
private:
std::atomic<uint64_t> m_useCount;
uint64_t m_cookie;
static constexpr uint64_t getIncrement(DxvkAccess access) {
uint64_t increment = RefcountInc;
if (access != DxvkAccess::None) {
increment |= (access == DxvkAccess::Read)
? RdAccessInc : WrAccessInc;
}
return increment;
}
static std::atomic<uint64_t> s_cookie;
};
}