AI/rocm-systems
2
0
Fork 0
Código Incidencias Pull Requests Acciones Paquetes Proyectos Lanzamientos Wiki Actividad
Files
ac1e6d59c27ca24ae43dfaf97ffdec4bb25fc6cf
rocm-systems/runtime/hsa-runtime/image/image_manager_kv.cpp
T
Khatri, Shweta 0984a1f0fd rocr: GFX9, GFX10, GFX11: Use view3dAs2dArray flag, for thick/3D swizzle modes. (#58) 
A HSA_IMAGE_ENABLE_3D_SWIZZLE_DEBUG environment flag exists already to
enable/disable this. Default value is false (view3dAs2dArray = 1)
Enabling this flag will enable support for swizzles that do 3D
interleaving on GFX9, GF10 and GFX11. By default support for swizzles that
do 3D interleaving is disabled.
2025-02-26 09:38:17 -05:00

1028 líneas
36 KiB
C++
Original Blame Histórico

////////////////////////////////////////////////////////////////////////////////
//
// The University of Illinois/NCSA
// Open Source License (NCSA)
//
// Copyright (c) 2014-2020, Advanced Micro Devices, Inc. All rights reserved.
//
// Developed by:
//
// AMD Research and AMD HSA Software Development
//
// Advanced Micro Devices, Inc.
//
// www.amd.com
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to
// deal with the Software without restriction, including without limitation
// the rights to use, copy, modify, merge, publish, distribute, sublicense,
// and/or sell copies of the Software, and to permit persons to whom the
// Software is furnished to do so, subject to the following conditions:
//
// - Redistributions of source code must retain the above copyright notice,
// this list of conditions and the following disclaimers.
// - Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimers in
// the documentation and/or other materials provided with the distribution.
// - Neither the names of Advanced Micro Devices, Inc,
// nor the names of its contributors may be used to endorse or promote
// products derived from this Software without specific prior written
// permission.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
// THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
// OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
// DEALINGS WITH THE SOFTWARE.
//
////////////////////////////////////////////////////////////////////////////////
#define NOMINMAX
#include "image_manager_kv.h"
#include <assert.h>
#include <algorithm>
#include <climits>
#include "core/inc/runtime.h"
#include "hsakmt/hsakmt.h"
#include "inc/hsa_ext_amd.h"
#include "core/inc/hsa_internal.h"
#include "core/inc/hsa_ext_amd_impl.h"
#include "addrlib/inc/addrinterface.h"
#include "addrlib/src/core/addrlib.h"
#include "image_runtime.h"
#include "resource.h"
#include "resource_kv.h"
#include "util.h"
#include "device_info.h"
namespace rocr {
namespace image {
ASSERT_SIZE_UINT32(SQ_BUF_RSRC_WORD0)
ASSERT_SIZE_UINT32(SQ_BUF_RSRC_WORD1)
ASSERT_SIZE_UINT32(SQ_BUF_RSRC_WORD2)
ASSERT_SIZE_UINT32(SQ_BUF_RSRC_WORD3)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD0)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD1)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD2)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD3)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD4)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD5)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD6)
ASSERT_SIZE_UINT32(SQ_IMG_RSRC_WORD7)
ASSERT_SIZE_UINT32(SQ_IMG_SAMP_WORD0)
ASSERT_SIZE_UINT32(SQ_IMG_SAMP_WORD1)
ASSERT_SIZE_UINT32(SQ_IMG_SAMP_WORD2)
ASSERT_SIZE_UINT32(SQ_IMG_SAMP_WORD3)
ImageManagerKv::ImageManagerKv() : ImageManager() {}
ImageManagerKv::~ImageManagerKv() {}
hsa_status_t ImageManagerKv::Initialize(hsa_agent_t agent_handle) {
agent_ = agent_handle;
hsa_status_t status = GetGPUAsicID(agent_, &chip_id_);
uint32_t major_ver = MajorVerFromDevID(chip_id_);
assert(status == HSA_STATUS_SUCCESS);
status = HSA::hsa_agent_get_info(
agent_, static_cast<hsa_agent_info_t>(HSA_AMD_AGENT_INFO_ASIC_FAMILY_ID), &family_type_);
assert(status == HSA_STATUS_SUCCESS);
HsaGpuTileConfig tileConfig = {0};
unsigned int tc[40];
unsigned int mtc[40];
tileConfig.TileConfig = &tc[0];
tileConfig.NumTileConfigs = 40;
tileConfig.MacroTileConfig = &mtc[0];
tileConfig.NumMacroTileConfigs = 40;
uint32_t node_id = 0;
status = HSA::hsa_agent_get_info(
agent_, static_cast<hsa_agent_info_t>(HSA_AMD_AGENT_INFO_DRIVER_NODE_ID), &node_id);
assert(status == HSA_STATUS_SUCCESS);
HSAKMT_STATUS stat = hsaKmtGetTileConfig(node_id, &tileConfig);
assert(stat == HSAKMT_STATUS_SUCCESS);
// Initialize address library.
// TODO(bwicakso) hard coded based on UGL parameters.
// Need to get this information from KMD.
addr_lib_ = NULL;
ADDR_CREATE_INPUT addr_create_input = {0};
ADDR_CREATE_OUTPUT addr_create_output = {0};
if (major_ver >= 9) {
addr_create_input.chipEngine = CIASICIDGFXENGINE_ARCTICISLAND;
} else {
addr_create_input.chipEngine = CIASICIDGFXENGINE_SOUTHERNISLAND;
}
addr_create_input.chipFamily = family_type_;
addr_create_input.chipRevision = 0; // TODO(bwicakso): find how to get this.
