[RISCV] Add fixed-length patterns for disjoint or patterns for vwadd[u].v{v,x} #136824
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…u].v{v,x}
This is the fixed-length equivalent of llvm#136716.
The pattern we need to match is ({s,z}ext_vl (or_vl disjoint a, b)). This only allows or_vls with an undef passthru, which allows us to ignore its mask and vl and just take it from the {s,z}ext_vl.
A riscv_or_vl_is_add_oneuse PatFrag is added to mirror or_is_add in RISCVInstrInfo.td.
Member
|
@llvm/pr-subscribers-backend-risc-v Author: Luke Lau (lukel97) ChangesThis is the fixed-length equivalent of #136716. The pattern we need to match is ({s,z}ext_vl (or_vl disjoint a, b)). This only allows or_vls with an undef passthru, which allows us to ignore its mask and vl and just take it from the {s,z}ext_vl. A riscv_or_vl_is_add_oneuse PatFrag is added to mirror or_is_add in RISCVInstrInfo.td. Full diff: https://github.com/llvm/llvm-project/pull/136824.diff 2 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index f80cbc9e2fb5e..068402c59d6e2 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -497,6 +497,16 @@ let HasOneUse = 1 in {
node:$E),
(riscv_add_vl node:$A, node:$B, node:$C,
node:$D, node:$E)>;
+ def riscv_or_vl_is_add_oneuse : PatFrag<(ops node:$A, node:$B, node:$C, node:$D,
+ node:$E),
+ (riscv_or_vl node:$A, node:$B, node:$C,
+ node:$D, node:$E), [{
+ if (N->getFlags().hasDisjoint())
+ return true;
+ KnownBits Known0 = CurDAG->computeKnownBits(N->getOperand(0), 0);
+ KnownBits Known1 = CurDAG->computeKnownBits(N->getOperand(1), 0);
+ return KnownBits::haveNoCommonBitsSet(Known0, Known1);
+ }]>;
def riscv_sub_vl_oneuse : PatFrag<(ops node:$A, node:$B, node:$C, node:$D,
node:$E),
(riscv_sub_vl node:$A, node:$B, node:$C,
@@ -2016,6 +2026,37 @@ foreach vtiToWti = AllWidenableIntVectors in {
}
}
+// DAGCombiner::hoistLogicOpWithSameOpcodeHands may hoist disjoint ors
+// to (ext (or disjoint (a, b)))
+multiclass VPatWidenOrDisjointVL_VV_VX<SDNode extop, string instruction_name> {
+ foreach vtiToWti = AllWidenableIntVectors in {
+ defvar vti = vtiToWti.Vti;
+ defvar wti = vtiToWti.Wti;
+ let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
+ GetVTypePredicates<wti>.Predicates) in {
+ def : Pat<(wti.Vector (extop (vti.Vector
+ (riscv_or_vl_is_add_oneuse
+ vti.RegClass:$rs2, vti.RegClass:$rs1,
+ undef, srcvalue, srcvalue)),
+ VMV0:$vm, VLOpFrag)),
+ (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX#"_MASK")
+ (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2,
+ vti.RegClass:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>;
+ def : Pat<(wti.Vector (extop (vti.Vector
+ (riscv_or_vl_is_add_oneuse
+ vti.RegClass:$rs2, (SplatPat (XLenVT GPR:$rs1)),
+ undef, srcvalue, srcvalue)),
+ VMV0:$vm, VLOpFrag)),
+ (!cast<Instruction>(instruction_name#"_VX_"#vti.LMul.MX#"_MASK")
+ (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2,
+ GPR:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>;
+ }
+ }
+}
+
+defm : VPatWidenOrDisjointVL_VV_VX<riscv_sext_vl, "PseudoVWADD">;
+defm : VPatWidenOrDisjointVL_VV_VX<riscv_zext_vl, "PseudoVWADDU">;
+
// 11.3. Vector Integer Extension
defm : VPatExtendVL_V<riscv_zext_vl, "PseudoVZEXT", "VF2",
AllFractionableVF2IntVectors>;
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vwadd.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vwadd.ll
index 5e7d1b91d7892..7f8c8258803fc 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vwadd.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vwadd.ll
@@ -901,9 +901,8 @@ define <4 x i32> @vwaddu_vv_disjoint_or(<4 x i16> %x.i16, <4 x i16> %y.i16) {
; CHECK-LABEL: vwaddu_vv_disjoint_or:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
-; CHECK-NEXT: vor.vv v9, v8, v9
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
-; CHECK-NEXT: vzext.vf2 v8, v9
+; CHECK-NEXT: vwaddu.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
; CHECK-NEXT: ret
%x.i32 = zext <4 x i16> %x.i16 to <4 x i32>
%y.i32 = zext <4 x i16> %y.i16 to <4 x i32>
@@ -915,9 +914,8 @@ define <4 x i32> @vwadd_vv_disjoint_or(<4 x i16> %x.i16, <4 x i16> %y.i16) {
; CHECK-LABEL: vwadd_vv_disjoint_or:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
-; CHECK-NEXT: vor.vv v9, v8, v9
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
-; CHECK-NEXT: vsext.vf2 v8, v9
+; CHECK-NEXT: vwadd.vv v10, v8, v9
+; CHECK-NEXT: vmv1r.v v8, v10
; CHECK-NEXT: ret
%x.i32 = sext <4 x i16> %x.i16 to <4 x i32>
%y.i32 = sext <4 x i16> %y.i16 to <4 x i32>
