[AArch64] Runtime-unroll small multi-exit loops on Apple Silicon. #124751
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
|
@llvm/pr-subscribers-backend-aarch64 @llvm/pr-subscribers-llvm-transforms Author: Florian Hahn (fhahn) ChangesExtend unrolling preferences to allow more aggressive unrolling of In combination with eac23a5 this enables unrolling loops like Patch is 30.66 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/124751.diff 2 Files Affected:
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index aae2fdaf5bec37..18b5a5beb62387 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -4102,15 +4102,14 @@ getAppleRuntimeUnrollPreferences(Loop *L, ScalarEvolution &SE,
TargetTransformInfo::UnrollingPreferences &UP,
AArch64TTIImpl &TTI) {
// Limit loops with structure that is highly likely to benefit from runtime
- // unrolling; that is we exclude outer loops, loops with multiple exits and
- // many blocks (i.e. likely with complex control flow). Note that the
- // heuristics here may be overly conservative and we err on the side of
- // avoiding runtime unrolling rather than unroll excessively. They are all
- // subject to further refinement.
- if (!L->isInnermost() || !L->getExitBlock() || L->getNumBlocks() > 8)
+ // unrolling; that is we exclude outer loops and loops with many blocks (i.e.
+ // likely with complex control flow). Note that the heuristics here may be
+ // overly conservative and we err on the side of avoiding runtime unrolling
+ // rather than unroll excessively. They are all subject to further refinement.
+ if (!L->isInnermost() || L->getNumBlocks() > 8)
return;
- const SCEV *BTC = SE.getBackedgeTakenCount(L);
+ const SCEV *BTC = SE.getSymbolicMaxBackedgeTakenCount(L);
if (isa<SCEVConstant>(BTC) || isa<SCEVCouldNotCompute>(BTC) ||
(SE.getSmallConstantMaxTripCount(L) > 0 &&
SE.getSmallConstantMaxTripCount(L) <= 32))
@@ -4129,6 +4128,28 @@ getAppleRuntimeUnrollPreferences(Loop *L, ScalarEvolution &SE,
}
}
+ // Small search loops with multiple exits can be highly beneficial to unroll.
+ if (!L->getExitBlock()) {
+ if (L->getNumBlocks() == 2 && Size < 6 &&
+ all_of(
+ L->getBlocks(),
+ [](BasicBlock *BB) {
+ return isa<BranchInst>(BB->getTerminator());
+ })) {
+ UP.RuntimeUnrollMultiExit = true;
+ UP.Runtime = true;
+ // Limit unroll count.
+ UP.DefaultUnrollRuntimeCount = 4;
+ // Allow slightly more costly trip-count expansion to catch search loops
+ // with pointer inductions.
+ UP.SCEVExpansionBudget = 5;
+ }
+ return;
+ }
+
+ if (SE.getSymbolicMaxBackedgeTakenCount(L) != SE.getBackedgeTakenCount(L))
+ return;
+
// Limit to loops with trip counts that are cheap to expand.
