maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/OpenMP/target_codegen.cpp

769 lines
44 KiB
C++
Raw Normal View History

[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
// Test host codegen.
[OpenMP] Use fopenmp prefix for all options introduced by the offloading implementation. Summary: This patch changes the options used by offloading to start with -fopenmp instead of -fomp. This makes the option naming more consistent and materializes a suggestion by Richard Smith in http://reviews.llvm.org/D9888. Reviewers: hfinkel, carlo.bertolli, arpith-jacob, ABataev Subscribers: kkwli0, cfe-commits, caomhin Differential Revision: http://reviews.llvm.org/D21841 llvm-svn: 274283
2016-07-01 05:22:08 +08:00
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
[OPENMP] Support for -fopenmp-simd option with compilation of simd loops only. Added support for -fopenmp-simd option that allows compilation of simd-based constructs without emission of OpenMP runtime calls. llvm-svn: 321560
2017-12-30 02:07:07 +08:00
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
// Test target codegen - host bc file has to be created first.
[OpenMP] Use fopenmp prefix for all options introduced by the offloading implementation. Summary: This patch changes the options used by offloading to start with -fopenmp instead of -fomp. This makes the option naming more consistent and materializes a suggestion by Richard Smith in http://reviews.llvm.org/D9888. Reviewers: hfinkel, carlo.bertolli, arpith-jacob, ABataev Subscribers: kkwli0, cfe-commits, caomhin Differential Revision: http://reviews.llvm.org/D21841 llvm-svn: 274283
2016-07-01 05:22:08 +08:00
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix TCHECK --check-prefix TCHECK-64
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix TCHECK --check-prefix TCHECK-64
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix TCHECK --check-prefix TCHECK-32
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix TCHECK --check-prefix TCHECK-32
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
[OPENMP] Support for -fopenmp-simd option with compilation of simd loops only. Added support for -fopenmp-simd option that allows compilation of simd-based constructs without emission of OpenMP runtime calls. llvm-svn: 321560
2017-12-30 02:07:07 +08:00
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// SIMD-ONLY1-NOT: {{__kmpc|__tgt}}
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// expected-no-diagnostics
#ifndef HEADER
#define HEADER
// CHECK-DAG: [[TT:%.+]] = type { i64, i8 }
// CHECK-DAG: [[S1:%.+]] = type { double }
[OpenMP] Add fields for flags in the offload entry descriptor. Summary: This patch adds two fields to the offload entry descriptor. One field is meant to signal Ctors/Dtors and `link` global variables, and the other is reserved for runtime library use. Currently, these fields are only filled with zeros in the current code generation, but that will change when `declare target` is added. The reason, we are adding these fields now is to make the code generation consistent with the runtime library proposal under review in https://reviews.llvm.org/D14031. Reviewers: ABataev, hfinkel, carlo.bertolli, kkwli0, arpith-jacob, Hahnfeld Subscribers: cfe-commits, caomhin, jholewinski Differential Revision: https://reviews.llvm.org/D28298 llvm-svn: 291124
2017-01-06 00:02:49 +08:00
// CHECK-DAG: [[ENTTY:%.+]] = type { i8*, i8*, i[[SZ:32|64]], i32, i32 }
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
// CHECK-DAG: [[DEVTY:%.+]] = type { i8*, i8*, [[ENTTY]]*, [[ENTTY]]* }
// CHECK-DAG: [[DSCTY:%.+]] = type { i32, [[DEVTY]]*, [[ENTTY]]*, [[ENTTY]]* }
[OpenMP] Add fields for flags in the offload entry descriptor. Summary: This patch adds two fields to the offload entry descriptor. One field is meant to signal Ctors/Dtors and `link` global variables, and the other is reserved for runtime library use. Currently, these fields are only filled with zeros in the current code generation, but that will change when `declare target` is added. The reason, we are adding these fields now is to make the code generation consistent with the runtime library proposal under review in https://reviews.llvm.org/D14031. Reviewers: ABataev, hfinkel, carlo.bertolli, kkwli0, arpith-jacob, Hahnfeld Subscribers: cfe-commits, caomhin, jholewinski Differential Revision: https://reviews.llvm.org/D28298 llvm-svn: 291124
2017-01-06 00:02:49 +08:00
// TCHECK: [[ENTTY:%.+]] = type { i8*, i8*, i{{32|64}}, i32, i32 }
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OpenMP] Create COMDAT group for OpenMP offload registration code to avoid multiple copies Thanks to Sergey Dmitriev for submitting the patch. Differential Revision: https://reviews.llvm.org/D33509 llvm-svn: 304056
2017-05-27 11:03:13 +08:00
// CHECK-DAG: $[[REGFN:\.omp_offloading\..+]] = comdat
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// We have 8 target regions, but only 7 that actually will generate offloading
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// code and have mapped arguments, and only 5 have all-constant map sizes.
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Do not capture local static variables. Previously we may erroneously try to capture locally declared static variables, which will lead to crash for target-based constructs. Patch fixes this problem. llvm-svn: 315076
2017-10-07 01:00:28 +08:00
// CHECK-DAG: [[SIZET:@.+]] = private unnamed_addr constant [2 x i[[SZ]]] [i[[SZ]] 0, i[[SZ]] 4]
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[MAPT:@.+]] = private unnamed_addr constant [2 x i64] [i64 544, i64 800]
[OPENMP] Do not capture local static variables. Previously we may erroneously try to capture locally declared static variables, which will lead to crash for target-based constructs. Patch fixes this problem. llvm-svn: 315076
2017-10-07 01:00:28 +08:00
// CHECK-DAG: [[SIZET2:@.+]] = private unnamed_addr constant [1 x i{{32|64}}] [i[[SZ]] 2]
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[MAPT2:@.+]] = private unnamed_addr constant [1 x i64] [i64 800]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[SIZET3:@.+]] = private unnamed_addr constant [2 x i[[SZ]]] [i[[SZ]] 4, i[[SZ]] 2]
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[MAPT3:@.+]] = private unnamed_addr constant [2 x i64] [i64 800, i64 800]
// CHECK-DAG: [[MAPT4:@.+]] = private unnamed_addr constant [9 x i64] [i64 800, i64 547, i64 288, i64 547, i64 547, i64 288, i64 288, i64 547, i64 547]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[SIZET5:@.+]] = private unnamed_addr constant [3 x i[[SZ]]] [i[[SZ]] 4, i[[SZ]] 2, i[[SZ]] 40]
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[MAPT5:@.+]] = private unnamed_addr constant [3 x i64] [i64 800, i64 800, i64 547]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[SIZET6:@.+]] = private unnamed_addr constant [4 x i[[SZ]]] [i[[SZ]] 4, i[[SZ]] 2, i[[SZ]] 1, i[[SZ]] 40]
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[MAPT6:@.+]] = private unnamed_addr constant [4 x i64] [i64 800, i64 800, i64 800, i64 547]
// CHECK-DAG: [[MAPT7:@.+]] = private unnamed_addr constant [6 x i64] [i64 32, i64 281474976711171, i64 800, i64 288, i64 288, i64 547]
[OPENMP, NVPTX] Fix linkage of the global entries. The linkage of the global entries must be weak to enable support of redefinition of the same target regions in multiple compilation units. llvm-svn: 331768
2018-05-08 22:16:57 +08:00
// CHECK-DAG: @{{.*}} = weak constant i8 0
// CHECK-DAG: @{{.*}} = weak constant i8 0
// CHECK-DAG: @{{.*}} = weak constant i8 0
// CHECK-DAG: @{{.*}} = weak constant i8 0
// CHECK-DAG: @{{.*}} = weak constant i8 0
// CHECK-DAG: @{{.*}} = weak constant i8 0
// CHECK-DAG: @{{.*}} = weak constant i8 0
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Added emission of offloading data sections for declare target variables. Added emission of the offloading data sections for the variables within declare target regions + fixes emission of the declare target variables marked as declare target not within the declare target region. llvm-svn: 328888
2018-03-31 02:31:07 +08:00
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
// TCHECK: @{{.+}} = weak constant [[ENTTY]]
[OPENMP] Added emission of offloading data sections for declare target variables. Added emission of the offloading data sections for the variables within declare target regions + fixes emission of the declare target variables marked as declare target not within the declare target region. llvm-svn: 328888
2018-03-31 02:31:07 +08:00
// TCHECK-NOT: @{{.+}} = weak constant [[ENTTY]]
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
// Check if offloading descriptor is created.
