Skip to content

[SYCL] Replace spec constants 2020 with RT-provided values in sycl-post-link #3430

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Apr 2, 2021
Merged

[SYCL] Replace spec constants 2020 with RT-provided values in sycl-post-link #3430

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
8dfbdbe
[SYCL] Replace spec constants 2020 with RT-provided values in sycl-po…
vmaksimo Mar 26, 2021
6e55e60
Enable support for composite spec constants
vmaksimo Mar 26, 2021
6185598
Apply comments
vmaksimo Mar 30, 2021
d12ffc5
Fix format
vmaksimo Mar 30, 2021
029274b
Apply comments
vmaksimo Mar 31, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 41 additions & 1 deletion llvm/test/tools/sycl-post-link/composite-spec-constant-default-value.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
; RUN: sycl-post-link -spec-const=default --ir-output-only %s -S -o - \
; RUN: | FileCheck %s --implicit-check-not __sycl_getCompositeSpecConstantValue
; RUN: | FileCheck %s --implicit-check-not __sycl_getCompositeSpecConstantValue --implicit-check-not __sycl_getComposite2020SpecConstantValue
;
; This test checks that composite specialization constants can be correctly
; initialized by sycl-post-link tool for AOT use-case (default initialization
Expand All @@ -24,8 +24,10 @@ target triple = "spir64-unknown-unknown-sycldevice"
%"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id" = type { %"class._ZTSN2cl4sycl6detail5arrayILi1EEE.cl::sycl::detail::array" }

$_ZTS4Test = comdat any
$_ZTS17SpecializedKernel = comdat any

@__builtin_unique_stable_name._ZNK2cl4sycl6ONEAPI12experimental13spec_constantI3PODS4_E3getIS4_EENSt9enable_ifIXsr3std6is_podIT_EE5valueES8_E4typeEv = private unnamed_addr addrspace(1) constant [9 x i8] c"_ZTS3POD\00", align 1
@__builtin_unique_stable_name._ZNK2cl4sycl6ONEAPI12experimental13spec_constantI13MyComposConstE3getIS4_EENSt9enable_ifIXaasr3std8is_classIT_EE5valuesr3std6is_podIS9_EE5valueES9_E4typeEv = private unnamed_addr addrspace(1) constant [20 x i8] c"_ZTS13MyComposConst\00", align 1

; Function Attrs: convergent norecurse uwtable
define weak_odr dso_local spir_kernel void @_ZTS4Test(%struct._ZTS3POD.POD addrspace(1)* %_arg_, %"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range"* byval(%"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range") align 8 %_arg_1, %"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range"* byval(%"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range") align 8 %_arg_2, %"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id"* byval(%"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id") align 8 %_arg_3) local_unnamed_addr #0 comdat !kernel_arg_buffer_location !4 {
Expand All @@ -45,6 +47,38 @@ entry:
ret void
}

