Media
/
compute-runtime
mirror of https://github.com/intel/compute-runtime.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
compute-runtime/shared/source/command_container/command_encoder.inl

474 lines
20 KiB
C++
Raw Blame History

/*
* Copyright (C) 2020 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#pragma once
#include "shared/source/command_container/command_encoder.h"
#include "shared/source/command_stream/linear_stream.h"
#include "shared/source/device/device.h"
#include "shared/source/execution_environment/execution_environment.h"
#include "shared/source/helpers/hw_helper.h"
#include "shared/source/helpers/preamble.h"
#include "shared/source/helpers/register_offsets.h"
#include "shared/source/helpers/simd_helper.h"
#include "shared/source/helpers/string.h"
#include "shared/source/kernel/dispatch_kernel_encoder_interface.h"
#include "opencl/source/helpers/hardware_commands_helper.h"
#include <algorithm>
namespace NEO {
template <typename Family>
uint32_t EncodeStates<Family>::copySamplerState(IndirectHeap *dsh,
uint32_t samplerStateOffset,
uint32_t samplerCount,
uint32_t borderColorOffset,
const void *fnDynamicStateHeap) {
auto sizeSamplerState = sizeof(SAMPLER_STATE) * samplerCount;
auto borderColorSize = samplerStateOffset - borderColorOffset;
dsh->align(alignIndirectStatePointer);
auto borderColorOffsetInDsh = static_cast<uint32_t>(dsh->getUsed());
auto borderColor = dsh->getSpace(borderColorSize);
memcpy_s(borderColor, borderColorSize, ptrOffset(fnDynamicStateHeap, borderColorOffset),
borderColorSize);
dsh->align(INTERFACE_DESCRIPTOR_DATA::SAMPLERSTATEPOINTER_ALIGN_SIZE);
auto samplerStateOffsetInDsh = static_cast<uint32_t>(dsh->getUsed());
auto dstSamplerState = reinterpret_cast<SAMPLER_STATE *>(dsh->getSpace(sizeSamplerState));
auto srcSamplerState = reinterpret_cast<const SAMPLER_STATE *>(ptrOffset(fnDynamicStateHeap, samplerStateOffset));
SAMPLER_STATE state = {};
for (uint32_t i = 0; i < samplerCount; i++) {
state = srcSamplerState[i];
state.setIndirectStatePointer(static_cast<uint32_t>(borderColorOffsetInDsh));
dstSamplerState[i] = state;
}
return samplerStateOffsetInDsh;
}
template <typename Family>
void EncodeMathMMIO<Family>::encodeMulRegVal(CommandContainer &container, uint32_t offset, uint32_t val, uint64_t dstAddress) {
int logLws = 0;
int i = val;
while (val >> logLws) {
logLws++;
}
EncodeSetMMIO<Family>::encodeREG(container, CS_GPR_R0, offset);
EncodeSetMMIO<Family>::encodeIMM(container, CS_GPR_R1, 0);
i = 0;
while (i < logLws) {
if (val & (1 << i)) {
EncodeMath<Family>::addition(container, AluRegisters::R_1,
AluRegisters::R_0, AluRegisters::R_2);
EncodeSetMMIO<Family>::encodeREG(container, CS_GPR_R1, CS_GPR_R2);
}
EncodeMath<Family>::addition(container, AluRegisters::R_0,
AluRegisters::R_0, AluRegisters::R_2);
EncodeSetMMIO<Family>::encodeREG(container, CS_GPR_R0, CS_GPR_R2);
i++;
}
EncodeStoreMMIO<Family>::encode(*container.getCommandStream(), CS_GPR_R1, dstAddress);
}
/*
* Compute *firstOperand > secondOperand and store the result in
* MI_PREDICATE_RESULT where firstOperand is an device memory address.
*
* To calculate the "greater than" operation in the device,
* (secondOperand - *firstOperand) is used, and if the carry flag register is
* set, then (*firstOperand) is greater than secondOperand.
