Skip to content

Commit e78be31

Browse files
mikhailramalhomikhailramalho
authored
[RISCV] Added cost model for fmuladd (#125683)
This patch updates the cost model for fmuladd on vector types to scale with LMUL. This was found when analyzing a hot loop in 519.lbm_r that was unprofitably vectorized, but doesn't directly impact that case and is split off so it doesn't get forgotten. Unlike other FP arithmetic ops, it's not scaled by 2 because the scalar cost isn't scaled by 2.
1 parent f77f777 commit e78be31

File tree

4 files changed

+48
-53
lines changed

4 files changed

+48
-53
lines changed

‎llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1176,6 +1176,14 @@ RISCVTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
11761176
}
11771177
break;
11781178
}
1179+
case Intrinsic::fmuladd: {
1180+
// TODO: handle promotion with f16/bf16 with zvfhmin/zvfbfmin
1181+
auto LT = getTypeLegalizationCost(RetTy);
1182+
if (ST->hasVInstructions() && LT.second.isVector())
1183+
return LT.first *
1184+
getRISCVInstructionCost(RISCV::VFMADD_VV, LT.second, CostKind);
1185+
break;
1186+
}
11791187
case Intrinsic::fabs: {
11801188
auto LT = getTypeLegalizationCost(RetTy);
11811189
if (ST->hasVInstructions() && LT.second.isVector()) {

‎llvm/test/Analysis/CostModel/RISCV/arith-fp.ll

Lines changed: 28 additions & 41 deletions
Original file line numberDiff line numberDiff line change
@@ -1508,30 +1508,30 @@ define void @fmuladd() {
15081508
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %4 = call <2 x bfloat> @llvm.fmuladd.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x bfloat> undef)
15091509
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %5 = call <4 x bfloat> @llvm.fmuladd.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x bfloat> undef)
15101510
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %6 = call <8 x bfloat> @llvm.fmuladd.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x bfloat> undef)
1511-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %7 = call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x bfloat> undef)
1512-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %8 = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> undef, <2 x float> undef, <2 x float> undef)
1513-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %9 = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
1511+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %7 = call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x bfloat> undef)
1512+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %8 = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> undef, <2 x float> undef, <2 x float> undef)
1513+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %9 = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
15141514
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %10 = call <8 x float> @llvm.fmuladd.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1515-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %11 = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
1516-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %12 = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1515+
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %11 = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
1516+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %12 = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
15171517
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %13 = call <4 x double> @llvm.fmuladd.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1518-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %14 = call <8 x double> @llvm.fmuladd.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1519-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %15 = call <16 x double> @llvm.fmuladd.v16f64(<16 x double> undef, <16 x double> undef, <16 x double> undef)
1518+
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %14 = call <8 x double> @llvm.fmuladd.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1519+
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %15 = call <16 x double> @llvm.fmuladd.v16f64(<16 x double> undef, <16 x double> undef, <16 x double> undef)
15201520
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %16 = call <vscale x 1 x bfloat> @llvm.fmuladd.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
15211521
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %17 = call <vscale x 2 x bfloat> @llvm.fmuladd.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
15221522
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %18 = call <vscale x 4 x bfloat> @llvm.fmuladd.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
1523-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %19 = call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
1524-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %20 = call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
1523+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %19 = call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
1524+
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %20 = call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
15251525
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call <vscale x 1 x float> @llvm.fmuladd.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x float> undef)
15261526
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <vscale x 2 x float> @llvm.fmuladd.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x float> undef)
1527-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x float> undef)
1528-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %24 = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x float> undef)
1529-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %25 = call <vscale x 16 x float> @llvm.fmuladd.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x float> undef)
1527+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x float> undef)
1528+
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x float> undef)
1529+
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <vscale x 16 x float> @llvm.fmuladd.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x float> undef)
15301530
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x double> @llvm.fmuladd.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x double> undef)
1531-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x double> @llvm.fmuladd.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x double> undef)
1532-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %28 = call <vscale x 4 x double> @llvm.fmuladd.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x double> undef)
1533-
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %29 = call <vscale x 8 x double> @llvm.fmuladd.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x double> undef)
1534-
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %30 = call <vscale x 16 x double> @llvm.fmuladd.nxv16f64(<vscale x 16 x double> undef, <vscale x 16 x double> undef, <vscale x 16 x double> undef)
1531+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x double> @llvm.fmuladd.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x double> undef)
1532+
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x double> @llvm.fmuladd.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x double> undef)
1533+
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x double> @llvm.fmuladd.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x double> undef)
1534+
; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %30 = call <vscale x 16 x double> @llvm.fmuladd.nxv16f64(<vscale x 16 x double> undef, <vscale x 16 x double> undef, <vscale x 16 x double> undef)
15351535
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
15361536
;
15371537
call bfloat @llvm.fmuladd.bf16(bfloat undef, bfloat undef, bfloat undef)
@@ -1568,31 +1568,18 @@ define void @fmuladd() {
15681568
}
15691569

15701570
define void @fmuladd_f16() {
1571-
; ZVFH-LABEL: 'fmuladd_f16'
1572-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half undef, half undef, half undef)
1573-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
1574-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
1575-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
1576-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
1577-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
1578-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
1579-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
1580-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
1581-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
1582-
; ZVFH-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
1583-
;
1584-
; ZVFHMIN-LABEL: 'fmuladd_f16'
1585-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half undef, half undef, half undef)
1586-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
1587-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
1588-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
1589-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
1590-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
1591-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
1592-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
1593-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
1594-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
1595-
; ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
1571+
; CHECK-LABEL: 'fmuladd_f16'
1572+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half undef, half undef, half undef)
1573+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
1574+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
1575+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
1576+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
1577+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
1578+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
1579+
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
1580+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
1581+
; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
1582+
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
15961583
;
15971584
call half @llvm.fmuladd.f16(half undef, half undef, half undef)
15981585
call <2 x half> @llvm.fmuladd.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)