; Function Attrs: convergent norecurse
define weak_odr dso_local spir_kernel void @_ZTS17SpecializedKernel(float addrspace(1)* %_arg_, %"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range"* byval(%"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range") align 8 %_arg_1, %"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range"* byval(%"class._ZTSN2cl4sycl5rangeILi1EEE.cl::sycl::range") align 8 %_arg_2, %"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id"* byval(%"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id") align 8 %_arg_3) local_unnamed_addr #0 comdat !kernel_arg_buffer_location !4 {
entry:
%c.i = alloca %struct._ZTS1A.A, align 4
%0 = getelementptr inbounds %"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id", %"class._ZTSN2cl4sycl2idILi1EEE.cl::sycl::id"* %_arg_3, i64 0, i32 0, i32 0, i64 0
%1 = addrspacecast i64* %0 to i64 addrspace(4)*
%2 = load i64, i64 addrspace(4)* %1, align 8
%add.ptr.i = getelementptr inbounds float, float addrspace(1)* %_arg_, i64 %2
; CHECK: %[[CAST1:[0-9a-z.]+]] = addrspacecast %struct._ZTS1A.A* %c.i
%c.ascast.i = addrspacecast %struct._ZTS1A.A* %c.i to %struct._ZTS1A.A addrspace(4)*
%3 = bitcast %struct._ZTS1A.A* %c.i to i8*
call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull %3) #3
%a.i.i = alloca i32, align 4
%bc = bitcast i32* %a.i.i to i8*
%tmp = addrspacecast i8* %bc to i8 addrspace(4)*
call spir_func void @_Z40__sycl_getComposite2020SpecConstantValueI13MyComposConstET_PKcPvS4_(%struct._ZTS1A.A addrspace(4)* sret(%struct._ZTS1A.A) align 4 %c.ascast.i, i8 addrspace(4)* getelementptr inbounds ([20 x i8], [20 x i8] addrspace(4)* addrspacecast ([20 x i8] addrspace(1)* @__builtin_unique_stable_name._ZNK2cl4sycl6ONEAPI12experimental13spec_constantI13MyComposConstE3getIS4_EENSt9enable_ifIXaasr3std8is_classIT_EE5valuesr3std6is_podIS9_EE5valueES9_E4typeEv to [20 x i8] addrspace(4)*), i64 0, i64 0), i8 addrspace(4)* null, i8 addrspace(4)* %tmp) #4
; CHECK: %[[GEP:[0-9a-z]+]] = getelementptr i8, i8 addrspace(4)* %tmp, i32 0
; CHECK: %[[BITCAST:[0-9a-z]+]] = bitcast i8 addrspace(4)* %[[GEP]] to %struct._ZTS1A.A addrspace(4)*
; CHECK: %[[LOAD:[0-9a-z]+]] = load %struct._ZTS1A.A, %struct._ZTS1A.A addrspace(4)* %[[BITCAST]], align 4
; CHECK: store %struct._ZTS1A.A %[[LOAD]], %struct._ZTS1A.A {{.*}}* %[[CAST1]]
%a.i = getelementptr inbounds %struct._ZTS1A.A, %struct._ZTS1A.A addrspace(4)* %c.ascast.i, i64 0, i32 0
%4 = load i32, i32 addrspace(4)* %a.i, align 4
%conv.i = sitofp i32 %4 to float
%b.i = getelementptr inbounds %struct._ZTS1A.A, %struct._ZTS1A.A addrspace(4)* %c.ascast.i, i64 0, i32 1
%5 = load float, float addrspace(4)* %b.i, align 4
%add.i = fadd float %5, %conv.i
%ptridx.ascast.i.i = addrspacecast float addrspace(1)* %add.ptr.i to float addrspace(4)*
store float %add.i, float addrspace(4)* %ptridx.ascast.i.i, align 4, !tbaa !11
call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull %3) #3
ret void
}

; Function Attrs: argmemonly nounwind willreturn
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #1

Expand All @@ -57,6 +91,9 @@ declare void @llvm.memcpy.p4i8.p0i8.i64(i8 addrspace(4)* noalias nocapture write
; Function Attrs: convergent
declare dso_local spir_func void @_Z36__sycl_getCompositeSpecConstantValueI3PODET_PKc(%struct._ZTS3POD.POD addrspace(4)* sret(%struct._ZTS3POD.POD) align 8, i8 addrspace(4)*) local_unnamed_addr #2

; Function Attrs: convergent
declare dso_local spir_func void @_Z40__sycl_getComposite2020SpecConstantValueI13MyComposConstET_PKcPvS4_(%struct._ZTS1A.A addrspace(4)* sret(%struct._ZTS1A.A) align 4, i8 addrspace(4)*, i8 addrspace(4)*, i8 addrspace(4)*) local_unnamed_addr #2

attributes #0 = { convergent norecurse uwtable "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "sycl-module-id"="../sycl/test/spec_const/composite.cpp" "tune-cpu"="generic" "uniform-work-group-size"="true" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { argmemonly nounwind willreturn }
attributes #2 = { convergent "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "tune-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
Expand All @@ -77,3 +114,6 @@ attributes #4 = { convergent }
!6 = !{!7, !7, i64 0}
!7 = !{!"omnipotent char", !8, i64 0}
!8 = !{!"Simple C++ TBAA"}
!9 = !{!"int", !7, i64 0}
!10 = !{!"float", !7, i64 0}
!11 = !{!10, !10, i64 0}
93 changes: 93 additions & 0 deletions llvm/test/tools/sycl-post-link/spec_const_2020.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,93 @@
; RUN: sycl-post-link --ir-output-only -spec-const=default %s -S -o - | \
; RUN: FileCheck %s -check-prefixes=CHECK,CHECK-DEF

; This test checks that the post link tool is able to correctly transform
; specialization constant intrinsics for different types in a device code
; compiled for AOT.

