Home
last modified time | relevance | path

Searched full:vectorized (Results 1 – 25 of 72) sorted by relevance

123

/freebsd/contrib/llvm-project/llvm/lib/Transforms/Vectorize/
H A DLoopVectorizationLegality.cpp150 // consider the loop to have been already vectorized because there's in LoopVectorizeHints()
192 LLVM_DEBUG(dbgs() << "LV: Not vectorizing: Disabled/already vectorized.\n"); in allowVectorization()
200 << "loop not vectorized: vectorization and interleaving are " in allowVectorization()
202 "vectorized"; in allowVectorization()
218 << "loop not vectorized: vectorization is explicitly disabled"; in emitRemarkWithHints()
222 R << "loop not vectorized"; in emitRemarkWithHints()
315 // \p OuterLp (i.e., if the outer loop is vectorized, all the vector lanes
480 /// vectorized loop, which can then be compared to detect their uniformity. This
687 // in the vectorized loop body, record them here. All casts could be recorded in addInductionPhi()
760 /// the sense that it should be vectorized and then expanded in
[all …]
H A DSLPVectorizer.cpp330 // Constant expressions and globals can't be vectorized like normal integer/FP in allConstant()
752 /// could be vectorized even if its structure is diverse.
921 /// possible scalar operand in vectorized instruction.
1170 /// Returns the vectorized root.
1217 /// Builds external uses of the vectorized scalars, i.e. the list of
1218 /// vectorized scalars to be extracted, their lanes and their scalar users. \p
1265 /// Gets reordering data for the given tree entry. If the entry is vectorized
1270 /// \param TopToBottom If true, include the order of vectorized stores and
1353 /// Checks if the given array of loads can be represented as a vectorized,
1452 /// Score if all users are vectorized
[all...]
H A DLoopVectorize.cpp177 STATISTIC(LoopsVectorized, "Number of loops vectorized");
179 STATISTIC(LoopsEpilogueVectorized, "Number of epilogues vectorized");
196 /// Loops with a known constant trip count below this number are vectorized only
201 "value are vectorized only if no scalar iteration overheads "
296 "vectorized loops."));
400 // Likelyhood of bypassing the vectorized loop because assumptions about SCEV
403 // Likelyhood of bypassing the vectorized loop because pointers overlap. See
406 // Likelyhood of bypassing the vectorized loop because there are zero trips left
497 /// Create a new empty loop that will contain vectorized instructions later
499 /// is generated around the vectorized (and scalar epilogue) loops consisting
[all …]
H A DLoadStoreVectorizer.cpp37 // multiple scalar registers, similar to a GPU vectorized load. In theory ARM
123 STATISTIC(NumScalarsVectorized, "Number of scalar accesses vectorized");
291 /// Converts the instrs in the chain into a single vectorized load or store.
773 << " is misaligned, and therefore can't be vectorized.\n"); in splitChainByAlignment()
787 << ". Therefore this access won't be vectorized.\n"); in splitChainByAlignment()
927 // vectorized the following code: in vectorizeChain()
934 // We will put the vectorized load at the location of the earliest load in in vectorizeChain()
1317 // be vectorized. in collectEquivalenceClasses()
1347 // functions are currently using an integer type for the vectorized in collectEquivalenceClasses()
H A DLoopVectorizationPlanner.h16 /// control-flow of the vectorized version, the replication of instructions that
371 /// Generate the IR code for the vectorized loop captured in VPlan \p BestPlan
461 /// epilogue, assuming the main loop is vectorized by \p VF.
H A DVPlanTransforms.cpp418 /// in the vectorized loop. There is no need to vectorize the cast - the same
427 // *must be bypassed* when the IV is vectorized, because the vectorized IV in removeRedundantInductionCasts()
1111 // If the value wasn't vectorized, we must maintain the original scalar in truncateToMinimalBitwidths()
/freebsd/contrib/llvm-project/llvm/include/llvm/Transforms/Vectorize/
H A DLoopVectorizationLegality.h91 /// Already Vectorized
128 /// Mark the loop L as already vectorized by setting the width to 1.
180 // Otherwise, a sequence of vectorized loops, even without reduction, in isPotentiallyUnsafe()