‎llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -91,10 +91,10 @@ define void @vfwmaccbf16.vv(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64
9191
; NO-ZVFBFMIN-LABEL: define void @vfwmaccbf16.vv(
9292
; NO-ZVFBFMIN-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
9393
; NO-ZVFBFMIN-NEXT: [[ENTRY:.*]]:
94-
; NO-ZVFBFMIN-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 8
94+
; NO-ZVFBFMIN-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
9595
; NO-ZVFBFMIN-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
9696
; NO-ZVFBFMIN: [[VECTOR_PH]]:
97-
; NO-ZVFBFMIN-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
97+
; NO-ZVFBFMIN-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
9898
; NO-ZVFBFMIN-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
9999
; NO-ZVFBFMIN-NEXT: br label %[[VECTOR_BODY:.*]]
100100
; NO-ZVFBFMIN: [[VECTOR_BODY]]:
@@ -104,16 +104,16 @@ define void @vfwmaccbf16.vv(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64
104104
; NO-ZVFBFMIN-NEXT: [[TMP2:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[TMP0]]
105105
; NO-ZVFBFMIN-NEXT: [[TMP3:%.*]] = getelementptr float, ptr [[C]], i64 [[TMP0]]
106106
; NO-ZVFBFMIN-NEXT: [[TMP4:%.*]] = getelementptr bfloat, ptr [[TMP1]], i32 0
107-
; NO-ZVFBFMIN-NEXT: [[WIDE_LOAD:%.*]] = load <8 x bfloat>, ptr [[TMP4]], align 2
107+
; NO-ZVFBFMIN-NEXT: [[WIDE_LOAD:%.*]] = load <4 x bfloat>, ptr [[TMP4]], align 2
108108
; NO-ZVFBFMIN-NEXT: [[TMP5:%.*]] = getelementptr bfloat, ptr [[TMP2]], i32 0
109-
; NO-ZVFBFMIN-NEXT: [[WIDE_LOAD1:%.*]] = load <8 x bfloat>, ptr [[TMP5]], align 2
109+
; NO-ZVFBFMIN-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x bfloat>, ptr [[TMP5]], align 2
110110
; NO-ZVFBFMIN-NEXT: [[TMP6:%.*]] = getelementptr float, ptr [[TMP3]], i32 0
111-
; NO-ZVFBFMIN-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP6]], align 4
112-
; NO-ZVFBFMIN-NEXT: [[TMP7:%.*]] = fpext <8 x bfloat> [[WIDE_LOAD]] to <8 x float>
113-
; NO-ZVFBFMIN-NEXT: [[TMP8:%.*]] = fpext <8 x bfloat> [[WIDE_LOAD1]] to <8 x float>
114-
; NO-ZVFBFMIN-NEXT: [[TMP9:%.*]] = call <8 x float> @llvm.fmuladd.v8f32(<8 x float> [[TMP7]], <8 x float> [[TMP8]], <8 x float> [[WIDE_LOAD2]])
115-
; NO-ZVFBFMIN-NEXT: store <8 x float> [[TMP9]], ptr [[TMP6]], align 4
116-
; NO-ZVFBFMIN-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
111+
; NO-ZVFBFMIN-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
112+
; NO-ZVFBFMIN-NEXT: [[TMP7:%.*]] = fpext <4 x bfloat> [[WIDE_LOAD]] to <4 x float>
113+
; NO-ZVFBFMIN-NEXT: [[TMP8:%.*]] = fpext <4 x bfloat> [[WIDE_LOAD1]] to <4 x float>
114+
; NO-ZVFBFMIN-NEXT: [[TMP9:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[TMP7]], <4 x float> [[TMP8]], <4 x float> [[WIDE_LOAD2]])
115+
; NO-ZVFBFMIN-NEXT: store <4 x float> [[TMP9]], ptr [[TMP6]], align 4
116+
; NO-ZVFBFMIN-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
117117
; NO-ZVFBFMIN-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
118118
; NO-ZVFBFMIN-NEXT: br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
119119
; NO-ZVFBFMIN: [[MIDDLE_BLOCK]]:

‎llvm/test/Transforms/SLPVectorizer/RISCV/revec-getGatherCost.ll

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@
88
; YAML: Function: test1
99
; YAML: Args:
1010
; YAML: - String: 'Stores SLP vectorized with cost '
11-
; YAML: - Cost: '4'
11+
; YAML: - Cost: '5'
1212
; YAML: - String: ' and with tree size '
1313
; YAML: - TreeSize: '5'
1414

@@ -46,7 +46,7 @@ declare <4 x float> @llvm.fmuladd.v4f32(<4 x float>, <4 x float>, <4 x float>)
4646
; YAML: Function: test2
4747
; YAML: Args:
4848
; YAML: - String: 'Stores SLP vectorized with cost '
49-
; YAML: - Cost: '12'
49+
; YAML: - Cost: '14'
5050
; YAML: - String: ' and with tree size '
5151
; YAML: - TreeSize: '5'
5252

0 commit comments

Comments
 (0)