[meta-intel] [PATCH v2 1/4] llvm-project-source: update SPIRV-LLVM-Translator 8.0.0 -> 9.0.0
Mittal, Anuj
anuj.mittal at intel.com
Wed Sep 11 02:12:21 PDT 2019
Hi Naveen,
On Wed, 2019-09-11 at 10:57 +0800, Naveen Saini wrote:
> Remove all the backported patches which are available in 9.0.0
> release.
>
> Few patches were recommended from llvm-patches repo:
> https://github.com/intel/intel-graphics-compiler/blob/master/documentation/build_ubuntu.md
> https://github.com/intel/llvm-patches/commit/3906cc086f675847ca99b08107d18e083803d53c
>
> Build failure observed with latest meta-clang layer (llvm9/clang9):
> Build error:
> > -- Found PythonInterp: ../poky/build-opencl/tmp/hosttools/python
> > (found version "2.7.12")
> > -- Found PythonLibs: ../poky/build-opencl/tmp/work/corei7-64-poky-
> > linux/clang/9.0.0-r0/recipe-sysroot-native/usr/lib/libpython3.7m.so
> > (found version "3.7.4")
> > CMake Error at ../poky/build-opencl/tmp/work-shared/llvm-project-
> > source-9.0.0-r0/git/lldb/cmake/modules/LLDBConfig.cmake:228
> > (message):
> > Found incompatible Python interpreter (2.7) and Python libraries
> > (3.7)
>
> Workaround patch is applied.
>
> Signed-off-by: Naveen Saini <naveen.kumar.saini at intel.com>
> ---
> ...DBConfig.cmake-skip-python-version-checks.patch | 41 +
I removed this patch and don't see any issues on my setup. We should
probably debug this further.
> ...L-Change-type-of-block-pointer-for-OpenCL.patch | 156 ----
> ...rOpenCL-pass-to-handle-new-blocks-represn.patch | 986 ---------
> ------------
> .../0001-dont-export-targets-for-binaries.patch | 28 +-
> .../clang/files/0001-skip-building-tests.patch | 17 +-
> ...plify-LLVM-IR-generated-for-OpenCL-blocks.patch | 294 ------
> .../clang/files/0002-Remove-extra-semicolon.patch | 27 -
> .../0003-OpenCL-Fix-assertion-due-to-blocks.patch | 61 --
> ...tils-Add-metadata-fixing-in-SplitBlockPre.patch | 108 +++
> ...ify-Do-not-use-SCEV-expander-for-IVCount-.patch | 143 +++
> .../clang/llvm-project-source.bbappend | 12 +-
The LLVM patches don't seem like backports. If they are from llvm-
patches repository and aren't present upstream yet, we should mark them
as Pending or Submitted as appropriate.
Thanks,
Anuj
> 11 files changed, 321 insertions(+), 1552 deletions(-)
> create mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-LLDBConfig.cmake-skip-python-version-
> checks.patch
> delete mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-OpenCL-Change-type-of-block-pointer-for-
> OpenCL.patch
> delete mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-Update-LowerOpenCL-pass-to-handle-new-
> blocks-represn.patch
> delete mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0002-OpenCL-Simplify-LLVM-IR-generated-for-
> OpenCL-blocks.patch
> delete mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0002-Remove-extra-semicolon.patch
> delete mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0003-OpenCL-Fix-assertion-due-to-blocks.patch
> create mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/BasicBlockUtils-Add-metadata-fixing-in-
> SplitBlockPre.patch
> create mode 100644 dynamic-layers/clang-layer/recipes-
> devtools/clang/files/IndVarSimplify-Do-not-use-SCEV-expander-for-
> IVCount-.patch
>
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-LLDBConfig.cmake-skip-python-version-
> checks.patch b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-LLDBConfig.cmake-skip-python-version-
> checks.patch
> new file mode 100644
> index 0000000..a94f62b
> --- /dev/null
> +++ b/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> LLDBConfig.cmake-skip-python-version-checks.patch
> @@ -0,0 +1,41 @@
> +From 57d019a7937c9564ebf3c043020bdbdd0367fae1 Mon Sep 17 00:00:00
> 2001
> +From: Naveen Saini <naveen.kumar.saini at intel.com>
> +Date: Fri, 6 Sep 2019 10:36:49 +0800
> +Subject: [PATCH] LLDBConfig.cmake: skip python version checks
> +
> +Build error:
> +| -- Found PythonInterp: ../poky/build-opencl/tmp/hosttools/python
> (found version "2.7.12")
> +| -- Found PythonLibs: ../poky/build-opencl/tmp/work/corei7-64-poky-
> linux/clang/9.0.0-r0/recipe-sysroot-native/usr/lib/libpython3.7m.so
> (found version "3.7.4")
> +| CMake Error at ../poky/build-opencl/tmp/work-shared/llvm-project-
> source-9.0.0-r0/git/lldb/cmake/modules/LLDBConfig.cmake:228
> (message):
> +| Found incompatible Python interpreter (2.7) and Python libraries
> (3.7)
> +
> +Upstream-Status: Inappropriate [configuration specific]
> +
> +Signed-off-by: Naveen Saini <naveen.kumar.saini at intel.com>
> +---
> + lldb/cmake/modules/LLDBConfig.cmake | 2 ++
> + 1 file changed, 2 insertions(+)
> +
> +diff --git a/lldb/cmake/modules/LLDBConfig.cmake
> b/lldb/cmake/modules/LLDBConfig.cmake
> +index 0dca557601f..4fb9ae1ef62 100644
> +--- a/lldb/cmake/modules/LLDBConfig.cmake
> ++++ b/lldb/cmake/modules/LLDBConfig.cmake
> +@@ -217,6 +217,7 @@ if (NOT LLDB_DISABLE_PYTHON)
> + find_package(PythonLibs REQUIRED)
> + endif()
> +
> ++if(FALSE)
> + if (NOT CMAKE_CROSSCOMPILING)
> + string(REPLACE "." ";" pythonlibs_version_list
> ${PYTHONLIBS_VERSION_STRING})
> + list(GET pythonlibs_version_list 0 pythonlibs_major)
> +@@ -230,6 +231,7 @@ if (NOT LLDB_DISABLE_PYTHON)
> + " and Python libraries
> (${pythonlibs_major}.${pythonlibs_minor})")
> + endif()
> + endif()
> ++endif()
> +
> + if (PYTHON_INCLUDE_DIR)
> + include_directories(${PYTHON_INCLUDE_DIR})
> +--
> +2.17.1
> +
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-OpenCL-Change-type-of-block-pointer-for-
> OpenCL.patch b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-OpenCL-Change-type-of-block-pointer-for-
> OpenCL.patch
> deleted file mode 100644
> index 1c49140..0000000
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> OpenCL-Change-type-of-block-pointer-for-OpenCL.patch
> +++ /dev/null
> @@ -1,156 +0,0 @@
> -From 39a3ac0065c23d1e2d55dfd8792cc28a146a4307 Mon Sep 17 00:00:00
> 2001
> -From: Alexey Bader <alexey.bader at intel.com>
> -Date: Tue, 19 Feb 2019 15:19:06 +0000
> -Subject: [PATCH 1/2] [OpenCL] Change type of block pointer for
> OpenCL
> -
> -Summary:
> -
> -For some reason OpenCL blocks in LLVM IR are represented as function
> pointers.
