compute-runtime/shared/source/helpers/preamble.h

100 lines
3.9 KiB
C++

/*
* Copyright (C) 2018-2020 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#pragma once
#include "shared/source/helpers/pipeline_select_helper.h"
#include "engine_node.h"
#include "igfxfmid.h"
#include <cstddef>
#include <cstdint>
namespace NEO {
struct HardwareInfo;
class Device;
struct DispatchFlags;
class GraphicsAllocation;
class LinearStream;
struct PipelineSelectArgs;
template <typename GfxFamily>
struct PreambleHelper {
using MI_LOAD_REGISTER_IMM = typename GfxFamily::MI_LOAD_REGISTER_IMM;
using PIPE_CONTROL = typename GfxFamily::PIPE_CONTROL;
using VFE_STATE_TYPE = typename GfxFamily::VFE_STATE_TYPE;
static void programL3(LinearStream *pCommandStream, uint32_t l3Config);
static void programPipelineSelect(LinearStream *pCommandStream,
const PipelineSelectArgs &pipelineSelectArgs,
const HardwareInfo &hwInfo);
static void programThreadArbitration(LinearStream *pCommandStream, uint32_t requiredThreadArbitrationPolicy);
static void programPreemption(LinearStream *pCommandStream, Device &device, GraphicsAllocation *preemptionCsr);
static void addPipeControlBeforeVfeCmd(LinearStream *pCommandStream, const HardwareInfo *hwInfo, aub_stream::EngineType engineType);
static uint64_t programVFEState(LinearStream *pCommandStream,
const HardwareInfo &hwInfo,
int scratchSize,
uint64_t scratchAddress,
uint32_t maxFrontEndThreads,
aub_stream::EngineType engineType);
static void programAdditionalFieldsInVfeState(VFE_STATE_TYPE *mediaVfeState, const HardwareInfo &hwInfo);
static void programPreamble(LinearStream *pCommandStream, Device &device, uint32_t l3Config,
uint32_t requiredThreadArbitrationPolicy, GraphicsAllocation *preemptionCsr, GraphicsAllocation *perDssBackedBuffer);
static void programKernelDebugging(LinearStream *pCommandStream);
static void programPerDssBackedBuffer(LinearStream *pCommandStream, const HardwareInfo &hwInfo, GraphicsAllocation *perDssBackBufferOffset);
static void programSemaphoreDelay(LinearStream *pCommandStream);
static uint32_t getL3Config(const HardwareInfo &hwInfo, bool useSLM);
static bool isL3Configurable(const HardwareInfo &hwInfo);
static size_t getAdditionalCommandsSize(const Device &device);
static size_t getThreadArbitrationCommandsSize();
static size_t getVFECommandsSize();
static size_t getKernelDebuggingCommandsSize(bool debuggingActive);
static void programGenSpecificPreambleWorkArounds(LinearStream *pCommandStream, const HardwareInfo &hwInfo);
static uint32_t getUrbEntryAllocationSize();
static size_t getPerDssBackedBufferCommandsSize(const HardwareInfo &hwInfo);
static size_t getCmdSizeForPipelineSelect(const HardwareInfo &hwInfo);
static size_t getSemaphoreDelayCommandSize();
};
template <PRODUCT_FAMILY ProductFamily>
static uint32_t getL3ConfigHelper(bool useSLM);
template <PRODUCT_FAMILY ProductFamily>
struct L3CNTLREGConfig {
static const uint32_t valueForSLM;
static const uint32_t valueForNoSLM;
};
template <PRODUCT_FAMILY ProductFamily>
uint32_t getL3ConfigHelper(bool useSLM) {
if (!useSLM) {
return L3CNTLREGConfig<ProductFamily>::valueForNoSLM;
}
return L3CNTLREGConfig<ProductFamily>::valueForSLM;
}
template <typename GfxFamily>
struct L3CNTLRegisterOffset {
static const uint32_t registerOffset;
};
template <typename GfxFamily>
struct DebugModeRegisterOffset {
enum {
registerOffset = 0x20ec,
debugEnabledValue = (1 << 6) | (1 << 22)
};
};
namespace TdDebugControlRegisterOffset {
static constexpr uint32_t registerOffset = 0xe400;
static constexpr uint32_t debugEnabledValue = (1 << 4) | (1 << 7);
}; // namespace TdDebugControlRegisterOffset
} // namespace NEO