// CHECK: [[ENTBEGIN:@.+]] = external constant [[ENTTY]]
// CHECK: [[ENTEND:@.+]] = external constant [[ENTTY]]
[OPENMP] Change linkage of offloading symbols to support dropping offload targets. Changed the linkage of omp_offloading.img_start.<triple> and omp_offloading.img_end.<triple> symbols from external to external weak to allow dropping of some targets during linking. llvm-svn: 338413
2018-08-01 02:27:42 +08:00
// CHECK: [[DEVBEGIN:@.+]] = extern_weak constant i8
// CHECK: [[DEVEND:@.+]] = extern_weak constant i8
[OpenMP] Create COMDAT group for OpenMP offload registration code to avoid multiple copies Thanks to Sergey Dmitriev for submitting the patch. Differential Revision: https://reviews.llvm.org/D33509 llvm-svn: 304056
2017-05-27 11:03:13 +08:00
// CHECK: [[IMAGES:@.+]] = internal unnamed_addr constant [1 x [[DEVTY]]] [{{.+}} { i8* [[DEVBEGIN]], i8* [[DEVEND]], [[ENTTY]]* [[ENTBEGIN]], [[ENTTY]]* [[ENTEND]] }], comdat($[[REGFN]])
// CHECK: [[DESC:@.+]] = internal constant [[DSCTY]] { i32 1, [[DEVTY]]* getelementptr inbounds ([1 x [[DEVTY]]], [1 x [[DEVTY]]]* [[IMAGES]], i32 0, i32 0), [[ENTTY]]* [[ENTBEGIN]], [[ENTTY]]* [[ENTEND]] }, comdat($[[REGFN]])
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
// Check target registration is registered as a Ctor.
[OPENMP] Added emission of offloading data sections for declare target variables. Added emission of the offloading data sections for the variables within declare target regions + fixes emission of the declare target variables marked as declare target not within the declare target region. llvm-svn: 328888
2018-03-31 02:31:07 +08:00
// CHECK: appending global [1 x { i32, void ()*, i8* }] [{ i32, void ()*, i8* } { i32 0, void ()* @[[REGFN]], i8* bitcast (void ()* @[[REGFN]] to i8*) }]
[OpenMP] Reapply rL256842: [OpenMP] Offloading descriptor registration and device codegen. This patch attempts to fix the regressions identified when the patch was committed initially. Thanks to Michael Liao for identifying the fix in the offloading metadata generation related with side effects in evaluation of function arguments. llvm-svn: 256933
2016-01-06 21:42:12 +08:00
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
template<typename tx, typename ty>
struct TT{
tx X;
ty Y;
};
[OPENMP] Support for re-declarations when checking captured variables. Need to check for variables re-declarations when checking that the variable was already captured in the captured region. llvm-svn: 313805
2017-09-21 04:11:31 +08:00
int global;
extern int global;
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: define {{.*}}[[FOO:@.+]](
int foo(int n) {
int a = 0;
short aa = 0;
float b[10];
float bn[n];
double c[5][10];
double cn[5][n];
TT<long long, char> d;
[OPENMP] Do not capture local static variables. Previously we may erroneously try to capture locally declared static variables, which will lead to crash for target-based constructs. Patch fixes this problem. llvm-svn: 315076
2017-10-07 01:00:28 +08:00
static long *plocal;
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Capture argument of `device` clause for target-based directives. The argument of the `device` clause in target-based executable directives must be captured to support codegen for the `target` directives with the `depend` clauses. llvm-svn: 314686
2017-10-03 00:32:39 +08:00
// CHECK: [[ADD:%.+]] = add nsw i32
[OPENMP] Add codegen for `depend` clauses on `target` directive. Added basic support for codegen of `depend` clauses on `target` directive. llvm-svn: 322501
2018-01-16 03:06:12 +08:00
// CHECK: store i32 [[ADD]], i32* [[DEVICE_CAP:%.+]],
// CHECK: [[DEV:%.+]] = load i32, i32* [[DEVICE_CAP]],
// CHECK: [[DEVICE:%.+]] = sext i32 [[DEV]] to i64
[Clang][OpenMP] New clang/libomptarget map interface: new function signatures, clang-side This clang patch changes the __tgt_* API function signatures in preparation for the new map interface. Changes are: Device IDs 32bits --> 64bits, Flags 32bits --> 64bits Differential revision: https://reviews.llvm.org/D40281 llvm-svn: 318789
2017-11-22 02:25:12 +08:00
// CHECK: [[RET:%.+]] = call i32 @__tgt_target(i64 [[DEVICE]], i8* @{{[^,]+}}, i32 0, i8** null, i8** null, i[[SZ]]* null, i64* null)
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK-NEXT: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:[^,]+]], label %[[END:[^,]+]]
// CHECK: [[FAIL]]
// CHECK: call void [[HVT0:@.+]]()
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Capture argument of `device` clause for target-based directives. The argument of the `device` clause in target-based executable directives must be captured to support codegen for the `target` directives with the `depend` clauses. llvm-svn: 314686
2017-10-03 00:32:39 +08:00
#pragma omp target device(global + a)
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
{
}
[OPENMP] Generalize capturing of clauses expressions. The handling and capturing of the non-constant expressions of some of the capturable clauses in combined directives is generalized. llvm-svn: 319227
2017-11-29 05:11:44 +08:00
// CHECK-DAG: [[ADD:%.+]] = add nsw i32
[OPENMP] Add codegen for `depend` clauses on `target` directive. Added basic support for codegen of `depend` clauses on `target` directive. llvm-svn: 322501
2018-01-16 03:06:12 +08:00
// CHECK-DAG: store i32 [[ADD]], i32* [[DEVICE_CAP:%.+]],
// CHECK-DAG: [[DEV:%.+]] = load i32, i32* [[DEVICE_CAP]],
// CHECK-DAG: [[DEVICE:%.+]] = sext i32 [[DEV]] to i64
[OPENMP] Add codegen for `nowait` clause in target directives. Added basic codegen for `nowait` clauses in target-based directives. llvm-svn: 320613
2017-12-14 05:04:20 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target_nowait(i64 [[DEVICE]], i8* @{{[^,]+}}, i32 2, i8** [[BPR:%[^,]+]], i8** [[PR:%[^,]+]], i[[SZ]]* getelementptr inbounds ([2 x i[[SZ]]], [2 x i[[SZ]]]* [[SIZET]], i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* [[MAPT]], i32 0, i32 0)
[OPENMP] Do not capture local static variables. Previously we may erroneously try to capture locally declared static variables, which will lead to crash for target-based constructs. Patch fixes this problem. llvm-svn: 315076
2017-10-07 01:00:28 +08:00
// CHECK-DAG: [[BPR]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[PR]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[P:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP]], i32 0, i32 0
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[P]], i32 0, i32 0
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to i[[SZ]]**
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to i[[SZ]]**
// CHECK-DAG: store i[[SZ]]* [[BP0:%[^,]+]], i[[SZ]]** [[CBPADDR0]]
// CHECK-DAG: store i[[SZ]]* [[BP0]], i[[SZ]]** [[CPADDR0]]
// CHECK-DAG: [[BPADDR1:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP]], i32 0, i32 1
// CHECK-DAG: [[PADDR1:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[P]], i32 0, i32 1
