Commit Graph

46 Commits

Author SHA1 Message Date
Krzysztof Parzyszek e8a0ae7346 [Hexagon] Mark HVX vector predicate bitwise ops as legal, add patterns
llvm-svn: 332525
2018-05-16 21:00:24 +00:00
Krzysztof Parzyszek cff73a2118 [Hexagon] Add patterns for vector shift-and-accumulate
llvm-svn: 331918
2018-05-09 21:10:41 +00:00
Krzysztof Parzyszek 41a24b7b13 [Hexagon] Improve HVX instruction selection (bitcast, vsplat)
There was some unfortunate interaction between VSPLAT and BITCAST
related to the selection of constant vectors (coming from selecting
shuffles). Introduce VSPLATW that always splats a 32-bit word, and
can have arbitrary result type (to avoid BITCASTs of VSPLAT).
Clean up the previous selection of BITCAST/VSPLAT.

llvm-svn: 330471
2018-04-20 19:38:37 +00:00
Krzysztof Parzyszek 2a9a83cd3f [Hexagon] Use legal types when lowering CONCAT_VECTORS via BUILD_VECTOR
llvm-svn: 330344
2018-04-19 17:11:58 +00:00
Krzysztof Parzyszek d92c37e090 [Hexagon] Generate code for vector bswap intrinsics
llvm-svn: 330333
2018-04-19 14:46:44 +00:00
Krzysztof Parzyszek 0375cd46ef [Hexagon] Implement TTI::shouldMaximizeVectorBandwidth
llvm-svn: 328648
2018-03-27 18:10:47 +00:00
Krzysztof Parzyszek 65059ee284 [Hexagon] Add heuristic to exclude critical path cost for scheduling
Patch by Brendon Cahoon.

llvm-svn: 328022
2018-03-20 19:26:27 +00:00
Krzysztof Parzyszek dca383123f [Hexagon] Improve scheduling based on register pressure
Patch by Brendon Cahoon.

llvm-svn: 327975
2018-03-20 12:28:43 +00:00
Krzysztof Parzyszek 480ab2bbc4 [Hexagon] Ignore indexed loads when handling unaligned loads
llvm-svn: 327037
2018-03-08 18:15:13 +00:00
Krzysztof Parzyszek 2c3edf0567 [Hexagon] Rewrite non-HVX unaligned loads as pairs of aligned ones
This is a follow-up to r325169, this time for all types, not just HVX
vector types.

Disable this by default, since it's not always safe. 

llvm-svn: 326915
2018-03-07 17:27:18 +00:00
Krzysztof Parzyszek e3e963236a [Hexagon] Generate valignb for shifting shuffles (instead of vdelta)
llvm-svn: 326627
2018-03-02 22:22:19 +00:00
Jonas Paulsson 77cdf3881c [Hexagon] Return true in enableMultipleCopyHints().
Enable multiple COPY hints to eliminate more COPYs during register allocation.

Note that this is something all targets should do, see
https://reviews.llvm.org/D38128.

Review: Krzysztof Parzyszek
llvm-svn: 325697
2018-02-21 16:37:45 +00:00
Krzysztof Parzyszek ad83ce4cb4 [Hexagon] Split HVX vector pair loads/stores, expand unaligned loads
llvm-svn: 325169
2018-02-14 20:46:06 +00:00
Krzysztof Parzyszek 9b48e8d233 [Hexagon] Add code to select QTRUE and QFALSE
Fixes http://llvm.org/PR36320.

llvm-svn: 324763
2018-02-09 19:10:46 +00:00
Krzysztof Parzyszek 97a5095db6 [Hexagon] Lower concat of more than 2 vectors into build_vector
llvm-svn: 324391
2018-02-06 20:18:58 +00:00
Krzysztof Parzyszek 88f11003a0 [Hexagon] Split HVX operations on vector pairs
Vector pairs are legal types, but not every operation can work on pairs.
For those operations that are legal for single vectors, generate a concat
of their results on pair halves.

llvm-svn: 324350
2018-02-06 14:24:57 +00:00
Krzysztof Parzyszek 69f1d7e370 [Hexagon] Handle lowering of SETCC via setCondCodeAction
It was expanded directly into instructions earlier. That was to avoid
loads from a constant pool for a vector negation: "xor x, splat(i1 -1)".
Implement ISD opcodes QTRUE and QFALSE to denote logical vectors of
all true and all false values, and handle setcc with negations through
selection patterns.

llvm-svn: 324348
2018-02-06 14:16:52 +00:00
Krzysztof Parzyszek 02947b7112 [Hexagon] Use V6_vmpyih for halfword multiplication
Unlike V6_vmpyhv, it produces the result in the exact form that is
expected without the need for a shuffle.

llvm-svn: 324241
2018-02-05 15:40:06 +00:00
Krzysztof Parzyszek 1108ee2496 [Hexagon] Implement HVX codegen for vector shifts
llvm-svn: 323914
2018-01-31 20:49:24 +00:00
Krzysztof Parzyszek 9eb085e6cf [Hexagon] Handle ANY_EXTEND_VECTOR_INREG in lowering
llvm-svn: 323912
2018-01-31 20:48:11 +00:00
Krzysztof Parzyszek b843f75179 [Hexagon] Handle SETCC on vector pairs in lowering
llvm-svn: 323911
2018-01-31 20:46:55 +00:00
Krzysztof Parzyszek 82a83391d3 [Hexagon] Handle BUILD_VECTOR from undef values in buildHvxVectorReg
llvm-svn: 323889
2018-01-31 16:52:15 +00:00
Krzysztof Parzyszek 8cc636c592 [Hexagon] Only process bitcasts of vsplats when selecting const vectors
Selecting of constant HVX vectors involves some "manual processing",
which mishandled an unrelated BITCAST operation causing a selection
error.

