release/21.x: [RISCV] Fix issues in ORI to QC.INSBI transformation (#148809) #148933
+65
−4
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
|
@topperc What do you think about merging this PR to the release branch? |
|
@llvm/pr-subscribers-backend-risc-v Author: None (llvmbot) ChangesBackport d67d91a Requested by: @lenary Full diff: https://github.com/llvm/llvm-project/pull/148933.diff 4 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
index 666c76b21e631..ad5ecaea85448 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
@@ -689,10 +689,16 @@ bool RISCVDAGToDAGISel::trySignedBitfieldInsertInMask(SDNode *Node) {
if (!isShiftedMask_32(C1) || isInt<12>(C1))
return false;
+ // INSBI will clobber the input register in N0. Bail out if we need a copy to
+ // preserve this value.
+ SDValue N0 = Node->getOperand(0);
+ if (!N0.hasOneUse())
+ return false;
+
// If C1 is a shifted mask (but can't be formed as an ORI),
// use a bitfield insert of -1.
// Transform (or x, C1)
- // -> (qc.insbi x, width, shift)
+ // -> (qc.insbi x, -1, width, shift)
const unsigned Leading = llvm::countl_zero((uint32_t)C1);
const unsigned Trailing = llvm::countr_zero((uint32_t)C1);
const unsigned Width = 32 - Leading - Trailing;
@@ -705,7 +711,7 @@ bool RISCVDAGToDAGISel::trySignedBitfieldInsertInMask(SDNode *Node) {
SDLoc DL(Node);
MVT VT = Node->getSimpleValueType(0);
- SDValue Ops[] = {CurDAG->getSignedTargetConstant(-1, DL, VT),
+ SDValue Ops[] = {N0, CurDAG->getSignedTargetConstant(-1, DL, VT),
CurDAG->getTargetConstant(Width, DL, VT),
CurDAG->getTargetConstant(Trailing, DL, VT)};
SDNode *BitIns = CurDAG->getMachineNode(RISCV::QC_INSBI, DL, VT, Ops);
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoXqci.td b/llvm/lib/Target/RISCV/RISCVInstrInfoXqci.td
index 26bb1e8d17857..c7cb6e237aeac 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoXqci.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoXqci.td
@@ -845,10 +845,11 @@ let Predicates = [HasVendorXqcibi, IsRV32] in {
let Predicates = [HasVendorXqcibm, IsRV32] in {
let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in {
def QC_INSBRI : QCIRVInstRI<0b1, simm11, "qc.insbri">;
- def QC_INSBI : RVInstIBase<0b001, OPC_CUSTOM_0, (outs GPRNoX0:$rd),
- (ins simm5:$imm5, uimm5_plus1:$width,
+ def QC_INSBI : RVInstIBase<0b001, OPC_CUSTOM_0, (outs GPRNoX0:$rd_wb),
+ (ins GPRNoX0:$rd, simm5:$imm5, uimm5_plus1:$width,
uimm5:$shamt), "qc.insbi",
"$rd, $imm5, $width, $shamt"> {
+ let Constraints = "$rd = $rd_wb";
bits<5> imm5;
bits<5> shamt;
bits<5> width;
diff --git a/llvm/test/CodeGen/RISCV/xqcibm-cto-clo-brev.ll b/llvm/test/CodeGen/RISCV/xqcibm-cto-clo-brev.ll
index f227fa9aa423d..2fa06517508ce 100644
--- a/llvm/test/CodeGen/RISCV/xqcibm-cto-clo-brev.ll
+++ b/llvm/test/CodeGen/RISCV/xqcibm-cto-clo-brev.ll
@@ -105,6 +105,7 @@ define i16 @test_cttz_i16(i16 %a) nounwind {
;
; RV32ZBBXQCIBM-LABEL: test_cttz_i16:
; RV32ZBBXQCIBM: # %bb.0:
+; RV32ZBBXQCIBM-NEXT: not a0, a0
; RV32ZBBXQCIBM-NEXT: qc.insbi a0, -1, 1, 16
; RV32ZBBXQCIBM-NEXT: ctz a0, a0
; RV32ZBBXQCIBM-NEXT: ret
diff --git a/llvm/test/CodeGen/RISCV/xqcibm-insert.ll b/llvm/test/CodeGen/RISCV/xqcibm-insert.ll
index 6b7f9ae856625..88054a691bad1 100644