%class.specialization_id = type { double }
%class.specialization_id.0 = type { i32 }
%class.specialization_id.1 = type { %struct.ComposConst }
%struct.ComposConst = type { i32, double, %struct.myConst }
%struct.myConst = type { i32, float }
%class.specialization_id.2 = type { %struct.ComposConst2 }
%struct.ComposConst2 = type { i8, %struct.myConst, double }

@id_double = dso_local global %class.specialization_id { double 3.140000e+00 }, align 8
@id_int = dso_local global %class.specialization_id.0 { i32 42 }, align 4
@id_compos = dso_local global %class.specialization_id.1 { %struct.ComposConst { i32 1, double 2.000000e+00, %struct.myConst { i32 13, float 0x4020666660000000 } } }, align 8
@id_compos2 = dso_local global %class.specialization_id.2 { %struct.ComposConst2 { i8 1, %struct.myConst { i32 52, float 0x40479999A0000000 }, double 2.000000e+00 } }, align 8

; check that no longer used globals are removed:
@__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_doubleE17specialization_idIdEdET1_v = private unnamed_addr constant [37 x i8] c"_ZTS14name_generatorIL_Z9id_doubleEE\00", align 1
; CHECK-NOT: @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_doubleE17specialization_idIdEdET1_v
@__builtin_unique_stable_name._Z27get_specialization_constantIL_Z6id_intE17specialization_idIiEiET1_v = private unnamed_addr constant [34 x i8] c"_ZTS14name_generatorIL_Z6id_intEE\00", align 1
; CHECK-NOT: @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z6id_intE17specialization_idIiEiET1_v
@__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_composE17specialization_idI11ComposConstES1_ET1_v = private unnamed_addr constant [37 x i8] c"_ZTS14name_generatorIL_Z9id_composEE\00", align 1
; CHECK-NOT: @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_composE17specialization_idI11ComposConstES1_ET1_v
@__builtin_unique_stable_name._Z27get_specialization_constantIL_Z10id_compos2E17specialization_idI12ComposConst2ES1_ET1_v = private unnamed_addr constant [39 x i8] c"_ZTS14name_generatorIL_Z10id_compos2EE\00", align 1
; CHECK-NOT: @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z10id_compos2E17specialization_idI12ComposConst2ES1_ET1_v

define dso_local void @_Z4testv() local_unnamed_addr #0 {
entry:
%call.i = tail call fast double @_Z37__sycl_getScalar2020SpecConstantValueIdET_PKcPvS3_(i8* getelementptr inbounds ([37 x i8], [37 x i8]* @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_doubleE17specialization_idIdEdET1_v, i64 0, i64 0), i8* bitcast (%class.specialization_id* @id_double to i8*), i8* null)
; CHECK-DEF: %[[GEP:[0-9a-z]+]] = getelementptr i8, i8* null, i32 0
; CHECK-DEF: %[[BITCAST:[0-9a-z]+]] = bitcast i8* %[[GEP]] to double*
; CHECK-DEF: %[[LOAD:[0-9a-z]+]] = load double, double* %[[BITCAST]], align 8
%call.i3 = tail call i32 @_Z37__sycl_getScalar2020SpecConstantValueIiET_PKcPvS3_(i8* getelementptr inbounds ([34 x i8], [34 x i8]* @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z6id_intE17specialization_idIiEiET1_v, i64 0, i64 0), i8* bitcast (%class.specialization_id.0* @id_int to i8*), i8* null)
; CHECK-DEF: %[[GEP1:[0-9a-z]+]] = getelementptr i8, i8* null, i32 8
; CHECK-DEF: %[[BITCAST1:[0-9a-z]+]] = bitcast i8* %[[GEP1]] to i32*
; CHECK-DEF: %[[LOAD1:[0-9a-z]+]] = load i32, i32* %[[BITCAST1]], align 4
%call.i4 = tail call fast double @_Z37__sycl_getScalar2020SpecConstantValueIdET_PKcPvS3_(i8* getelementptr inbounds ([37 x i8], [37 x i8]* @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_doubleE17specialization_idIdEdET1_v, i64 0, i64 0), i8* bitcast (%class.specialization_id* @id_double to i8*), i8* null)
; CHECK-DEF: %[[GEP2:[0-9a-z]+]] = getelementptr i8, i8* null, i32 0
; CHECK-DEF: %[[BITCAST2:[0-9a-z]+]] = bitcast i8* %[[GEP2]] to double*
; CHECK-DEF: %[[LOAD2:[0-9a-z]+]] = load double, double* %[[BITCAST2]], align 8
ret void
}