// CHECK-DAG: [[CBPADDR1:%.+]] = bitcast i8** [[BPADDR1]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR1:%.+]] = bitcast i8** [[PADDR1]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[BP1:%[^,]+]], i[[SZ]]* [[CBPADDR1]]
// CHECK-DAG: store i[[SZ]] [[BP1]], i[[SZ]]* [[CPADDR1]]
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:[^,]+]], label %[[END:[^,]+]]
// CHECK: [[FAIL]]
// CHECK: call void [[HVT0_:@.+]](i[[SZ]]* [[BP0]], i[[SZ]] [[BP1]])
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Add codegen for `nowait` clause in target directives. Added basic codegen for `nowait` clauses in target-based directives. llvm-svn: 320613
2017-12-14 05:04:20 +08:00
#pragma omp target device(global + a) nowait
[OPENMP] Do not capture local static variables. Previously we may erroneously try to capture locally declared static variables, which will lead to crash for target-based constructs. Patch fixes this problem. llvm-svn: 315076
2017-10-07 01:00:28 +08:00
{
static int local1;
*plocal = global;
local1 = global;
}
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: call void [[HVT1:@.+]](i[[SZ]] {{[^,]+}})
[OPENMP] Handle re-declaration of captured variables in CodeGen. If the captured variable has re-declaration we may end up with the situation where the captured variable is the re-declaration while the referenced variable is the canonical declaration (or vice versa). In this case we may generate wrong code. Patch fixes this situation. llvm-svn: 313995
2017-09-23 00:56:13 +08:00
#pragma omp target if(0) firstprivate(global)
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
{
[OPENMP] Support for re-declarations when checking captured variables. Need to check for variables re-declarations when checking that the variable was already captured in the captured region. llvm-svn: 313805
2017-09-21 04:11:31 +08:00
global += 1;
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
}
[Clang][OpenMP] New clang/libomptarget map interface: new function signatures, clang-side This clang patch changes the __tgt_* API function signatures in preparation for the new map interface. Changes are: Device IDs 32bits --> 64bits, Flags 32bits --> 64bits Differential revision: https://reviews.llvm.org/D40281 llvm-svn: 318789
2017-11-22 02:25:12 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target(i64 -1, i8* @{{[^,]+}}, i32 1, i8** [[BP:%[^,]+]], i8** [[P:%[^,]+]], i[[SZ]]* getelementptr inbounds ([1 x i[[SZ]]], [1 x i[[SZ]]]* [[SIZET2]], i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* [[MAPT2]], i32 0, i32 0))
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BP]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[BPR:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[P]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[PR:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[BPR]], i32 0, i32 [[IDX0:[0-9]+]]
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[PR]], i32 0, i32 [[IDX0]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[BP0:%[^,]+]], i[[SZ]]* [[CBPADDR0]]
// CHECK-DAG: store i[[SZ]] [[P0:%[^,]+]], i[[SZ]]* [[CPADDR0]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:[^,]+]], label %[[END:[^,]+]]
// CHECK: [[FAIL]]
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: call void [[HVT2:@.+]](i[[SZ]] {{[^,]+}})
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
#pragma omp target if(1)
{
aa += 1;
}
// CHECK: [[IF:%.+]] = icmp sgt i32 {{[^,]+}}, 10
// CHECK: br i1 [[IF]], label %[[IFTHEN:[^,]+]], label %[[IFELSE:[^,]+]]
// CHECK: [[IFTHEN]]
[Clang][OpenMP] New clang/libomptarget map interface: new function signatures, clang-side This clang patch changes the __tgt_* API function signatures in preparation for the new map interface. Changes are: Device IDs 32bits --> 64bits, Flags 32bits --> 64bits Differential revision: https://reviews.llvm.org/D40281 llvm-svn: 318789
2017-11-22 02:25:12 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target(i64 -1, i8* @{{[^,]+}}, i32 2, i8** [[BPR:%[^,]+]], i8** [[PR:%[^,]+]], i[[SZ]]* getelementptr inbounds ([2 x i[[SZ]]], [2 x i[[SZ]]]* [[SIZET3]], i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* [[MAPT3]], i32 0, i32 0))
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPR]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[PR]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[P:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP]], i32 0, i32 0
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[P]], i32 0, i32 0
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[BP0:%[^,]+]], i[[SZ]]* [[CBPADDR0]]
// CHECK-DAG: store i[[SZ]] [[P0:%[^,]+]], i[[SZ]]* [[CPADDR0]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR1:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP]], i32 0, i32 1
// CHECK-DAG: [[PADDR1:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[P]], i32 0, i32 1
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR1:%.+]] = bitcast i8** [[BPADDR1]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR1:%.+]] = bitcast i8** [[PADDR1]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[BP1:%[^,]+]], i[[SZ]]* [[CBPADDR1]]
// CHECK-DAG: store i[[SZ]] [[P1:%[^,]+]], i[[SZ]]* [[CPADDR1]]
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
// CHECK: [[FAIL]]
// CHECK: call void [[HVT3:@.+]]({{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK-NEXT: br label %[[IFEND:.+]]
// CHECK: [[IFELSE]]
// CHECK: call void [[HVT3]]({{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[IFEND]]
// CHECK: [[IFEND]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
#pragma omp target if(n>10)
{
a += 1;
aa += 1;
}
// We capture 3 VLA sizes in this target region
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64: [[A_VAL:%.+]] = load i32, i32* %{{.+}},
// CHECK-64: [[A_ADDR:%.+]] = bitcast i[[SZ]]* [[A_CADDR:%.+]] to i32*
// CHECK-64: store i32 [[A_VAL]], i32* [[A_ADDR]],
// CHECK-64: [[A_CVAL:%.+]] = load i[[SZ]], i[[SZ]]* [[A_CADDR]],
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-32: [[A_VAL:%.+]] = load i32, i32* %{{.+}},
// CHECK-32: store i32 [[A_VAL]], i32* [[A_CADDR:%.+]],
// CHECK-32: [[A_CVAL:%.+]] = load i[[SZ]], i[[SZ]]* [[A_CADDR]],
[OPENMP] Add codegen for `depend` clauses on `target` directive. Added basic support for codegen of `depend` clauses on `target` directive. llvm-svn: 322501
2018-01-16 03:06:12 +08:00
// CHECK: [[IF:%.+]] = icmp sgt i32 {{[^,]+}}, 20
// CHECK: br i1 [[IF]], label %[[TRY:[^,]+]], label %[[IFELSE:[^,]+]]
// CHECK: [[TRY]]
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[BNSIZE:%.+]] = mul nuw i[[SZ]] [[VLA0:%.+]], 4
// CHECK: [[CNELEMSIZE2:%.+]] = mul nuw i[[SZ]] 5, [[VLA1:%.+]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: [[CNSIZE:%.+]] = mul nuw i[[SZ]] [[CNELEMSIZE2]], 8
[Clang][OpenMP] New clang/libomptarget map interface: new function signatures, clang-side This clang patch changes the __tgt_* API function signatures in preparation for the new map interface. Changes are: Device IDs 32bits --> 64bits, Flags 32bits --> 64bits Differential revision: https://reviews.llvm.org/D40281 llvm-svn: 318789
