120 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py2; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output < %s | FileCheck %s --check-prefix=CHECK-VSCALE-13; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mcpu=neoverse-v1 < %s | FileCheck %s --check-prefix=CHECK-VSCALE-24; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mcpu=neoverse-n2 < %s | FileCheck %s --check-prefix=CHECK-VSCALE-15; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mcpu=cortex-a510 < %s | FileCheck %s --check-prefix=CHECK-VSCALE-16; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mcpu=neoverse-v2 < %s | FileCheck %s --check-prefix=CHECK-VSCALE-17 8target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"9target triple="aarch64--linux-gnu"10 11define void @masked_gathers(<vscale x 4 x i1> %nxv4i1mask, <vscale x 8 x i1> %nxv8i1mask, <4 x i1> %v4i1mask, <1 x i1> %v1i1mask, <vscale x 1 x i1> %nxv1i1mask) #0 {12; CHECK-VSCALE-1-LABEL: 'masked_gathers'13; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv4i32 = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> align 4 undef, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i32> zeroinitializer)14; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv8i32 = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> align 4 undef, <vscale x 8 x i1> %nxv8i1mask, <vscale x 8 x i32> zeroinitializer)15; CHECK-VSCALE-1-NEXT: Cost Model: Invalid cost for instruction: %res.nxv1i64 = call <vscale x 1 x i64> @llvm.masked.gather.nxv1i64.nxv1p0(<vscale x 1 x ptr> align 8 undef, <vscale x 1 x i1> %nxv1i1mask, <vscale x 1 x i64> zeroinitializer)16; CHECK-VSCALE-1-NEXT: Cost Model: Invalid cost for instruction: %res.nxv4i1 = call <vscale x 4 x i1> @llvm.masked.gather.nxv4i1.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i1> zeroinitializer)17; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void18;19; CHECK-VSCALE-2-LABEL: 'masked_gathers'20; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv4i32 = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> align 4 undef, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i32> zeroinitializer)21; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 160 for instruction: %res.nxv8i32 = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> align 4 undef, <vscale x 8 x i1> %nxv8i1mask, <vscale x 8 x i32> zeroinitializer)22; CHECK-VSCALE-2-NEXT: Cost Model: Invalid cost for instruction: %res.nxv1i64 = call <vscale x 1 x i64> @llvm.masked.gather.nxv1i64.nxv1p0(<vscale x 1 x ptr> align 8 undef, <vscale x 1 x i1> %nxv1i1mask, <vscale x 1 x i64> zeroinitializer)23; CHECK-VSCALE-2-NEXT: Cost Model: Invalid cost for instruction: %res.nxv4i1 = call <vscale x 4 x i1> @llvm.masked.gather.nxv4i1.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i1> zeroinitializer)24; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void25;26 %res.nxv4i32 = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32(<vscale x 4 x ptr> undef, i32 0, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i32> zeroinitializer)27 %res.nxv8i32 = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32(<vscale x 8 x ptr> undef, i32 0, <vscale x 8 x i1> %nxv8i1mask, <vscale x 8 x i32> zeroinitializer)28 %res.nxv1i64 = call <vscale x 1 x i64> @llvm.masked.gather.nxv1i64(<vscale x 1 x ptr> undef, i32 0, <vscale x 1 x i1> %nxv1i1mask, <vscale x 1 x i64> zeroinitializer)29 %res.nxv4i1 = call <vscale x 4 x i1> @llvm.masked.gather.nxv4i1(<vscale x 4 x ptr> undef, i32 0, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i1> zeroinitializer)30 ret void31}32 33define void @masked_gathers_tune_generic(<vscale x 4 x i1> %nxv4i1mask, <vscale x 8 x i1> %nxv8i1mask, <4 x i1> %v4i1mask, <1 x i1> %v1i1mask, <vscale x 1 x i1> %nxv1i1mask) #1 {34; CHECK-VSCALE-1-LABEL: 'masked_gathers_tune_generic'35; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv4i32 = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> align 4 undef, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i32> zeroinitializer)36; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv8i32 = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> align 4 undef, <vscale x 8 x i1> %nxv8i1mask, <vscale x 8 x i32> zeroinitializer)37; CHECK-VSCALE-1-NEXT: Cost Model: Invalid cost for instruction: %res.nxv1i64 = call <vscale x 1 x i64> @llvm.masked.gather.nxv1i64.nxv1p0(<vscale x 1 x ptr> align 8 undef, <vscale x 1 x i1> %nxv1i1mask, <vscale x 1 x i64> zeroinitializer)38; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void39;40; CHECK-VSCALE-2-LABEL: 