ADDR_CREATE_FLAGS create_flags = {};
create_flags.value = 0;
create_flags.useTileIndex = 1;
addr_create_input.createFlags = create_flags;
addr_create_input.callbacks.allocSysMem = AllocSysMem;
addr_create_input.callbacks.freeSysMem = FreeSysMem;
addr_create_input.callbacks.debugPrint = 0;
ADDR_REGISTER_VALUE reg_val = {0};
reg_val.gbAddrConfig = tileConfig.GbAddrConfig;
reg_val.noOfBanks = tileConfig.NumBanks;
reg_val.noOfRanks = tileConfig.NumRanks;
reg_val.pTileConfig = tileConfig.TileConfig;
reg_val.noOfEntries = tileConfig.NumTileConfigs;
reg_val.noOfMacroEntries = tileConfig.NumMacroTileConfigs;
reg_val.pMacroTileConfig = tileConfig.MacroTileConfig;
addr_create_input.regValue = reg_val;
addr_create_input.minPitchAlignPixels = 0;
ADDR_E_RETURNCODE addr_ret =
AddrCreate(&addr_create_input, &addr_create_output);
if (addr_ret == ADDR_OK) {
addr_lib_ = addr_create_output.hLib;
} else {
return HSA_STATUS_ERROR;
}
// The ImageManagerKv::Initialize is called on the first call to
// hsa_ext_image_*, so checking the coherency mode here is fine as long as
// the change to the coherency mode happens before a call to
// hsa_ext_image_create.
hsa_amd_coherency_type_t coherency_type;
status = AMD::hsa_amd_coherency_get_type(agent_, &coherency_type);
assert(status == HSA_STATUS_SUCCESS);
mtype_ = (coherency_type == HSA_AMD_COHERENCY_TYPE_COHERENT) ? 3 : 1;
// TODO: handle the case where the call to hsa_set_memory_type happens after
// hsa_ext_image_create.
hsa_region_t local_region = {0};
status = HSA::hsa_agent_iterate_regions(agent_, GetLocalMemoryRegion, &local_region);
assert(status == HSA_STATUS_SUCCESS);
local_memory_base_address_ = 0;
if (local_region.handle != 0) {
status = HSA::hsa_region_get_info(local_region,
static_cast<hsa_region_info_t>(HSA_AMD_REGION_INFO_BASE),
&local_memory_base_address_);
assert(status == HSA_STATUS_SUCCESS);
}
// Zeroed the queue object so it can be created on demand.
blit_queue_.queue_ = NULL;
blit_queue_.cached_index_ = 0;
return HSA_STATUS_SUCCESS;
}
void ImageManagerKv::Cleanup() {
if (blit_queue_.queue_ != NULL) {
HSA::hsa_queue_destroy(blit_queue_.queue_);
}
if (addr_lib_ != NULL) {
AddrDestroy(addr_lib_);
}
}
ImageProperty ImageManagerKv::GetImageProperty(
hsa_agent_t component, const hsa_ext_image_format_t& format,
hsa_ext_image_geometry_t geometry) const {
return ImageLut().MapFormat(format, geometry);
}
void ImageManagerKv::GetImageInfoMaxDimension(hsa_agent_t component,
hsa_ext_image_geometry_t geometry,
uint32_t& width, uint32_t& height,
uint32_t& depth,
uint32_t& array_size) const {
width = ImageLut().GetMaxWidth(geometry);
height = ImageLut().GetMaxHeight(geometry);
depth = ImageLut().GetMaxDepth(geometry);
array_size = ImageLut().GetMaxArraySize(geometry);
}
hsa_status_t ImageManagerKv::CalculateImageSizeAndAlignment(
hsa_agent_t component, const hsa_ext_image_descriptor_t& desc,
hsa_ext_image_data_layout_t image_data_layout,
size_t image_data_row_pitch,
size_t image_data_slice_pitch,
hsa_ext_image_data_info_t& image_info) const {
ADDR_COMPUTE_SURFACE_INFO_OUTPUT out = {0};
hsa_profile_t profile;
hsa_status_t status = HSA::hsa_agent_get_info(component, HSA_AGENT_INFO_PROFILE, &profile);
if (status != HSA_STATUS_SUCCESS) return status;
Image::TileMode tileMode = Image::TileMode::LINEAR;
if (image_data_layout == HSA_EXT_IMAGE_DATA_LAYOUT_OPAQUE) {
tileMode = (profile == HSA_PROFILE_BASE &&
desc.geometry != HSA_EXT_IMAGE_GEOMETRY_1DB)?
Image::TileMode::TILED : Image::TileMode::LINEAR;
}
if (!GetAddrlibSurfaceInfo(component, desc, tileMode,
image_data_row_pitch, image_data_slice_pitch, out)) {
return HSA_STATUS_ERROR;
}
size_t rowPitch = (out.bpp >> 3) * out.pitch;
size_t slicePitch = rowPitch * out.height;
if (desc.geometry != HSA_EXT_IMAGE_GEOMETRY_1DB &&
image_data_layout == HSA_EXT_IMAGE_DATA_LAYOUT_LINEAR &&
((image_data_row_pitch && (rowPitch != image_data_row_pitch)) ||
(image_data_slice_pitch && (slicePitch != image_data_slice_pitch)))) {
return static_cast<hsa_status_t>(HSA_EXT_STATUS_ERROR_IMAGE_PITCH_UNSUPPORTED);
}
image_info.size = out.surfSize;
assert(image_info.size != 0);
image_info.alignment = out.baseAlign;
assert(image_info.alignment != 0);
return HSA_STATUS_SUCCESS;
}
static const uint64_t kLimitSystem = 1ULL << 48;
bool ImageManagerKv::IsLocalMemory(const void* address) const {
uintptr_t u_address = reinterpret_cast<uintptr_t>(address);
uint32_t major_ver = MajorVerFromDevID(chip_id_);
if (major_ver >= 8) {
return true;
}
#ifdef HSA_LARGE_MODEL
// Fast path without querying local memory region info.