> -These pointers do not point to any real function and never get
> called. Actually
> -they point to some structure, which in turn contains pointer to the
> real block
> -invoke function.
> -This patch changes represntation of OpenCL blocks in LLVM IR from
> function
> -pointers to pointers to `%struct.__block_literal_generic`.
> -Such representation allows to avoid unnecessary bitcasts and
> simplifies
> -further processing (e.g. translation to SPIR-V ) of the module for
> targets
> -which do not support function pointers.
> -
> -Patch by: Alexey Sotkin.
> -
> -Reviewers: Anastasia, yaxunl, svenvh
> -
> -Reviewed By: Anastasia
> -
> -Subscribers: alexbatashev, cfe-commits
> -
> -Tags: #clang
> -
> -Differential Revision: https://reviews.llvm.org/D58277
> -
> -git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@354337
> 91177308-0d34-0410-b5e6-96231b3b80d8
> -
> -Upstream-Status: Backport
> -[
> https://github.com/llvm-mirror/clang/commit/283f308bdb5893bab1f36791711346e746045f94
> ]
> -Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> ----
> - lib/CodeGen/CodeGenTypes.cpp | 4 +++-
> - test/CodeGenOpenCL/blocks.cl | 18 ++++++++-------
> ---
> - test/CodeGenOpenCL/cl20-device-side-enqueue.cl | 18 +++++++++----
> -----
> - 3 files changed, 20 insertions(+), 20 deletions(-)
> -
> -diff --git a/lib/CodeGen/CodeGenTypes.cpp
> b/lib/CodeGen/CodeGenTypes.cpp
> -index 2acf1ac..93b3ebf 100644
> ---- a/lib/CodeGen/CodeGenTypes.cpp
> -+++ b/lib/CodeGen/CodeGenTypes.cpp
> -@@ -637,7 +637,9 @@ llvm::Type *CodeGenTypes::ConvertType(QualType
> T) {
> -
> - case Type::BlockPointer: {
> - const QualType FTy = cast<BlockPointerType>(Ty)-
> >getPointeeType();
> -- llvm::Type *PointeeType = ConvertTypeForMem(FTy);
> -+ llvm::Type *PointeeType = CGM.getLangOpts().OpenCL
> -+ ?
> CGM.getGenericBlockLiteralType()
> -+ : ConvertTypeForMem(FTy);
> - unsigned AS = Context.getTargetAddressSpace(FTy);
> - ResultType = llvm::PointerType::get(PointeeType, AS);
> - break;
> -diff --git a/test/CodeGenOpenCL/blocks.cl
> b/test/CodeGenOpenCL/blocks.cl
> -index 675240c..19aacc3 100644
> ---- a/test/CodeGenOpenCL/blocks.cl
> -+++ b/test/CodeGenOpenCL/blocks.cl
> -@@ -35,11 +35,10 @@ void foo(){
> - // SPIR: %[[block_captured:.*]] = getelementptr inbounds <{ i32,
> i32, i8 addrspace(4)*, i32 }>, <{ i32, i32, i8 addrspace(4)*, i32 }>*
> %[[block]], i32 0, i32 3
> - // SPIR: %[[i_value:.*]] = load i32, i32* %i
> - // SPIR: store i32 %[[i_value]], i32* %[[block_captured]],
> -- // SPIR: %[[blk_ptr:.*]] = bitcast <{ i32, i32, i8 addrspace(4)*,
> i32 }>* %[[block]] to i32 ()*
> -- // SPIR: %[[blk_gen_ptr:.*]] = addrspacecast i32 ()* %[[blk_ptr]]
> to i32 () addrspace(4)*
> -- // SPIR: store i32 () addrspace(4)* %[[blk_gen_ptr]], i32 ()
> addrspace(4)** %[[block_B:.*]],
> -- // SPIR: %[[blk_gen_ptr:.*]] = load i32 () addrspace(4)*, i32 ()
> addrspace(4)** %[[block_B]]
> -- // SPIR: %[[block_literal:.*]] = bitcast i32 () addrspace(4)*
> %[[blk_gen_ptr]] to %struct.__opencl_block_literal_generic
> addrspace(4)*
> -+ // SPIR: %[[blk_ptr:.*]] = bitcast <{ i32, i32, i8 addrspace(4)*,
> i32 }>* %[[block]] to %struct.__opencl_block_literal_generic*
> -+ // SPIR: %[[blk_gen_ptr:.*]] = addrspacecast
> %struct.__opencl_block_literal_generic* %[[blk_ptr]] to
> %struct.__opencl_block_literal_generic addrspace(4)*
> -+ // SPIR: store %struct.__opencl_block_literal_generic
> addrspace(4)* %[[blk_gen_ptr]],
> %struct.__opencl_block_literal_generic addrspace(4)**
> %[[block_B:.*]],
> -+ // SPIR: %[[block_literal:.*]] = load
> %struct.__opencl_block_literal_generic addrspace(4)*,
> %struct.__opencl_block_literal_generic addrspace(4)** %[[block_B]]
> - // SPIR: %[[invoke_addr:.*]] = getelementptr inbounds
> %struct.__opencl_block_literal_generic,
> %struct.__opencl_block_literal_generic addrspace(4)*
> %[[block_literal]], i32 0, i32 2
> - // SPIR: %[[blk_gen_ptr:.*]] = bitcast
> %struct.__opencl_block_literal_generic addrspace(4)*
> %[[block_literal]] to i8 addrspace(4)*
> - // SPIR: %[[invoke_func_ptr:.*]] = load i8 addrspace(4)*, i8
> addrspace(4)* addrspace(4)* %[[invoke_addr]]
> -@@ -50,11 +49,10 @@ void foo(){
> - // AMDGCN: %[[block_captured:.*]] = getelementptr inbounds <{
> i32, i32, i8*, i32 }>, <{ i32, i32, i8*, i32 }> addrspace(5)*
> %[[block]], i32 0, i32 3
> - // AMDGCN: %[[i_value:.*]] = load i32, i32 addrspace(5)* %i
> - // AMDGCN: store i32 %[[i_value]], i32 addrspace(5)*
> %[[block_captured]],
> -- // AMDGCN: %[[blk_ptr:.*]] = bitcast <{ i32, i32, i8*, i32 }>
> addrspace(5)* %[[block]] to i32 () addrspace(5)*
> -- // AMDGCN: %[[blk_gen_ptr:.*]] = addrspacecast i32 ()
> addrspace(5)* %[[blk_ptr]] to i32 ()*
> -- // AMDGCN: store i32 ()* %[[blk_gen_ptr]], i32 ()* addrspace(5)*
> %[[block_B:.*]],
> -- // AMDGCN: %[[blk_gen_ptr:.*]] = load i32 ()*, i32 ()*
> addrspace(5)* %[[block_B]]
> -- // AMDGCN: %[[block_literal:.*]] = bitcast i32 ()*
> %[[blk_gen_ptr]] to %struct.__opencl_block_literal_generic*
> -+ // AMDGCN: %[[blk_ptr:.*]] = bitcast <{ i32, i32, i8*, i32 }>
> addrspace(5)* %[[block]] to %struct.__opencl_block_literal_generic
> addrspace(5)*
> -+ // AMDGCN: %[[blk_gen_ptr:.*]] = addrspacecast
> %struct.__opencl_block_literal_generic addrspace(5)* %[[blk_ptr]] to
> %struct.__opencl_block_literal_generic*
> -+ // AMDGCN: store %struct.__opencl_block_literal_generic*
> %[[blk_gen_ptr]], %struct.__opencl_block_literal_generic*
> addrspace(5)* %[[block_B:.*]],