337 /// to be vectorized.
387 /// has a vectorized variant available.
454 /// calls to the appropriate masked intrinsic when the loop is vectorized
512 /// runtime guard). These casts can be ignored when creating the vectorized
550 /// vectorized variant, record that fact so that LoopVectorize can
H A DSLPVectorizer.h96 /// \returns true if a value was vectorized.
/freebsd/contrib/llvm-project/llvm/lib/Target/NVPTX/
H A DNVPTXTargetTransformInfo.h60 // Loads and stores can be vectorized if the alignment is at least as big as
78 // Only <2 x half> should be vectorized, so always return 32 for the vector
/freebsd/sys/contrib/device-tree/Bindings/interrupt-controller/
H A Dloongson,htvec.yaml14 receiving vectorized interrupts from PCH's interrupt controller.
H A Dloongson,pch-pic.yaml14 transforming interrupts from on-chip devices into HyperTransport vectorized
H A Dloongson,pch-msi.yaml14 transforming interrupts from PCIe MSI into HyperTransport vectorized
/freebsd/contrib/llvm-project/llvm/include/llvm/Transforms/Scalar/
H A DLoopDistribute.h10 // distribute loops that cannot be vectorized due to dependence cycles. It
/freebsd/contrib/llvm-project/llvm/include/llvm/Analysis/
H A DTargetLibraryInfo.h29 /// vectorized by a factor 'VectorizationFactor'.
211 /// Return the name of the equivalent of F, vectorized with factor VF. If no
217 /// mappings in TLI for the equivalent of F, vectorized with factor VF.
H A DLoopAccessAnalysis.h138 // Forward, but if vectorized, is likely to prevent store-to-load
654 /// to be vectorized.
714 /// vectorized.
/freebsd/contrib/llvm-project/llvm/lib/Target/RISCV/
H A DRISCVTargetTransformInfo.h335 // Don't interleave if the loop has been vectorized with scalable vectors. in getMaxInterleaveFactor()
338 // If the loop will not be vectorized, don't interleave the loop. in getMaxInterleaveFactor()
/freebsd/contrib/llvm-project/llvm/lib/CodeGen/
H A DReplaceWithVeclib.cpp98 /// vectorized intrinsic, with a suitable function taking vector arguments,
157 // There is no guarantee that the vectorized instructions followed the VFABI in replaceWithCallToVeclib()
/freebsd/contrib/llvm-project/llvm/lib/Transforms/Scalar/
H A DWarnMissedTransforms.cpp60 << "loop not vectorized: the optimizer was unable to perform the " in warnAboutLeftoverTransformations()
/freebsd/crypto/openssl/crypto/sha/asm/
H A Dsha1-alpha.pl21 # 8.7 cycles per processed byte. Implementation features vectorized
57 srl @X[$i],24,$t0 # vectorized byte swap
H A Dsha1-s390x.pl84 return if ($i&1); # Xupdate is vectorized and executed every 2nd cycle
/freebsd/contrib/llvm-project/llvm/lib/Target/AMDGPU/
H A DR600TargetTransformInfo.cpp86 // Disable unrolling if the loop is not vectorized. in getMaxInterleaveFactor()
/freebsd/contrib/llvm-project/libcxx/include/__algorithm/
H A Dsimd_utils.h76 …on-simd workloads and don't provide any performance improvement for currently vectorized algorithms
/freebsd/contrib/llvm-project/llvm/lib/Transforms/AggressiveInstCombine/
H A DAggressiveInstCombineInternal.h39 // fit more values in one vectorized instruction. In addition, this optimization
/freebsd/contrib/llvm-project/llvm/lib/Analysis/
H A DLoopAccessAnalysis.cpp1754 // prevent store-load forwarding making vectorized code run a lot slower. in couldPreventStoreLoadForward()
1828 // the distance against the loop iteration count. Since the vectorized code in isSafeDependenceDistance()
2140 // The minimum number of iterations for a vectorized/unrolled version. in isDependent()
2211 // both A and B could be vectorized by 2 iterations. in isDependent()
2455 // If the function has an explicit vectorized counterpart, we can safely in analyzeLoop()
2456 // assume that it can be vectorized. in analyzeLoop()
2464 << "instruction cannot be vectorized"; in analyzeLoop()
2488 << "instruction cannot be vectorized"; in analyzeLoop()
/freebsd/contrib/llvm-project/clang/lib/CodeGen/
H A DCGLoopInfo.cpp238 // Apply all loop properties to the vectorized loop. in createLoopVectorizeMetadata()
242 // Don't vectorize an already vectorized loop. in createLoopVectorizeMetadata()

123