summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen/SelectionDAG
Commit message (Collapse)AuthorAgeFilesLines
* Fix PR15632: No support for ppcf128 floating-point remainder on PowerPC.Bill Schmidt2013-04-032-0/+12
| | | | | | | | For this we need to use a libcall. Previously LLVM didn't implement libcall support for frem, so I've added it in the usual straightforward manner. A test case from the bug report is included. llvm-svn: 178639
* DAGCombiner: Merge store/loads when we have extload/truncstoresArnold Schwaighofer2013-04-021-0/+19
| | | | | | | | | | | | | | | | This is helps on architectures where i8,i16 are not legal but we have byte, and short loads/stores. Allowing us to merge copies like the one below on ARM. copy(char *a, char *b, int n) { do { int t0 = a[0]; int t1 = a[1]; b[0] = t0; b[1] = t1; radar://13536387 llvm-svn: 178546
* Merge load/store sequences with adresses: base + index + offsetArnold Schwaighofer2013-04-011-25/+88
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | We would also like to merge sequences that involve a variable index like in the example below. int index = *idx++ int i0 = c[index+0]; int i1 = c[index+1]; b[0] = i0; b[1] = i1; By extending the parsing of the base pointer to handle dags that contain a base, index, and offset we can handle examples like the one above. The dag for the code above will look something like: (load (i64 add (i64 copyfromreg %c) (i64 signextend (i8 load %index)))) (load (i64 add (i64 copyfromreg %c) (i64 signextend (i32 add (i32 signextend (i8 load %index)) (i32 1))))) The code that parses the tree ignores the intermediate sign extensions. However, if there is a sign extension it needs to be on all indexes. (load (i64 add (i64 copyfromreg %c) (i64 signextend (add (i8 load %index) (i8 1)))) vs (load (i64 add (i64 copyfromreg %c) (i64 signextend (i32 add (i32 signextend (i8 load %index)) (i32 1))))) radar://13536387 llvm-svn: 178483
* DAGCombine: visitXOR can replace a node without returning it, bail out in ↵Benjamin Kramer2013-03-301-3/+3
| | | | | | | | that case. Fixes the crash reported in PR15608. llvm-svn: 178429
* [fast-isel] Add a preemptive fix for the case where we fail to materialize anChad Rosier2013-03-281-0/+2
| | | | | | | | | | | immediate in a register. I don't believe this should ever fail, but I see no harm in trying to make this code bullet proof. I've added an assert to ensure my assumtion is correct. If the assertion fires something is wrong and we should fix it, rather then just silently fall back to SelectionDAG isel. llvm-svn: 178305
* Enhance folding of (extract_subvec (insert_subvec V1, V2, IIdx), EIdx)Michael Liao2013-03-251-21/+29
| | | | | | | | - Handle the case where the result of 'insert_subvect' is bitcasted before 'extract_subvec'. This removes the redundant insertf128/extractf128 pair on unaligned 256-bit vector load/store on vectors of non 64-bit integer. llvm-svn: 177945
* Disable some unsafe-fp-math DAG-combine transformation after legalization.Shuxin Yang2013-03-251-5/+17
| | | | | | | | | | | | | | For instance, following transformation will be disabled: x + x + x => 3.0f * x; The problem of these transformations is that it introduces a FP constant, which following Instruction-Selection pass cannot handle. Reviewed by Nadav, thanks a lot! rdar://13445387 llvm-svn: 177933
* Remove the type legality check from the SelectionDAGBuilder when it lowers ↵Owen Anderson2013-03-231-1/+0
| | | | | | | | | | @llvm.fmuladd to ISD::FMA nodes. Performing this check unilaterally prevented us from generating FMAs when the incoming IR contained illegal vector types which would eventually be legalized to underlying types that *did* support FMA. For example, an @llvm.fmuladd on an OpenCL float16 should become a sequence of float4 FMAs, not float4 fmul+fadd's. NOTE: Because we still call the target-specific profitability hook, individual targets can reinstate the old behavior, if desired, by simply performing the legality check inside their callback hook. They can also perform more sophisticated legality checks, if, for example, some illegal vector types can be productively implemented as FMAs, but not others. llvm-svn: 177820