llvm-svn: 323887
2018-01-31 16:48:20 +00:00
Krzysztof Parzyszek 90ca4e8b0c [Hexagon] Generate constant splats instead of loads from constant pool
llvm-svn: 323568
2018-01-26 21:54:56 +00:00
Krzysztof Parzyszek cf3ad5841b [Hexagon] Run late copy propagation and dead code elimination passes
llvm-svn: 323346
2018-01-24 17:48:11 +00:00
Krzysztof Parzyszek d5e8a260bb [Hexagon] Add patterns for sext_inreg of HVX vector types
llvm-svn: 323250
2018-01-23 19:56:16 +00:00
Krzysztof Parzyszek 3780a0e1fa [Hexagon] Implement basic vector operations on vectors vNi1
In addition to that, make sure that there are no boolean vector types that
are associated with multiple register classes. Specifically, remove v32i1
and v64i1 from integer register classes. These types will correspond to
results of vector comparisons, and as such should belong to the vector
predicate class. Having them in scalar registers as well makes legalization
ambiguous.

llvm-svn: 323229
2018-01-23 17:53:59 +00:00
Krzysztof Parzyszek 4ef6cfff6a [Hexagon] Cast elements to correct type when creating constant vector
llvm-svn: 322301
2018-01-11 18:03:23 +00:00
Krzysztof Parzyszek b0b52618c0 [Hexagon] Even simpler patterns for sign- and zero-extending HVX vectors
Recommit r321897 with updated testcases.

llvm-svn: 321908
2018-01-05 22:31:11 +00:00
Krzysztof Parzyszek 4ed8ef6f8e Revert r321894: it requires a part of another commit that is not ready yet
Commit message:
[Hexagon] Add patterns for sext_inreg of HVX vector types

llvm-svn: 321904
2018-01-05 21:57:43 +00:00
Krzysztof Parzyszek f9d01a12d1 [Hexagon] Add patterns for truncating HVX vector types
Only non-bool vectors.

llvm-svn: 321895
2018-01-05 20:48:03 +00:00
Krzysztof Parzyszek 9d0c6355a0 [Hexagon] Add patterns for sext_inreg of HVX vector types
llvm-svn: 321894
2018-01-05 20:46:41 +00:00
Krzysztof Parzyszek cfe4a3616f [Hexagon] Fix generation of vector sign extensions
llvm-svn: 321650
2018-01-02 15:28:49 +00:00
Krzysztof Parzyszek e4ce92cabf [Hexagon] Allow construction of HVX vector predicates
Handle BUILD_VECTOR of boolean values.

llvm-svn: 321220
2017-12-20 20:49:43 +00:00
Krzysztof Parzyszek e704583f23 [Hexagon] Cache loads to select to avoid traversing mutating DAG
llvm-svn: 321034
2017-12-18 23:13:27 +00:00
Krzysztof Parzyszek 6b589e593d [Hexagon] Generate HVX code for vector sign-, zero- and any-extends
Implement any-extend as zero-extend.

llvm-svn: 321004
2017-12-18 18:32:27 +00:00
Krzysztof Parzyszek 266d6f03a1 [Hexagon] Handle concat_vectors of all allowed HVX types
llvm-svn: 320865
2017-12-15 21:23:12 +00:00
Krzysztof Parzyszek 470760533a [Hexagon] Generate HVX code for comparisons and selects
llvm-svn: 320744
2017-12-14 21:28:48 +00:00
Krzysztof Parzyszek 2eda05db87 [Hexagon] Relax some checks in testcases, NFC
llvm-svn: 320529
2017-12-12 21:44:04 +00:00
Krzysztof Parzyszek edcd9dcbc4 [Hexagon] Better detection of identity and undef masks in shuffles
llvm-svn: 320523
2017-12-12 20:23:12 +00:00
Krzysztof Parzyszek 40a605f1be [Hexagon] Fix wrong order of operands for vmux
Shuffle generation uses vmux to collapse vectors resulting from two
individual shuffles into one. The indexes of the elements selected
from the first operand were indicated by 0xFF in the constant vector
used in the compare instruction, but the compare (veqb) set the bits
corresponding to the 0x00 elements, thus inverting the selection.

Reverse the order of operands to vmux to get the correct output.

llvm-svn: 320516
2017-12-12 19:32:41 +00:00
Krzysztof Parzyszek 152414595b [Hexagon] Crash in instruction selection for insert_vector_elt for HVX
A wrong type was passed to insertVector, causing an out-of-bounds value
to be added an an operand to HexagonISD::INSERT. This later failed in
instruction selection.

llvm-svn: 320369
2017-12-11 14:46:06 +00:00
Krzysztof Parzyszek 039d4d9286 [Hexagon] Generate HVX code for basic arithmetic operations
Handle and, or, xor, add, sub, mul for vectors of i8, i16, and i32.

llvm-svn: 320063
2017-12-07 17:37:28 +00:00
Krzysztof Parzyszek d2967868be [Hexagon] Recognize vdealb, vdealh, vshuffb and vshuffh specifically
llvm-svn: 319978
2017-12-06 22:41:49 +00:00
Krzysztof Parzyszek 64533cf630 [Hexagon] Handle perfect shuffles on single vectors
llvm-svn: 319965
2017-12-06 21:25:03 +00:00
Krzysztof Parzyszek 7d37dd8902 [Hexagon] Generate HVX code for vector construction and access
Support for:
  - build vector,
  - extract vector element, subvector,
  - insert vector element, subvector,
  - shuffle.

llvm-svn: 319901
2017-12-06 16:40:37 +00:00