--- a/llvm/test/CodeGen/RISCV/xqcibm-insert.ll
+++ b/llvm/test/CodeGen/RISCV/xqcibm-insert.ll
@@ -47,6 +47,29 @@ define i32 @test_insbi_mask(i32 %a) nounwind {
ret i32 %or
}
+define i32 @test_insbi_mask_mv(i32 %a, i32 %b) nounwind {
+; RV32I-LABEL: test_insbi_mask_mv:
+; RV32I: # %bb.0:
+; RV32I-NEXT: lui a0, 16
+; RV32I-NEXT: addi a0, a0, -1
+; RV32I-NEXT: or a0, a1, a0
+; RV32I-NEXT: ret
+;
+; RV32IXQCIBM-LABEL: test_insbi_mask_mv:
+; RV32IXQCIBM: # %bb.0:
+; RV32IXQCIBM-NEXT: mv a0, a1
+; RV32IXQCIBM-NEXT: qc.insbi a0, -1, 16, 0
+; RV32IXQCIBM-NEXT: ret
+;
+; RV32IXQCIBMZBS-LABEL: test_insbi_mask_mv:
+; RV32IXQCIBMZBS: # %bb.0:
+; RV32IXQCIBMZBS-NEXT: mv a0, a1
+; RV32IXQCIBMZBS-NEXT: qc.insbi a0, -1, 16, 0
+; RV32IXQCIBMZBS-NEXT: ret
+ %or = or i32 %b, 65535
+ ret i32 %or
+}
+
define i32 @test_insbi_shifted_mask(i32 %a) nounwind {
; RV32I-LABEL: test_insbi_shifted_mask:
; RV32I: # %bb.0:
@@ -67,6 +90,36 @@ define i32 @test_insbi_shifted_mask(i32 %a) nounwind {
ret i32 %or
}
+define i32 @test_insbi_shifted_mask_multiple_uses(i32 %a) nounwind {
+; RV32I-LABEL: test_insbi_shifted_mask_multiple_uses:
+; RV32I: # %bb.0:
+; RV32I-NEXT: lui a1, 15
+; RV32I-NEXT: or a1, a0, a1
+; RV32I-NEXT: addi a0, a0, 10
+; RV32I-NEXT: xor a0, a1, a0
+; RV32I-NEXT: ret
+;
+; RV32IXQCIBM-LABEL: test_insbi_shifted_mask_multiple_uses:
+; RV32IXQCIBM: # %bb.0:
+; RV32IXQCIBM-NEXT: lui a1, 15
+; RV32IXQCIBM-NEXT: or a1, a1, a0
+; RV32IXQCIBM-NEXT: addi a0, a0, 10
+; RV32IXQCIBM-NEXT: xor a0, a0, a1
+; RV32IXQCIBM-NEXT: ret
+;
+; RV32IXQCIBMZBS-LABEL: test_insbi_shifted_mask_multiple_uses:
+; RV32IXQCIBMZBS: # %bb.0:
+; RV32IXQCIBMZBS-NEXT: lui a1, 15
+; RV32IXQCIBMZBS-NEXT: or a1, a1, a0
+; RV32IXQCIBMZBS-NEXT: addi a0, a0, 10
+; RV32IXQCIBMZBS-NEXT: xor a0, a0, a1
+; RV32IXQCIBMZBS-NEXT: ret
+ %or = or i32 %a, 61440
+ %add = add i32 %a, 10
+ %xor = xor i32 %or, %add
+ ret i32 %xor
+}
+
define i32 @test_single_bit_set(i32 %a) nounwind {
; RV32I-LABEL: test_single_bit_set:
; RV32I: # %bb.0:
|
github-project-automation
bot
moved this from Needs Review
to Needs Merge
in LLVM Release Status
The transformation done in llvm#147349 was incorrect since we were not passing the input node of the `OR` instruction to the `QC.INSBI` instruction leading to the generated instruction doing the wrong thing. In order to do this we first needed to add the output register to `QC.INSBI` as being both an input and output. The code produced after the above fix will need a copy (mv) to preserve the register input to the OR instruction if it has more than one use making the transformation net neutral ( `6-byte QC.E.ORI/ORAI` vs `2-byte C.MV + 4-byte QC.INSB`I). Avoid doing the transformation if there is more than one use of the input register to the OR instruction. (cherry picked from commit d67d91a)
github-project-automation
bot
moved this from Needs Merge
to Done
in LLVM Release Status
|
@lenary (or anyone else). If you would like to add a note about this fix in the release notes (completely optional). Please reply to this comment with a one or two sentence description of the fix. When you are done, please add the release:note label to this PR. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Backport d67d91a
Requested by: @lenary