shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/SLPVectorizer/AArch64/getelementptr2.ll
Florian Hahn 02d3738be9
[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526) 
getVectorInstrCostHelper would return costs of zero for vector
inserts/extracts that move data between GPR and vector registers, if
there was no 'real' use, i.e. there was no corresponding existing
instruction.

This meant that passes like LoopVectorize and SLPVectorizer, which
likely are the main users of the interface, would understimate the cost
of insert/extracts that move data between GPR and vector registers,
which has non-trivial costs.

The patch removes the special case and only returns costs of zero for
lane 0 if it there is no need to transfer between integer and vector
registers.

This impacts a number of SLP test, and most of them look like general
improvements.I think the change should make things more accurate for any
AArch64 target, but if not it could also just be Apple CPU specific.

I am seeing +2% end-to-end improvements on SLP-heavy workloads.

PR: https://github.com/llvm/llvm-project/pull/146526
2025-07-15 15:19:27 +01:00

174 lines
7.7 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
;test_i16_extend NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -S -mtriple=aarch64--linux-gnu -passes=slp-vectorizer -slp-threshold=-5 -pass-remarks-output=%t < %s | FileCheck %s
; RUN: cat %t | FileCheck -check-prefix=YAML %s
; RUN: opt -S -mtriple=aarch64--linux-gnu -passes=slp-vectorizer -slp-threshold=-5 -pass-remarks-output=%t < %s | FileCheck %s
; RUN: cat %t | FileCheck -check-prefix=YAML %s
@global = internal global { ptr } zeroinitializer, align 8
; YAML-LABEL: --- !Passed
; YAML-NEXT: Pass: slp-vectorizer
; YAML-NEXT: Name: VectorizedList
; YAML-NEXT: Function: test_i16_extend
; YAML-NEXT: Args:
; YAML-NEXT: - String: 'SLP vectorized with cost '
; YAML-NEXT: - Cost: '-16
; YAML-NEXT: - String: ' and with tree size '
; YAML-NEXT: - TreeSize: '5'
; YAML-NEXT: ...
; Make sure we vectorize to maximize the load with when loading i16 and
; extending it for compute operations.
define void @test_i16_extend(ptr %p.1, ptr %p.2, i32 %idx.i32) {
; CHECK-LABEL: @test_i16_extend(
; CHECK-NEXT: [[P_0:%.*]] = load ptr, ptr @global, align 8
; CHECK-NEXT: [[IDX_0:%.*]] = zext i32 [[IDX_I32:%.*]] to i64
; CHECK-NEXT: [[T53:%.*]] = getelementptr inbounds i16, ptr [[P_1:%.*]], i64 [[IDX_0]]
; CHECK-NEXT: [[T56:%.*]] = getelementptr inbounds i16, ptr [[P_2:%.*]], i64 [[IDX_0]]
; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, ptr [[T53]], align 2
; CHECK-NEXT: [[TMP2:%.*]] = zext <8 x i16> [[TMP1]] to <8 x i32>
; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i16>, ptr [[T56]], align 2
; CHECK-NEXT: [[TMP4:%.*]] = zext <8 x i16> [[TMP3]] to <8 x i32>
; CHECK-NEXT: [[TMP5:%.*]] = sub <8 x i32> [[TMP2]], [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <8 x i32> [[TMP5]], i32 0