// User mode system memory addressable by CPU is 0 to 2^48.
return (u_address >= kLimitSystem);
#else
// No local memory on 32 bit.
return false;
#endif
}
hsa_status_t ImageManagerKv::PopulateImageSrd(Image& image, const metadata_amd_t* descriptor) const {
metadata_amd_ci_vi_t* desc = (metadata_amd_ci_vi_t*)descriptor;
bool atc_access = true;
uint32_t mtype = mtype_;
const void* image_data_addr = image.data;
ImageProperty image_prop = ImageLut().MapFormat(image.desc.format, image.desc.geometry);
if((image_prop.cap == HSA_EXT_IMAGE_CAPABILITY_NOT_SUPPORTED) ||
(image_prop.element_size == 0))
return (hsa_status_t)HSA_EXT_STATUS_ERROR_IMAGE_FORMAT_UNSUPPORTED;
uint32_t hwPixelSize =
ImageLut().GetPixelSize(desc->word1.bitfields.data_format, desc->word1.bitfields.num_format);
if(image_prop.element_size!=hwPixelSize)
return (hsa_status_t)HSA_EXT_STATUS_ERROR_IMAGE_FORMAT_UNSUPPORTED;
const Swizzle swizzle = ImageLut().MapSwizzle(image.desc.format.channel_order);
if (IsLocalMemory(image.data)) {
atc_access = false;
mtype = 1;
image_data_addr = reinterpret_cast<const void*>(
reinterpret_cast<uintptr_t>(image.data) - local_memory_base_address_);
}
image.srd[0]=desc->word0.u32_all;
image.srd[1]=desc->word1.u32_all;
image.srd[2]=desc->word2.u32_all;
image.srd[3]=desc->word3.u32_all;
image.srd[4]=desc->word4.u32_all;
image.srd[5]=desc->word5.u32_all;
image.srd[6]=desc->word6.u32_all;
image.srd[7]=desc->word7.u32_all;
((SQ_IMG_RSRC_WORD0*)(&image.srd[0]))->bits.base_address = PtrLow40Shift8(image_data_addr);
((SQ_IMG_RSRC_WORD1*)(&image.srd[1]))->bits.base_address_hi = PtrHigh64Shift40(image_data_addr);
((SQ_IMG_RSRC_WORD1*)(&image.srd[1]))->bits.data_format = image_prop.data_format;
((SQ_IMG_RSRC_WORD1*)(&image.srd[1]))->bits.num_format = image_prop.data_type;
((SQ_IMG_RSRC_WORD1*)(&image.srd[1]))->bits.mtype = mtype;
((SQ_IMG_RSRC_WORD3*)(&image.srd[3]))->bits.atc=atc_access;
((SQ_IMG_RSRC_WORD3*)(&image.srd[3]))->bits.dst_sel_x = swizzle.x;
((SQ_IMG_RSRC_WORD3*)(&image.srd[3]))->bits.dst_sel_y = swizzle.y;
((SQ_IMG_RSRC_WORD3*)(&image.srd[3]))->bits.dst_sel_z = swizzle.z;
((SQ_IMG_RSRC_WORD3*)(&image.srd[3]))->bits.dst_sel_w = swizzle.w;
((SQ_IMG_RSRC_WORD7*)(&image.srd[7]))->bits.meta_data_address += PtrLow40Shift8(image_data_addr);
//Looks like this is only used for CPU copies.
image.row_pitch = (desc->word4.bits.pitch+1)*image_prop.element_size;
image.slice_pitch = image.row_pitch * (desc->word2.bits.height+1);
//Used by HSAIL shader ABI
image.srd[8] = image.desc.format.channel_type;
image.srd[9] = image.desc.format.channel_order;
image.srd[10] = static_cast<uint32_t>(image.desc.width);
return HSA_STATUS_SUCCESS;
}
hsa_status_t ImageManagerKv::PopulateImageSrd(Image& image) const {
ImageProperty image_prop = ImageLut().MapFormat(image.desc.format, image.desc.geometry);
assert(image_prop.cap != HSA_EXT_IMAGE_CAPABILITY_NOT_SUPPORTED);
assert(image_prop.element_size != 0);
bool atc_access = true;
uint32_t mtype = mtype_;
const void* image_data_addr = image.data;
if (IsLocalMemory(image.data)) {
atc_access = false;
mtype = 1;
image_data_addr = reinterpret_cast<const void*>(
reinterpret_cast<uintptr_t>(image.data) - local_memory_base_address_);
}
if (image.desc.geometry == HSA_EXT_IMAGE_GEOMETRY_1DB) {
SQ_BUF_RSRC_WORD0 word0;
SQ_BUF_RSRC_WORD1 word1;
SQ_BUF_RSRC_WORD2 word2;
SQ_BUF_RSRC_WORD3 word3;
word0.u32_all = 0;
word0.bits.base_address = PtrLow32(image_data_addr);
word1.u32_all = 0;
word1.bits.base_address_hi = PtrHigh32(image_data_addr);
word1.bits.stride = image_prop.element_size;
word1.bits.swizzle_enable = false;
word1.bits.cache_swizzle = false;
uint32_t major_ver = MajorVerFromDevID(chip_id_);
word2.bits.num_records = (major_ver < 8) ?
