[RISCV][NFC] Match 3/5/9 * 3/5/9 * 2^N without a loop
#165547
Merged
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
Member
|
@llvm/pr-subscribers-backend-risc-v Author: Piotr Fusik (pfusik) Changes#158851 matches Full diff: https://github.com/llvm/llvm-project/pull/165547.diff 1 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 1c930acd9c4a0..426ac74f1454d 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -16485,6 +16485,34 @@ static SDValue expandMulToAddOrSubOfShl(SDNode *N, SelectionDAG &DAG,
return DAG.getNode(Op, DL, VT, Shift1, Shift2);
}
+static SDValue getShlAddShlAdd(SDNode *N, SelectionDAG &DAG, int ShX, int ShY) {
+ SDLoc DL(N);
+ EVT VT = N->getValueType(0);
+ SDValue X = N->getOperand(0);
+ SDValue Mul359 = DAG.getNode(RISCVISD::SHL_ADD, DL, VT, X,
+ DAG.getConstant(ShY, DL, VT), X);
+ return DAG.getNode(RISCVISD::SHL_ADD, DL, VT, Mul359,
+ DAG.getConstant(ShX, DL, VT), Mul359);
+}
+
+static SDValue expandMulToShlAddShlAdd(SDNode *N, SelectionDAG &DAG,
+ uint64_t MulAmt) {
+ switch (MulAmt) {
+ case 5 * 3:
+ return getShlAddShlAdd(N, DAG, 2, 1);
+ case 9 * 3:
+ return getShlAddShlAdd(N, DAG, 3, 1);
+ case 5 * 5:
+ return getShlAddShlAdd(N, DAG, 2, 2);
+ case 9 * 5:
+ return getShlAddShlAdd(N, DAG, 3, 2);
+ case 9 * 9:
+ return getShlAddShlAdd(N, DAG, 3, 3);
+ default:
+ return SDValue();
+ }
+}
+
// Try to expand a scalar multiply to a faster sequence.
static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
TargetLowering::DAGCombinerInfo &DCI,
@@ -16518,14 +16546,12 @@ static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
// We're adding additional uses of X here, and in principle, we should be freezing
// X before doing so. However, adding freeze here causes real regressions, and no
// other target properly freezes X in these cases either.
- SDValue X = N->getOperand(0);
-
if (Subtarget.hasShlAdd(3)) {
+ SDValue X = N->getOperand(0);
int Shift;
if (int ShXAmount = isShifted359(MulAmt, Shift)) {
// 3/5/9 * 2^N -> shl (shXadd X, X), N
SDLoc DL(N);
- SDValue X = N->getOperand(0);
// Put the shift first if we can fold a zext into the shift forming
// a slli.uw.
if (X.getOpcode() == ISD::AND && isa<ConstantSDNode>(X.getOperand(1)) &&
@@ -16544,38 +16570,8 @@ static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
}
// 3/5/9 * 3/5/9 -> shXadd (shYadd X, X), (shYadd X, X)
- int ShX;
- int ShY;
- switch (MulAmt) {
- case 3 * 5:
- ShY = 1;
- ShX = 2;
- break;
- case 3 * 9:
- ShY = 1;
- ShX = 3;
- break;
- case 5 * 5:
- ShX = ShY = 2;
- break;
- case 5 * 9:
- ShY = 2;
- ShX = 3;
- break;
- case 9 * 9:
- ShX = ShY = 3;
- break;
- default:
- ShX = ShY = 0;
- break;
- }
- if (ShX) {
- SDLoc DL(N);
- SDValue Mul359 = DAG.getNode(RISCVISD::SHL_ADD, DL, VT, X,
- DAG.getConstant(ShY, DL, VT), X);
- return DAG.getNode(RISCVISD::SHL_ADD, DL, VT, Mul359,
- DAG.getConstant(ShX, DL, VT), Mul359);
- }
+ if (SDValue V = expandMulToShlAddShlAdd(N, DAG, MulAmt))
+ return V;
// If this is a power 2 + 2/4/8, we can use a shift followed by a single
// shXadd. First check if this a sum of two power of 2s because that's
@@ -16638,23 +16634,12 @@ static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
}
}
- for (uint64_t Divisor : {3, 5, 9}) {
- if (MulAmt % Divisor != 0)
- continue;
- uint64_t MulAmt2 = MulAmt / Divisor;
- // 3/5/9 * 3/5/9 * 2^N - In particular, this covers multiples
- // of 25 which happen to be quite common.
- if (int ShBAmount = isShifted359(MulAmt2, Shift)) {
- SDLoc DL(N);
- SDValue Mul359A =
- DAG.getNode(RISCVISD::SHL_ADD, DL, VT, X,
- DAG.getConstant(Log2_64(Divisor - 1), DL, VT), X);
- SDValue Mul359B =
- DAG.getNode(RISCVISD::SHL_ADD, DL, VT, Mul359A,
- DAG.getConstant(ShBAmount, DL, VT), Mul359A);
- return DAG.getNode(ISD::SHL, DL, VT, Mul359B,
- DAG.getConstant(Shift, DL, VT));
- }
+ // 3/5/9 * 3/5/9 * 2^N - In particular, this covers multiples
+ // of 25 which happen to be quite common.
+ Shift = llvm::countr_zero(MulAmt);
+ if (SDValue V = expandMulToShlAddShlAdd(N, DAG, MulAmt >> Shift)) {
+ SDLoc DL(N);
+ return DAG.getNode(ISD::SHL, DL, VT, V, DAG.getConstant(Shift, DL, VT));
}
}
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
topperc
reviewed
Nov 4, 2025
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/23/builds/15249 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/186/builds/13702 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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.
#158851 matches
3/5/9 * 3/5/9with aswitch.Reuse it for the shifted case to improve compilation time.