define dso_local void @_Z5test2v() local_unnamed_addr #0 {
entry:
%tmp = alloca %struct.ComposConst, align 8
%tmp1 = alloca %struct.ComposConst2, align 8
%tmp2 = alloca %struct.ComposConst, align 8
%0 = bitcast %struct.ComposConst* %tmp to i8*
call void @llvm.lifetime.start.p0i8(i64 24, i8* nonnull %0) #3
call void @_Z40__sycl_getComposite2020SpecConstantValueI11ComposConstET_PKcPvS4_(%struct.ComposConst* nonnull sret(%struct.ComposConst) align 8 %tmp, i8* getelementptr inbounds ([37 x i8], [37 x i8]* @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_composE17specialization_idI11ComposConstES1_ET1_v, i64 0, i64 0), i8* bitcast (%class.specialization_id.1* @id_compos to i8*), i8* null)
; CHECK: %[[GEP:[0-9a-z]+]] = getelementptr i8, i8* null, i32 12
; CHECK: %[[BITCAST:[0-9a-z]+]] = bitcast i8* %[[GEP]] to %struct.ComposConst*
; CHECK: %[[LOAD:[0-9a-z]+]] = load %struct.ComposConst, %struct.ComposConst* %[[BITCAST]], align 8
; CHECK: store %struct.ComposConst %[[LOAD]], %struct.ComposConst*
call void @llvm.lifetime.end.p0i8(i64 24, i8* nonnull %0) #3
%1 = getelementptr inbounds %struct.ComposConst2, %struct.ComposConst2* %tmp1, i64 0, i32 0
call void @llvm.lifetime.start.p0i8(i64 24, i8* nonnull %1) #3
call void @_Z40__sycl_getComposite2020SpecConstantValueI12ComposConst2ET_PKcPvS4_(%struct.ComposConst2* nonnull sret(%struct.ComposConst2) align 8 %tmp1, i8* getelementptr inbounds ([39 x i8], [39 x i8]* @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z10id_compos2E17specialization_idI12ComposConst2ES1_ET1_v, i64 0, i64 0), i8* getelementptr inbounds (%class.specialization_id.2, %class.specialization_id.2* @id_compos2, i64 0, i32 0, i32 0), i8* null) call void @llvm.lifetime.end.p0i8(i64 24, i8* nonnull %1) #3
; CHECK: %[[GEP1:[0-9a-z]+]] = getelementptr i8, i8* null, i32 36
; CHECK: %[[BITCAST1:[0-9a-z]+]] = bitcast i8* %[[GEP1]] to %struct.ComposConst2*
; CHECK: %[[LOAD1:[0-9a-z]+]] = load %struct.ComposConst2, %struct.ComposConst2* %[[BITCAST1]], align 8
; CHECK: store %struct.ComposConst2 %[[LOAD1]], %struct.ComposConst2*
%2 = bitcast %struct.ComposConst* %tmp2 to i8*
call void @llvm.lifetime.start.p0i8(i64 24, i8* nonnull %2) #3
call void @_Z40__sycl_getComposite2020SpecConstantValueI11ComposConstET_PKcPvS4_(%struct.ComposConst* nonnull sret(%struct.ComposConst) align 8 %tmp2, i8* getelementptr inbounds ([37 x i8], [37 x i8]* @__builtin_unique_stable_name._Z27get_specialization_constantIL_Z9id_composE17specialization_idI11ComposConstES1_ET1_v, i64 0, i64 0), i8* bitcast (%class.specialization_id.1* @id_compos to i8*), i8* null)
; CHECK: %[[GEP2:[0-9a-z]+]] = getelementptr i8, i8* null, i32 12
; CHECK: %[[BITCAST2:[0-9a-z]+]] = bitcast i8* %[[GEP2]] to %struct.ComposConst*
; CHECK: %[[LOAD2:[0-9a-z]+]] = load %struct.ComposConst, %struct.ComposConst* %[[BITCAST2]], align 8
; CHECK: store %struct.ComposConst %[[LOAD2]], %struct.ComposConst*
call void @llvm.lifetime.end.p0i8(i64 24, i8* nonnull %2) #3
ret void
}
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #1

declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture) #1

declare dso_local double @_Z37__sycl_getScalar2020SpecConstantValueIdET_PKcPvS3_(i8*, i8*, i8*) local_unnamed_addr #1

declare dso_local i32 @_Z37__sycl_getScalar2020SpecConstantValueIiET_PKcPvS3_(i8*, i8*, i8*) local_unnamed_addr #1

declare dso_local void @_Z40__sycl_getComposite2020SpecConstantValueI11ComposConstET_PKcPvS4_(%struct.ComposConst* sret(%struct.ComposConst) align 8, i8*, i8*, i8*) local_unnamed_addr #2

declare dso_local void @_Z40__sycl_getComposite2020SpecConstantValueI12ComposConst2ET_PKcPvS4_(%struct.ComposConst2* sret(%struct.ComposConst2) align 8, i8*, i8*, i8*) local_unnamed_addr #2

attributes #0 = { uwtable mustprogress "denormal-fp-math"="preserve-sign,preserve-sign" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="true" "no-jump-tables"="false" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="true" "use-soft-float"="false" }
attributes #1 = { argmemonly nofree nosync nounwind willreturn }
attributes #2 = { "denormal-fp-math"="preserve-sign,preserve-sign" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="true" "use-soft-float"="false" }
attributes #3 = { nounwind }
5 changes: 4 additions & 1 deletion llvm/test/tools/sycl-post-link/spec_const_O2.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,7 +116,10 @@ define weak_odr dso_local spir_kernel void @_ZTS17SpecializedKernel(float addrsp
%37 = tail call spir_func double @_Z37__sycl_getScalar2020SpecConstantValueIdET_PKcPvS3_(i8 addrspace(4)* addrspacecast (i8* getelementptr inbounds ([21 x i8], [21 x i8]* @__unique_stable_name.SC_Id14MyDoubleConst2E3getEv, i64 0, i64 0) to i8 addrspace(4)*), i8 addrspace(4)* null, i8 addrspace(4)* null)
; CHECK-RT: %{{[0-9]+}} = call double @_Z20__spirv_SpecConstantid(i32 11, double 0.000000e+00), !SYCL_SPEC_CONST_SYM_ID ![[ID11:[0-9]+]]
%38 = fadd double %37, %36
; CHECK-DEF: %[[SUM10:[0-9]+]] = fadd double 0.000000e+00, %[[SUM9]]
; CHECK-DEF: %[[GEP:[0-9a-z]+]] = getelementptr i8, i8 addrspace(4)* null, i32 0
; CHECK-DEF: %[[BITCAST:[0-9a-z]+]] = bitcast i8 addrspace(4)* %[[GEP]] to double addrspace(4)*
; CHECK-DEF: %[[LOAD:[0-9a-z]+]] = load double, double addrspace(4)* %[[BITCAST]], align 8
; CHECK-DEF: %[[SUM10:[0-9]+]] = fadd double %[[LOAD]], %[[SUM9]]
%39 = fptrunc double %38 to float
; CHECK-DEF: %[[VAL8:[0-9]+]] = fptrunc double %[[SUM10]] to float
%40 = addrspacecast float addrspace(1)* %7 to float addrspace(4)*
Expand Down
Loading