image.desc.width : image.desc.width * image_prop.element_size;
const Swizzle swizzle = ImageLut().MapSwizzle(image.desc.format.channel_order);
word3.u32_all = 0;
word3.bits.dst_sel_x = swizzle.x;
word3.bits.dst_sel_y = swizzle.y;
word3.bits.dst_sel_z = swizzle.z;
word3.bits.dst_sel_w = swizzle.w;
word3.bits.num_format = image_prop.data_type;
word3.bits.data_format = image_prop.data_format;
word3.bits.atc = atc_access;
word3.bits.element_size = image_prop.element_size;
word3.bits.type = ImageLut().MapGeometry(image.desc.geometry);
word3.bits.mtype = mtype;
image.srd[0] = word0.u32_all;
image.srd[1] = word1.u32_all;
image.srd[2] = word2.u32_all;
image.srd[3] = word3.u32_all;
image.row_pitch = image.desc.width * image_prop.element_size;
image.slice_pitch = image.row_pitch;
} else {
SQ_IMG_RSRC_WORD0 word0;
SQ_IMG_RSRC_WORD1 word1;
SQ_IMG_RSRC_WORD2 word2;
SQ_IMG_RSRC_WORD3 word3;
SQ_IMG_RSRC_WORD4 word4;
SQ_IMG_RSRC_WORD5 word5;
SQ_IMG_RSRC_WORD6 word6;
SQ_IMG_RSRC_WORD7 word7;
ADDR_COMPUTE_SURFACE_INFO_OUTPUT out = {0};
if (!GetAddrlibSurfaceInfo(image.component, image.desc, image.tile_mode,
image.row_pitch, image.slice_pitch, out)) {
return HSA_STATUS_ERROR;
}
assert((out.bpp / 8) == image_prop.element_size);
const size_t row_pitch_size = out.pitch * image_prop.element_size;
word0.bits.base_address = PtrLow40Shift8(image_data_addr);
word1.u32_all = 0;
word1.bits.base_address_hi = PtrHigh64Shift40(image_data_addr);
word1.bits.min_lod = 0;
word1.bits.data_format = image_prop.data_format;
word1.bits.num_format = image_prop.data_type;
word1.bits.mtype = mtype;
word2.u32_all = 0;
word2.bits.width = image.desc.width - 1;
word2.bits.height = image.desc.height - 1;
word2.bits.perf_mod = 0;
word2.bits.interlaced = 0;
const Swizzle swizzle = ImageLut().MapSwizzle(image.desc.format.channel_order);
word3.u32_all = 0;
word3.bits.dst_sel_x = swizzle.x;
word3.bits.dst_sel_y = swizzle.y;
word3.bits.dst_sel_z = swizzle.z;
word3.bits.dst_sel_w = swizzle.w;
word3.bits.tiling_index = out.tileIndex;
word3.bits.pow2_pad = (IsPowerOfTwo(row_pitch_size) && IsPowerOfTwo(image.desc.height)) ? 1 : 0;
word3.bits.type = ImageLut().MapGeometry(image.desc.geometry);
word3.bits.atc = atc_access;
const bool image_array =
(image.desc.geometry == HSA_EXT_IMAGE_GEOMETRY_1DA ||
image.desc.geometry == HSA_EXT_IMAGE_GEOMETRY_2DA ||
image.desc.geometry == HSA_EXT_IMAGE_GEOMETRY_2DADEPTH);
const bool image_3d = (image.desc.geometry == HSA_EXT_IMAGE_GEOMETRY_3D);
word4.u32_all = 0;
word4.bits.depth =
(image_array)
? std::max(image.desc.array_size, static_cast<size_t>(1)) - 1
: (image_3d) ? image.desc.depth - 1 : 0;
word4.bits.pitch = out.pitch - 1;
word5.u32_all = 0;
word5.bits.last_array =
(image_array)
? (std::max(image.desc.array_size, static_cast<size_t>(1)) - 1)
: 0;
word6.u32_all = 0;
word7.u32_all = 0;
image.srd[0] = word0.u32_all;
image.srd[1] = word1.u32_all;
image.srd[2] = word2.u32_all;
image.srd[3] = word3.u32_all;
image.srd[4] = word4.u32_all;
image.srd[5] = word5.u32_all;
image.srd[6] = word6.u32_all;
image.srd[7] = word7.u32_all;
image.row_pitch = row_pitch_size;
image.slice_pitch = out.sliceSize;
}
image.srd[8] = image.desc.format.channel_type;
image.srd[9] = image.desc.format.channel_order;
image.srd[10] = static_cast<uint32_t>(image.desc.width);
return HSA_STATUS_SUCCESS;
}
hsa_status_t ImageManagerKv::ModifyImageSrd(
Image& image, hsa_ext_image_format_t& new_format) const {
image.desc.format = new_format;
ImageProperty image_prop = ImageLut().MapFormat(image.desc.format, image.desc.geometry);
assert(image_prop.cap != HSA_EXT_IMAGE_CAPABILITY_NOT_SUPPORTED);
assert(image_prop.element_size != 0);
if (image.desc.geometry == HSA_EXT_IMAGE_GEOMETRY_1DB) {
const Swizzle swizzle = ImageLut().MapSwizzle(image.desc.format.channel_order);
SQ_BUF_RSRC_WORD3* word3 =
reinterpret_cast<SQ_BUF_RSRC_WORD3*>(&image.srd[3]);
word3->bits.dst_sel_x = swizzle.x;
word3->bits.dst_sel_y = swizzle.y;
word3->bits.dst_sel_z = swizzle.z;
word3->bits.dst_sel_w = swizzle.w;
word3->bits.num_format = image_prop.data_type;
word3->bits.data_format = image_prop.data_format;
} else {
SQ_IMG_RSRC_WORD1* word1 =
reinterpret_cast<SQ_IMG_RSRC_WORD1*>(&image.srd[1]);
word1->bits.data_format = image_prop.data_format;
word1->bits.num_format = image_prop.data_type;
const Swizzle swizzle = ImageLut().MapSwizzle(image.desc.format.channel_order);
