Media/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-09-15 13:01:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

Zebin improvements

Browse Source 
* cleaning-up per-thread buffer size calculations
* fixing default slm alignment to 16
* adding scratch space slots
* renaming slm addressing mode
* adding support for zeinfo version

Change-Id: I0450353788b8488a685976d7f1dd26a44360383c
This commit is contained in:
Jaroslaw Chodor
2020-10-04 19:18:49 +02:00
committed by sys_ocldev
parent 20f4786423
commit 345e56a151
19 changed files with 553 additions and 84 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
opencl/source/kernel/kernel.cpp
View File

@ -10,6 +10,7 @@
#include "shared/source/built_ins/built_ins.h"
#include "shared/source/command_stream/command_stream_receiver.h"
#include "shared/source/debug_settings/debug_settings_manager.h"
#include "shared/source/device_binary_format/patchtokens_decoder.h"
#include "shared/source/gmm_helper/gmm_helper.h"
#include "shared/source/helpers/aligned_memory.h"
#include "shared/source/helpers/api_specific_config.h"
@ -271,13 +272,10 @@ cl_int Kernel::initialize() {
localBindingTableOffset = (patchInfo.bindingTableState != nullptr) ? patchInfo.bindingTableState->Offset : 0;
// patch crossthread data and ssh with inline surfaces, if necessary
auto perThreadPrivateMemorySize = patchInfo.pAllocateStatelessPrivateSurface
? patchInfo.pAllocateStatelessPrivateSurface->PerThreadPrivateMemorySize
: 0;
auto perHwThreadPrivateMemorySize = PatchTokenBinary::getPerHwThreadPrivateSurfaceSize(patchInfo.pAllocateStatelessPrivateSurface, getKernelInfo().getMaxSimdSize());
if (perThreadPrivateMemorySize) {
privateSurfaceSize = KernelHelper::getPrivateSurfaceSize(perThreadPrivateMemorySize, device.getSharedDeviceInfo().computeUnitsUsedForScratch,
getKernelInfo().getMaxSimdSize(), patchInfo.pAllocateStatelessPrivateSurface->IsSimtThread);
if (perHwThreadPrivateMemorySize) {
privateSurfaceSize = KernelHelper::getPrivateSurfaceSize(perHwThreadPrivateMemorySize, device.getSharedDeviceInfo().computeUnitsUsedForScratch);
DEBUG_BREAK_IF(privateSurfaceSize == 0);
if (privateSurfaceSize > std::numeric_limits<uint32_t>::max()) {