> -+ // AMDGCN: %[[block_literal:.*]] = load
> %struct.__opencl_block_literal_generic*,
> %struct.__opencl_block_literal_generic* addrspace(5)* %[[block_B]]
> - // AMDGCN: %[[invoke_addr:.*]] = getelementptr inbounds
> %struct.__opencl_block_literal_generic,
> %struct.__opencl_block_literal_generic* %[[block_literal]], i32 0,
> i32 2
> - // AMDGCN: %[[blk_gen_ptr:.*]] = bitcast
> %struct.__opencl_block_literal_generic* %[[block_literal]] to i8*
> - // AMDGCN: %[[invoke_func_ptr:.*]] = load i8*, i8**
> %[[invoke_addr]]
> -diff --git a/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> b/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> -index 4732194..8445016 100644
> ---- a/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> -+++ b/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> -@@ -11,7 +11,7 @@ typedef struct {int a;} ndrange_t;
> -
> - // For a block global variable, first emit the block literal as a
> global variable, then emit the block variable itself.
> - // COMMON: [[BL_GLOBAL:@__block_literal_global[^ ]*]] = internal
> addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-
> 9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast
> (void (i8 addrspace(4)*, i8 addrspace(3)*)* [[INV_G:@[^ ]+]] to i8*)
> to i8 addrspace(4)*) }
> --// COMMON: @block_G = addrspace(1) constant void (i8
> addrspace(3)*) addrspace(4)* addrspacecast (void (i8 addrspace(3)*)
> addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> [[BL_GLOBAL]] to void (i8 addrspace(3)*) addrspace(1)*) to void (i8
> addrspace(3)*) addrspace(4)*)
> -+// COMMON: @block_G = addrspace(1) constant
> %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast
> (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32,
> i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to
> %struct.__opencl_block_literal_generic addrspace(1)*) to
> %struct.__opencl_block_literal_generic addrspace(4)*)
> -
> - // For anonymous blocks without captures, emit block literals as
> global variable.
> - // COMMON: [[BLG1:@__block_literal_global[^ ]*]] = internal
> addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-
> 9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast
> (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8
> addrspace(4)*) }
> -@@ -77,9 +77,9 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> %opencl.queue_t{{.*}}** %default_queue
> - // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> - // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast
> (void (i8 addrspace(4)*)*
> [[INVL1:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8
> addrspace(4)*), i8 addrspace(4)** %block.invoke
> -- // B32: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*,
> i32 addrspace(1)*, i32, i32 addrspace(1)* }>* %block to void ()*
> -- // B64: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*,
> i32 addrspace(1)*, i32 addrspace(1)*, i32 }>* %block to void ()*
> -- // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast void ()* [[BL]] to
> i8 addrspace(4)*
> -+ // B32: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*,
> i32 addrspace(1)*, i32, i32 addrspace(1)* }>* %block to
> %struct.__opencl_block_literal_generic*
> -+ // B64: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*,
> i32 addrspace(1)*, i32 addrspace(1)*, i32 }>* %block to
> %struct.__opencl_block_literal_generic*
> -+ // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast
> %struct.__opencl_block_literal_generic* [[BL]] to i8 addrspace(4)*
> - // COMMON-LABEL: call i32 @__enqueue_kernel_basic(
> - // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> %struct.ndrange_t* byval [[NDR]]{{([0-9]+)?}},
> - // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> ({{.*}} [[INVLK1:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> -@@ -95,8 +95,8 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - // COMMON: [[WAIT_EVNT:%[0-9]+]] = addrspacecast
> %opencl.clk_event_t{{.*}}** %event_wait_list to
> %opencl.clk_event_t{{.*}}* addrspace(4)*
> - // COMMON: [[EVNT:%[0-9]+]] = addrspacecast
> %opencl.clk_event_t{{.*}}** %clk_event to %opencl.clk_event_t{{.*}}*
> addrspace(4)*
> - // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast
> (void (i8 addrspace(4)*)*
> [[INVL2:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8
> addrspace(4)*), i8 addrspace(4)** %block.invoke
> -- // COMMON: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8
> addrspace(4)*, i32{{.*}}, i32{{.*}}, i32{{.*}} }>* %block3 to void
> ()*
> -- // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast void ()* [[BL]] to
> i8 addrspace(4)*
> -+ // COMMON: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8
> addrspace(4)*, i32{{.*}}, i32{{.*}}, i32{{.*}} }>* %block3 to
> %struct.__opencl_block_literal_generic*
> -+ // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast
> %struct.__opencl_block_literal_generic* [[BL]] to i8 addrspace(4)*
> - // COMMON-LABEL: call i32 @__enqueue_kernel_basic_events
> - // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32
> [[FLAGS]], %struct.ndrange_t* {{.*}}, i32 2,
> %opencl.clk_event_t{{.*}}* addrspace(4)* [[WAIT_EVNT]],
> %opencl.clk_event_t{{.*}}* addrspace(4)* [[EVNT]],
> - // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> ({{.*}} [[INVLK2:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> -@@ -300,13 +300,13 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - // Emits global block literal [[BLG8]] and invoke function
> [[INVG8]].
