Commit Graph

4251 Commits

Author SHA1 Message Date
f57ff2913c Optimize small buffers allocator
- Do not wait for GPU completion on pool exhaust if allocs are in use,
allocate new pool instead
- Reuse existing pool if allocs are not in use

Related-To: NEO-7769

Signed-off-by: Igor Venevtsev <igor.venevtsev@intel.com>
2023-03-15 19:12:30 +01:00
0bee81c0c0 refactor: Move isLinearStoragePreferred function from gfx to product helper
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2023-03-15 18:51:59 +01:00
5ffb93cf2a refactor(ocl): remove redundant check
Remove redundant check for graphicsAllocations.size() in set kernel arg
svm method.

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2023-03-15 16:20:53 +01:00
a23230ebcf refactor: Remove tiledImagesSupported value
The value is redundant and capabilityTable.supportsImages
can be used instead.

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2023-03-15 15:48:30 +01:00
686790274d Disable OCL RelaxedOrdering if tag update is required
Related-To: NEO-7458

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-03-15 14:44:05 +01:00
e62c5e25d5 refactor: change debugging enabled to debugging mode
Related-To: NEO-7630

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2023-03-15 13:41:41 +01:00
db719b2acb refactor: rename function getExtensions->getDeviceExtensions
Related-To: NEO-7800
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-15 12:46:46 +01:00
028e1bb1b6 fix ocloc: define extensions in one place
unify exposed device extensions between OCL and ocloc

Related-To: NEO-7800
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-15 11:14:46 +01:00
320deec53b Wait for barrier on every OCL split blitter
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2023-03-15 09:34:41 +01:00
faba16f657 refactor: move extensions string creation to CompilerProductHelper
Related-To: NEO-7800
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-14 16:27:21 +01:00
477448c097 RelaxedOrdering dispatch for OCL 2/n
Related-To: NEO-7458

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-03-14 15:48:06 +01:00
340f932ca2 refactor: move GfxCoreHelper::getExtensions to CompilerProductHelper
Related-To: NEO-7800
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-14 13:56:19 +01:00
15485fb1c8 Register CSR clients for OCL CommandQueues
Related-To: NEO-7458

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-03-13 18:29:14 +01:00
97a35debde RelaxedOrdering dispatch for OCL 1/n
Related-To: NEO-7458

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-03-13 17:37:12 +01:00
3cfba5b73d refactor(cmake): check images support per platform instead of per core
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2023-03-13 16:14:39 +01:00
00b675643e fix: missed error handler of SubmissionStatus::FAILED
Related-To: NEO-7802

Signed-off-by: Pan Zhenjie <zhenjie.pan@intel.com>
2023-03-13 16:08:27 +01:00
91ced4edcc fix: Correct SCS programming for CL_DEPTH format
Resolves: HSD-14018806608

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2023-03-13 10:48:28 +01:00
f3324964f6 [perf] initialize stream properties only once without further check
Related-To: NEO-5055

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2023-03-10 17:50:49 +01:00
24c8f089ed [perf] add state compute mode dirty flag to allow selective properties update
- full properties update is time intesive task and must be done only once
- selective update can be done after initial update
- dirty flag will allow to distinguish initial update is done

Related-To: NEO-5055

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2023-03-10 17:27:08 +01:00
5a99635232 Don't check addressing mode for gtpin surface
If gtpin is used, then don't check addressing mode
of the last explicit arg, which is
gtpin's surface.