2017-11-22 02:25:12 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target(i64 -1, i8* @{{[^,]+}}, i32 9, i8** [[BPR:%[^,]+]], i8** [[PR:%[^,]+]], i[[SZ]]* [[SR:%[^,]+]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* [[MAPT4]], i32 0, i32 0))
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPR]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[PR]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P:%[^,]+]], i32 0, i32 0
// CHECK-DAG: [[SR]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S:%[^,]+]], i32 0, i32 0
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR0:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX0:0]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX0]]
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX0]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR1:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX1:1]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR1:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX1]]
// CHECK-DAG: [[PADDR1:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX1]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR2:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX2:2]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR2:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX2]]
// CHECK-DAG: [[PADDR2:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX2]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR3:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX3:3]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR3:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX3]]
// CHECK-DAG: [[PADDR3:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX3]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR4:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX4:4]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR4:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX4]]
// CHECK-DAG: [[PADDR4:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX4]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR5:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX5:5]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR5:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX5]]
// CHECK-DAG: [[PADDR5:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX5]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR6:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX6:6]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR6:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX6]]
// CHECK-DAG: [[PADDR6:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX6]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR7:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX7:7]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR7:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX7]]
// CHECK-DAG: [[PADDR7:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX7]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[SADDR8:%.+]] = getelementptr inbounds [9 x i[[SZ]]], [9 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX8:8]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR8:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[BP]], i32 0, i32 [[IDX8]]
// CHECK-DAG: [[PADDR8:%.+]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[P]], i32 0, i32 [[IDX8]]
// The names below are not necessarily consistent with the names used for the
// addresses above as some are repeated.
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR2:%.+]] = bitcast i8** [[BPADDR2]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR2:%.+]] = bitcast i8** [[PADDR2]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VLA0]], i[[SZ]]* [[CBPADDR2]]
// CHECK-DAG: store i[[SZ]] [[VLA0]], i[[SZ]]* [[CPADDR2]]
// CHECK-DAG: store i[[SZ]] {{4|8}}, i[[SZ]]* [[SADDR2]]
// CHECK-DAG: [[CBPADDR6:%.+]] = bitcast i8** [[BPADDR6]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR6:%.+]] = bitcast i8** [[PADDR6]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VLA1]], i[[SZ]]* [[CBPADDR6]]
// CHECK-DAG: store i[[SZ]] [[VLA1]], i[[SZ]]* [[CPADDR6]]
// CHECK-DAG: store i[[SZ]] {{4|8}}, i[[SZ]]* [[SADDR6]]
// CHECK-DAG: [[CBPADDR5:%.+]] = bitcast i8** [[BPADDR5]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR5:%.+]] = bitcast i8** [[PADDR5]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] 5, i[[SZ]]* [[CBPADDR5]]
// CHECK-DAG: store i[[SZ]] 5, i[[SZ]]* [[CPADDR5]]
// CHECK-DAG: store i[[SZ]] {{4|8}}, i[[SZ]]* [[SADDR5]]
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[A_CVAL]], i[[SZ]]* [[CBPADDR0]]
// CHECK-DAG: store i[[SZ]] [[A_CVAL]], i[[SZ]]* [[CPADDR0]]
// CHECK-DAG: store i[[SZ]] 4, i[[SZ]]* [[SADDR0]]
// CHECK-DAG: [[CBPADDR1:%.+]] = bitcast i8** [[BPADDR1]] to [10 x float]**
// CHECK-DAG: [[CPADDR1:%.+]] = bitcast i8** [[PADDR1]] to [10 x float]**
// CHECK-DAG: store [10 x float]* %{{.+}}, [10 x float]** [[CBPADDR1]]
// CHECK-DAG: store [10 x float]* %{{.+}}, [10 x float]** [[CPADDR1]]
// CHECK-DAG: store i[[SZ]] 40, i[[SZ]]* [[SADDR1]]
// CHECK-DAG: [[CBPADDR3:%.+]] = bitcast i8** [[BPADDR3]] to float**
// CHECK-DAG: [[CPADDR3:%.+]] = bitcast i8** [[PADDR3]] to float**
// CHECK-DAG: store float* %{{.+}}, float** [[CBPADDR3]]
// CHECK-DAG: store float* %{{.+}}, float** [[CPADDR3]]
// CHECK-DAG: store i[[SZ]] [[BNSIZE]], i[[SZ]]* [[SADDR3]]
// CHECK-DAG: [[CBPADDR4:%.+]] = bitcast i8** [[BPADDR4]] to [5 x [10 x double]]**
// CHECK-DAG: [[CPADDR4:%.+]] = bitcast i8** [[PADDR4]] to [5 x [10 x double]]**
// CHECK-DAG: store [5 x [10 x double]]* %{{.+}}, [5 x [10 x double]]** [[CBPADDR4]]
// CHECK-DAG: store [5 x [10 x double]]* %{{.+}}, [5 x [10 x double]]** [[CPADDR4]]
// CHECK-DAG: store i[[SZ]] 400, i[[SZ]]* [[SADDR4]]
// CHECK-DAG: [[CBPADDR7:%.+]] = bitcast i8** [[BPADDR7]] to double**
// CHECK-DAG: [[CPADDR7:%.+]] = bitcast i8** [[PADDR7]] to double**
// CHECK-DAG: store double* %{{.+}}, double** [[CBPADDR7]]
// CHECK-DAG: store double* %{{.+}}, double** [[CPADDR7]]
// CHECK-DAG: store i[[SZ]] [[CNSIZE]], i[[SZ]]* [[SADDR7]]
// CHECK-DAG: [[CBPADDR8:%.+]] = bitcast i8** [[BPADDR8]] to [[TT]]**
// CHECK-DAG: [[CPADDR8:%.+]] = bitcast i8** [[PADDR8]] to [[TT]]**
// CHECK-DAG: store [[TT]]* %{{.+}}, [[TT]]** [[CBPADDR8]]
// CHECK-DAG: store [[TT]]* %{{.+}}, [[TT]]** [[CPADDR8]]
// CHECK-DAG: store i[[SZ]] {{12|16}}, i[[SZ]]* [[SADDR8]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: [[FAIL]]
// CHECK: call void [[HVT4:@.+]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK-NEXT: br label %[[IFEND:.+]]
// CHECK: [[IFELSE]]
// CHECK: call void [[HVT4]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[IFEND]]
// CHECK: [[IFEND]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
#pragma omp target if(n>20)
{
a += 1;
b[2] += 1.0;
bn[3] += 1.0;
c[1][2] += 1.0;
cn[1][3] += 1.0;
d.X += 1;
d.Y += 1;
}
return a;
}
// Check that the offloading functions are emitted and that the arguments are
// correct and loaded correctly for the target regions in foo().