> - // The full type of these expressions are long (and repeated
> elsewhere), so we
> - // capture it as part of the regex for convenience and clarity.
> -- // COMMON: store void () addrspace(4)* addrspacecast (void ()
> addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> [[BLG8]] to void () addrspace(1)*) to void () addrspace(4)*), void ()
> addrspace(4)** %block_A
> -+ // COMMON: store %struct.__opencl_block_literal_generic
> addrspace(4)* addrspacecast (%struct.__opencl_block_literal_generic
> addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> [[BLG8]] to %struct.__opencl_block_literal_generic addrspace(1)*) to
> %struct.__opencl_block_literal_generic addrspace(4)*),
> %struct.__opencl_block_literal_generic addrspace(4)** %block_A
> - void (^const block_A)(void) = ^{
> - return;
> - };
> -
> - // Emits global block literal [[BLG9]] and invoke function
> [[INVG9]].
> -- // COMMON: store void (i8 addrspace(3)*) addrspace(4)*
> addrspacecast (void (i8 addrspace(3)*) addrspace(1)* bitcast ({ i32,
> i32, i8 addrspace(4)* } addrspace(1)* [[BLG9]] to void (i8
> addrspace(3)*) addrspace(1)*) to void (i8 addrspace(3)*)
> addrspace(4)*), void (i8 addrspace(3)*) addrspace(4)** %block_B
> -+ // COMMON: store %struct.__opencl_block_literal_generic
> addrspace(4)* addrspacecast (%struct.__opencl_block_literal_generic
> addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> [[BLG9]] to %struct.__opencl_block_literal_generic addrspace(1)*) to
> %struct.__opencl_block_literal_generic addrspace(4)*),
> %struct.__opencl_block_literal_generic addrspace(4)** %block_B
> - void (^const block_B)(local void *) = ^(local void *a) {
> - return;
> - };
> -@@ -346,7 +346,7 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast
> (void (i8 addrspace(4)*)*
> [[INVL3:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8
> addrspace(4)*), i8 addrspace(4)** %block.invoke
> - // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> %opencl.queue_t{{.*}}** %default_queue
> - // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> -- // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast void ()* {{.*}} to
> i8 addrspace(4)*
> -+ // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast
> %struct.__opencl_block_literal_generic* {{.*}} to i8 addrspace(4)*
> - // COMMON-LABEL: call i32 @__enqueue_kernel_basic(
> - // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> %struct.ndrange_t* byval [[NDR]]{{([0-9]+)?}},
> - // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> ({{.*}} [[INVLK3:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> ---
> -1.8.3.1
> -
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-Update-LowerOpenCL-pass-to-handle-new-
> blocks-represn.patch b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-Update-LowerOpenCL-pass-to-handle-new-
> blocks-represn.patch
> deleted file mode 100644
> index 2037421..0000000
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> Update-LowerOpenCL-pass-to-handle-new-blocks-represn.patch
> +++ /dev/null
> @@ -1,986 +0,0 @@
> -From 177cce531fd3665bb964a03db51890e0241e3e72 Mon Sep 17 00:00:00
> 2001
> -From: Alexey Sotkin <alexey.sotkin at intel.com>
> -Date: Thu, 21 Feb 2019 17:14:36 +0300
> -Subject: [PATCH] Update LowerOpenCL pass to handle new blocks
> represntation in
> - LLVM IR
> -
> -Upstream-Status: Backport [
> https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/bd6ddfaf7232cd81c7f2fe9877e66f286731bd8e
> ]
> -Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> -
> ----
> - lib/SPIRV/SPIRVLowerOCLBlocks.cpp | 249 ++++---------------
> -----------
> - test/global_block.ll | 71 ++++-----
> - test/literal-struct.ll | 31 ++--
> - test/transcoding/block_w_struct_return.ll | 47 +++---
> - test/transcoding/enqueue_kernel.ll | 237 ++++++++++++++++---
> ---------
> - 5 files changed, 235 insertions(+), 400 deletions(-)
> -
> -diff --git a/lib/SPIRV/SPIRVLowerOCLBlocks.cpp
> b/lib/SPIRV/SPIRVLowerOCLBlocks.cpp
> -index c80bf04..b42a4ec 100644
> ---- a/lib/SPIRV/SPIRVLowerOCLBlocks.cpp
> -+++ b/lib/SPIRV/SPIRVLowerOCLBlocks.cpp
> -@@ -40,207 +40,34 @@
> - // In both cases values with function type used as intermediate
> representation
> - // for block literal structure.
> - //
> --// This pass is designed to find such cases and simplify them to
> avoid any
> --// function pointer types occurrences in LLVM IR in 4 steps.
> --//
> --// 1. Find all function pointer allocas, like
> --// %block = alloca void () *
> --//
> --// Then find a single store to that alloca:
> --// %blockLit = alloca <{ i32, i32, ...}>, align 4
> --// %0 = bitcast <{ i32, i32, ... }>* %blockLit to void ()*
> --// > store void ()* %0, void ()** %block, align 4
> --//
> --// And replace the alloca users by new instructions which used
> stored value
> --// %blockLit itself instead of function pointer alloca %block.