SQ_IMG_RSRC_WORD3* word3 =
reinterpret_cast<SQ_IMG_RSRC_WORD3*>(&image.srd[3]);
word3->bits.dst_sel_x = swizzle.x;
word3->bits.dst_sel_y = swizzle.y;
word3->bits.dst_sel_z = swizzle.z;
word3->bits.dst_sel_w = swizzle.w;
}
image.srd[8] = image.desc.format.channel_type;
image.srd[9] = image.desc.format.channel_order;
image.srd[10] = static_cast<uint32_t>(image.desc.width);
return HSA_STATUS_SUCCESS;
}
hsa_status_t ImageManagerKv::PopulateSamplerSrd(Sampler& sampler) const {
const hsa_ext_sampler_descriptor_v2_t &sampler_descriptor = sampler.desc;
SQ_IMG_SAMP_WORD0 word0;
SQ_IMG_SAMP_WORD1 word1;
SQ_IMG_SAMP_WORD2 word2;
SQ_IMG_SAMP_WORD3 word3;
word0.u32_all = 0;
hsa_status_t status = convertAddressMode<SQ_IMG_SAMP_WORD0, SQ_TEX_CLAMP>
(word0, sampler_descriptor.address_modes);
if (status != HSA_STATUS_SUCCESS) return status;
word0.bits.force_unormalized = (sampler_descriptor.coordinate_mode ==
HSA_EXT_SAMPLER_COORDINATE_MODE_UNNORMALIZED);
word1.u32_all = 0;
word1.bits.max_lod = 4095;
word2.u32_all = 0;
switch (sampler_descriptor.filter_mode) {
case HSA_EXT_SAMPLER_FILTER_MODE_NEAREST:
word2.bits.xy_mag_filter = static_cast<int>(SQ_TEX_XY_FILTER_POINT);
break;
case HSA_EXT_SAMPLER_FILTER_MODE_LINEAR:
word2.bits.xy_mag_filter = static_cast<int>(SQ_TEX_XY_FILTER_BILINEAR);
break;
default:
return HSA_STATUS_ERROR_INVALID_ARGUMENT;
}
word2.bits.xy_min_filter = word2.bits.xy_mag_filter;
word2.bits.z_filter = SQ_TEX_Z_FILTER_NONE;
word2.bits.mip_filter = SQ_TEX_MIP_FILTER_NONE;
word3.u32_all = 0;
// TODO: check this bit with HSAIL spec.
word3.bits.border_color_type = SQ_TEX_BORDER_COLOR_TRANS_BLACK;
sampler.srd[0] = word0.u32_all;
sampler.srd[1] = word1.u32_all;
sampler.srd[2] = word2.u32_all;
sampler.srd[3] = word3.u32_all;
return HSA_STATUS_SUCCESS;
}
hsa_status_t ImageManagerKv::CopyBufferToImage(
const void* src_memory, size_t src_row_pitch, size_t src_slice_pitch,
const Image& dst_image, const hsa_ext_image_region_t& image_region) {
if (BlitQueueInit().queue_ == NULL) {
return HSA_STATUS_ERROR_OUT_OF_RESOURCES;
}
return ImageRuntime::instance()->blit_kernel().CopyBufferToImage(
blit_queue_, blit_code_catalog_, src_memory, src_row_pitch, src_slice_pitch, dst_image,
image_region);
}
hsa_status_t ImageManagerKv::CopyImageToBuffer(
const Image& src_image, void* dst_memory, size_t dst_row_pitch,
size_t dst_slice_pitch, const hsa_ext_image_region_t& image_region) {
if (BlitQueueInit().queue_ == NULL) {
return HSA_STATUS_ERROR_OUT_OF_RESOURCES;
}
return ImageRuntime::instance()->blit_kernel().CopyImageToBuffer(
blit_queue_, blit_code_catalog_, src_image, dst_memory, dst_row_pitch, dst_slice_pitch,
image_region);
}
hsa_status_t ImageManagerKv::CopyImage(const Image& dst_image,
const Image& src_image,
const hsa_dim3_t& dst_origin,
const hsa_dim3_t& src_origin,
const hsa_dim3_t size) {
if (BlitQueueInit().queue_ == NULL) {
return HSA_STATUS_ERROR_OUT_OF_RESOURCES;
}
const hsa_ext_image_format_t src_format = src_image.desc.format;
const hsa_ext_image_channel_order32_t src_order = src_format.channel_order;
const hsa_ext_image_channel_type32_t src_type = src_format.channel_type;
const hsa_ext_image_format_t dst_format = dst_image.desc.format;
const hsa_ext_image_channel_order32_t dst_order = dst_format.channel_order;
const hsa_ext_image_channel_type32_t dst_type = dst_format.channel_type;
BlitKernel::KernelOp copy_type = BlitKernel::KERNEL_OP_COPY_IMAGE_DEFAULT;
if ((src_order == dst_order) && (src_type == dst_type)) {
return ImageRuntime::instance()->blit_kernel().CopyImage(blit_queue_, blit_code_catalog_,
dst_image, src_image, dst_origin,
src_origin, size, copy_type);
}
// Source and destination format must be the same, except for
// SRGBA <--> RGBA images.
if ((src_type == HSA_EXT_IMAGE_CHANNEL_TYPE_UNORM_INT8) &&
(dst_type == HSA_EXT_IMAGE_CHANNEL_TYPE_UNORM_INT8)) {
if ((src_order == HSA_EXT_IMAGE_CHANNEL_ORDER_SRGBA) &&
(dst_order == HSA_EXT_IMAGE_CHANNEL_ORDER_RGBA)) {
copy_type = BlitKernel::KERNEL_OP_COPY_IMAGE_STANDARD_TO_LINEAR;
} else if ((src_order == HSA_EXT_IMAGE_CHANNEL_ORDER_RGBA) &&
(dst_order == HSA_EXT_IMAGE_CHANNEL_ORDER_SRGBA)) {
copy_type = BlitKernel::KERNEL_OP_COPY_IMAGE_LINEAR_TO_STANDARD;
}
if (copy_type != BlitKernel::KERNEL_OP_COPY_IMAGE_DEFAULT) {
// KV and CZ don't have write support for SRGBA image, so treat the
// destination image as RGBA image.