*/
template <typename Family>
void EncodeMathMMIO<Family>::encodeGreaterThanPredicate(CommandContainer &container, uint64_t firstOperand, uint32_t secondOperand) {
EncodeSetMMIO<Family>::encodeMEM(container, CS_GPR_R0, firstOperand);
EncodeSetMMIO<Family>::encodeIMM(container, CS_GPR_R1, secondOperand);
/* CS_GPR_R* registers map to AluRegisters::R_* registers */
EncodeMath<Family>::greaterThan(container, AluRegisters::R_0,
AluRegisters::R_1, AluRegisters::R_2);
EncodeSetMMIO<Family>::encodeREG(container, CS_PREDICATE_RESULT, CS_GPR_R2);
}
/*
* encodeAlu() performs operations that leave a state including the result of
* an operation such as the carry flag, and the accu flag with subtraction and
* addition result.
*
* Parameter "postOperationStateRegister" is the ALU register with the result
* from the operation that the function caller is interested in obtaining.
*
* Parameter "finalResultRegister" is the final destination register where
* data from "postOperationStateRegister" will be copied.
*/
template <typename Family>
void EncodeMathMMIO<Family>::encodeAlu(MI_MATH_ALU_INST_INLINE *pAluParam, AluRegisters srcA, AluRegisters srcB, AluRegisters op, AluRegisters finalResultRegister, AluRegisters postOperationStateRegister) {
MI_MATH_ALU_INST_INLINE aluParam;
aluParam.DW0.Value = 0x0;
aluParam.DW0.BitField.ALUOpcode = static_cast<uint32_t>(AluRegisters::OPCODE_LOAD);
aluParam.DW0.BitField.Operand1 = static_cast<uint32_t>(AluRegisters::R_SRCA);
aluParam.DW0.BitField.Operand2 = static_cast<uint32_t>(srcA);
*pAluParam = aluParam;
pAluParam++;
aluParam.DW0.Value = 0x0;
aluParam.DW0.BitField.ALUOpcode = static_cast<uint32_t>(AluRegisters::OPCODE_LOAD);
aluParam.DW0.BitField.Operand1 = static_cast<uint32_t>(AluRegisters::R_SRCB);
aluParam.DW0.BitField.Operand2 = static_cast<uint32_t>(srcB);
*pAluParam = aluParam;
pAluParam++;
/* Order of operation: Operand1 <ALUOpcode> Operand2 */
aluParam.DW0.Value = 0x0;
aluParam.DW0.BitField.ALUOpcode = static_cast<uint32_t>(op);
aluParam.DW0.BitField.Operand1 = 0;
aluParam.DW0.BitField.Operand2 = 0;
*pAluParam = aluParam;
pAluParam++;
aluParam.DW0.Value = 0x0;
aluParam.DW0.BitField.ALUOpcode = static_cast<uint32_t>(AluRegisters::OPCODE_STORE);
aluParam.DW0.BitField.Operand1 = static_cast<uint32_t>(finalResultRegister);
aluParam.DW0.BitField.Operand2 = static_cast<uint32_t>(postOperationStateRegister);
*pAluParam = aluParam;
pAluParam++;
}
template <typename Family>
uint32_t *EncodeMath<Family>::commandReserve(CommandContainer &container) {
size_t size = sizeof(MI_MATH) + sizeof(MI_MATH_ALU_INST_INLINE) * NUM_ALU_INST_FOR_READ_MODIFY_WRITE;
auto cmd = reinterpret_cast<uint32_t *>(container.getCommandStream()->getSpace(size));
MI_MATH mathBuffer;
mathBuffer.DW0.Value = 0x0;
mathBuffer.DW0.BitField.InstructionType = MI_MATH::COMMAND_TYPE_MI_COMMAND;
mathBuffer.DW0.BitField.InstructionOpcode = MI_MATH::MI_COMMAND_OPCODE_MI_MATH;
mathBuffer.DW0.BitField.DwordLength = NUM_ALU_INST_FOR_READ_MODIFY_WRITE - 1;
*reinterpret_cast<MI_MATH *>(cmd) = mathBuffer;
cmd++;
return cmd;
}
template <typename Family>
void EncodeMathMMIO<Family>::encodeAluAdd(MI_MATH_ALU_INST_INLINE *pAluParam,
AluRegisters firstOperandRegister,
AluRegisters secondOperandRegister,
AluRegisters finalResultRegister) {
encodeAlu(pAluParam, firstOperandRegister, secondOperandRegister, AluRegisters::OPCODE_ADD, finalResultRegister, AluRegisters::R_ACCU);
}
template <typename Family>
void EncodeMathMMIO<Family>::encodeAluSubStoreCarry(MI_MATH_ALU_INST_INLINE *pAluParam, AluRegisters regA, AluRegisters regB, AluRegisters finalResultRegister) {
/* regB is subtracted from regA */
encodeAlu(pAluParam, regA, regB, AluRegisters::OPCODE_SUB, finalResultRegister, AluRegisters::R_CF);
}
/*
* greaterThan() tests if firstOperandRegister is greater than
* secondOperandRegister.