> --//
> --// 2. Find consecutive casts from block literal type to i8
> addrspace(4)*
> --// used function pointers as an intermediate type:
> --// %0 = bitcast <{ i32, i32 }> %block to void() *
> --// %1 = addrspacecast void() * %0 to i8 addrspace(4)*
> --// And simplify them:
> --// %2 = addrspacecast <{ i32, i32 }> %block to i8
> addrspace(4)*
> --//
> --// 3. Find all unused instructions with function pointer type
> occured after
> --// pp.1-2 and remove them.
> --//
> --// 4. Find unused globals with function pointer type, like
> --// @block = constant void ()*
> --// bitcast ({ i32, i32 }* @__block_literal_global to
> void ()*
> --//
> --// And remove them.
> -+// In LLVM IR produced by clang, blocks are represented with the
> following
> -+// structure:
> -+// %struct.__opencl_block_literal_generic = type { i32, i32, i8
> addrspace(4)* }
> -+// Pointers to block invoke functions are stored in the third
> field. Clang
> -+// replaces inderect function calls in all cases except if block is
> passed as a
> -+// function argument. Note that it is somewhat unclear if the
> OpenCL C spec
> -+// should allow passing blocks as function argumernts. This pass is
> not supposed
> -+// to work correctly with such functions.
> -+// Clang though has to store function pointers to this structure.
> Purpose of
> -+// this pass is to replace store of function pointers(not allowed
> in SPIR-V)
> -+// with null pointers.
> - //
> - //===------------------------------------------------------------
> ----------===//
> - #define DEBUG_TYPE "spv-lower-ocl-blocks"
> -
> --#include "OCLUtil.h"
> - #include "SPIRVInternal.h"
> -
> --#include "llvm/ADT/SetVector.h"
> --#include "llvm/Analysis/ValueTracking.h"
> --#include "llvm/IR/GlobalVariable.h"
> --#include "llvm/IR/InstIterator.h"
> - #include "llvm/IR/Module.h"
> - #include "llvm/Pass.h"
> --#include "llvm/PassSupport.h"
> --#include "llvm/Support/Casting.h"
> -+#include "llvm/Support/Regex.h"
> -
> - using namespace llvm;
> -
> - namespace {
> -
> --static void
> --removeUnusedFunctionPtrInst(Instruction *I,
> -- SmallSetVector<Instruction *, 16>
> &FuncPtrInsts) {
> -- for (unsigned OpIdx = 0, Ops = I->getNumOperands(); OpIdx != Ops;
> ++OpIdx) {
> -- Instruction *OpI = dyn_cast<Instruction>(I->getOperand(OpIdx));
> -- I->setOperand(OpIdx, nullptr);
> -- if (OpI && OpI != I && OpI->user_empty())
> -- FuncPtrInsts.insert(OpI);
> -- }
> -- I->eraseFromParent();
> --}
> --
> --static bool isFuncPtrAlloca(const AllocaInst *AI) {
> -- auto *ET = dyn_cast<PointerType>(AI->getAllocatedType());
> -- return ET && ET->getElementType()->isFunctionTy();
> --}
> --
> --static bool hasFuncPtrType(const Value *V) {
> -- auto *PT = dyn_cast<PointerType>(V->getType());
> -- return PT && PT->getElementType()->isFunctionTy();
> --}
> --
> --static bool isFuncPtrInst(const Instruction *I) {
> -- if (auto *AI = dyn_cast<AllocaInst>(I))
> -- return isFuncPtrAlloca(AI);
> --
> -- for (auto &Op : I->operands()) {
> -- if (auto *AI = dyn_cast<AllocaInst>(Op))
> -- return isFuncPtrAlloca(AI);
> --
> -- auto *OpI = dyn_cast<Instruction>(&Op);
> -- if (OpI && OpI != I && hasFuncPtrType(OpI))
> -- return true;
> -- }
> -- return false;
> --}
> --
> --static StoreInst *findSingleStore(AllocaInst *AI) {
> -- StoreInst *Store = nullptr;
> -- for (auto *U : AI->users()) {
> -- if (!isa<StoreInst>(U))
> -- continue; // not a store
> -- if (Store)
> -- return nullptr; // there are more than one stores
> -- Store = dyn_cast<StoreInst>(U);
> -- }
> -- return Store;
> --}
> --
> --static void fixFunctionPtrAllocaUsers(AllocaInst *AI) {
> -- // Find and remove a single store to alloca
> -- auto *SingleStore = findSingleStore(AI);
> -- assert(SingleStore && "More than one store to the function
> pointer alloca");
> -- auto *StoredVal = SingleStore->getValueOperand();
> -- SingleStore->eraseFromParent();
> --
> -- // Find loads from the alloca and replace thier users
> -- for (auto *U : AI->users()) {
> -- auto *LI = dyn_cast<LoadInst>(U);
> -- if (!LI)
> -- continue;
> --
> -- for (auto *U : LI->users()) {
> -- auto *UInst = cast<Instruction>(U);
> -- auto *Cast = CastInst::CreatePointerBitCastOrAddrSpaceCast(
> -- StoredVal, UInst->getType(), "", UInst);
> -- UInst->replaceAllUsesWith(Cast);
> -- }
> -- }
> --}
> --
> --static int getBlockLiteralIdx(const Function &F) {
> -- StringRef FName = F.getName();
> -- if (isEnqueueKernelBI(FName))
> -- return FName.contains("events") ? 7 : 4;
> -- if (isKernelQueryBI(FName))
> -- return FName.contains("for_ndrange") ? 2 : 1;
> -- if (FName.startswith("__") && FName.contains("_block_invoke"))
> -- return F.hasStructRetAttr() ? 1 : 0;
> --
> -- return -1; // No block literal argument
> --}
> --
> --static bool hasBlockLiteralArg(const Function &F) {
> -- return getBlockLiteralIdx(F) != -1;
> --}
> --
> --static bool simplifyFunctionPtrCasts(Function &F) {
> -- bool Changed = false;
> -- int BlockLiteralIdx = getBlockLiteralIdx(F);
> -- for (auto *U : F.users()) {
> -- auto *Call = dyn_cast<CallInst>(U);
> -- if (!Call)
> -- continue;
> -- if (Call->getFunction()->getName() == F.getName().str() +
> "_kernel")
> -- continue; // Skip block invoke function calls inside block