Related-To: NEO-6075
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2023-03-10 16:01:24 +01:00
398c7b2d29 refactor, remove typo in struct name
change name of EncodeSempahore to EncodeSemaphore
Signed-off-by: Cencelewska, Katarzyna <katarzyna.cencelewska@intel.com>
2023-03-10 15:44:25 +01:00
89cd4bd4a5 refactor: move getters for device extension support to compiler product helper
isMatrixMultiplyAccumulateSupported
isBFloat16ConversionSupported

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-10 15:41:26 +01:00
fa8579602f refactor: rename product helper files n/n
Related-To: NEO-7703
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2023-03-10 13:24:38 +01:00
c274309d7b wa: add dummy blits before command MI_FLUSH_DW
to guarantee that all subblt got complete for previous copy
affect xe hpg

temporary changes under flag ForceDummyBlitWa

Related-To: NEO-7450

Signed-off-by: Cencelewska, Katarzyna <katarzyna.cencelewska@intel.com>
2023-03-09 10:40:35 +01:00
69a16fd3ed feature: check indirect access for kernel
Do not make indirect allocations resident if kernel does not use
indirect access.
For both level zero and opencl.
Currently disabled by default, enable with debug flag
DetectIndirectAccessInKernel

Related-To: NEO-7712

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2023-03-08 16:58:26 +01:00
553dd7f21f refactor: return thread per eu from compiler product helper
Related-To: NEO-7442
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-08 16:25:20 +01:00
0c5264dfeb Ocloc - fix binary output filename with "-output_no_suffix"
- .bin extension should be added to produced binary file when
"-output_no_suffix" flag is passed

Related-To: NEO-7474

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2023-03-08 15:51:20 +01:00
6ab9037c35 Use FusedEu Dpas W/A only in ACM
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2023-03-08 14:14:58 +01:00
67b14e2034 refactor: rename test product helper files
Related-To: NEO-7703
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2023-03-08 13:21:34 +01:00
01db7df52b refactor: rename raiiProductHelper file
Related-To: NEO-7703
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2023-03-08 13:02:54 +01:00
86e739e9dc feature: Relaxed Virtual Memory reservation with pStart
Related-To: LOCI-3871

- Relaxed the Virtual Memory Reservation to allow pStart and not fail if
the pStart value is not obtained.
- Moves checks on pStart to the user to check and determine if they want
to re-reserve or use the address allocated.
- Changed reserveGpuAddress to use unit64_t type to allow internal
address range structure assignment without cast.

Signed-off-by: Spruit, Neil R <neil.r.spruit@intel.com>
2023-03-08 08:06:50 +01:00
55952d33df feature: Enable use KMD calls in prefetch APIs
If applications call Prefetch APIs, like
zeCommandListAppendMemoryPrefetch and
clEnqueueMigrateMemINTEL, then enable the use of KMD calls
by default.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2023-03-07 16:31:50 +01:00
27636f7258 fix(gtpin/ocl): Pass debug zebin to gtpin
In order to unify behaviour between L0/OCL with GTPin,
OCL callbacks should return debug zebin (if instrumented binary is
zebin). This commit fixes currently exisitng issue with no debug data being
passed to GTPin if instrumented binary is zebin.
Signed-off-by: Kacper Nowak <kacper.nowak@intel.com>
2023-03-07 15:35:42 +01:00
efba242570 fix(zebin): Extend oneDNN WA for whole application context
When a dummy kernel "kernel void_(){}" is passed in sources - specific
for workloads with ngen backend - enforce fallback to CTNI for the whole
application context (mark the context as non-zebinary).

Related-To: NEO-7772
Signed-off-by: Kacper Nowak <kacper.nowak@intel.com>
2023-03-07 14:21:57 +01:00
6c59953072 Add FP64 emulation support for ATS-M
This patch adds FP64 emulation support for ATS-M.
Introducing new environment variable - NEO_FP64_EMULATION - which provides
an option to allow the opt-in emulation of FP64.
When emulation is enabled, we pass -cl-fp64-gen-emu (ocl) /
-ze-fp64-gen-emu (L0) as an internal option to IGC.