// CHECK: define internal void [[HVT0]]()
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: define internal void [[HVT1]](i[[SZ]] %{{.+}})
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// Create stack storage and store argument in there.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[AA_ADDR:%.+]] = alloca i[[SZ]], align
// CHECK: store i[[SZ]] %{{.+}}, i[[SZ]]* [[AA_ADDR]], align
// CHECK-64: [[AA_CADDR:%.+]] = bitcast i[[SZ]]* [[AA_ADDR]] to i32*
// CHECK-64: load i32, i32* [[AA_CADDR]], align
// CHECK-32: load i32, i32* [[AA_ADDR]], align
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: define internal void [[HVT2]](i[[SZ]] %{{.+}})
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// Create stack storage and store argument in there.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[AA_ADDR:%.+]] = alloca i[[SZ]], align
// CHECK: store i[[SZ]] %{{.+}}, i[[SZ]]* [[AA_ADDR]], align
// CHECK: [[AA_CADDR:%.+]] = bitcast i[[SZ]]* [[AA_ADDR]] to i16*
// CHECK: load i16, i16* [[AA_CADDR]], align
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: define internal void [[HVT3]]
// Create stack storage and store argument in there.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[A_ADDR:%.+]] = alloca i[[SZ]], align
// CHECK: [[AA_ADDR:%.+]] = alloca i[[SZ]], align
// CHECK-DAG: store i[[SZ]] %{{.+}}, i[[SZ]]* [[A_ADDR]], align
// CHECK-DAG: store i[[SZ]] %{{.+}}, i[[SZ]]* [[AA_ADDR]], align
// CHECK-64-DAG:[[A_CADDR:%.+]] = bitcast i[[SZ]]* [[A_ADDR]] to i32*
// CHECK-DAG: [[AA_CADDR:%.+]] = bitcast i[[SZ]]* [[AA_ADDR]] to i16*
// CHECK-64-DAG:load i32, i32* [[A_CADDR]], align
// CHECK-32-DAG:load i32, i32* [[A_ADDR]], align
// CHECK-DAG: load i16, i16* [[AA_CADDR]], align
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: define internal void [[HVT4]]
// Create local storage for each capture.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[LOCAL_A:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_B:%.+]] = alloca [10 x float]*
// CHECK: [[LOCAL_VLA1:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_BN:%.+]] = alloca float*
// CHECK: [[LOCAL_C:%.+]] = alloca [5 x [10 x double]]*
// CHECK: [[LOCAL_VLA2:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_VLA3:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_CN:%.+]] = alloca double*
// CHECK: [[LOCAL_D:%.+]] = alloca [[TT]]*
// CHECK-DAG: store i[[SZ]] [[ARG_A:%.+]], i[[SZ]]* [[LOCAL_A]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store [10 x float]* [[ARG_B:%.+]], [10 x float]** [[LOCAL_B]]
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-DAG: store i[[SZ]] [[ARG_VLA1:%.+]], i[[SZ]]* [[LOCAL_VLA1]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store float* [[ARG_BN:%.+]], float** [[LOCAL_BN]]
// CHECK-DAG: store [5 x [10 x double]]* [[ARG_C:%.+]], [5 x [10 x double]]** [[LOCAL_C]]
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-DAG: store i[[SZ]] [[ARG_VLA2:%.+]], i[[SZ]]* [[LOCAL_VLA2]]
// CHECK-DAG: store i[[SZ]] [[ARG_VLA3:%.+]], i[[SZ]]* [[LOCAL_VLA3]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store double* [[ARG_CN:%.+]], double** [[LOCAL_CN]]
// CHECK-DAG: store [[TT]]* [[ARG_D:%.+]], [[TT]]** [[LOCAL_D]]
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG:[[REF_A:%.+]] = bitcast i[[SZ]]* [[LOCAL_A]] to i32*
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[REF_B:%.+]] = load [10 x float]*, [10 x float]** [[LOCAL_B]],
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-DAG: [[VAL_VLA1:%.+]] = load i[[SZ]], i[[SZ]]* [[LOCAL_VLA1]],
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[REF_BN:%.+]] = load float*, float** [[LOCAL_BN]],
// CHECK-DAG: [[REF_C:%.+]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[LOCAL_C]],
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-DAG: [[VAL_VLA2:%.+]] = load i[[SZ]], i[[SZ]]* [[LOCAL_VLA2]],
// CHECK-DAG: [[VAL_VLA3:%.+]] = load i[[SZ]], i[[SZ]]* [[LOCAL_VLA3]],
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[REF_CN:%.+]] = load double*, double** [[LOCAL_CN]],
// CHECK-DAG: [[REF_D:%.+]] = load [[TT]]*, [[TT]]** [[LOCAL_D]],
// Use captures.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG: load i32, i32* [[REF_A]]
// CHECK-32-DAG: load i32, i32* [[LOCAL_A]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: getelementptr inbounds [10 x float], [10 x float]* [[REF_B]], i[[SZ]] 0, i[[SZ]] 2
// CHECK-DAG: getelementptr inbounds float, float* [[REF_BN]], i[[SZ]] 3
// CHECK-DAG: getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[REF_C]], i[[SZ]] 0, i[[SZ]] 1
// CHECK-DAG: getelementptr inbounds double, double* [[REF_CN]], i[[SZ]] %{{.+}}
// CHECK-DAG: getelementptr inbounds [[TT]], [[TT]]* [[REF_D]], i32 0, i32 0
template<typename tx>
tx ftemplate(int n) {
tx a = 0;
short aa = 0;
tx b[10];
#pragma omp target if(n>40)
{
a += 1;
aa += 1;
b[2] += 1;
}
return a;
}
static
int fstatic(int n) {
int a = 0;
short aa = 0;
char aaa = 0;
int b[10];
#pragma omp target if(n>50)
{
a += 1;
aa += 1;
aaa += 1;
b[2] += 1;
}
return a;
}
struct S1 {
double a;
int r1(int n){
int b = n+1;
short int c[2][n];
#pragma omp target if(n>60)
{
this->a = (double)b + 1.5;
c[1][1] = ++a;
}
return c[1][1] + (int)b;
}
};
// CHECK: define {{.*}}@{{.*}}bar{{.*}}
int bar(int n){
int a = 0;
// CHECK: call {{.*}}i32 [[FOO]](i32 {{.*}})
a += foo(n);
S1 S;
// CHECK: call {{.*}}i32 [[FS1:@.+]]([[S1]]* {{.*}}, i32 {{.*}})
a += S.r1(n);
// CHECK: call {{.*}}i32 [[FSTATIC:@.+]](i32 {{.*}})
a += fstatic(n);
// CHECK: call {{.*}}i32 [[FTEMPLATE:@.+]](i32 {{.*}})
a += ftemplate<int>(n);
return a;
}
//
// CHECK: define {{.*}}[[FS1]]
//
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: i8* @llvm.stacksave()
// CHECK-64: [[B_ADDR:%.+]] = bitcast i[[SZ]]* [[B_CADDR:%.+]] to i32*
// CHECK-64: store i32 %{{.+}}, i32* [[B_ADDR]],
// CHECK-64: [[B_CVAL:%.+]] = load i[[SZ]], i[[SZ]]* [[B_CADDR]],
[DebugInfo] Avoid name conflict of generated VLA expression variable. Summary: This patch also adds the 'DW_AT_artificial' flag to the generated variable. Addresses the issues mentioned in http://llvm.org/PR30553. Reviewers: CarlosAlbertoEnciso, probinson, aprantl Reviewed By: aprantl Subscribers: JDevlieghere, cfe-commits Differential Revision: https://reviews.llvm.org/D43189 llvm-svn: 324988
2018-02-13 15:49:34 +08:00
// CHECK-32: store i32 %{{.+}}, i32* %__vla_expr
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-32: store i32 %{{.+}}, i32* [[B_ADDR:%.+]],
// CHECK-32: [[B_CVAL:%.+]] = load i[[SZ]], i[[SZ]]* [[B_ADDR]],
[OPENMP] Add codegen for `depend` clauses on `target` directive. Added basic support for codegen of `depend` clauses on `target` directive. llvm-svn: 322501
2018-01-16 03:06:12 +08:00
// CHECK: [[IF:%.+]] = icmp sgt i32 {{[^,]+}}, 60
// CHECK: br i1 [[IF]], label %[[TRY:[^,]+]], label %[[IFELSE:[^,]+]]
// CHECK: [[TRY]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// We capture 2 VLA sizes in this target region