UP.SCEVExpansionBudget = 1;
diff --git a/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling-multi-exit.ll b/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling-multi-exit.ll
new file mode 100644
index 00000000000000..31b23eae0f8660
--- /dev/null
+++ b/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling-multi-exit.ll
@@ -0,0 +1,515 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -p loop-unroll -mcpu=apple-m1 -S %s | FileCheck --check-prefix=APPLE %s
+; RUN: opt -p loop-unroll -mcpu=apple-m2 -S %s | FileCheck --check-prefix=APPLE %s
+; RUN: opt -p loop-unroll -mcpu=apple-m3 -S %s | FileCheck --check-prefix=APPLE %s
+; RUN: opt -p loop-unroll -mcpu=apple-m4 -S %s | FileCheck --check-prefix=APPLE %s
+; RUN: opt -p loop-unroll -mcpu=cortex-a57 -S %s | FileCheck --check-prefix=OTHER %s
+
+target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-n32:64-S128-Fn32"
+target triple = "arm64-apple-macosx15.0.0"
+
+define i1 @multi_2_exit_find_i8_loop(ptr %vec, i8 %tgt) {
+; APPLE-LABEL: define i1 @multi_2_exit_find_i8_loop(
+; APPLE-SAME: ptr [[VEC:%.*]], i8 [[TGT:%.*]]) #[[ATTR0:[0-9]+]] {
+; APPLE-NEXT: [[ENTRY:.*]]:
+; APPLE-NEXT: [[START:%.*]] = load ptr, ptr [[VEC]], align 8
+; APPLE-NEXT: [[START2:%.*]] = ptrtoint ptr [[START]] to i64
+; APPLE-NEXT: [[GEP_END:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 1
+; APPLE-NEXT: [[END:%.*]] = load ptr, ptr [[GEP_END]], align 8
+; APPLE-NEXT: [[END1:%.*]] = ptrtoint ptr [[END]] to i64
+; APPLE-NEXT: [[TMP0:%.*]] = sub i64 [[END1]], [[START2]]
+; APPLE-NEXT: [[TMP1:%.*]] = freeze i64 [[TMP0]]
+; APPLE-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], -1
+; APPLE-NEXT: [[XTRAITER:%.*]] = and i64 [[TMP1]], 3
+; APPLE-NEXT: [[LCMP_MOD:%.*]] = icmp ne i64 [[XTRAITER]], 0
+; APPLE-NEXT: br i1 [[LCMP_MOD]], label %[[LOOP_HEADER_PROL_PREHEADER:.*]], label %[[LOOP_HEADER_PROL_LOOPEXIT:.*]]
+; APPLE: [[LOOP_HEADER_PROL_PREHEADER]]:
+; APPLE-NEXT: br label %[[LOOP_HEADER_PROL:.*]]
+; APPLE: [[LOOP_HEADER_PROL]]:
+; APPLE-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[PTR_IV_NEXT:%.*]], %[[LOOP_LATCH_PROL:.*]] ], [ [[START]], %[[LOOP_HEADER_PROL_PREHEADER]] ]
+; APPLE-NEXT: [[PROL_ITER:%.*]] = phi i64 [ 0, %[[LOOP_HEADER_PROL_PREHEADER]] ], [ [[PROL_ITER_NEXT:%.*]], %[[LOOP_LATCH_PROL]] ]
+; APPLE-NEXT: [[L:%.*]] = load i8, ptr [[PTR_IV]], align 8
+; APPLE-NEXT: [[C_1:%.*]] = icmp eq i8 [[L]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1]], label %[[EXIT_UNR_LCSSA_LOOPEXIT3:.*]], label %[[LOOP_LATCH_PROL]]
+; APPLE: [[LOOP_LATCH_PROL]]:
+; APPLE-NEXT: [[PTR_IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR_IV]], i64 1
+; APPLE-NEXT: [[C_2:%.*]] = icmp eq ptr [[PTR_IV_NEXT]], [[END]]
+; APPLE-NEXT: [[PROL_ITER_NEXT]] = add i64 [[PROL_ITER]], 1
+; APPLE-NEXT: [[PROL_ITER_CMP:%.*]] = icmp ne i64 [[PROL_ITER_NEXT]], [[XTRAITER]]