Related-To: NEO-7611
Signed-off-by: Fabian Zwolinski <fabian.zwolinski@intel.com>
2023-03-07 14:21:01 +01:00
3e116ea378 refactor: use same paths when add command mi_semaphore_wait
Signed-off-by: Cencelewska, Katarzyna <katarzyna.cencelewska@intel.com>
2023-03-07 10:35:26 +01:00
50da32ffb1 wa: add dummy blits before command MI_ARB_CHECK
to guarantee that all subblt got complete for previous copy
affect xe hpg

Related-To: NEO-7450

Signed-off-by: Cencelewska, Katarzyna <katarzyna.cencelewska@intel.com>
2023-03-07 10:21:05 +01:00
2e872c1871 fix ocloc: don't fail when help parameter is passed to ocloc compile
https://github.com/intel/compute-runtime/issues/621

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-06 10:08:00 +01:00
58d3f892a9 refactor(zebin): move files to seperate directory
* Moved zebin related files to zebin directory.
* Moved zebin related code to Zebin namespace.
* Separated zeInfo from zebin elf.
* Seperated zeInfo decoding from zebin decoder.
* Refactored populateKernelPayloadArgument function.

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
2023-03-03 21:48:19 +01:00
52d322e738 Move barrier flush property from csr to cmdQueue
Related-To: NEO-6982, HSD-15010621906

Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2023-03-03 14:33:53 +01:00
84a7438ff2 Revert "refactor(zeInfo): move zeinfo to seperate file"
This reverts commit e3ce887662.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2023-03-03 11:56:52 +01:00
72ada8b41f Add missing pollForCompletion on clEnqueueVerifyMemoryINTEL
Related-To: NEO-7618

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-03-03 10:55:33 +01:00
0b7817eecf feature: Experimental OCL debugging
- notify module create/destroy

Related-To: NEO-7075

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2023-03-02 15:59:15 +01:00
e3ce887662 refactor(zeInfo): move zeinfo to seperate file
* Moved zebin related files to zebin directory.
* Separated zeInfo from zebin elf.
* Seperated zeInfo decoding from zebin decoder.
* Refactored populateKernelPayloadArgument function.

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
2023-03-02 12:46:56 +01:00
5caf7f7290 feat(linker): allow use of local symbols
This commit adds support for handling local symbols.
* Added 2 fields to SymbolInfo - binding, and associated
instructions segment id.
* Simplified code for decoding elf symbols and relocations.
* Simplified code for patching instruction segments.
* Changed logic of decodeElfSymbolTableAndRelocations:
* Add every global symbol to symbol map.
* Add any local symbol used by relocation to symbol map.
* Changed logic of link:
* After performing relocations remove local symbols from map.
* Replaced UNRECOVERABLE_IF with returning error.
* Removed LocalSymbolInfo structure used before for local kernel jumps.
* Removed old tests.

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
2023-03-02 12:39:21 +01:00
9f7f2318bd fix(ocloc): fallback to deprecated acronyms
When revision_id argument is passed fallback to using
initHardwareInfoForDeprecatedAcronyms.
* initHardwareInfoForProductConfig returns INVALID_DEVICE if device
acronym and revision id is passed.
* initHardwareInfoForDeprecatedAcronyms sets aot config revision
to revisionId.

Resolves: NEO-7738

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
2023-03-02 11:43:34 +01:00
bd32b21f7c refactor: add method to get hw config from CompilerProductHelper
Related-To: NEO-7622

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-03-02 08:25:50 +01:00
4320fd6785 Fix LUID and node mask deviceInfo on Windows
Resolves: NEO-7761

Signed-off-by: Kacper Kasper <kacper.k.kasper@intel.com>
2023-03-01 08:30:03 +01:00
aece8195eb Add support for cl_ext_float_atomics
Resolves: NEO-6596

Signed-off-by: Kacper Kasper <kacper.k.kasper@intel.com>
2023-02-28 10:26:48 +01:00
c28f0c72ea test: fix cleanup order in ULTs
- program using device must be released before device is destroyed.
Reset program unique_ptr in fixtures before device fixture tearDown()
- reorder fixtures members to ensure correct order of destructors

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2023-02-27 18:23:16 +01:00