SQ_IMG_RSRC_WORD1* word1 = reinterpret_cast<SQ_IMG_RSRC_WORD1*>(
&const_cast<Image&>(dst_image).srd[1]);
// Destination can be linear or standard, preserve the original value.
uint32_t num_format_original = word1->bits.num_format;
word1->bits.num_format = TYPE_UNORM;
hsa_status_t status = ImageRuntime::instance()->blit_kernel().CopyImage(
blit_queue_, blit_code_catalog_, dst_image, src_image, dst_origin, src_origin, size,
copy_type);
// Revert to the original format after the copy operation is finished.
word1->bits.num_format = num_format_original;
return status;
}
}
return HSA_STATUS_ERROR_INVALID_ARGUMENT;
}
hsa_status_t ImageManagerKv::FillImage(const Image& image, const void* pattern,
const hsa_ext_image_region_t& region) {
if (BlitQueueInit().queue_ == NULL) {
return HSA_STATUS_ERROR_OUT_OF_RESOURCES;
}
Image* image_view = const_cast<Image*>(&image);
SQ_BUF_RSRC_WORD3* word3_buff = NULL;
SQ_IMG_RSRC_WORD3* word3_image = NULL;
uint32_t dst_sel_w_original = 0;
if (image_view->desc.format.channel_type ==
HSA_EXT_IMAGE_CHANNEL_TYPE_UNORM_SHORT_101010) {
// Force GPU to ignore the last two bits (alpha bits).
if (image_view->desc.geometry == HSA_EXT_IMAGE_GEOMETRY_1DB) {
word3_buff = reinterpret_cast<SQ_BUF_RSRC_WORD3*>(&image_view->srd[3]);
dst_sel_w_original = word3_buff->bits.dst_sel_w;
word3_buff->bits.dst_sel_w = SEL_0;
} else {
word3_image = reinterpret_cast<SQ_IMG_RSRC_WORD3*>(&image_view->srd[3]);
dst_sel_w_original = word3_image->bits.dst_sel_w;
word3_image->bits.dst_sel_w = SEL_0;
}
}
SQ_IMG_RSRC_WORD1* word1 = NULL;
uint32_t num_format_original = 0;
const void* new_pattern = pattern;
float fill_value[4] = {0};
switch (image_view->desc.format.channel_order) {
case HSA_EXT_IMAGE_CHANNEL_ORDER_SRGBA:
case HSA_EXT_IMAGE_CHANNEL_ORDER_SRGB:
case HSA_EXT_IMAGE_CHANNEL_ORDER_SRGBX:
case HSA_EXT_IMAGE_CHANNEL_ORDER_SBGRA: {
// KV and CZ don't have write support for SRGBA image, so convert pattern
// to standard form and treat the image as RGBA image.
const float* pattern_f = reinterpret_cast<const float*>(pattern);
fill_value[0] = LinearToStandardRGB(pattern_f[0]);
fill_value[1] = LinearToStandardRGB(pattern_f[1]);
fill_value[2] = LinearToStandardRGB(pattern_f[2]);
fill_value[3] = pattern_f[3];
new_pattern = fill_value;
word1 = reinterpret_cast<SQ_IMG_RSRC_WORD1*>(&image_view->srd[1]);
num_format_original = word1->bits.num_format;
word1->bits.num_format = TYPE_UNORM;
} break;
default:
break;
}
hsa_status_t status = ImageRuntime::instance()->blit_kernel().FillImage(
blit_queue_, blit_code_catalog_, *image_view, new_pattern, region);
// Revert back original configuration.