+; APPLE-NEXT: br i1 [[PROL_ITER_CMP]], label %[[LOOP_HEADER_PROL]], label %[[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA:.*]], !llvm.loop [[LOOP0:![0-9]+]]
+; APPLE: [[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA]]:
+; APPLE-NEXT: [[RES_UNR_PH:%.*]] = phi ptr [ [[END]], %[[LOOP_LATCH_PROL]] ]
+; APPLE-NEXT: [[PTR_IV_UNR_PH:%.*]] = phi ptr [ [[PTR_IV_NEXT]], %[[LOOP_LATCH_PROL]] ]
+; APPLE-NEXT: br label %[[LOOP_HEADER_PROL_LOOPEXIT]]
+; APPLE: [[LOOP_HEADER_PROL_LOOPEXIT]]:
+; APPLE-NEXT: [[RES_UNR:%.*]] = phi ptr [ poison, %[[ENTRY]] ], [ [[RES_UNR_PH]], %[[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA]] ]
+; APPLE-NEXT: [[PTR_IV_UNR:%.*]] = phi ptr [ [[START]], %[[ENTRY]] ], [ [[PTR_IV_UNR_PH]], %[[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA]] ]
+; APPLE-NEXT: [[TMP3:%.*]] = icmp ult i64 [[TMP2]], 3
+; APPLE-NEXT: br i1 [[TMP3]], label %[[EXIT:.*]], label %[[ENTRY_NEW:.*]]
+; APPLE: [[ENTRY_NEW]]:
+; APPLE-NEXT: br label %[[LOOP_HEADER:.*]]
+; APPLE: [[LOOP_HEADER]]:
+; APPLE-NEXT: [[PTR_IV1:%.*]] = phi ptr [ [[PTR_IV_UNR]], %[[ENTRY_NEW]] ], [ [[RES:%.*]], %[[LOOP_LATCH_3:.*]] ]
+; APPLE-NEXT: [[L1:%.*]] = load i8, ptr [[PTR_IV1]], align 8
+; APPLE-NEXT: [[C_4:%.*]] = icmp eq i8 [[L1]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_4]], label %[[EXIT_UNR_LCSSA_LOOPEXIT:.*]], label %[[LOOP_LATCH:.*]]
+; APPLE: [[LOOP_LATCH]]:
+; APPLE-NEXT: [[PTR_IV_NEXT1:%.*]] = getelementptr inbounds nuw i8, ptr [[PTR_IV1]], i64 1
+; APPLE-NEXT: [[L_1:%.*]] = load i8, ptr [[PTR_IV_NEXT1]], align 8
+; APPLE-NEXT: [[C_1_1:%.*]] = icmp eq i8 [[L_1]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1_1]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_LATCH_1:.*]]
+; APPLE: [[LOOP_LATCH_1]]:
+; APPLE-NEXT: [[PTR_IV_NEXT_1:%.*]] = getelementptr inbounds nuw i8, ptr [[PTR_IV_NEXT1]], i64 1
+; APPLE-NEXT: [[L_2:%.*]] = load i8, ptr [[PTR_IV_NEXT_1]], align 8
+; APPLE-NEXT: [[C_1_2:%.*]] = icmp eq i8 [[L_2]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1_2]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_LATCH_2:.*]]
+; APPLE: [[LOOP_LATCH_2]]:
+; APPLE-NEXT: [[PTR_IV_NEXT_2:%.*]] = getelementptr inbounds nuw i8, ptr [[PTR_IV_NEXT_1]], i64 1
+; APPLE-NEXT: [[L_3:%.*]] = load i8, ptr [[PTR_IV_NEXT_2]], align 8
+; APPLE-NEXT: [[C_1_3:%.*]] = icmp eq i8 [[L_3]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1_3]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_LATCH_3]]
+; APPLE: [[LOOP_LATCH_3]]:
+; APPLE-NEXT: [[RES]] = getelementptr inbounds nuw i8, ptr [[PTR_IV_NEXT_2]], i64 1
+; APPLE-NEXT: [[C_3:%.*]] = icmp eq ptr [[RES]], [[END]]
+; APPLE-NEXT: br i1 [[C_3]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_HEADER]]
+; APPLE: [[EXIT_UNR_LCSSA_LOOPEXIT]]:
+; APPLE-NEXT: [[RES_PH_PH:%.*]] = phi ptr [ [[PTR_IV1]], %[[LOOP_HEADER]] ], [ [[PTR_IV_NEXT1]], %[[LOOP_LATCH]] ], [ [[PTR_IV_NEXT_1]], %[[LOOP_LATCH_1]] ], [ [[PTR_IV_NEXT_2]], %[[LOOP_LATCH_2]] ], [ [[END]], %[[LOOP_LATCH_3]] ]