'masked_gathers_tune_generic'41; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv4i32 = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> align 4 undef, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i32> zeroinitializer)42; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv8i32 = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> align 4 undef, <vscale x 8 x i1> %nxv8i1mask, <vscale x 8 x i32> zeroinitializer)43; CHECK-VSCALE-2-NEXT: Cost Model: Invalid cost for instruction: %res.nxv1i64 = call <vscale x 1 x i64> @llvm.masked.gather.nxv1i64.nxv1p0(<vscale x 1 x ptr> align 8 undef, <vscale x 1 x i1> %nxv1i1mask, <vscale x 1 x i64> zeroinitializer)44; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void45;46 %res.nxv4i32 = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32(<vscale x 4 x ptr> undef, i32 0, <vscale x 4 x i1> %nxv4i1mask, <vscale x 4 x i32> zeroinitializer)47 %res.nxv8i32 = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32(<vscale x 8 x ptr> undef, i32 0, <vscale x 8 x i1> %nxv8i1mask, <vscale x 8 x i32> zeroinitializer)48 %res.nxv1i64 = call <vscale x 1 x i64> @llvm.masked.gather.nxv1i64(<vscale x 1 x ptr> undef, i32 0, <vscale x 1 x i1> %nxv1i1mask, <vscale x 1 x i64> zeroinitializer)49 ret void50}51 52define void @masked_gathers_no_vscale_range() #2 {53; CHECK-VSCALE-1-LABEL: 'masked_gathers_no_vscale_range'54; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv4f64 = call <vscale x 4 x double> @llvm.masked.gather.nxv4f64.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> undef, <vscale x 4 x double> undef)55; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %res.nxv2f64 = call <vscale x 2 x double> @llvm.masked.gather.nxv2f64.nxv2p0(<vscale x 2 x ptr> align 1 undef, <vscale x 2 x i1> undef, <vscale x 2 x double> undef)56; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv8f32 = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> align 1 undef, <vscale x 8 x i1> undef, <vscale x 8 x float> undef)57; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv4f32 = call <vscale x 4 x float> @llvm.masked.gather.nxv4f32.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> undef, <vscale x 4 x float> undef)58; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %res.nxv2f32 = call <vscale x 2 x float> @llvm.masked.gather.nxv2f32.nxv2p0(<vscale x 2 x ptr> align 1 undef, <vscale x 2 x i1> undef, <vscale x 2 x float> undef)59; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 160 for instruction: %res.nxv16i16 = call <vscale x 16 x i16> @llvm.masked.gather.nxv16i16.nxv16p0(<vscale x 16 x ptr> align 1 undef, <vscale x 16 x i1> undef, <vscale x 16 x i16> undef)60; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv8i16 = call <vscale x 8 x i16> @llvm.masked.gather.nxv8i16.nxv8p0(<vscale x 8 x ptr> align 1 undef, <vscale x 8 x i1> undef, <vscale x 8 x i16> undef)61; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv4i16 = call <vscale x 4 x i16> @llvm.masked.gather.nxv4i16.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> undef, <vscale x 4 x i16> undef)62; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void63;64; CHECK-VSCALE-2-LABEL: 'masked_gathers_no_vscale_range'65; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv4f64 = call <vscale x 4 x double> @llvm.masked.gather.nxv4f64.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> undef, <vscale x 4 x double> undef)66; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv2f64 = call <vscale x 2 x double> @llvm.masked.gather.nxv2f64.nxv2p0(<vscale x 2 x ptr> align 1 undef, <vscale x 2 x i1> undef, <vscale x 2 x double> undef)67; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 160 for instruction: %res.nxv8f32 = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> align 1 undef, <vscale x 8 x i1> undef, <vscale x 8 x float> undef)68; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv4f32 = call <vscale x 4 x float> @llvm.masked.gather.nxv4f32.