* Make variable name more explicit and eliminate redundant lookup in ↵Justin Holewinski2013-03-201-4/+4
| | | | | | SDNodeOrdering llvm-svn: 177600
* When computing the demanded bits of Load SDNodes, make sure that we are ↵Nadav Rotem2013-03-201-12/+16
| | | | | | | | looking at the loaded-value operand and not the ptr result (in case of pre-inc loads). rdar://13348420 llvm-svn: 177596
* Revert "pre-RA-sched: fix TargetOpcode usage"Christian Konig2013-03-201-18/+12
| | | | | | | | | | | This reverts commit 06091513c283c863296f01cc7c2e86b56bb50d02. The code is obviously wrong, but the trivial fix causes inefficient code generation on X86. Somebody with more knowledge of the code needs to take a look here. Signed-off-by: Christian König <christian.koenig@amd.com> llvm-svn: 177529
* Move SDNode order propagation to SDNodeOrdering, which also fixes a missedJustin Holewinski2013-03-204-23/+15
| | | | | | | | case of order propagation during isel. Thanks Owen for the suggestion! llvm-svn: 177525
* pre-RA-sched: fix TargetOpcode usageChristian Konig2013-03-201-12/+18
| | | | | | | TargetOpcodes need to be treaded as Machine- and not ISD-Opcodes. Signed-off-by: Christian König <christian.koenig@amd.com> llvm-svn: 177518
* Propagate DAG node ordering during type legalization and instruction selectionJustin Holewinski2013-03-203-2/+43
| | | | | | | | A node's ordering is only propagated during legalization if (a) the new node does not have an ordering (is not a CSE'd node), or (b) the new node has an ordering that is higher than the node being legalized. llvm-svn: 177465
* Reset some of the target options which affect code generation.Bill Wendling2013-03-131-0/+1
| | | | | | | | This doesn't reset all of the target options within the TargetOptions object. This is because some of those are ABI-specific and must be determined if it's okay to change those on the fly. llvm-svn: 176986
* Avoid generating ISD::SELECT for vector operands to SIGN_EXTENDRichard Relph2013-03-121-2/+2
| | | | llvm-svn: 176881
* Fix a crasher newly introduced in r176659/r176649, where fast-isel tries toNick Lewycky2013-03-111-0/+2
| | | | | | lower an expect intrinsic that is a constant expression. llvm-svn: 176830
* Disable statistics on Release builds and move tests that depend on -stats.Jan Wen Voung2013-03-082-18/+14
| | | | | | | | | | | | | | | | | Summary: Statistics are still available in Release+Asserts (any +Asserts builds), and stats can also be turned on with LLVM_ENABLE_STATS. Move some of the FastISel stats that were moved under DEBUG() back out of DEBUG(), since stats are disabled across the board now. Many tests depend on grepping "-stats" output. Move those into a orig_dir/Stats/. so that they can be marked as unsupported when building without statistics. Differential Revision: http://llvm-reviews.chandlerc.com/D486 llvm-svn: 176733
* Remove default from fully covered switch.Benjamin Kramer2013-03-081-2/+1
| | | | llvm-svn: 176703
* LegalizeDAG: Respect the result of TLI.getBooleanContents() when expanding SETCCTom Stellard2013-03-081-1/+13
| | | | llvm-svn: 176695
* DAGCombiner: Use correct value type for checking legality of BR_CC v3Tom Stellard2013-03-081-1/+2
| | | | | | | | | | | | | LegalizeDAG.cpp uses the value of the comparison operands when checking the legality of BR_CC, so DAGCombiner should do the same. v2: - Expand more BR_CC value types for NVPTX v3: - Expand correct BR_CC value types for Hexagon, Mips, and XCore. llvm-svn: 176694
* Revert r176154 in favor of a better approach.Bill Wendling2013-03-081-7/+0