*/
template <typename Family>
void EncodeMath<Family>::greaterThan(CommandContainer &container,
AluRegisters firstOperandRegister,
AluRegisters secondOperandRegister,
AluRegisters finalResultRegister) {
uint32_t *cmd = EncodeMath<Family>::commandReserve(container);
/* firstOperandRegister will be subtracted from secondOperandRegister */
EncodeMathMMIO<Family>::encodeAluSubStoreCarry(reinterpret_cast<MI_MATH_ALU_INST_INLINE *>(cmd),
secondOperandRegister,
firstOperandRegister,
finalResultRegister);
}
template <typename Family>
void EncodeMath<Family>::addition(CommandContainer &container,
AluRegisters firstOperandRegister,
AluRegisters secondOperandRegister,
AluRegisters finalResultRegister) {
uint32_t *cmd = EncodeMath<Family>::commandReserve(container);
EncodeMathMMIO<Family>::encodeAluAdd(reinterpret_cast<MI_MATH_ALU_INST_INLINE *>(cmd),
firstOperandRegister,
secondOperandRegister,
finalResultRegister);
}
template <typename Family>
void EncodeIndirectParams<Family>::setGroupCountIndirect(CommandContainer &container, const NEO::CrossThreadDataOffset offsets[3], void *crossThreadAddress) {
for (int i = 0; i < 3; ++i) {
if (NEO::isUndefinedOffset(offsets[i])) {
continue;
}
EncodeStoreMMIO<Family>::encode(*container.getCommandStream(), GPUGPU_DISPATCHDIM[i], ptrOffset(reinterpret_cast<uint64_t>(crossThreadAddress), offsets[i]));
}
}
template <typename Family>
void EncodeIndirectParams<Family>::setGlobalWorkSizeIndirect(CommandContainer &container, const NEO::CrossThreadDataOffset offsets[3], void *crossThreadAddress, const uint32_t *lws) {
for (int i = 0; i < 3; ++i) {
if (NEO::isUndefinedOffset(offsets[i])) {
continue;
}
EncodeMathMMIO<Family>::encodeMulRegVal(container, GPUGPU_DISPATCHDIM[i], lws[i], ptrOffset(reinterpret_cast<uint64_t>(crossThreadAddress), offsets[i]));
}
}
template <typename Family>
void EncodeSetMMIO<Family>::encodeIMM(CommandContainer &container, uint32_t offset, uint32_t data) {
MI_LOAD_REGISTER_IMM cmd = Family::cmdInitLoadRegisterImm;
cmd.setRegisterOffset(offset);
cmd.setDataDword(data);
auto buffer = container.getCommandStream()->getSpaceForCmd<MI_LOAD_REGISTER_IMM>();
*buffer = cmd;
}
template <typename Family>
void EncodeSetMMIO<Family>::encodeMEM(CommandContainer &container, uint32_t offset, uint64_t address) {
MI_LOAD_REGISTER_MEM cmd = Family::cmdInitLoadRegisterMem;
cmd.setRegisterAddress(offset);
cmd.setMemoryAddress(address);
auto buffer = container.getCommandStream()->getSpaceForCmd<MI_LOAD_REGISTER_MEM>();
*buffer = cmd;
}
template <typename Family>
void EncodeSetMMIO<Family>::encodeREG(CommandContainer &container, uint32_t dstOffset, uint32_t srcOffset) {
MI_LOAD_REGISTER_REG cmd = Family::cmdInitLoadRegisterReg;
cmd.setSourceRegisterAddress(srcOffset);