; CHECK-NEXT: [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
; CHECK-NEXT: [[T60:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP7]]
; CHECK-NEXT: [[L_1:%.*]] = load i32, ptr [[T60]], align 4
; CHECK-NEXT: [[TMP8:%.*]] = extractelement <8 x i32> [[TMP5]], i32 1
; CHECK-NEXT: [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
; CHECK-NEXT: [[T71:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP9]]
; CHECK-NEXT: [[L_2:%.*]] = load i32, ptr [[T71]], align 4
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <8 x i32> [[TMP5]], i32 2
; CHECK-NEXT: [[TMP11:%.*]] = sext i32 [[TMP10]] to i64
; CHECK-NEXT: [[T82:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP11]]
; CHECK-NEXT: [[L_3:%.*]] = load i32, ptr [[T82]], align 4
; CHECK-NEXT: [[TMP12:%.*]] = extractelement <8 x i32> [[TMP5]], i32 3
; CHECK-NEXT: [[TMP13:%.*]] = sext i32 [[TMP12]] to i64
; CHECK-NEXT: [[T93:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP13]]
; CHECK-NEXT: [[L_4:%.*]] = load i32, ptr [[T93]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = extractelement <8 x i32> [[TMP5]], i32 4
; CHECK-NEXT: [[TMP15:%.*]] = sext i32 [[TMP14]] to i64
; CHECK-NEXT: [[T104:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP15]]
; CHECK-NEXT: [[L_5:%.*]] = load i32, ptr [[T104]], align 4
; CHECK-NEXT: [[TMP16:%.*]] = extractelement <8 x i32> [[TMP5]], i32 5
; CHECK-NEXT: [[TMP17:%.*]] = sext i32 [[TMP16]] to i64
; CHECK-NEXT: [[T115:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP17]]
; CHECK-NEXT: [[L_6:%.*]] = load i32, ptr [[T115]], align 4
; CHECK-NEXT: [[TMP18:%.*]] = extractelement <8 x i32> [[TMP5]], i32 6
; CHECK-NEXT: [[TMP19:%.*]] = sext i32 [[TMP18]] to i64
; CHECK-NEXT: [[T126:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP19]]
; CHECK-NEXT: [[L_7:%.*]] = load i32, ptr [[T126]], align 4
; CHECK-NEXT: [[TMP20:%.*]] = extractelement <8 x i32> [[TMP5]], i32 7
; CHECK-NEXT: [[TMP21:%.*]] = sext i32 [[TMP20]] to i64
; CHECK-NEXT: [[T137:%.*]] = getelementptr inbounds i32, ptr [[P_0]], i64 [[TMP21]]
; CHECK-NEXT: [[L_8:%.*]] = load i32, ptr [[T137]], align 4
; CHECK-NEXT: call void @use(i32 [[L_1]], i32 [[L_2]], i32 [[L_3]], i32 [[L_4]], i32 [[L_5]], i32 [[L_6]], i32 [[L_7]], i32 [[L_8]])
; CHECK-NEXT: ret void
;
%p.0 = load ptr, ptr @global, align 8
%idx.0 = zext i32 %idx.i32 to i64
%idx.1 = add nsw i64 %idx.0, 1
%idx.2 = add nsw i64 %idx.0, 2
%idx.3 = add nsw i64 %idx.0, 3
%idx.4 = add nsw i64 %idx.0, 4
%idx.5 = add nsw i64 %idx.0, 5
%idx.6 = add nsw i64 %idx.0, 6
%idx.7 = add nsw i64 %idx.0, 7
%t53 = getelementptr inbounds i16, ptr %p.1, i64 %idx.0
%op1.l = load i16, ptr %t53, align 2
%op1.ext = zext i16 %op1.l to i64
%t56 = getelementptr inbounds i16, ptr %p.2, i64 %idx.0
%op2.l = load i16, ptr %t56, align 2
%op2.ext = zext i16 %op2.l to i64
%sub.1 = sub nsw i64 %op1.ext, %op2.ext
