Hash :
968041b5
Author :
Date :
2022-08-19T12:11:23
Metal: Optimized BufferSubData per device Adds a staging buffer path which means there are 4 paths for bufferSubData. 1. direct copy * get a pointer to the buffer * copy the new data to the buffer * if the buffer is managed, tell metal which part was updated 2. use a shadow copy * copy the data to a shadow copy * copy the entire shadow to a new buffer * start using the new buffer 3. use a new buffer * get a new buffer (or unused) * put the new data in the new buffer * blit any unchanged data from the old buffer to the new buffer * start using the new buffer 4. use a staging buffer * get a staging buffer * put the new data in the staging buffer * blit from the staging buffer to the existing buffer. Further, there are 3 types of memory storage modes. Managed, Staged, Private. Based on the GPU type different storage modes and different paths in different sitatutions are more performant. So, add feature flags to select paths by GPU. Bug: angleproject:7544 Change-Id: I741dd1874201043416374194bd2001ded8dbd9b4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3842641 Reviewed-by: Kyle Piddington <kpiddington@apple.com> Reviewed-by: Kenneth Russell <kbr@chromium.org> Reviewed-by: Quyen Le <lehoangquyen@chromium.org> Commit-Queue: Gregg Tavares <gman@chromium.org>
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
//
// Copyright 2022 The ANGLE Project Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
//
// mtl_buffer_manager.mm:
// Implements the class methods for BufferManager.
//
#include "libANGLE/renderer/metal/mtl_buffer_manager.h"
#include "libANGLE/renderer/metal/ContextMtl.h"
#include "libANGLE/renderer/metal/DisplayMtl.h"
namespace rx
{
namespace mtl
{
namespace
{
constexpr size_t Log2(size_t num)
{
return num <= 1 ? 0 : (1 + Log2(num / 2));
}
constexpr size_t Log2Ceil(size_t num)
{
size_t l = Log2(num);
size_t size = size_t(1) << l;
return num == size ? l : l + 1;
}
#ifdef ANGLE_MTL_TRACK_BUFFER_MEM
const char *memUnitSuffix(size_t powerOf2)
{
if (powerOf2 < 10)
{
return "b";
}
if (powerOf2 < 20)
{
return "k";
}
if (powerOf2 < 30)
{
return "M";
}
return "G";
}
size_t memUnitValue(size_t powerOf2)
{
if (powerOf2 < 10)
{
return 1u << powerOf2;
}
if (powerOf2 < 20)
{
return 1u << (powerOf2 - 10);
}
if (powerOf2 < 30)
{
return 1u << (powerOf2 - 20);
}
return 1u << (powerOf2 - 30);
}
#endif // ANGLE_MTL_TRACK_BUFFER_MEM
int sharedMemToIndex(bool useSharedMem)
{
return useSharedMem ? 1 : 0;
}
} // namespace
BufferManager::BufferManager()
#ifdef ANGLE_MTL_TRACK_BUFFER_MEM
: mAllocations(kMaxSizePowerOf2, 0)
#endif
{}
void BufferManager::freeUnusedBuffers(ContextMtl *contextMtl)
{
// Scan for the first buffer still in use.
BufferList::iterator firstInUseIter =
std::find_if(mInUseBuffers.begin(), mInUseBuffers.end(),
[&contextMtl](auto ref) { return ref->isBeingUsedByGPU(contextMtl); });
// Move unused buffers to the free lists
for (BufferList::iterator it = mInUseBuffers.begin(); it != firstInUseIter; ++it)
{
addBufferRefToFreeLists(*it);
}
mInUseBuffers.erase(mInUseBuffers.begin(), firstInUseIter);
}
void BufferManager::addBufferRefToFreeLists(mtl::BufferRef &bufferRef)
{
const size_t bucketNdx = Log2Ceil(bufferRef->size());
ASSERT(bucketNdx < kMaxSizePowerOf2);
int sharedNdx = sharedMemToIndex(bufferRef->get().storageMode == MTLStorageModeShared);
mFreeBuffers[sharedNdx][bucketNdx].push_back(bufferRef);
}
void BufferManager::returnBuffer(ContextMtl *contextMtl, BufferRef &bufferRef)
{
if (bufferRef->isBeingUsedByGPU(contextMtl))
{
mInUseBuffers.push_back(bufferRef);
}
else
{
addBufferRefToFreeLists(bufferRef);
}
}
angle::Result BufferManager::getBuffer(ContextMtl *contextMtl,
size_t size,
bool useSharedMem,
BufferRef &bufferRef)
{
freeUnusedBuffers(contextMtl);
const size_t bucketNdx = Log2Ceil(size);
const int sharedNdx = sharedMemToIndex(useSharedMem);
BufferList &freeBuffers = mFreeBuffers[sharedNdx][bucketNdx];
// If there are free buffers grab one
if (!freeBuffers.empty())
{
bufferRef = freeBuffers.back();
freeBuffers.pop_back();
return angle::Result::Continue;
}
// Create a new one
mtl::BufferRef newBufferRef;
size_t allocSize = size_t(1) << bucketNdx;
ASSERT(allocSize >= size);
ANGLE_TRY(mtl::Buffer::MakeBufferWithSharedMemOpt(contextMtl, useSharedMem, allocSize, nullptr,
&newBufferRef));
#ifdef ANGLE_MTL_TRACK_BUFFER_MEM
{
mTotalMem += allocSize;
mAllocations[bucketNdx]++;
fprintf(stderr, "totalMem: %zu, ", mTotalMem);
size_t numBuffers = 0;
for (size_t i = 0; i < kMaxSizePowerOf2; ++i)
{
if (mAllocations[i])
{
numBuffers += mAllocations[i];
fprintf(stderr, "%zu%s: %zu, ", memUnitValue(i), memUnitSuffix(i), mAllocations[i]);
}
}
fprintf(stderr, " total: %zu\n", numBuffers);
}
#endif
bufferRef = newBufferRef;
return angle::Result::Continue;
}
angle::Result BufferManager::queueBlitCopyDataToBuffer(ContextMtl *contextMtl,
const void *srcPtr,
size_t sizeToCopy,
size_t offset,
mtl::BufferRef &dstMetalBuffer)
{
const uint8 *src = reinterpret_cast<const uint8 *>(srcPtr);
bool useShared =
!contextMtl->getDisplay()->getFeatures().alwaysUseManagedStorageModeForBuffers.enabled;
for (size_t srcOffset = 0; srcOffset < sizeToCopy; srcOffset += kMaxStagingBufferSize)
{
size_t subSizeToCopy = std::min(kMaxStagingBufferSize, sizeToCopy - srcOffset);
mtl::BufferRef bufferRef;
ANGLE_TRY(getBuffer(contextMtl, subSizeToCopy, useShared, bufferRef));
// copy data to buffer
uint8_t *ptr = bufferRef->mapWithOpt(contextMtl, false, true);
std::copy(src + srcOffset, src + srcOffset + subSizeToCopy, ptr);
bufferRef->unmapAndFlushSubset(contextMtl, 0, subSizeToCopy);
// queue blit
mtl::BlitCommandEncoder *blitEncoder = contextMtl->getBlitCommandEncoder();
blitEncoder->copyBuffer(bufferRef, 0, dstMetalBuffer, offset + srcOffset, subSizeToCopy);
returnBuffer(contextMtl, bufferRef);
}
return angle::Result::Continue;
}
} // namespace mtl
} // namespace rx