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[CELEMSIZE2:%.+]] = mul nuw i[[SZ]] 2, [[VLA0:%.+]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: [[CSIZE:%.+]] = mul nuw i[[SZ]] [[CELEMSIZE2]], 2
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target(i64 -1, i8* @{{[^,]+}}, i32 6, i8** [[BPR:%[^,]+]], i8** [[PR:%[^,]+]], i[[SZ]]* [[SR:%[^,]+]], i64* getelementptr inbounds ([6 x i64], [6 x i64]* [[MAPT7]], i32 0, i32 0))
// CHECK-DAG: [[BPR]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP:%.+]], i32 0, i32 0
// CHECK-DAG: [[PR]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P:%.+]], i32 0, i32 0
// CHECK-DAG: [[SR]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S:%.+]], i32 0, i32 0
// CHECK-DAG: [[SADDR0:%.+]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX0:0]]
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP]], i32 0, i32 [[IDX0]]
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P]], i32 0, i32 [[IDX0]]
// CHECK-DAG: [[SADDR1:%.+]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX1:1]]
// CHECK-DAG: [[BPADDR1:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP]], i32 0, i32 [[IDX1]]
// CHECK-DAG: [[PADDR1:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P]], i32 0, i32 [[IDX1]]
// CHECK-DAG: [[SADDR2:%.+]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX2:2]]
// CHECK-DAG: [[BPADDR2:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP]], i32 0, i32 [[IDX2]]
// CHECK-DAG: [[PADDR2:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P]], i32 0, i32 [[IDX2]]
// CHECK-DAG: [[SADDR3:%.+]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX3:3]]
// CHECK-DAG: [[BPADDR3:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP]], i32 0, i32 [[IDX3]]
// CHECK-DAG: [[PADDR3:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P]], i32 0, i32 [[IDX3]]
// CHECK-DAG: [[SADDR4:%.+]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX4:4]]
// CHECK-DAG: [[BPADDR4:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP]], i32 0, i32 [[IDX4]]
// CHECK-DAG: [[PADDR4:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P]], i32 0, i32 [[IDX4]]
// CHECK-DAG: [[SADDR5:%.+]] = getelementptr inbounds [6 x i[[SZ]]], [6 x i[[SZ]]]* [[S]], i32 0, i32 [[IDX5:5]]
// CHECK-DAG: [[BPADDR5:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[BP]], i32 0, i32 [[IDX5]]
// CHECK-DAG: [[PADDR5:%.+]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[P]], i32 0, i32 [[IDX5]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// The names below are not necessarily consistent with the names used for the
// addresses above as some are repeated.
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: [[CBPADDR5:%.+]] = bitcast i8** [[BPADDR5]] to i16**
// CHECK-DAG: [[CPADDR5:%.+]] = bitcast i8** [[PADDR5]] to i16**
// CHECK-DAG: store i16* %{{.+}}, i16** [[CBPADDR5]]
// CHECK-DAG: store i16* %{{.+}}, i16** [[CPADDR5]]
// CHECK-DAG: store i[[SZ]] [[CSIZE]], i[[SZ]]* [[SADDR5]]
// CHECK-DAG: [[CBPADDR4:%.+]] = bitcast i8** [[BPADDR4]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR4:%.+]] = bitcast i8** [[PADDR4]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VLA0]], i[[SZ]]* [[CBPADDR4]]
// CHECK-DAG: store i[[SZ]] [[VLA0]], i[[SZ]]* [[CPADDR4]]
// CHECK-DAG: store i[[SZ]] {{4|8}}, i[[SZ]]* [[SADDR4]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR3:%.+]] = bitcast i8** [[BPADDR3]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR3:%.+]] = bitcast i8** [[PADDR3]] to i[[SZ]]*
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: store i[[SZ]] 2, i[[SZ]]* [[CBPADDR3]]
// CHECK-DAG: store i[[SZ]] 2, i[[SZ]]* [[CPADDR3]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: store i[[SZ]] {{4|8}}, i[[SZ]]* [[SADDR3]]
// CHECK-DAG: [[CBPADDR2:%.+]] = bitcast i8** [[BPADDR2]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR2:%.+]] = bitcast i8** [[PADDR2]] to i[[SZ]]*
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: store i[[SZ]] [[B_CVAL]], i[[SZ]]* [[CBPADDR2]]
// CHECK-DAG: store i[[SZ]] [[B_CVAL]], i[[SZ]]* [[CPADDR2]]
// CHECK-DAG: store i[[SZ]] 4, i[[SZ]]* [[SADDR2]]
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to [[S1]]**
[OPENMP] Generate implicit map|firstprivate clauses for target-based directives. If the variable is used in the target-based region but is not found in any private|mapping clause, then generate implicit firstprivate|map clauses for these implicitly mapped variables. llvm-svn: 314205
2017-09-26 21:47:31 +08:00
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to double**
The patch adds support for the new map interface between clang and libomptarget. The changes in the interface are the following: device IDs are now 64-bit integers (as opposed to 32-bit) map flags are 64-bit long (used to be 32-bit) mappings for partially mapped structs are now calculated at compile time and members of partially mapped structs are flagged using the MEMBER_OF field Support for is_device_ptr on struct members was dropped - this functionality is not supported by the OpenMP standard and its implementation is technically infeasible (however, use_device_ptr on struct members works as a non-standard extension of the compiler) llvm-svn: 337468
2018-07-20 00:34:13 +08:00
// CHECK-DAG: store [[S1]]* [[THIS:%.+]], [[S1]]** [[CBPADDR0]]
// CHECK-DAG: store double* [[A:%.+]], double** [[CPADDR0]]
// CHECK-DAG: store i[[SZ]] %{{.+}}, i[[SZ]]* [[SADDR0]]
// CHECK-DAG: [[CBPADDR1:%.+]] = bitcast i8** [[BPADDR1]] to [[S1]]**
// CHECK-DAG: [[CPADDR1:%.+]] = bitcast i8** [[PADDR1]] to double**
// CHECK-DAG: store [[S1]]* [[THIS]], [[S1]]** [[CBPADDR1]]
// CHECK-DAG: store double* [[A]], double** [[CPADDR1]]
// CHECK-DAG: store i[[SZ]] 8, i[[SZ]]* [[SADDR1]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: [[FAIL]]
// CHECK: call void [[HVT7:@.+]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK-NEXT: br label %[[IFEND:.+]]
// CHECK: [[IFELSE]]
// CHECK: call void [[HVT7]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[IFEND]]
// CHECK: [[IFEND]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
//
// CHECK: define {{.*}}[[FSTATIC]]
//
// CHECK: [[IF:%.+]] = icmp sgt i32 {{[^,]+}}, 50
// CHECK: br i1 [[IF]], label %[[IFTHEN:[^,]+]], label %[[IFELSE:[^,]+]]
// CHECK: [[IFTHEN]]
[Clang][OpenMP] New clang/libomptarget map interface: new function signatures, clang-side This clang patch changes the __tgt_* API function signatures in preparation for the new map interface. Changes are: Device IDs 32bits --> 64bits, Flags 32bits --> 64bits Differential revision: https://reviews.llvm.org/D40281 llvm-svn: 318789
2017-11-22 02:25:12 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target(i64 -1, i8* @{{[^,]+}}, i32 4, i8** [[BPR:%[^,]+]], i8** [[PR:%[^,]+]], i[[SZ]]* getelementptr inbounds ([4 x i[[SZ]]], [4 x i[[SZ]]]* [[SIZET6]], i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* [[MAPT6]], i32 0, i32 0))
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPR]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[BP:%.+]], i32 0, i32 0