+; APPLE-NEXT: br label %[[EXIT_UNR_LCSSA:.*]]
+; APPLE: [[EXIT_UNR_LCSSA_LOOPEXIT3]]:
+; APPLE-NEXT: [[RES_PH_PH4:%.*]] = phi ptr [ [[PTR_IV]], %[[LOOP_HEADER_PROL]] ]
+; APPLE-NEXT: br label %[[EXIT_UNR_LCSSA]]
+; APPLE: [[EXIT_UNR_LCSSA]]:
+; APPLE-NEXT: [[RES_PH:%.*]] = phi ptr [ [[RES_PH_PH]], %[[EXIT_UNR_LCSSA_LOOPEXIT]] ], [ [[RES_PH_PH4]], %[[EXIT_UNR_LCSSA_LOOPEXIT3]] ]
+; APPLE-NEXT: br label %[[EXIT]]
+; APPLE: [[EXIT]]:
+; APPLE-NEXT: [[RES1:%.*]] = phi ptr [ [[RES_UNR]], %[[LOOP_HEADER_PROL_LOOPEXIT]] ], [ [[RES_PH]], %[[EXIT_UNR_LCSSA]] ]
+; APPLE-NEXT: [[C_5:%.*]] = icmp eq ptr [[RES1]], [[END]]
+; APPLE-NEXT: ret i1 [[C_5]]
+;
+; OTHER-LABEL: define i1 @multi_2_exit_find_i8_loop(
+; OTHER-SAME: ptr [[VEC:%.*]], i8 [[TGT:%.*]]) #[[ATTR0:[0-9]+]] {
+; OTHER-NEXT: [[ENTRY:.*]]:
+; OTHER-NEXT: [[START:%.*]] = load ptr, ptr [[VEC]], align 8
+; OTHER-NEXT: [[GEP_END:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 1
+; OTHER-NEXT: [[END:%.*]] = load ptr, ptr [[GEP_END]], align 8
+; OTHER-NEXT: br label %[[LOOP_HEADER:.*]]
+; OTHER: [[LOOP_HEADER]]:
+; OTHER-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[PTR_IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[START]], %[[ENTRY]] ]
+; OTHER-NEXT: [[L:%.*]] = load i8, ptr [[PTR_IV]], align 8
+; OTHER-NEXT: [[C_1:%.*]] = icmp eq i8 [[L]], [[TGT]]
+; OTHER-NEXT: br i1 [[C_1]], label %[[EXIT:.*]], label %[[LOOP_LATCH]]
+; OTHER: [[LOOP_LATCH]]:
+; OTHER-NEXT: [[PTR_IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR_IV]], i64 1
+; OTHER-NEXT: [[C_2:%.*]] = icmp eq ptr [[PTR_IV_NEXT]], [[END]]
+; OTHER-NEXT: br i1 [[C_2]], label %[[EXIT]], label %[[LOOP_HEADER]]
+; OTHER: [[EXIT]]:
+; OTHER-NEXT: [[RES:%.*]] = phi ptr [ [[PTR_IV]], %[[LOOP_HEADER]] ], [ [[END]], %[[LOOP_LATCH]] ]
+; OTHER-NEXT: [[C_3:%.*]] = icmp eq ptr [[RES]], [[END]]
+; OTHER-NEXT: ret i1 [[C_3]]
+;
+entry:
+ %start = load ptr, ptr %vec, align 8
+ %gep.end = getelementptr inbounds nuw i8, ptr %vec, i64 1
+ %end = load ptr, ptr %gep.end, align 8
+ br label %loop.header
+
+loop.header:
+ %ptr.iv = phi ptr [ %ptr.iv.next, %loop.latch ], [ %start, %entry ]
+ %l = load i8, ptr %ptr.iv, align 8
+ %c.1 = icmp eq i8 %l, %tgt
+ br i1 %c.1, label %exit, label %loop.latch
+
+loop.latch:
+ %ptr.iv.next = getelementptr inbounds nuw i8, ptr %ptr.iv, i64 1
+ %c.2 = icmp eq ptr %ptr.iv.next, %end
+ br i1 %c.2, label %exit, label %loop.header
+
+exit:
+ %res = phi ptr [ %ptr.iv, %loop.header ], [ %end, %loop.latch ]
+ %c.3 = icmp eq ptr %res, %end
+ ret i1 %c.3
+}
+
+
+define i1 @multi_2_exit_find_ptr_loop(ptr %vec, ptr %tgt) {
+; APPLE-LABEL: define i1 @multi_2_exit_find_ptr_loop(
+; APPLE-SAME: ptr [[VEC:%.*]], ptr [[TGT:%.*]]) #[[ATTR0]] {
+; APPLE-NEXT: [[ENTRY:.*]]:
+; APPLE-NEXT: [[START:%.*]] = load ptr, ptr [[VEC]], align 8
+; APPLE-NEXT: [[START2:%.*]] = ptrtoint ptr [[START]] to i64
+; APPLE-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[START]], i64 8) ]