%t60 = getelementptr inbounds i32, ptr %p.0, i64 %sub.1
%l.1 = load i32, ptr %t60, align 4
%t64 = getelementptr inbounds i16, ptr %p.1, i64 %idx.1
%t65 = load i16, ptr %t64, align 2
%t66 = zext i16 %t65 to i64
%t67 = getelementptr inbounds i16, ptr %p.2, i64 %idx.1
%t68 = load i16, ptr %t67, align 2
%t69 = zext i16 %t68 to i64
%sub.2 = sub nsw i64 %t66, %t69
%t71 = getelementptr inbounds i32, ptr %p.0, i64 %sub.2
%l.2 = load i32, ptr %t71, align 4
%t75 = getelementptr inbounds i16, ptr %p.1, i64 %idx.2
%t76 = load i16, ptr %t75, align 2
%t77 = zext i16 %t76 to i64
%t78 = getelementptr inbounds i16, ptr %p.2, i64 %idx.2
%t79 = load i16, ptr %t78, align 2
%t80 = zext i16 %t79 to i64
%sub.3 = sub nsw i64 %t77, %t80
%t82 = getelementptr inbounds i32, ptr %p.0, i64 %sub.3
%l.3 = load i32, ptr %t82, align 4
%t86 = getelementptr inbounds i16, ptr %p.1, i64 %idx.3
%t87 = load i16, ptr %t86, align 2
%t88 = zext i16 %t87 to i64
%t89 = getelementptr inbounds i16, ptr %p.2, i64 %idx.3
%t90 = load i16, ptr %t89, align 2
%t91 = zext i16 %t90 to i64
%sub.4 = sub nsw i64 %t88, %t91
%t93 = getelementptr inbounds i32, ptr %p.0, i64 %sub.4
%l.4 = load i32, ptr %t93, align 4
%t97 = getelementptr inbounds i16, ptr %p.1, i64 %idx.4
%t98 = load i16, ptr %t97, align 2
%t99 = zext i16 %t98 to i64
%t100 = getelementptr inbounds i16, ptr %p.2, i64 %idx.4
%t101 = load i16, ptr %t100, align 2
%t102 = zext i16 %t101 to i64
%sub.5 = sub nsw i64 %t99, %t102
%t104 = getelementptr inbounds i32, ptr %p.0, i64 %sub.5
%l.5 = load i32, ptr %t104, align 4
%t108 = getelementptr inbounds i16, ptr %p.1, i64 %idx.5
%t109 = load i16, ptr %t108, align 2
%t110 = zext i16 %t109 to i64
%t111 = getelementptr inbounds i16, ptr %p.2, i64 %idx.5
%t112 = load i16, ptr %t111, align 2
%t113 = zext i16 %t112 to i64
%sub.6 = sub nsw i64 %t110, %t113
%t115 = getelementptr inbounds i32, ptr %p.0, i64 %sub.6
%l.6 = load i32, ptr %t115, align 4
%t119 = getelementptr inbounds i16, ptr %p.1, i64 %idx.6
%t120 = load i16, ptr %t119, align 2
%t121 = zext i16 %t120 to i64
%t122 = getelementptr inbounds i16, ptr %p.2, i64 %idx.6
%t123 = load i16, ptr %t122, align 2
%t124 = zext i16 %t123 to i64
%sub.7 = sub nsw i64 %t121, %t124
%t126 = getelementptr inbounds i32, ptr %p.0, i64 %sub.7
%l.7 = load i32, ptr %t126, align 4
%t130 = getelementptr inbounds i16, ptr %p.1, i64 %idx.7
%t131 = load i16, ptr %t130, align 2
%t132 = zext i16 %t131 to i64
%t133 = getelementptr inbounds i16, ptr %p.2, i64 %idx.7
%t134 = load i16, ptr %t133, align 2
%t135 = zext i16 %t134 to i64
%sub.8 = sub nsw i64 %t132, %t135
%t137 = getelementptr inbounds i32, ptr %p.0, i64 %sub.8
%l.8 = load i32, ptr %t137, align 4
call void @use(i32 %l.1, i32 %l.2, i32 %l.3, i32 %l.4, i32 %l.5, i32 %l.6, i32 %l.7, i32 %l.8)
ret void
}
declare void @use(i32, i32, i32, i32, i32, i32, i32, i32)
Reference in New Issue View Git Blame Copy Permalink