> invoke kernels
> --
> -- const DataLayout &DL = F.getParent()->getDataLayout();
> -- auto *BlockLiteral = Call->getOperand(BlockLiteralIdx);
> -- auto *BlockLiteralVal = GetUnderlyingObject(BlockLiteral, DL);
> -- if (isa<GlobalVariable>(BlockLiteralVal))
> -- continue; // nothing to do with globals
> --
> -- auto *BlockLiteralAlloca = cast<AllocaInst>(BlockLiteralVal);
> -- assert(!BlockLiteralAlloca->getAllocatedType()->isFunctionTy()
> &&
> -- "Function type shouldn't be there");
> --
> -- auto *NewBlockLiteral =
> CastInst::CreatePointerBitCastOrAddrSpaceCast(
> -- BlockLiteralAlloca, BlockLiteral->getType(), "", Call);
> -- BlockLiteral->replaceAllUsesWith(NewBlockLiteral);
> -- Changed |= true;
> -- }
> -- return Changed;
> --}
> --
> --static void
> --findFunctionPtrAllocas(Module &M,
> -- SmallVectorImpl<AllocaInst *>
> &FuncPtrAllocas) {
> -- for (auto &F : M) {
> -- if (F.isDeclaration())
> -- continue;
> -- for (auto &I : instructions(F)) {
> -- auto *AI = dyn_cast<AllocaInst>(&I);
> -- if (!AI || !isFuncPtrAlloca(AI))
> -- continue;
> -- FuncPtrAllocas.push_back(AI);
> -- }
> -- }
> --}
> --
> --static void
> --findUnusedFunctionPtrInsts(Module &M,
> -- SmallSetVector<Instruction *, 16>
> &FuncPtrInsts) {
> -- for (auto &F : M) {
> -- if (F.isDeclaration())
> -- continue;
> -- for (auto &I : instructions(F))
> -- if (I.user_empty() && isFuncPtrInst(&I))
> -- FuncPtrInsts.insert(&I);
> -- }
> --}
> --
> --static void
> --findUnusedFunctionPtrGlbs(Module &M,
> -- SmallVectorImpl<GlobalVariable *>
> &FuncPtrGlbs) {
> -- for (auto &GV : M.globals()) {
> -- if (!GV.user_empty())
> -- continue;
> -- auto *GVType = dyn_cast<PointerType>(GV.getType()-
> >getElementType());
> -- if (GVType && GVType->getElementType()->isFunctionTy())
> -- FuncPtrGlbs.push_back(&GV);
> -- }
> -+static bool isBlockInvoke(Function &F) {
> -+ static Regex BlockInvokeRegex("_block_invoke_?[0-9]*$");
> -+ return BlockInvokeRegex.match(F.getName());
> - }
> -
> - class SPIRVLowerOCLBlocks : public ModulePass {
> -@@ -250,44 +77,24 @@ public:
> -
> - bool runOnModule(Module &M) {
> - bool Changed = false;
> --
> -- // 1. Find function pointer allocas and fix their users
> -- SmallVector<AllocaInst *, 16> FuncPtrAllocas;
> -- findFunctionPtrAllocas(M, FuncPtrAllocas);
> --
> -- Changed |= !FuncPtrAllocas.empty();
> -- for (auto *AI : FuncPtrAllocas)
> -- fixFunctionPtrAllocaUsers(AI);
> --
> -- // 2. Simplify consecutive casts which use function pointer
> types
> -- for (auto &F : M)
> -- if (hasBlockLiteralArg(F))
> -- Changed |= simplifyFunctionPtrCasts(F);
> --
> -- // 3. Cleanup unused instructions with function pointer type
> -- // which are occured after pp. 1-2
> -- SmallSetVector<Instruction *, 16> FuncPtrInsts;
> -- findUnusedFunctionPtrInsts(M, FuncPtrInsts);
> --
> -- Changed |= !FuncPtrInsts.empty();
> -- while (!FuncPtrInsts.empty()) {
> -- Instruction *I = FuncPtrInsts.pop_back_val();
> -- removeUnusedFunctionPtrInst(I, FuncPtrInsts);
> -+ for (Function &F : M) {
> -+ if (!isBlockInvoke(F))
> -+ continue;
> -+ for (User *U : F.users()) {
> -+ if (!isa<Constant>(U))
> -+ continue;
> -+ Constant *Null = Constant::getNullValue(U->getType());
> -+ if (U != Null) {
> -+ U->replaceAllUsesWith(Null);
> -+ Changed = true;
> -+ }
> -+ }
> - }
> --
> -- // 4. Find and remove unused global variables with function
> pointer type
> -- SmallVector<GlobalVariable *, 16> FuncPtrGlbs;
> -- findUnusedFunctionPtrGlbs(M, FuncPtrGlbs);
> --
> -- Changed |= !FuncPtrGlbs.empty();
> -- for (auto *GV : FuncPtrGlbs)
> -- GV->eraseFromParent();
> --
> - return Changed;
> - }
> -
> - static char ID;
> --}; // class SPIRVLowerOCLBlocks
> -+};
> -
> - char SPIRVLowerOCLBlocks::ID = 0;
> -
> -diff --git a/test/global_block.ll b/test/global_block.ll
> -index 4fc453b..b558213 100644
> ---- a/test/global_block.ll
> -+++ b/test/global_block.ll
> -@@ -17,7 +17,7 @@
> - ; RUN: spirv-val %t.spv
> - ; RUN: llvm-spirv -r %t.spv -o - | llvm-dis | FileCheck %s --check-
> prefix=CHECK-LLVM
> -
> --target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-
> v96:128-v192:256-v256:256-v512:512-v1024:1024-n8:16:32:64"
> -+target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-
> v96:128-v192:256-v256:256-v512:512-v1024:1024"
> - target triple = "spir-unknown-unknown"
> -
> - ; CHECK-SPIRV: Name [[block_invoke:[0-9]+]] "_block_invoke"
> -@@ -27,71 +27,56 @@ target triple = "spir-unknown-unknown"
> - ; CHECK-SPIRV: TypePointer [[int8Ptr:[0-9]+]] 8 [[int8]]
> - ; CHECK-SPIRV: TypeFunction [[block_invoke_type:[0-9]+]] [[int]]
> [[int8Ptr]] [[int]]
> -
> --;; This variable is not needed in SPIRV
> --; CHECK-SPIRV-NOT: Name {{[0-9]+}} block_kernel.b1
> --; CHECK-LLVM-NOT: @block_kernel.b1
> -- at block_kernel.b1 = internal addrspace(2) constant i32 (i32)
> addrspace(4)* addrspacecast (i32 (i32) addrspace(1)* bitcast ({ i32,
> i32 } addrspace(1)* @__block_literal_global to i32 (i32)
> addrspace(1)*) to i32 (i32) addrspace(4)*), align 8
> -+%struct.__opencl_block_literal_generic = type { i32, i32, i8
> addrspace(4)* }
> -
> -- at __block_literal_global = internal addrspace(1) constant { i32, i32
> } { i32 8, i32 4 }, align 4