+; APPLE-NEXT: [[GEP_END:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 8
+; APPLE-NEXT: [[END:%.*]] = load ptr, ptr [[GEP_END]], align 8
+; APPLE-NEXT: [[END1:%.*]] = ptrtoint ptr [[END]] to i64
+; APPLE-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 8) ]
+; APPLE-NEXT: [[TMP0:%.*]] = add i64 [[END1]], -8
+; APPLE-NEXT: [[TMP1:%.*]] = sub i64 [[TMP0]], [[START2]]
+; APPLE-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 3
+; APPLE-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; APPLE-NEXT: [[TMP4:%.*]] = freeze i64 [[TMP3]]
+; APPLE-NEXT: [[TMP5:%.*]] = add i64 [[TMP4]], -1
+; APPLE-NEXT: [[XTRAITER:%.*]] = and i64 [[TMP4]], 3
+; APPLE-NEXT: [[LCMP_MOD:%.*]] = icmp ne i64 [[XTRAITER]], 0
+; APPLE-NEXT: br i1 [[LCMP_MOD]], label %[[LOOP_HEADER_PROL_PREHEADER:.*]], label %[[LOOP_HEADER_PROL_LOOPEXIT:.*]]
+; APPLE: [[LOOP_HEADER_PROL_PREHEADER]]:
+; APPLE-NEXT: br label %[[LOOP_HEADER_PROL:.*]]
+; APPLE: [[LOOP_HEADER_PROL]]:
+; APPLE-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[PTR_IV_NEXT:%.*]], %[[LOOP_LATCH_PROL:.*]] ], [ [[START]], %[[LOOP_HEADER_PROL_PREHEADER]] ]
+; APPLE-NEXT: [[PROL_ITER:%.*]] = phi i64 [ 0, %[[LOOP_HEADER_PROL_PREHEADER]] ], [ [[PROL_ITER_NEXT:%.*]], %[[LOOP_LATCH_PROL]] ]
+; APPLE-NEXT: [[L:%.*]] = load ptr, ptr [[PTR_IV]], align 8
+; APPLE-NEXT: [[C_1:%.*]] = icmp eq ptr [[L]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1]], label %[[EXIT_UNR_LCSSA_LOOPEXIT3:.*]], label %[[LOOP_LATCH_PROL]]
+; APPLE: [[LOOP_LATCH_PROL]]:
+; APPLE-NEXT: [[PTR_IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR_IV]], i64 8
+; APPLE-NEXT: [[C_2:%.*]] = icmp eq ptr [[PTR_IV_NEXT]], [[END]]
+; APPLE-NEXT: [[PROL_ITER_NEXT]] = add i64 [[PROL_ITER]], 1
+; APPLE-NEXT: [[PROL_ITER_CMP:%.*]] = icmp ne i64 [[PROL_ITER_NEXT]], [[XTRAITER]]
+; APPLE-NEXT: br i1 [[PROL_ITER_CMP]], label %[[LOOP_HEADER_PROL]], label %[[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA:.*]], !llvm.loop [[LOOP2:![0-9]+]]
+; APPLE: [[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA]]:
+; APPLE-NEXT: [[RES_UNR_PH:%.*]] = phi ptr [ [[END]], %[[LOOP_LATCH_PROL]] ]
+; APPLE-NEXT: [[PTR_IV_UNR_PH:%.*]] = phi ptr [ [[PTR_IV_NEXT]], %[[LOOP_LATCH_PROL]] ]
+; APPLE-NEXT: br label %[[LOOP_HEADER_PROL_LOOPEXIT]]
+; APPLE: [[LOOP_HEADER_PROL_LOOPEXIT]]:
+; APPLE-NEXT: [[RES_UNR:%.*]] = phi ptr [ poison, %[[ENTRY]] ], [ [[RES_UNR_PH]], %[[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA]] ]
+; APPLE-NEXT: [[PTR_IV_UNR:%.*]] = phi ptr [ [[START]], %[[ENTRY]] ], [ [[PTR_IV_UNR_PH]], %[[LOOP_HEADER_PROL_LOOPEXIT_UNR_LCSSA]] ]
+; APPLE-NEXT: [[TMP6:%.*]] = icmp ult i64 [[TMP5]], 3
+; APPLE-NEXT: br i1 [[TMP6]], label %[[EXIT:.*]], label %[[ENTRY_NEW:.*]]
+; APPLE: [[ENTRY_NEW]]:
+; APPLE-NEXT: br label %[[LOOP_HEADER:.*]]
+; APPLE: [[LOOP_HEADER]]:
+; APPLE-NEXT: [[PTR_IV1:%.*]] = phi ptr [ [[PTR_IV_UNR]], %[[ENTRY_NEW]] ], [ [[PTR_IV_NEXT_3:%.*]], %[[LOOP_LATCH_3:.*]] ]