| | | | | | | | | | | | | | | | | | | Code generation makes some basic assumptions about the IR it's been given. In particular, if there is only one 'invoke' in the function, then that invoke won't be going away. However, with the advent of the `llvm.donothing' intrinsic, those invokes may go away. If all of them go away, the landing pad no longer has any users. This confuses the back-end, which asserts. This happens with SjLj exceptions, because that's the model that modifies the IR based on there being invokes, etc. in the function. Remove any invokes of `llvm.donothing' during SjLj EH preparation. This will give us a CFG that the back-end won't be confused about. If all of the invokes in a function are removed, then the SjLj EH prepare pass won't insert the bogus code the relies upon the invokes being there. <rdar://problem/13228754&13316637> llvm-svn: 176677
* [fast-isel] Seriously, add support for the expect intrinsic.Chad Rosier2013-03-071-3/+3
| | | | | | rdar://13370942 llvm-svn: 176659
* [fast-isel] Add support for the expect intrinsic.Chad Rosier2013-03-071-0/+5
| | | | | | rdar://13370942 llvm-svn: 176649
* ArrayRefize some code. No functionality change.Benjamin Kramer2013-03-072-5/+3
| | | | llvm-svn: 176648
* pre-RA-sched debug-only fixAndrew Trick2013-03-071-2/+4
| | | | llvm-svn: 176638
* pre-RA-sched assertion fix. This bug was exposed by r176037.Andrew Trick2013-03-071-3/+0
| | | | | | | | | | | | | rdar:13370002 [pre-RA-sched] assertion: released too many times I tracked this down to an earlier hack that is no longer applicable and interfered with normal scheduler logic. With the changes in r176037, it was causing an instruction to be scheduled multiple times. I have an external test case that I tried hard to reduce and failed. I can't even reproduce with llc. llvm-svn: 176636
* No need to go through int64 and APInt when generating a new constant.Nadav Rotem2013-03-071-3/+3
| | | | llvm-svn: 176615
* SDAG: Handle scalarizing an extend of a <1 x iN> vector.Jim Grosbach2013-03-072-0/+21
| | | | | | | | | Just scalarize the element and rebuild a vector of the result type from that. rdar://13281568 llvm-svn: 176614
* Remove duplicate line and move another closer to its actual useEli Bendersky2013-03-011-3/+1
| | | | llvm-svn: 176391
* Set properties for f128 type.Akira Hatanaka2013-03-011-17/+62
| | | | llvm-svn: 176378
* Generate an error message instead of asserting or segfaulting when we can'tChad Rosier2013-03-011-0/+1
| | | | | | | handle indirect register inputs. rdar://13322011 llvm-svn: 176367
* Fix PR10475Michael Liao2013-03-013-12/+17
| | | | | | | | | | | | | | - ISD::SHL/SRL/SRA must have either both scalar or both vector operands but TLI.getShiftAmountTy() so far only return scalar type. As a result, backend logic assuming that breaks. - Rename the original TLI.getShiftAmountTy() to TLI.getScalarShiftAmountTy() and re-define TLI.getShiftAmountTy() to return target-specificed scalar type or the same vector type as the 1st operand. - Fix most TICG logic assuming TLI.getShiftAmountTy() a simple scalar type. llvm-svn: 176364
* A small refactoring + adding comments.Eli Bendersky2013-02-282-10/+5
| | | | | | | | | | | SelectionDAGIsel::LowerArguments needs a function, not a basic block. So it makes sense to pass it the function instead of extracting a basic-block from the function and then tossing it. This is also more self-documenting (functions have arguments, BBs don't). In addition, added comments to a couple of Select* methods. llvm-svn: 176305
* Put some per-instruction statistics of fast isel under NDEBUG, together withEli Bendersky2013-02-281-3/+5
| | | | | | other per-instruction statistics. llvm-svn: 176273
* Remove unnecessary cast to void.Eric Christopher2013-02-271-1/+0
| | | | llvm-svn: 176222
* Silence the unused variable warning. Nadav Rotem2013-02-271-5/+7
