Searched refs:LDS (Results 1 – 25 of 61) sorted by relevance
123
| /external/llvm/test/CodeGen/AMDGPU/ |
| D | lds-oqap-crash.ll | 5 ; if an LDS instruction could fit in the current clause, but never finding
8 ; The LDS instruction should not have been defining OQAP in the first place,
9 ; because the LDS instructions are pseudo instructions and the OQAP
|
| D | gep-address-space.ll | 15 ; The LDS offset will be 65536 bytes, which is larger than the size of LDS on
|
| D | promote-alloca-padding-size-estimate.ll | 3 ; This shows that the amount of LDS estimate is sensitive to the order
4 ; of the LDS globals.
6 ; Both of these functions use the same amount of LDS, but the total
|
| D | private-memory-atomics.ll | 4 ; This works because promote allocas pass replaces these with LDS atomics.
|
| D | local-64.ll | 36 ; The LDS offset will be 65536 bytes, which is larger than the size of LDS on
|
| D | promote-alloca-stored-pointer-value.ll | 4 ; Pointer value is stored in a candidate for LDS usage.
|
| D | llvm.SI.fs.interp.ll | 23 ; on 16 bank LDS chips.
|
| D | local-memory-two-objects.ll | 9 ; Check that the LDS size emitted correctly
|
| D | local-memory.ll | 8 ; Check that the LDS size emitted correctly
|
| /external/swiftshader/third_party/llvm-7.0/llvm/test/CodeGen/AMDGPU/ |
| D | lds-oqap-crash.ll | 5 ; if an LDS instruction could fit in the current clause, but never finding
8 ; The LDS instruction should not have been defining OQAP in the first place,
9 ; because the LDS instructions are pseudo instructions and the OQAP
|
| D | gep-address-space.ll | 15 ; The LDS offset will be 65536 bytes, which is larger than the size of LDS on
|
| D | local-memory.amdgcn.ll | 6 ; Check that the LDS size emitted correctly
39 ; Check that the LDS size emitted correctly
|
| D | local-memory.r600.ll | 5 ; Check that the LDS size emitted correctly
37 ; Check that the LDS size emitted correctly
|
| D | promote-alloca-padding-size-estimate.ll | 3 ; This shows that the amount of LDS estimate is sensitive to the order
4 ; of the LDS globals.
6 ; Both of these functions use the same amount of LDS, but the total
|
| D | private-memory-atomics.ll | 4 ; This works because promote allocas pass replaces these with LDS atomics.
|
| D | local-64.ll | 49 ; The LDS offset will be 65536 bytes, which is larger than the size of LDS on
|
| D | promote-alloca-stored-pointer-value.ll | 4 ; Pointer value is stored in a candidate for LDS usage.
|
| D | coalescing-with-subregs-in-loop-bug.mir | 6 # in an LDS read. %29 is a 128 bit value that is linked by copies to
|
| /external/swiftshader/third_party/llvm-7.0/llvm/lib/Target/AMDGPU/ |
| D | SIMemoryLegalizer.cpp | 85 LDS = 1u << 1, enumerator
91 FLAT = GLOBAL | LDS | SCRATCH,
94 ATOMIC = GLOBAL | LDS | SCRATCH | GDS,
97 ALL = GLOBAL | LDS | SCRATCH | GDS | OTHER,
461 return SIAtomicAddrSpace::LDS; in toSIAtomicAddrSpace()
747 if ((AddrSpace & SIAtomicAddrSpace::LDS) != SIAtomicAddrSpace::NONE) { in insertWait()
|
| D | SIInstrFormats.td | 308 // VINTRP instructions read parameter values from LDS, but these parameter
309 // values are stored outside of the LDS memory that is allocated to the
313 // the parameter values in LDS, this would essentially be an out-of-bounds
|
| /external/swiftshader/third_party/llvm-7.0/llvm/docs/ |
| D | AMDGPUOperandSyntax.rst | 122 Specifies whether to use GDS or LDS memory (LDS is the default).
511 Specifies where to store the result: VGPRs or LDS (VGPRs by default).
516 lds Store result in LDS.
559 Specifies which half of the LDS word to use. Low half of LDS word is used by default.
565 high Use high half of LDS word.
|
| D | AMDGPUUsage.rst | 284 3 Local (group/LDS)
848 2 Local (group/LDS)
1168 to dynamically allocated LDS
1500 Local group LDS 32 0xFFFFFFFF
1520 The local memory space uses the hardware Local Data Store (LDS) which is
1549 and group (LDS) memory depending in if the address is within one of the
1552 access to LDS requires hardware aperture setup and M0 (GFX7-GFX8) register setup
2001 L1 or LDS
2003 mis-aligned atomic, LDS
2024 (LDS) to allocate for each
[all …]
|
| /external/mesa3d/src/amd/compiler/ |
| D | README.md | 172 * LS and HS share the same LDS space, so LS can store its output to LDS, where HS can read it
185 * HW LS and HS stages are merged, and the merged shader still uses LDS in the same way as before
186 * HW ES and GS stages are merged, so ES outputs can go to LDS instead of VRAM
|
| /external/mesa3d/docs/relnotes/ |
| D | 17.3.8.rst | 66 - radv: get correct offset into LDS for indexed vars.
|
| D | 20.3.1.rst | 146 - radv: don't count unusable vertices to the NGG LDS size
|
123