+; APPLE-NEXT: [[L1:%.*]] = load ptr, ptr [[PTR_IV1]], align 8
+; APPLE-NEXT: [[C_4:%.*]] = icmp eq ptr [[L1]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_4]], label %[[EXIT_UNR_LCSSA_LOOPEXIT:.*]], label %[[LOOP_LATCH:.*]]
+; APPLE: [[LOOP_LATCH]]:
+; APPLE-NEXT: [[PTR_IV_NEXT1:%.*]] = getelementptr inbounds nuw i8, ptr [[PTR_IV1]], i64 8
+; APPLE-NEXT: [[L_1:%.*]] = load ptr, ptr [[PTR_IV_NEXT1]], align 8
+; APPLE-NEXT: [[C_1_1:%.*]] = icmp eq ptr [[L_1]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1_1]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_LATCH_1:.*]]
+; APPLE: [[LOOP_LATCH_1]]:
+; APPLE-NEXT: [[PTR_IV_NEXT_1:%.*]] = getelementptr inbounds nuw i8, ptr [[PTR_IV_NEXT1]], i64 8
+; APPLE-NEXT: [[L_2:%.*]] = load ptr, ptr [[PTR_IV_NEXT_1]], align 8
+; APPLE-NEXT: [[C_1_2:%.*]] = icmp eq ptr [[L_2]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1_2]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_LATCH_2:.*]]
+; APPLE: [[LOOP_LATCH_2]]:
+; APPLE-NEXT: [[PTR_IV_NEXT_2:%.*]] = getelementptr inbounds nuw i8, ptr [[PTR_IV_NEXT_1]], i64 8
+; APPLE-NEXT: [[L_3:%.*]] = load ptr, ptr [[PTR_IV_NEXT_2]], align 8
+; APPLE-NEXT: [[C_1_3:%.*]] = icmp eq ptr [[L_3]], [[TGT]]
+; APPLE-NEXT: br i1 [[C_1_3]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_LATCH_3]]
+; APPLE: [[LOOP_LATCH_3]]:
+; APPLE-NEXT: [[PTR_IV_NEXT_3]] = getelementptr inbounds nuw i8, ptr [[PTR_IV_NEXT_2]], i64 8
+; APPLE-NEXT: [[C_2_3:%.*]] = icmp eq ptr [[PTR_IV_NEXT_3]], [[END]]
+; APPLE-NEXT: br i1 [[C_2_3]], label %[[EXIT_UNR_LCSSA_LOOPEXIT]], label %[[LOOP_HEADER]]
+; APPLE: [[EXIT_UNR_LCSSA_LOOPEXIT]]:
+; APPLE-NEXT: [[RES_PH_PH:%.*]] = phi ptr [ [[PTR_IV1]], %[[LOOP_HEADER]] ], [ [[PTR_IV_NEXT1]], %[[LOOP_LATCH]] ], [ [[PTR_IV_NEXT_1]], %[[LOOP_LATCH_1]] ], [ [[PTR_IV_NEXT_2]], %[[LOOP_LATCH_2]] ], [ [[END]], %[[LOOP_LATCH_3]] ]
+; APPLE-NEXT: br label %[[EXIT_UNR_LCSSA:.*]]
+; APPLE: [[EXIT_UNR_LCSSA_LOOPEXIT3]]:
+; APPLE-NEXT: [[RES_PH_PH4:%.*]] = phi ptr [ [[PTR_IV]], %[[LOOP_HEADER_PROL]] ]
+; APPLE-NEXT: br label %[[EXIT_UNR_LCSSA]]
+; APPLE: [[EXIT_UNR_LCSSA]]:
+; APPLE-NEXT: [[RES_PH:%.*]] = phi ptr [ [[RES_PH_PH]], %[[EXIT_UNR_LCSSA_LOOPEXIT]] ], [ [[RES_PH_PH4]], %[[EXIT_UNR_LCSSA_LOOPEXIT3]] ]
+; APPLE-NEXT: br label %[[EXIT]]
+; APPLE: [[EXIT]]:
+; APPLE-NEXT: [[RES:%.*]] = phi ptr [ [[RES_UNR]], %[[LOOP_HEADER_PROL_LOOPEXIT]] ], [ [[RES_PH]], %[[EXIT_UNR_LCSSA]] ]
+; APPLE-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 8) ]
+; APPLE-NEXT: [[C_3:%.*]] = icmp eq ptr [[RES]], [[END]]
+; APPLE-NEXT: ret i1 [[C_3]]
+;
+; OTHER-LABEL: define i1 @multi_2_exit_find_ptr_loop(
+; OTHER-SAME: ptr [[VEC:%.*]], ptr [[TGT:%.*]]) #[[ATTR0]] {
+; OTHER-NEXT: [[ENTRY:.*]]:
+; OTHER-NEXT: [[START:%.*]] = load ptr, ptr [[VEC]], align 8
+; OTHER-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[START]], i64 8) ]