// CHECK-DAG: [[PR]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[P:%.+]], i32 0, i32 0
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[BP]], i32 0, i32 0
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[P]], i32 0, i32 0
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VAL0:%[^,]+]], i[[SZ]]* [[CBPADDR0]]
// CHECK-DAG: store i[[SZ]] [[VAL0]], i[[SZ]]* [[CPADDR0]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR1:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[BP]], i32 0, i32 1
// CHECK-DAG: [[PADDR1:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[P]], i32 0, i32 1
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR1:%.+]] = bitcast i8** [[BPADDR1]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR1:%.+]] = bitcast i8** [[PADDR1]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VAL1:%[^,]+]], i[[SZ]]* [[CBPADDR1]]
// CHECK-DAG: store i[[SZ]] [[VAL1]], i[[SZ]]* [[CPADDR1]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR2:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[BP]], i32 0, i32 2
// CHECK-DAG: [[PADDR2:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[P]], i32 0, i32 2
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR2:%.+]] = bitcast i8** [[BPADDR2]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR2:%.+]] = bitcast i8** [[PADDR2]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VAL2:%[^,]+]], i[[SZ]]* [[CBPADDR2]]
// CHECK-DAG: store i[[SZ]] [[VAL2]], i[[SZ]]* [[CPADDR2]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR3:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[BP]], i32 0, i32 3
// CHECK-DAG: [[PADDR3:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[P]], i32 0, i32 3
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR3:%.+]] = bitcast i8** [[BPADDR3]] to [10 x i32]**
// CHECK-DAG: [[CPADDR3:%.+]] = bitcast i8** [[PADDR3]] to [10 x i32]**
// CHECK-DAG: store [10 x i32]* [[VAL3:%[^,]+]], [10 x i32]** [[CBPADDR3]]
// CHECK-DAG: store [10 x i32]* [[VAL3]], [10 x i32]** [[CPADDR3]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
// CHECK: [[FAIL]]
// CHECK: call void [[HVT6:@.+]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK-NEXT: br label %[[IFEND:.+]]
// CHECK: [[IFELSE]]
// CHECK: call void [[HVT6]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[IFEND]]
// CHECK: [[IFEND]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
//
// CHECK: define {{.*}}[[FTEMPLATE]]
//
// CHECK: [[IF:%.+]] = icmp sgt i32 {{[^,]+}}, 40
// CHECK: br i1 [[IF]], label %[[IFTHEN:[^,]+]], label %[[IFELSE:[^,]+]]
// CHECK: [[IFTHEN]]
[Clang][OpenMP] New clang/libomptarget map interface: new function signatures, clang-side This clang patch changes the __tgt_* API function signatures in preparation for the new map interface. Changes are: Device IDs 32bits --> 64bits, Flags 32bits --> 64bits Differential revision: https://reviews.llvm.org/D40281 llvm-svn: 318789
2017-11-22 02:25:12 +08:00
// CHECK-DAG: [[RET:%.+]] = call i32 @__tgt_target(i64 -1, i8* @{{[^,]+}}, i32 3, i8** [[BPR:%[^,]+]], i8** [[PR:%[^,]+]], i[[SZ]]* getelementptr inbounds ([3 x i[[SZ]]], [3 x i[[SZ]]]* [[SIZET5]], i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* [[MAPT5]], i32 0, i32 0))
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPR]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[BP:%.+]], i32 0, i32 0
// CHECK-DAG: [[PR]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[P:%.+]], i32 0, i32 0
// CHECK-DAG: [[BPADDR0:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[BP]], i32 0, i32 0
// CHECK-DAG: [[PADDR0:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[P]], i32 0, i32 0
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR0:%.+]] = bitcast i8** [[BPADDR0]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR0:%.+]] = bitcast i8** [[PADDR0]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VAL0:%[^,]+]], i[[SZ]]* [[CBPADDR0]]
// CHECK-DAG: store i[[SZ]] [[VAL0]], i[[SZ]]* [[CPADDR0]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR1:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[BP]], i32 0, i32 1
// CHECK-DAG: [[PADDR1:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[P]], i32 0, i32 1
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR1:%.+]] = bitcast i8** [[BPADDR1]] to i[[SZ]]*
// CHECK-DAG: [[CPADDR1:%.+]] = bitcast i8** [[PADDR1]] to i[[SZ]]*
// CHECK-DAG: store i[[SZ]] [[VAL1:%[^,]+]], i[[SZ]]* [[CBPADDR1]]
// CHECK-DAG: store i[[SZ]] [[VAL1]], i[[SZ]]* [[CPADDR1]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[BPADDR2:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[BP]], i32 0, i32 2
// CHECK-DAG: [[PADDR2:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[P]], i32 0, i32 2
[OPENMP][DEBUG] Generate second function with correct arg types. Currently, if the some of the parameters are captured by value, this argument is converted to uintptr_t type and thus we loosing the debug info about real type of the argument (captured variable): ``` void @.outlined_function.(uintptr %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.outlined_function.(uintptr %a.casted.val) ... ``` To resolve this problem, in debug mode a speciall external wrapper function is generated, that calls the outlined function with the correct parameters types: ``` void @.wrapper.(uintptr %par) { %a = alloca i32 %cast = bitcast i32* %a to uintptr* store uintptr %par, uintptr *%cast %a.val = load i32, i32* %a call void @.outlined_function.(i32 %a) ret void } void @.outlined_function.(i32 %par); ... %a = alloca i32 %a.casted = alloca uintptr %cast = bitcast uintptr* %a.casted to i32* %a.val = load i32, i32 *%a store i32 %a.val, i32 *%cast %a.casted.val = load uintptr, uintptr* %a.casted call void @.wrapper.(uintptr %a.casted.val) ... ``` llvm-svn: 306697
2017-06-30 00:43:05 +08:00
// CHECK-DAG: [[CBPADDR2:%.+]] = bitcast i8** [[BPADDR2]] to [10 x i32]**
// CHECK-DAG: [[CPADDR2:%.+]] = bitcast i8** [[PADDR2]] to [10 x i32]**
// CHECK-DAG: store [10 x i32]* [[VAL2:%[^,]+]], [10 x i32]** [[CBPADDR2]]
// CHECK-DAG: store [10 x i32]* [[VAL2]], [10 x i32]** [[CPADDR2]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-NEXT: br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
// CHECK: [[FAIL]]
// CHECK: call void [[HVT5:@.+]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[END]]
// CHECK: [[END]]
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK-NEXT: br label %[[IFEND:.+]]
// CHECK: [[IFELSE]]
// CHECK: call void [[HVT5]]({{[^,]+}}, {{[^,]+}}, {{[^,]+}})
// CHECK-NEXT: br label %[[IFEND]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
[OPENMP] Simplify codegen for non-offloading code. Simplified and generalized codegen for non-offloading part that works if offloading is failed or condition of the `if` clause is `false`. llvm-svn: 314670
2017-10-02 22:20:58 +08:00
// CHECK: [[IFEND]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// Check that the offloading functions are emitted and that the arguments are
// correct and loaded correctly for the target regions of the callees of bar().