cmd.setDestinationRegisterAddress(dstOffset);
auto buffer = container.getCommandStream()->getSpaceForCmd<MI_LOAD_REGISTER_REG>();
*buffer = cmd;
}
template <typename Family>
void EncodeStoreMMIO<Family>::encode(LinearStream &csr, uint32_t offset, uint64_t address) {
MI_STORE_REGISTER_MEM cmd = Family::cmdInitStoreRegisterMem;
cmd.setRegisterAddress(offset);
cmd.setMemoryAddress(address);
remapOffset(&cmd);
auto buffer = csr.getSpaceForCmd<MI_STORE_REGISTER_MEM>();
*buffer = cmd;
}
template <typename Family>
void EncodeSurfaceState<Family>::encodeBuffer(void *dst, void *address, size_t size, uint32_t mocs,
bool cpuCoherent) {
auto ss = reinterpret_cast<R_SURFACE_STATE *>(dst);
UNRECOVERABLE_IF(!isAligned<getSurfaceBaseAddressAlignment()>(size));
SURFACE_STATE_BUFFER_LENGTH Length = {0};
Length.Length = static_cast<uint32_t>(size - 1);
ss->setWidth(Length.SurfaceState.Width + 1);
ss->setHeight(Length.SurfaceState.Height + 1);
ss->setDepth(Length.SurfaceState.Depth + 1);
ss->setSurfaceType((address != nullptr) ? R_SURFACE_STATE::SURFACE_TYPE_SURFTYPE_BUFFER
: R_SURFACE_STATE::SURFACE_TYPE_SURFTYPE_NULL);
ss->setSurfaceFormat(SURFACE_FORMAT::SURFACE_FORMAT_RAW);
ss->setSurfaceVerticalAlignment(R_SURFACE_STATE::SURFACE_VERTICAL_ALIGNMENT_VALIGN_4);
ss->setSurfaceHorizontalAlignment(R_SURFACE_STATE::SURFACE_HORIZONTAL_ALIGNMENT_HALIGN_4);
ss->setTileMode(R_SURFACE_STATE::TILE_MODE_LINEAR);
ss->setVerticalLineStride(0);
ss->setVerticalLineStrideOffset(0);
ss->setMemoryObjectControlState(mocs);
ss->setSurfaceBaseAddress(reinterpret_cast<uintptr_t>(address));
ss->setCoherencyType(cpuCoherent ? R_SURFACE_STATE::COHERENCY_TYPE_IA_COHERENT
: R_SURFACE_STATE::COHERENCY_TYPE_GPU_COHERENT);
ss->setAuxiliarySurfaceMode(AUXILIARY_SURFACE_MODE::AUXILIARY_SURFACE_MODE_AUX_NONE);
}
template <typename Family>
void *EncodeDispatchKernel<Family>::getInterfaceDescriptor(CommandContainer &container, uint32_t &iddOffset) {
if (container.nextIddInBlock == container.getNumIddPerBlock()) {
container.getIndirectHeap(HeapType::DYNAMIC_STATE)->align(HardwareCommandsHelper<Family>::alignInterfaceDescriptorData);
container.setIddBlock(container.getHeapSpaceAllowGrow(HeapType::DYNAMIC_STATE,
sizeof(INTERFACE_DESCRIPTOR_DATA) * container.getNumIddPerBlock()));
container.nextIddInBlock = 0;
EncodeMediaInterfaceDescriptorLoad<Family>::encode(container);
}
iddOffset = container.nextIddInBlock;
auto interfaceDescriptorData = static_cast<INTERFACE_DESCRIPTOR_DATA *>(container.getIddBlock());
return &interfaceDescriptorData[container.nextIddInBlock++];
}
template <typename Family>
void EncodeDispatchKernel<Family>::patchBindlessSurfaceStateOffsets(const size_t sshOffset, const KernelDescriptor &kernelDesc, uint8_t *crossThread) {