+; OTHER-NEXT: [[GEP_END:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 8
+; OTHER-NEXT: [[END:%.*]] = load ptr, ptr [[GEP_END]], align 8
+; OTHER-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 8) ]
+; OTHER-NEXT: br label %[[LOOP_HEADER:.*]]
+; OTHER: [[LOOP_HEADER]]:
+; OTHER-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[PTR_IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[START]], %[[ENTRY]] ]
+; OTHER-NEXT: [[L:%.*]] = load ptr, ptr [[PTR_IV]], align 8
+; OTHER-NEXT: [[C_1:%.*]] = icmp eq ptr [[L]], [[TGT]]
+; OTHER-NEXT: br i1 [[C_1]], label %[[EXIT:.*]], label %[[LOOP_LATCH]]
+; OTHER: [[LOOP_LATCH]]:
+; OTHER-NEXT: [[PTR_IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR_IV]], i64 8
+; OTHER-NEXT: [[C_2:%.*]] = icmp eq ptr [[PTR_IV_NEXT]], [[END]]
+; OTHER-NEXT: br i1 [[C_2]], label %[[EXIT]], label %[[LOOP_HEADER]]
+; OTHER: [[EXIT]]:
+; OTHER-NEXT: [[RES:%.*]] = phi ptr [ [[PTR_IV]], %[[LOOP_HEADER]] ], [ [[END]], %[[LOOP_LATCH]] ]
+; OTHER-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 8) ]
+; OTHER-NEXT: [[C_3:%.*]] = icmp eq ptr [[RES]], [[END]]
+; OTHER-NEXT: ret i1 [[C_3]]
+;
+entry:
+ %start = load ptr, ptr %vec, align 8
+ call void @llvm.assume(i1 true) [ "align"(ptr %start, i64 8) ]
+ %gep.end = getelementptr inbounds nuw i8, ptr %vec, i64 8
+ %end = load ptr, ptr %gep.end, align 8
+ call void @llvm.assume(i1 true) [ "align"(ptr %end, i64 8) ]
+ br label %loop.header
+
+loop.header:
+ %ptr.iv = phi ptr [ %ptr.iv.next, %loop.latch ], [ %start, %entry ]
+ %l = load ptr, ptr %ptr.iv, align 8
+ %c.1 = icmp eq ptr %l, %tgt
+ br i1 %c.1, label %exit, label %loop.latch
+
+loop.latch:
+ %ptr.iv.next = getelementptr inbounds nuw i8, ptr %ptr.iv, i64 8
+ %c.2 = icmp eq ptr %ptr.iv.next, %end
+ br i1 %c.2, label %exit, label %loop.header
+
+exit:
+ %res = phi ptr [ %ptr.iv, %loop.header ], [ %end, %loop.latch ]
+ call void @llvm.assume(i1 true) [ "align"(ptr %end, i64 8) ]
+ %c.3 = icmp eq ptr %res, %end
+ ret i1 %c.3
+}
+
+define i1 @multi_2_exit_find_i8_loop_too_large(ptr %vec, i8 %tgt) {
+; APPLE-LABEL: define i1 @multi_2_exit_find_i8_loop_too_large(
+; APPLE-SAME: ptr [[VEC:%.*]], i8 [[TGT:%.*]]) #[[ATTR0]] {
+; APPLE-NEXT: [[ENTRY:.*]]:
+; APPLE-NEXT: [[START:%.*]] = load ptr, ptr [[VEC]], align 8
+; APPLE-NEXT: [[GEP_END:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 1
+; APPLE-NEXT: [[END:%.*]] = load ptr, ptr [[GEP_END]], align 8
+; APPLE-NEXT: br label %[[LOOP_HEADER:.*]]
+; APPLE: [[LOOP_HEADER]]:
+; APPLE-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[PTR_IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[START]], %[[ENTRY]] ]
+; APPLE-NEXT: [[L:%.*]] = load i8, ptr [[PTR_IV]], align 8
+; APPLE-NEXT: [[UDIV:%.*]] = udiv i8 [[L]], [[TGT]]
+; APPLE-NEXT: [[UDIV_2:%.*]] = udiv i8 [[UDIV]], 10
+; APPLE-NEXT: [[C_1:%.*]] = icmp eq i8 [[UDIV_2]], 2
+; APPLE-NEXT: br i1 [[C_1]], la...
[truncated]
|
fhahn
force-pushed