@@ -929,9 +927,8 @@ define <4 x i32> @vwaddu_vx_disjoint_or(<4 x i16> %x.i16, i16 %y.i16) {
; CHECK-LABEL: vwaddu_vx_disjoint_or:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
-; CHECK-NEXT: vor.vx v9, v8, a0
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
-; CHECK-NEXT: vzext.vf2 v8, v9
+; CHECK-NEXT: vwaddu.vx v9, v8, a0
+; CHECK-NEXT: vmv1r.v v8, v9
; CHECK-NEXT: ret
%x.i32 = zext <4 x i16> %x.i16 to <4 x i32>
%y.head = insertelement <4 x i16> poison, i16 %y.i16, i32 0
@@ -945,9 +942,8 @@ define <4 x i32> @vwadd_vx_disjoint_or(<4 x i16> %x.i16, i16 %y.i16) {
; CHECK-LABEL: vwadd_vx_disjoint_or:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
-; CHECK-NEXT: vor.vx v9, v8, a0
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
-; CHECK-NEXT: vsext.vf2 v8, v9
+; CHECK-NEXT: vwadd.vx v9, v8, a0
+; CHECK-NEXT: vmv1r.v v8, v9
; CHECK-NEXT: ret
%x.i32 = sext <4 x i16> %x.i16 to <4 x i32>
%y.head = insertelement <4 x i16> poison, i16 %y.i16, i32 0
|
wangpc-pp
approved these changes
Apr 23, 2025
preames
approved these changes
Apr 23, 2025
tclin914
approved these changes
Apr 24, 2025
Comment on lines
2037
to
2041
| def : Pat<(wti.Vector (extop (vti.Vector | ||
| (riscv_or_vl_is_add_oneuse | ||
| vti.RegClass:$rs2, vti.RegClass:$rs1, | ||
| undef, srcvalue, srcvalue)), | ||
| VMV0:$vm, VLOpFrag)), |
Contributor
There was a problem hiding this comment.
Suggested change
| def : Pat<(wti.Vector (extop (vti.Vector | |
| (riscv_or_vl_is_add_oneuse | |
| vti.RegClass:$rs2, vti.RegClass:$rs1, | |
| undef, srcvalue, srcvalue)), | |
| VMV0:$vm, VLOpFrag)), | |
| def : Pat<(wti.Vector (extop (vti.Vector | |
| (riscv_or_vl_is_add_oneuse | |
| vti.RegClass:$rs2, vti.RegClass:$rs1, | |
| undef, srcvalue, srcvalue)), | |
| VMV0:$vm, VLOpFrag)), |
Comment on lines
2042
to
2044
| (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX#"_MASK") | ||
| (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2, | ||
| vti.RegClass:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>; |
Contributor
There was a problem hiding this comment.
Suggested change
| (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX#"_MASK") | |
| (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2, | |
| vti.RegClass:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>; | |
| (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX#"_MASK") | |
| (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2, | |
| vti.RegClass:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>; |
Comment on lines
2046
to
2048
| (riscv_or_vl_is_add_oneuse | ||
| vti.RegClass:$rs2, (SplatPat (XLenVT GPR:$rs1)), | ||
| undef, srcvalue, srcvalue)), |
Contributor
There was a problem hiding this comment.
Suggested change
| (riscv_or_vl_is_add_oneuse | |
| vti.RegClass:$rs2, (SplatPat (XLenVT GPR:$rs1)), | |
| undef, srcvalue, srcvalue)), | |
| (riscv_or_vl_is_add_oneuse | |
| vti.RegClass:$rs2, (SplatPat (XLenVT GPR:$rs1)), | |
| undef, srcvalue, srcvalue)), |
Comment on lines
2050
to
2052
| (!cast<Instruction>(instruction_name#"_VX_"#vti.LMul.MX#"_MASK") | ||
| (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2, | ||
| GPR:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>; |
Contributor
There was a problem hiding this comment.
Suggested change
| (!cast<Instruction>(instruction_name#"_VX_"#vti.LMul.MX#"_MASK") | |
| (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2, | |
| GPR:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>; | |
| (!cast<Instruction>(instruction_name#"_VX_"#vti.LMul.MX#"_MASK") | |
| (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2, | |
| GPR:$rs1, VMV0:$vm, GPR:$vl, vti.Log2SEW, TA_MA)>; |
IanWood1
pushed a commit
to IanWood1/llvm-project
that referenced
this pull request
May 6, 2025
…u].v{v,x} (llvm#136824)
This is the fixed-length equivalent of llvm#136716.
The pattern we need to match is ({s,z}ext_vl (or_vl disjoint a, b)).
This only allows or_vls with an undef passthru, which allows us to
ignore its mask and vl and just take it from the {s,z}ext_vl.
A riscv_or_vl_is_add_oneuse PatFrag is added to mirror or_is_add in
RISCVInstrInfo.td.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is the fixed-length equivalent of #136716.
The pattern we need to match is ({s,z}ext_vl (or_vl disjoint a, b)). This only allows or_vls with an undef passthru, which allows us to ignore its mask and vl and just take it from the {s,z}ext_vl.
A riscv_or_vl_is_add_oneuse PatFrag is added to mirror or_is_add in RISCVInstrInfo.td.