auto &hwHelper = HwHelperHw<Family>::get();
for (const auto &argT : kernelDesc.payloadMappings.explicitArgs) {
CrossThreadDataOffset bindless = undefined<CrossThreadDataOffset>;
SurfaceStateHeapOffset bindful = undefined<SurfaceStateHeapOffset>;
switch (argT.type) {
case ArgDescriptor::ArgTPointer: {
auto &arg = argT.as<NEO::ArgDescPointer>();
bindless = arg.bindless;
bindful = arg.bindful;
} break;
case ArgDescriptor::ArgTImage: {
auto &arg = argT.as<NEO::ArgDescImage>();
bindless = arg.bindless;
bindful = arg.bindful;
} break;
default:
break;
}
if (NEO::isValidOffset(bindless)) {
auto patchLocation = ptrOffset(crossThread, bindless);
auto bindlessOffset = static_cast<uint32_t>(sshOffset) + bindful;
auto patchValue = hwHelper.getBindlessSurfaceExtendedMessageDescriptorValue(bindlessOffset);
patchWithRequiredSize(patchLocation, sizeof(patchValue), patchValue);
}
}
}
template <typename Family>
size_t EncodeStates<Family>::getAdjustStateComputeModeSize() {
return 0;
}
template <typename Family>
size_t EncodeIndirectParams<Family>::getCmdsSizeForIndirectParams() {
return 3 * sizeof(typename Family::MI_LOAD_REGISTER_MEM);
}
template <typename Family>
size_t EncodeIndirectParams<Family>::getCmdsSizeForSetGroupCountIndirect() {
return 3 * (sizeof(MI_STORE_REGISTER_MEM));
}
template <typename Family>
size_t EncodeIndirectParams<Family>::getCmdsSizeForSetGroupSizeIndirect() {
return 3 * (sizeof(MI_LOAD_REGISTER_REG) + sizeof(MI_LOAD_REGISTER_IMM) + sizeof(MI_MATH) + sizeof(MI_MATH_ALU_INST_INLINE) + sizeof(MI_STORE_REGISTER_MEM));
}
template <typename Family>
void EncodeSempahore<Family>::programMiSemaphoreWait(MI_SEMAPHORE_WAIT *cmd,
uint64_t compareAddress,
uint32_t compareData,
COMPARE_OPERATION compareMode) {
MI_SEMAPHORE_WAIT localCmd = Family::cmdInitMiSemaphoreWait;
localCmd.setCompareOperation(compareMode);
localCmd.setSemaphoreDataDword(compareData);
localCmd.setSemaphoreGraphicsAddress(compareAddress);
localCmd.setWaitMode(MI_SEMAPHORE_WAIT::WAIT_MODE::WAIT_MODE_POLLING_MODE);
*cmd = localCmd;
}
template <typename Family>
void EncodeSempahore<Family>::addMiSemaphoreWaitCommand(LinearStream &commandStream,
uint64_t compareAddress,
uint32_t compareData,
COMPARE_OPERATION compareMode) {
auto semaphoreCommand = commandStream.getSpaceForCmd<MI_SEMAPHORE_WAIT>();
programMiSemaphoreWait(semaphoreCommand,
compareAddress,
compareData,
compareMode);
}
template <typename Family>
size_t EncodeSempahore<Family>::getSizeMiSemaphoreWait() {
return sizeof(MI_SEMAPHORE_WAIT);
}
template <typename Family>
void EncodeAtomic<Family>::programMiAtomic(MI_ATOMIC *atomic, uint64_t writeAddress,
ATOMIC_OPCODES opcode,
DATA_SIZE dataSize) {
MI_ATOMIC cmd = Family::cmdInitAtomic;