| | | | llvm-svn: 176218
* The FastISEL should be fast. But when we record statistics we use atomic ↵Nadav Rotem2013-02-271-9/+10
| | | | | | | | operations to increment the counters. This patch disables the counters on non-debug builds. This reduces the runtime of SelectionDAGISel::SelectCodeCommon by ~5%. llvm-svn: 176214
* Reverted: r176136 - Have a way for a target to opt-out of target-independent ↵Michael Ilseman2013-02-271-1/+1
| | | | | | fast isel llvm-svn: 176204
* SelectionDAG: If llvm.donothing has a landingpad, we should clearManman Ren2013-02-271-0/+7
| | | | | | | | | CurrentCallSite to avoid an assertion failure: assert(MMI.getCurrentCallSite() == 0 && "Overlapping call sites!"); rdar://problem/13228754 llvm-svn: 176154
* Have a way for a target to opt-out of target-independent fast iselMichael Ilseman2013-02-261-1/+1
| | | | llvm-svn: 176136
* Fix wording.Chad Rosier2013-02-251-3/+4
| | | | llvm-svn: 176055
* [fast-isel] Add X86FastIsel::FastLowerArguments to handle functions with 6 orChad Rosier2013-02-251-0/+10
| | | | | | | | | | fewer scalar integer (i32 or i64) arguments. It completely eliminates the need for SDISel for trivial functions. Also, add the new llc -fast-isel-abort-args option, which is similar to -fast-isel-abort option, but for formal argument lowering. llvm-svn: 176052
* pre-RA-sched fix: only reevaluate physreg interferences when necessary.Andrew Trick2013-02-251-33/+61
| | | | | | Fixes rdar:13279013: scheduler was blowing up on select instructions. llvm-svn: 176037
* 'Hexadecimal' has two 'a's and only one 'i'.Matt Beaumont-Gay2013-02-251-2/+2
| | | | llvm-svn: 176031
* Fix spelling noticed by Duncan.Chandler Carruth2013-02-251-3/+3
| | | | llvm-svn: 176023
* Fix the root cause of PR15348 by correctly handling alignment 0 onChandler Carruth2013-02-252-0/+9
| | | | | | | | | | | | | | | | | | | | | | | | memory intrinsics in the SDAG builder. When alignment is zero, the lang ref says that *no* alignment assumptions can be made. This is the exact opposite of the internal API contracts of the DAG where alignment 0 indicates that the alignment can be made to be anything desired. There is another, more explicit alignment that is better suited for the role of "no alignment at all": an alignment of 1. Map the intrinsic alignment to this early so that we don't end up generating aligned DAGs. It is really terrifying that we've never seen this before, but we suddenly started generating a large number of alignment 0 memcpys due to the new code to do memcpy-based copying of POD class members. That patch contains a bug that rounds bitfield alignments down when they are the first field. This can in turn produce zero alignments. This fixes weird crashes I've seen in library users of LLVM on 32-bit hosts, etc. llvm-svn: 176022
* SelectionDAG compile time improvement.Nadav Rotem2013-02-221-0/+19
| | | | | | | | One of the phases of SelectionDAG is LegalizeVectors. We don't need to sort the DAG and copy nodes around if there are no vector ops. Speeds up the compilation time of SelectionDAG on a big scalar workload by ~8%. llvm-svn: 175929
* Fix isa<> check which could never be true.Pete Cooper2013-02-221-1/+1
| | | | | | | | | | | | It was incorrectly checking a Function* being an IntrinsicInst* which isn't possible. It should always have been checking the CallInst* instead. Added test case for x86 which ensures we only get one constant load. It was 2 before this change. rdar://problem/13267920 llvm-svn: 175853
* DAGCombiner: Make the post-legalize vector op optimization more aggressive.Benjamin Kramer2013-02-211-10/+0
| | | | | | | | A legal BUILD_VECTOR goes in and gets constant folded into another legal BUILD_VECTOR so we don't lose any legality here. The problematic PPC optimization that made this check necessary was fixed recently. llvm-svn: 175759
OpenPOWER on IntegriCloud