> -+ at block_kernel.b1 = internal addrspace(2) constant
> %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast
> (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32,
> i32, i8 addrspace(4)* } addrspace(1)* @__block_literal_global to
> %struct.__opencl_block_literal_generic addrspace(1)*) to
> %struct.__opencl_block_literal_generic addrspace(4)*), align 4
> -+ at __block_literal_global = internal addrspace(1) constant { i32,
> i32, i8 addrspace(4)* } { i32 12, i32 4, i8 addrspace(4)*
> addrspacecast (i8* bitcast (i32 (i8 addrspace(4)*, i32)*
> @_block_invoke to i8*) to i8 addrspace(4)*) }, align 4
> -
> --; Function Attrs: convergent nounwind
> --define spir_kernel void @block_kernel(i32 addrspace(1)* %res) #0
> !kernel_arg_addr_space !4 !kernel_arg_access_qual !5 !kernel_arg_type
> !6 !kernel_arg_base_type !6 !kernel_arg_type_qual !7 {
> -+; Function Attrs: convergent noinline nounwind optnone
> -+define spir_kernel void @block_kernel(i32 addrspace(1)* %res) #0
> !kernel_arg_addr_space !3 !kernel_arg_access_qual !4 !kernel_arg_type
> !5 !kernel_arg_base_type !5 !kernel_arg_type_qual !6 {
> - entry:
> -- %res.addr = alloca i32 addrspace(1)*, align 8
> -- store i32 addrspace(1)* %res, i32 addrspace(1)** %res.addr, align
> 8, !tbaa !10
> --
> -+ %res.addr = alloca i32 addrspace(1)*, align 4
> -+ store i32 addrspace(1)* %res, i32 addrspace(1)** %res.addr, align
> 4
> - ; CHECK-SPIRV: FunctionCall [[int]] {{[0-9]+}} [[block_invoke]]
> {{[0-9]+}} [[five]]
> - ; CHECK-LLVM: %call = call spir_func i32 @_block_invoke(i8
> addrspace(4)* {{.*}}, i32 5)
> -- %call = call spir_func i32 @_block_invoke(i8 addrspace(4)*
> addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32 } addrspace(1)*
> @__block_literal_global to i8 addrspace(1)*) to i8 addrspace(4)*),
> i32 5) #2
> --
> -- %0 = load i32 addrspace(1)*, i32 addrspace(1)** %res.addr, align
> 8, !tbaa !10
> -- store i32 %call, i32 addrspace(1)* %0, align 4, !tbaa !14
> -+ %call = call spir_func i32 @_block_invoke(i8 addrspace(4)*
> addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)*
> } addrspace(1)* @__block_literal_global to i8 addrspace(1)*) to i8
> addrspace(4)*), i32 5) #2
> -+ %0 = load i32 addrspace(1)*, i32 addrspace(1)** %res.addr, align
> 4
> -+ store i32 %call, i32 addrspace(1)* %0, align 4
> - ret void
> - }
> -
> --; CHECK-SPIRV: 5 Function [[int]] [[block_invoke]] 0
> [[block_invoke_type]]
> -+; CHECK-SPIRV: 5 Function [[int]] [[block_invoke]] 2
> [[block_invoke_type]]
> - ; CHECK-SPIRV-NEXT: 3 FunctionParameter [[int8Ptr]] {{[0-9]+}}
> - ; CHECK-SPIRV-NEXT: 3 FunctionParameter [[int]] {{[0-9]+}}
> - ; CHECK-LLVM: define internal spir_func i32 @_block_invoke(i8
> addrspace(4)* {{.*}}, i32 %{{.*}})
> --; Function Attrs: convergent nounwind
> -+; Function Attrs: convergent noinline nounwind optnone
> - define internal spir_func i32 @_block_invoke(i8 addrspace(4)*
> %.block_descriptor, i32 %i) #1 {
> - entry:
> -- %.block_descriptor.addr = alloca i8 addrspace(4)*, align 8
> -+ %.block_descriptor.addr = alloca i8 addrspace(4)*, align 4
> - %i.addr = alloca i32, align 4
> -- store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 8
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32 }> addrspace(4)*
> -- store i32 %i, i32* %i.addr, align 4, !tbaa !14
> -- %0 = load i32, i32* %i.addr, align 4, !tbaa !14
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)* }>
> addrspace(4)*, align 4
> -+ store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 4
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)* }> addrspace(4)*
> -+ store i32 %i, i32* %i.addr, align 4
> -+ store <{ i32, i32, i8 addrspace(4)* }> addrspace(4)* %block, <{
> i32, i32, i8 addrspace(4)* }> addrspace(4)** %block.addr, align 4
> -+ %0 = load i32, i32* %i.addr, align 4
> - %add = add nsw i32 %0, 1
> - ret i32 %add
> - }
> -
> --attributes #0 = { convergent nounwind "correctly-rounded-divide-
> sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-
> fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "uniform-work-group-size"="false" "unsafe-
> fp-math"="false" "use-soft-float"="false" }
> --attributes #1 = { convergent nounwind "correctly-rounded-divide-
> sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-
> fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-
> float"="false" }
> -+attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "uniform-work-group-size"="false" "unsafe-
> fp-math"="false" "use-soft-float"="false" }
> -+attributes #1 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-
> float"="false" }
> - attributes #2 = { convergent }
> -
> - !llvm.module.flags = !{!0}
> --!opencl.enable.FP_CONTRACT = !{}
> - !opencl.ocl.version = !{!1}
> - !opencl.spir.version = !{!1}
> --!opencl.used.extensions = !{!2}
> --!opencl.used.optional.core.features = !{!2}
> --!opencl.compiler.options = !{!2}
> --!llvm.ident = !{!3}
> -+!llvm.ident = !{!2}
> -
> - !0 = !{i32 1, !"wchar_size", i32 4}
> - !1 = !{i32 2, i32 0}
> --!2 = !{}
> --!3 = !{!"clang version 7.0.0"}
> --!4 = !{i32 1}
> --!5 = !{!"none"}
> --!6 = !{!"int*"}
> --!7 = !{!""}
> --!8 = !{i1 false}
> --!9 = !{i32 0}
> --!10 = !{!11, !11, i64 0}
> --!11 = !{!"any pointer", !12, i64 0}