cmd.setAtomicOpcode(opcode);
cmd.setDataSize(dataSize);
cmd.setMemoryAddress(static_cast<uint32_t>(writeAddress & 0x0000FFFFFFFFULL));
cmd.setMemoryAddressHigh(static_cast<uint32_t>(writeAddress >> 32));
*atomic = cmd;
}
template <typename Family>
void EncodeBatchBufferStartOrEnd<Family>::programBatchBufferStart(LinearStream *commandStream,
uint64_t address,
bool secondLevel) {
MI_BATCH_BUFFER_START cmd = Family::cmdInitBatchBufferStart;
if (secondLevel) {
cmd.setSecondLevelBatchBuffer(MI_BATCH_BUFFER_START::SECOND_LEVEL_BATCH_BUFFER_SECOND_LEVEL_BATCH);
}
cmd.setAddressSpaceIndicator(MI_BATCH_BUFFER_START::ADDRESS_SPACE_INDICATOR_PPGTT);
cmd.setBatchBufferStartAddressGraphicsaddress472(address);
auto buffer = commandStream->getSpaceForCmd<MI_BATCH_BUFFER_START>();
*buffer = cmd;
}
template <typename Family>
void EncodeBatchBufferStartOrEnd<Family>::programBatchBufferEnd(CommandContainer &container) {
MI_BATCH_BUFFER_END cmd = Family::cmdInitBatchBufferEnd;
auto buffer = container.getCommandStream()->getSpaceForCmd<MI_BATCH_BUFFER_END>();
*buffer = cmd;
}
template <typename Family>
void EncodeSurfaceState<Family>::getSshAlignedPointer(uintptr_t &ptr, size_t &offset) {
auto sshAlignmentMask =
getSurfaceBaseAddressAlignmentMask();
uintptr_t alignedPtr = ptr & sshAlignmentMask;
offset = 0;
if (ptr != alignedPtr) {
offset = ptrDiff(ptr, alignedPtr);
ptr = alignedPtr;
}
}
template <typename GfxFamily>
void EncodeMiFlushDW<GfxFamily>::programMiFlushDw(LinearStream &commandStream, uint64_t immediateDataGpuAddress, uint64_t immediateData, bool timeStampOperation, bool commandWithPostSync) {
programMiFlushDwWA(commandStream);
auto miFlushDwCmd = commandStream.getSpaceForCmd<MI_FLUSH_DW>();
MI_FLUSH_DW miFlush = GfxFamily::cmdInitMiFlushDw;
if (commandWithPostSync) {
auto postSyncType = timeStampOperation ? MI_FLUSH_DW::POST_SYNC_OPERATION_WRITE_TIMESTAMP_REGISTER : MI_FLUSH_DW::POST_SYNC_OPERATION_WRITE_IMMEDIATE_DATA_QWORD;
miFlush.setPostSyncOperation(postSyncType);
miFlush.setDestinationAddress(immediateDataGpuAddress);
miFlush.setImmediateData(immediateData);
}
appendMiFlushDw(&miFlush);
*miFlushDwCmd = miFlush;
}
template <typename GfxFamily>
size_t EncodeMiFlushDW<GfxFamily>::getMiFlushDwCmdSizeForDataWrite() {
return sizeof(typename GfxFamily::MI_FLUSH_DW) + EncodeMiFlushDW<GfxFamily>::getMiFlushDwWaSize();
}
template <typename GfxFamily>
void EncodeMemoryPrefetch<GfxFamily>::programMemoryPrefetch(LinearStream &commandStream, const GraphicsAllocation &graphicsAllocation, uint32_t size) {}
template <typename GfxFamily>
size_t EncodeMemoryPrefetch<GfxFamily>::getSizeForMemoryPrefetch() { return 0; }
} // namespace NEO
Reference in New Issue View Git Blame Copy Permalink