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> undef, <vscale x 4 x float> undef)69; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %res.nxv2f32 = call <vscale x 2 x float> @llvm.masked.gather.nxv2f32.nxv2p0(<vscale x 2 x ptr> align 1 undef, <vscale x 2 x i1> undef, <vscale x 2 x float> undef)70; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 320 for instruction: %res.nxv16i16 = call <vscale x 16 x i16> @llvm.masked.gather.nxv16i16.nxv16p0(<vscale x 16 x ptr> align 1 undef, <vscale x 16 x i1> undef, <vscale x 16 x i16> undef)71; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 160 for instruction: %res.nxv8i16 = call <vscale x 8 x i16> @llvm.masked.gather.nxv8i16.nxv8p0(<vscale x 8 x ptr> align 1 undef, <vscale x 8 x i1> undef, <vscale x 8 x i16> undef)72; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %res.nxv4i16 = call <vscale x 4 x i16> @llvm.masked.gather.nxv4i16.nxv4p0(<vscale x 4 x ptr> align 1 undef, <vscale x 4 x i1> undef, <vscale x 4 x i16> undef)73; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void74;75 %res.nxv4f64 = call <vscale x 4 x double> @llvm.masked.gather.nxv4f64(<vscale x 4 x ptr> undef, i32 1, <vscale x 4 x i1> undef, <vscale x 4 x double> undef)76 %res.nxv2f64 = call <vscale x 2 x double> @llvm.masked.gather.nxv2f64(<vscale x 2 x ptr> undef, i32 1, <vscale x 2 x i1> undef, <vscale x 2 x double> undef)77 78 %res.nxv8f32 = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32(<vscale x 8 x ptr> undef, i32 1, <vscale x 8 x i1> undef, <vscale x 8 x float> undef)79 %res.nxv4f32 = call <vscale x 4 x float> @llvm.masked.gather.nxv4f32(<vscale x 4 x ptr> undef, i32 1, <vscale x 4 x i1> undef, <vscale x 4 x float> undef)80 %res.nxv2f32 = call <vscale x 2 x float> @llvm.masked.gather.nxv2f32(<vscale x 2 x ptr> undef, i32 1, <vscale x 2 x i1> undef, <vscale x 2 x float> undef)81 82 %res.nxv16i16 = call <vscale x 16 x i16> @llvm.masked.gather.nxv16i16(<vscale x 16 x ptr> undef, i32 1, <vscale x 16 x i1> undef, <vscale x 16 x i16> undef)83 %res.nxv8i16 = call <vscale x 8 x i16> @llvm.masked.gather.nxv8i16(<vscale x 8 x ptr> undef, i32 1, <vscale x 8 x i1> undef, <vscale x 8 x i16> undef)84 %res.nxv4i16 = call <vscale x 4 x i16> @llvm.masked.gather.nxv4i16(<vscale x 4 x ptr> undef, i32 1, <vscale x 4 x i1> undef, <vscale x 4 x i16> undef)85 86 ret void87}88 89define <2 x i128> @masked_gather_v1i128(<2 x ptr> %ld, <2 x i1> %masks, <2 x i128> %passthru) #3 {90; CHECK-VSCALE-1-LABEL: 'masked_gather_v1i128'91; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <2 x i128> @llvm.masked.gather.v2i128.v2p0(<2 x ptr> align 16 %ld, <2 x i1> %masks, <2 x i128> %passthru)92; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i128> %res93;94; CHECK-VSCALE-2-LABEL: 'masked_gather_v1i128'95; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <2 x i128> @llvm.masked.gather.v2i128.v2p0(<2 x ptr> align 16 %ld, <2 x i1> %masks, <2 x i128> %passthru)96; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i128> %res97;98 %res = call <2 x i128> @llvm.masked.gather.v2i128.v2p0(<2 x ptr> %ld, i32 0, <2 x i1> %masks, <2 x i128> %passthru)99 ret <2 x i128> %res100}101 102attributes #0 = { "target-features"="+sve" vscale_range(1, 8) }103attributes #1 = { "target-features"="+sve" vscale_range(1, 16) "tune-cpu"="generic" }104attributes #2 = { "target-features"="+sve" }105attributes #3 = { "target-features"="+sve" vscale_range(2, 2) }106 107declare <vscale x 4 x i32> @llvm.masked.gather.nxv4i32(<vscale x 4 x ptr>, i32, <vscale x 4 x i1>, <vscale x 4 x i32>)108declare <vscale x 8 x i32> @llvm.masked.gather.nxv8i32(<vscale x 8 x ptr>, i32, <vscale x 8 x i1>, <vscale x 8 x i32>)109declare <vscale x 1 x i64> @llvm.masked.gather.nxv1i64(<vscale x 1 x ptr>, i32, <vscale x 1 x i1>, <vscale x 1 x i64>)110declare <vscale x 4 x i1> @llvm.masked.gather.nxv4i1(<vscale x 4 x ptr>, i32, <vscale x 4 x i1>, <vscale x 4 x i1>)111declare <vscale x 4 x double> @llvm.masked.gather.nxv4f64(<vscale x 4 x ptr>, i32, <vscale x 4 x i1>, <vscale x 4 x double>)112declare <vscale x 2 x double> @llvm.masked.gather.nxv2f64(<vscale x 2 x ptr>, i32, <vscale x 2 x i1>, <vscale x 2 x double>)113declare <vscale x 8 x float> @llvm.masked.gather.nxv8f32(<vscale x 8 x ptr>, i32, <vscale x 8 x i1>, <vscale x 8 x float>)114declare <vscale x 4 x float> @llvm.masked.gather.nxv4f32(<vscale x 4 x ptr>, i32, <vscale x 4 x i1>, <vscale x 4 x float>)115declare <vscale x 2 x float> @llvm.masked.gather.nxv2f32(<vscale x 2 x ptr>, i32, <vscale x 2 x i1>, <vscale x 2 x float>)116declare <vscale x 16 x i16> @llvm.masked.gather.nxv16i16(<vscale x 16 x ptr>, i32, <vscale x 16 x i1>, <vscale x 16 x i16>)117declare <vscale x 8 x i16> @llvm.masked.gather.nxv8i16(<vscale x 8 x ptr>, i32, <vscale x 8 x i1>, <vscale x 8 x i16>)118declare <vscale x 4 x i16> @llvm.masked.gather.nxv4i16(<vscale x 4 x ptr>, i32, <vscale x 4 x i1>, <vscale x 4 x i16>)119declare <2 x i128> @llvm.masked.gather.v2i128.v2p0(<2 x ptr>, i32, <2 x i1>, <2 x i128>)120