the
rt-multi-exit-apple-si
branch
from
47d7b73 to
d446135
Compare
github-actions bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Jan 28, 2025
Extend unrolling preferences to allow more aggressive unrolling of search loops with 2 exits, building on the TTI hook added in ad9da92. In combination with eac23a5 this enables unrolling loops like std::find, which can improve performance significantly (+15% end-to-end on a workload that makes heavy use of std::find). It increase the total number of unrolled loops by ~2.5% across a very large corpus of workloads.
fhahn
force-pushed
the
rt-multi-exit-apple-si
branch
from
d446135 to
2747463
Compare
jroelofs
approved these changes
Jan 28, 2025
fhahn
added a commit
to fhahn/llvm-project
that referenced
this pull request
Feb 25, 2025
Test coverage to llvm#124751.
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Feb 27, 2025
…ilicon. (#124751) Extend unrolling preferences to allow more aggressive unrolling of search loops with 2 exits, building on the TTI hook added in llvm/llvm-project@ad9da92. In combination with llvm/llvm-project@eac23a5 this enables unrolling loops like std::find, which can improve performance significantly (+15% end-to-end on a workload that makes heavy use of std::find). It increase the total number of unrolled loops by ~2.5% across a very large corpus of workloads. For SPEC2017, +1.6% more loops are unrolled and the following workloads increase in size (`__text`): workload base patch 500.perlbench_r 1682884.00 1694104.00 0.7% 523.xalancbmk_r 3001716.00 3003832.00 0.1% PR: llvm/llvm-project#124751
joaosaffran
pushed a commit
to joaosaffran/llvm-project
that referenced
this pull request
Mar 3, 2025
…vm#124751) Extend unrolling preferences to allow more aggressive unrolling of search loops with 2 exits, building on the TTI hook added in llvm@ad9da92. In combination with llvm@eac23a5 this enables unrolling loops like std::find, which can improve performance significantly (+15% end-to-end on a workload that makes heavy use of std::find). It increase the total number of unrolled loops by ~2.5% across a very large corpus of workloads. For SPEC2017, +1.6% more loops are unrolled and the following workloads increase in size (`__text`): workload base patch 500.perlbench_r 1682884.00 1694104.00 0.7% 523.xalancbmk_r 3001716.00 3003832.00 0.1% PR: llvm#124751
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.
Extend unrolling preferences to allow more aggressive unrolling of
search loops with 2 exits, building on the TTI hook added in
ad9da92.
In combination with eac23a5 this enables unrolling loops like
std::find, which can improve performance significantly (+15% end-to-end
on a workload that makes heavy use of std::find). It increase the total
number of unrolled loops by ~2.5% across a very large corpus of
workloads.
For SPEC2017, +1.6% more loops are unrolled and the following workloads increase in size (
__text):