if (word3_buff != NULL) {
word3_buff->bits.dst_sel_w = dst_sel_w_original;
}
if (word3_image != NULL) {
word3_image->bits.dst_sel_w = dst_sel_w_original;
}
if (word1 != NULL) {
word1->bits.num_format = num_format_original;
}
return status;
}
hsa_status_t ImageManagerKv::GetLocalMemoryRegion(hsa_region_t region,
void* data) {
if (data == NULL) {
return HSA_STATUS_ERROR_INVALID_ARGUMENT;
}
hsa_region_segment_t segment;
hsa_status_t stat = HSA::hsa_region_get_info(region, HSA_REGION_INFO_SEGMENT, &segment);
if (stat != HSA_STATUS_SUCCESS) {
return stat;
}
if (segment != HSA_REGION_SEGMENT_GLOBAL) {
return HSA_STATUS_SUCCESS;
}
uint32_t base = 0;
stat = HSA::hsa_region_get_info(region, HSA_REGION_INFO_GLOBAL_FLAGS, &base);
if (stat != HSA_STATUS_SUCCESS) {
return stat;
}
if ((base & HSA_REGION_GLOBAL_FLAG_COARSE_GRAINED) != 0) {
hsa_region_t* local_memory_region = (hsa_region_t*)data;
*local_memory_region = region;
}
return HSA_STATUS_SUCCESS;
}
AddrFormat ImageManagerKv::GetAddrlibFormat(const ImageProperty& image_prop) {
switch (image_prop.data_format) {
case FMT_8:
return ADDR_FMT_8;
break;
case FMT_16:
return (image_prop.data_type != TYPE_FLOAT) ? ADDR_FMT_16
: ADDR_FMT_16_FLOAT;
break;
case FMT_8_8:
return ADDR_FMT_8_8;
break;
case FMT_32:
return (image_prop.data_type != TYPE_FLOAT) ? ADDR_FMT_32
: ADDR_FMT_32_FLOAT;
break;
case FMT_16_16:
return (image_prop.data_type != TYPE_FLOAT) ? ADDR_FMT_16_16
: ADDR_FMT_16_16_FLOAT;
break;
case FMT_2_10_10_10:
return ADDR_FMT_2_10_10_10;
break;
case FMT_8_8_8_8:
return ADDR_FMT_8_8_8_8;
break;
case FMT_32_32:
return (image_prop.data_type != TYPE_FLOAT) ? ADDR_FMT_32_32
: ADDR_FMT_32_32_FLOAT;
break;
case FMT_16_16_16_16:
return (image_prop.data_type != TYPE_FLOAT) ? ADDR_FMT_16_16_16_16
: ADDR_FMT_16_16_16_16_FLOAT;
break;
case FMT_32_32_32_32:
return (image_prop.data_type != TYPE_FLOAT) ? ADDR_FMT_32_32_32_32
: ADDR_FMT_32_32_32_32_FLOAT;
break;
case FMT_5_6_5:
return ADDR_FMT_5_6_5;
break;
case FMT_1_5_5_5:
return ADDR_FMT_1_5_5_5;
break;
case FMT_8_24:
return ADDR_FMT_8_24;
break;
default:
assert(false && "Should not reach here");
return ADDR_FMT_INVALID;
break;
}
assert(false && "Should not reach here");
return ADDR_FMT_INVALID;
}
VOID* ADDR_API
ImageManagerKv::AllocSysMem(const ADDR_ALLOCSYSMEM_INPUT* input) {
return malloc(input->sizeInBytes);
}
ADDR_E_RETURNCODE ADDR_API
ImageManagerKv::FreeSysMem(const ADDR_FREESYSMEM_INPUT* input) {
free(input->pVirtAddr);
return ADDR_OK;
}
bool ImageManagerKv::GetAddrlibSurfaceInfo(
hsa_agent_t component, const hsa_ext_image_descriptor_t& desc,
Image::TileMode tileMode,
size_t image_data_row_pitch,
size_t image_data_slice_pitch,
ADDR_COMPUTE_SURFACE_INFO_OUTPUT& out) const {
const ImageProperty image_prop =
GetImageProperty(component, desc.format, desc.geometry);
const AddrFormat addrlib_format = GetAddrlibFormat(image_prop);
const uint32_t width = static_cast<uint32_t>(desc.width);
const uint32_t height = static_cast<uint32_t>(desc.height);
static const size_t kMinNumSlice = 1;
const uint32_t num_slice = static_cast<uint32_t>(
std::max(kMinNumSlice, std::max(desc.array_size, desc.depth)));
uint32_t major_ver = MajorVerFromDevID(chip_id_);
if (major_ver >= 9) {
ADDR2_COMPUTE_SURFACE_INFO_INPUT in = {0};
in.size = sizeof(ADDR2_COMPUTE_SURFACE_INFO_INPUT);
in.format = addrlib_format;
in.bpp = static_cast<unsigned int>(image_prop.element_size) * 8;
in.width = width;
in.height = height;
in.numSlices = num_slice;
in.pitchInElement = image_data_row_pitch / image_prop.element_size;
switch(desc.geometry) {
case HSA_EXT_IMAGE_GEOMETRY_1D:
case HSA_EXT_IMAGE_GEOMETRY_1DB:
in.resourceType = ADDR_RSRC_TEX_1D;
break;
case HSA_EXT_IMAGE_GEOMETRY_2D:
case HSA_EXT_IMAGE_GEOMETRY_2DDEPTH:
case HSA_EXT_IMAGE_GEOMETRY_1DA:
in.resourceType = ADDR_RSRC_TEX_2D;
break;
case HSA_EXT_IMAGE_GEOMETRY_3D:
case HSA_EXT_IMAGE_GEOMETRY_2DA:
case HSA_EXT_IMAGE_GEOMETRY_2DADEPTH:
{
in.resourceType = ADDR_RSRC_TEX_3D;
/*
* 3D swizzle modes enforce alignment
* of the number of slices to the block depth.
* If numSlices = 3 then the 3 slices are
* interleaved for 3D locality among the 8 slices
* that make up each block. This causes the memory
* footprint to jump to a 3x size of the ideal size
* 'enable3DSwizzleMode' flag tests for env variable
* HSA_IMAGE_ENABLE_3D_SWIZZLE_DEBUG to enable or disable
* 3D swizzle:
* true: Keep view3dAs2dArray = 0 for real 3D interleaving.
* false: Use view3dAs2dArray = 1 to avoid the alignment
* expansion.
* 2D swizzle modes can lower size overhead but may yield
* suboptimal cache behavior for fully 3D volumetric
* operations.