// CHECK: define internal void [[HVT7]]
// Create local storage for each capture.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[LOCAL_THIS:%.+]] = alloca [[S1]]*
// CHECK: [[LOCAL_B:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_VLA1:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_VLA2:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_C:%.+]] = alloca i16*
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store [[S1]]* [[ARG_THIS:%.+]], [[S1]]** [[LOCAL_THIS]]
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-DAG: store i[[SZ]] [[ARG_B:%.+]], i[[SZ]]* [[LOCAL_B]]
// CHECK-DAG: store i[[SZ]] [[ARG_VLA1:%.+]], i[[SZ]]* [[LOCAL_VLA1]]
// CHECK-DAG: store i[[SZ]] [[ARG_VLA2:%.+]], i[[SZ]]* [[LOCAL_VLA2]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store i16* [[ARG_C:%.+]], i16** [[LOCAL_C]]
// Store captures in the context.
// CHECK-DAG: [[REF_THIS:%.+]] = load [[S1]]*, [[S1]]** [[LOCAL_THIS]],
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG:[[REF_B:%.+]] = bitcast i[[SZ]]* [[LOCAL_B]] to i32*
// CHECK-DAG: [[VAL_VLA1:%.+]] = load i[[SZ]], i[[SZ]]* [[LOCAL_VLA1]],
// CHECK-DAG: [[VAL_VLA2:%.+]] = load i[[SZ]], i[[SZ]]* [[LOCAL_VLA2]],
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[REF_C:%.+]] = load i16*, i16** [[LOCAL_C]],
// Use captures.
// CHECK-DAG: getelementptr inbounds [[S1]], [[S1]]* [[REF_THIS]], i32 0, i32 0
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG:load i32, i32* [[REF_B]]
// CHECK-32-DAG:load i32, i32* [[LOCAL_B]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: getelementptr inbounds i16, i16* [[REF_C]], i[[SZ]] %{{.+}}
// CHECK: define internal void [[HVT6]]
// Create local storage for each capture.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[LOCAL_A:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_AA:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_AAA:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_B:%.+]] = alloca [10 x i32]*
// CHECK-DAG: store i[[SZ]] [[ARG_A:%.+]], i[[SZ]]* [[LOCAL_A]]
// CHECK-DAG: store i[[SZ]] [[ARG_AA:%.+]], i[[SZ]]* [[LOCAL_AA]]
// CHECK-DAG: store i[[SZ]] [[ARG_AAA:%.+]], i[[SZ]]* [[LOCAL_AAA]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store [10 x i32]* [[ARG_B:%.+]], [10 x i32]** [[LOCAL_B]]
// Store captures in the context.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG: [[REF_A:%.+]] = bitcast i[[SZ]]* [[LOCAL_A]] to i32*
// CHECK-DAG: [[REF_AA:%.+]] = bitcast i[[SZ]]* [[LOCAL_AA]] to i16*
// CHECK-DAG: [[REF_AAA:%.+]] = bitcast i[[SZ]]* [[LOCAL_AAA]] to i8*
// CHECK-DAG: [[REF_B:%.+]] = load [10 x i32]*, [10 x i32]** [[LOCAL_B]],
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// Use captures.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG: load i32, i32* [[REF_A]]
// CHECK-DAG: load i16, i16* [[REF_AA]]
// CHECK-DAG: load i8, i8* [[REF_AAA]]
// CHECK-32-DAG: load i32, i32* [[LOCAL_A]]
// CHECK-DAG: getelementptr inbounds [10 x i32], [10 x i32]* [[REF_B]], i[[SZ]] 0, i[[SZ]] 2
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK: define internal void [[HVT5]]
// Create local storage for each capture.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK: [[LOCAL_A:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_AA:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_B:%.+]] = alloca [10 x i32]*
// CHECK-DAG: store i[[SZ]] [[ARG_A:%.+]], i[[SZ]]* [[LOCAL_A]]
// CHECK-DAG: store i[[SZ]] [[ARG_AA:%.+]], i[[SZ]]* [[LOCAL_AA]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: store [10 x i32]* [[ARG_B:%.+]], [10 x i32]** [[LOCAL_B]]
// Store captures in the context.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG:[[REF_A:%.+]] = bitcast i[[SZ]]* [[LOCAL_A]] to i32*
// CHECK-DAG: [[REF_AA:%.+]] = bitcast i[[SZ]]* [[LOCAL_AA]] to i16*
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: [[REF_B:%.+]] = load [10 x i32]*, [10 x i32]** [[LOCAL_B]],
// Use captures.
[OpenMP] Update target directive codegen to use 4.5 implicit data mappings. Summary: This patch implements the 4.5 specification for the implicit data maps. OpenMP 4.5 specification changes the default way data is captured into a target region. All the non-aggregate kinds are passed by value by default. This required activating the capturing by value during SEMA for the target region. All the non-aggregate values that can be encoded in the size of a pointer are properly casted and forwarded to the runtime library. On top of fixing the previous weird behavior for mapping pointers in nested data regions (an explicit map was always required), this also improves performance as the number of allocations/transactions to the device per non-aggregate map are reduced from two to only one - instead of passing a reference and the value, only the value passed. Explicit maps will be added later on once firstprivate, private, and map clauses' SEMA and parsing are available. Reviewers: hfinkel, rjmccall, ABataev Subscribers: cfe-commits, carlo.bertolli Differential Revision: http://reviews.llvm.org/D14940 llvm-svn: 254521
2015-12-03 01:44:43 +08:00
// CHECK-64-DAG: load i32, i32* [[REF_A]]
// CHECK-32-DAG: load i32, i32* [[LOCAL_A]]
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
// CHECK-DAG: load i16, i16* [[REF_AA]]
// CHECK-DAG: getelementptr inbounds [10 x i32], [10 x i32]* [[REF_B]], i[[SZ]] 0, i[[SZ]] 2
[OPENMP] Do not crash on incorrect input data. Emit error messages instead of compiler crashing when the target region does not exist in the device code + fix crash when the location comes from macros. llvm-svn: 331195
2018-05-01 00:26:57 +08:00
void bar () {
#define pragma_target _Pragma("omp target")
pragma_target
{}
}
[OpenMP] Target directive host codegen. This patch implements the outlining for offloading functions for code annotated with the OpenMP target directive. It uses a temporary naming of the outlined functions that will have to be updated later on once target side codegen and registration of offloading libraries is implemented - the naming needs to be made unique in the produced library. llvm-svn: 249148
2015-10-03 00:14:20 +08:00
#endif