intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-21 09:14:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

feature: introduce heapless state init in L0

Browse Source 
Related-To: NEO-7824
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
This commit is contained in:
Kamil Kopryk
2024-03-29 03:22:39 +00:00
committed by Compute-Runtime-Automation
parent 8ef6cdbabd
commit 4eae28bd64
25 changed files with 291 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
shared/source/command_container/command_encoder_xehp_and_later.inl
View File

@@ -18,6 +18,7 @@
#include "shared/source/gmm_helper/client_context/gmm_client_context.h"
#include "shared/source/helpers/basic_math.h"
#include "shared/source/helpers/cache_policy.h"
#include "shared/source/helpers/compiler_product_helper.h"
#include "shared/source/helpers/constants.h"
#include "shared/source/helpers/gfx_core_helper.h"
#include "shared/source/helpers/hw_walk_order.h"
@@ -284,33 +285,37 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
}
}
if (container.isAnyHeapDirty() ||
args.requiresUncachedMocs) {
bool heaplessStateInitEnabled = rootDeviceEnvironment.getHelper<CompilerProductHelper>().isHeaplessStateInitEnabled();
PipeControlArgs syncArgs;
syncArgs.dcFlushEnable = args.dcFlushEnable;
MemorySynchronizationCommands<Family>::addSingleBarrier(*container.getCommandStream(), syncArgs);
STATE_BASE_ADDRESS sbaCmd;
auto gmmHelper = container.getDevice()->getGmmHelper();
uint32_t statelessMocsIndex =
args.requiresUncachedMocs ? (gmmHelper->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER_CACHELINE_MISALIGNED) >> 1) : (gmmHelper->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER) >> 1);
auto l1CachePolicy = container.l1CachePolicyDataRef()->getL1CacheValue(false);
auto l1CachePolicyDebuggerActive = container.l1CachePolicyDataRef()->getL1CacheValue(true);
if (heaplessStateInitEnabled == false) {
if (container.isAnyHeapDirty() ||
args.requiresUncachedMocs) {
EncodeStateBaseAddressArgs<Family> encodeStateBaseAddressArgs = {
&container, // container
sbaCmd, // sbaCmd
nullptr, // sbaProperties
statelessMocsIndex, // statelessMocsIndex
l1CachePolicy, // l1CachePolicy
l1CachePolicyDebuggerActive, // l1CachePolicyDebuggerActive
args.partitionCount > 1, // multiOsContextCapable
args.isRcs, // isRcs
container.doubleSbaWaRef(), // doubleSbaWa
heaplessModeEnabled, // heaplessModeEnabled
};
EncodeStateBaseAddress<Family>::encode(encodeStateBaseAddressArgs);
container.setDirtyStateForAllHeaps(false);
PipeControlArgs syncArgs;
syncArgs.dcFlushEnable = args.dcFlushEnable;
MemorySynchronizationCommands<Family>::addSingleBarrier(*container.getCommandStream(), syncArgs);
STATE_BASE_ADDRESS sbaCmd;
auto gmmHelper = container.getDevice()->getGmmHelper();
uint32_t statelessMocsIndex =
args.requiresUncachedMocs ? (gmmHelper->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER_CACHELINE_MISALIGNED) >> 1) : (gmmHelper->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER) >> 1);
auto l1CachePolicy = container.l1CachePolicyDataRef()->getL1CacheValue(false);
auto l1CachePolicyDebuggerActive = container.l1CachePolicyDataRef()->getL1CacheValue(true);
EncodeStateBaseAddressArgs<Family> encodeStateBaseAddressArgs = {
&container, // container
sbaCmd, // sbaCmd
nullptr, // sbaProperties
statelessMocsIndex, // statelessMocsIndex
l1CachePolicy, // l1CachePolicy
l1CachePolicyDebuggerActive, // l1CachePolicyDebuggerActive
args.partitionCount > 1, // multiOsContextCapable
args.isRcs, // isRcs
container.doubleSbaWaRef(), // doubleSbaWa
heaplessModeEnabled, // heaplessModeEnabled
};
EncodeStateBaseAddress<Family>::encode(encodeStateBaseAddressArgs);
container.setDirtyStateForAllHeaps(false);
}
}
if (NEO::PauseOnGpuProperties::pauseModeAllowed(NEO::debugManager.flags.PauseOnEnqueue.get(), args.device->debugExecutionCounter.load(), NEO::PauseOnGpuProperties::PauseMode::BeforeWorkload)) {