*/
bool enable3DSwizzleMode = core::Runtime::runtime_singleton_->flag().enable_3d_swizzle();
if (enable3DSwizzleMode)
in.flags.view3dAs2dArray = 0;
else
in.flags.view3dAs2dArray = 1;
break;
}
}
in.flags.texture = 1;
ADDR2_GET_PREFERRED_SURF_SETTING_INPUT prefSettingsInput = { 0 };
ADDR2_GET_PREFERRED_SURF_SETTING_OUTPUT prefSettingsOutput = { 0 };
prefSettingsInput.size = sizeof(prefSettingsInput);
prefSettingsInput.flags = in.flags;
prefSettingsInput.bpp = in.bpp;
prefSettingsInput.format = in.format;
prefSettingsInput.width = in.width;
prefSettingsInput.height = in.height;
prefSettingsInput.numFrags = in.numFrags;
prefSettingsInput.numSamples = in.numSamples;
prefSettingsInput.numMipLevels = in.numMipLevels;
prefSettingsInput.numSlices = in.numSlices;
prefSettingsInput.resourceLoction = ADDR_RSRC_LOC_UNDEF;
prefSettingsInput.resourceType = in.resourceType;
// Disallow all swizzles but linear.
if (tileMode == Image::TileMode::LINEAR)
{
prefSettingsInput.forbiddenBlock.macroThin4KB = 1;
prefSettingsInput.forbiddenBlock.macroThick4KB = 1;
prefSettingsInput.forbiddenBlock.macroThin64KB = 1;
prefSettingsInput.forbiddenBlock.macroThick64KB = 1;
}
prefSettingsInput.forbiddenBlock.micro = 1; // but don't ever allow the 256b swizzle modes
prefSettingsInput.forbiddenBlock.var = 1; // and don't allow variable-size block modes
if (ADDR_OK != Addr2GetPreferredSurfaceSetting(addr_lib_, &prefSettingsInput, &prefSettingsOutput)) {
return false;
}
in.swizzleMode = prefSettingsOutput.swizzleMode;
ADDR2_COMPUTE_SURFACE_INFO_OUTPUT out2 = {0};
out.size = sizeof(ADDR2_COMPUTE_SURFACE_INFO_OUTPUT);
if (ADDR_OK != Addr2ComputeSurfaceInfo(addr_lib_, &in, &out2)) {
return false;
}
out.pitch = out2.pitch;
out.height = out2.height;
out.surfSize = out2.surfSize;
out.bpp = out2.bpp;
out.baseAlign = out2.baseAlign;
out.tileIndex = in.swizzleMode;
out.sliceSize = out2.sliceSize;
return true;
}
ADDR_COMPUTE_SURFACE_INFO_INPUT in = {0};
in.size = sizeof(ADDR_COMPUTE_SURFACE_INFO_INPUT);
in.tileMode = (tileMode == Image::TileMode::LINEAR)?
ADDR_TM_LINEAR_ALIGNED : ADDR_TM_2D_TILED_THIN1;
in.format = addrlib_format;
in.bpp = static_cast<unsigned int>(image_prop.element_size) * 8;
in.numSamples = 1;
in.width = width;
in.height = height;
in.numSlices = num_slice;
in.flags.texture = 1;
in.flags.noStencil = 1;
in.flags.opt4Space = 0;
in.tileType = ADDR_NON_DISPLAYABLE;
in.tileIndex = -1;
if (image_data_row_pitch != 0) {
in.width = image_data_row_pitch / image_prop.element_size;
// in.pitchAlign = image_data_row_pitch / image_prop.element_size;
// in.heightAlign = image_data_slice_pitch / image_data_row_pitch;
}
if (ADDR_OK != AddrComputeSurfaceInfo(addr_lib_, &in, &out)) {
return false;
}
assert(out.tileIndex != -1);
return (out.tileIndex != -1) ? true : false;
}
size_t ImageManagerKv::CalWorkingSizeBytes(hsa_ext_image_geometry_t geometry,
hsa_dim3_t size_pixel,
uint32_t element_size) const {
switch (geometry) {
case HSA_EXT_IMAGE_GEOMETRY_1D:
case HSA_EXT_IMAGE_GEOMETRY_1DB:
return size_pixel.x * element_size;
case HSA_EXT_IMAGE_GEOMETRY_2D:
case HSA_EXT_IMAGE_GEOMETRY_2DDEPTH:
case HSA_EXT_IMAGE_GEOMETRY_1DA:
return size_pixel.x * size_pixel.y * element_size;
default:
return size_pixel.x * size_pixel.y * size_pixel.z * element_size;
}
}
BlitQueue& ImageManagerKv::BlitQueueInit() {
if (blit_queue_.queue_ == NULL) {
// Queue is a precious resource, so only create it when it is needed.
std::lock_guard<std::mutex> lock(lock_);
if (blit_queue_.queue_ == NULL) {
// Create the kernel queue.
blit_queue_.cached_index_ = 0;
uint32_t max_queue_size = 0;
hsa_status_t status =
HSA::hsa_agent_get_info(agent_, HSA_AGENT_INFO_QUEUE_MAX_SIZE, &max_queue_size);
status = HSA::hsa_queue_create(agent_, max_queue_size, HSA_QUEUE_TYPE_MULTI, NULL, NULL,
UINT_MAX, UINT_MAX, &blit_queue_.queue_);
if (HSA_STATUS_SUCCESS != status) {
blit_queue_.queue_ = NULL;
return blit_queue_;
}
// Get the kernel handles.
status = ImageRuntime::instance()->blit_kernel().BuildBlitCode(agent_, blit_code_catalog_);
if (HSA_STATUS_SUCCESS != status) {
blit_code_catalog_.clear();
HSA::hsa_queue_destroy(blit_queue_.queue_);
blit_queue_.queue_ = NULL;
return blit_queue_;
}
}
}
assert(blit_queue_.queue_ != NULL &&
blit_code_catalog_.size() == BlitKernel::KERNEL_OP_COUNT);
return blit_queue_;
}
} // namespace image
} // namespace rocr
Referencia en una nueva incidencia Ver la culpa de Git Copiar Permalink