> --!12 = !{!"omnipotent char", !13, i64 0}
> --!13 = !{!"Simple C/C++ TBAA"}
> --!14 = !{!15, !15, i64 0}
> --!15 = !{!"int", !12, i64 0}
> -+!2 = !{!"clang version 9.0.0 (https://llvm.org/git/clang
> 04fb8964a801a5c5d7baa5a22272243a7d183896) (https://llvm.org/git/llvm
384f64397f6ad95a361b72d62c07d7bac9f24163)"}
> -+!3 = !{i32 1}
> -+!4 = !{!"none"}
> -+!5 = !{!"int*"}
> -+!6 = !{!""}
> -diff --git a/test/literal-struct.ll b/test/literal-struct.ll
> -index b88187f..dec957a 100644
> ---- a/test/literal-struct.ll
> -+++ b/test/literal-struct.ll
> -@@ -2,7 +2,7 @@
> - ; structs, i.e. structs whose type has no name. Typicaly clang
> generate such
> - ; structs if the kernel contains OpenCL 2.0 blocks. The IR was
> produced with
> - ; the following command:
> --; clang -cc1 -triple spir -cl-std=cl2.0 -O0 -finclude-default-
> header literal-struct.cl -emit-llvm -o test/literal-struct.ll
> -+; clang -cc1 -triple spir -cl-std=cl2.0 -O0 literal-struct.cl
> -emit-llvm -o test/literal-struct.ll
> -
> - ; literal-struct.cl:
> - ; void foo()
> -@@ -17,25 +17,28 @@
> - ; RUN: llvm-spirv %t.bc -o %t.spv
> - ; RUN: spirv-val %t.spv
> -
> --; CHECK-DAG: TypeInt [[Int:[0-9]+]] 32 0
> --; CHECK-DAG: TypeStruct [[StructType:[0-9]+]] [[Int]] [[Int]] {{$}}
> -+; CHECK: TypeInt [[Int:[0-9]+]] 32 0
> -+; CHECK: TypeInt [[Int8:[0-9]+]] 8 0
> -+; CHECK: TypePointer [[Int8Ptr:[0-9]+]] 8 [[Int8]]
> -+; CHECK: TypeStruct [[StructType:[0-9]+]] [[Int]] [[Int]]
> [[Int8Ptr]]
> -
> - target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-
> v96:128-v192:256-v256:256-v512:512-v1024:1024"
> - target triple = "spir"
> -
> -- at __block_literal_global = internal addrspace(1) constant { i32, i32
> } { i32 8, i32 4 }, align 4
> -+%struct.__opencl_block_literal_generic = type { i32, i32, i8
> addrspace(4)* }
> -+
> -+ at __block_literal_global = internal addrspace(1) constant { i32,
> i32, i8 addrspace(4)* } { i32 12, i32 4, i8 addrspace(4)*
> addrspacecast (i8* bitcast (void (i8 addrspace(4)*)*
> @__foo_block_invoke to i8*) to i8 addrspace(4)*) }, align 4
> - ; CHECK: ConstantComposite [[StructType]]
> -
> --; This is artificial case is added to cover ConstantNull
> instrucitions with TypeStruct.
> -- at __block_literal_global.1 = internal addrspace(1) constant { i32,
> i32 } zeroinitializer, align 4
> -+ at __block_literal_global.1 = internal addrspace(1) constant { i32,
> i32, i8 addrspace(4)* } zeroinitializer, align 4
> - ; CHECK: ConstantNull [[StructType]]
> -
> - ; Function Attrs: convergent noinline nounwind optnone
> - define spir_func void @foo() #0 {
> - entry:
> -- %myBlock = alloca void () addrspace(4)*, align 4
> -- store void () addrspace(4)* addrspacecast (void () addrspace(1)*
> bitcast ({ i32, i32 } addrspace(1)* @__block_literal_global to void
> () addrspace(1)*) to void () addrspace(4)*), void () addrspace(4)**
> %myBlock, align 4
> -- call spir_func void @__foo_block_invoke(i8 addrspace(4)*
> addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32 } addrspace(1)*
> @__block_literal_global to i8 addrspace(1)*) to i8 addrspace(4)*)) #1
> -+ %myBlock = alloca %struct.__opencl_block_literal_generic
> addrspace(4)*, align 4
> -+ store %struct.__opencl_block_literal_generic addrspace(4)*
> addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)*
> bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> @__block_literal_global to %struct.__opencl_block_literal_generic
> addrspace(1)*) to %struct.__opencl_block_literal_generic
> addrspace(4)*), %struct.__opencl_block_literal_generic addrspace(4)**
> %myBlock, align 4
> -+ call spir_func void @__foo_block_invoke(i8 addrspace(4)*
> addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)*
> } addrspace(1)* @__block_literal_global to i8 addrspace(1)*) to i8
> addrspace(4)*)) #1
> - ret void
> - }
> -
> -@@ -43,14 +46,14 @@ entry:
> - define internal spir_func void @__foo_block_invoke(i8 addrspace(4)*
> %.block_descriptor) #0 {
> - entry:
> - %.block_descriptor.addr = alloca i8 addrspace(4)*, align 4
> -- %block.addr = alloca <{ i32, i32 }> addrspace(4)*, align 4
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)* }>
> addrspace(4)*, align 4
> - store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 4
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32 }> addrspace(4)*
> -- store <{ i32, i32 }> addrspace(4)* %block, <{ i32, i32 }>
> addrspace(4)** %block.addr, align 4
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)* }> addrspace(4)*
> -+ store <{ i32, i32, i8 addrspace(4)* }> addrspace(4)* %block, <{
> i32, i32, i8 addrspace(4)* }> addrspace(4)** %block.addr, align 4
> - ret void
> - }
> -
> --attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-
> pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-
> tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-
> math"="false" "no-trapping-math"="false" "stack-protector-buffer-
> size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
> -+attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-
> float"="false" }
> - attributes #1 = { convergent }
> -
> - !llvm.module.flags = !{!0}
> -@@ -60,4 +63,4 @@ attributes #1 = { convergent }
> -
> - !0 = !{i32 1, !"wchar_size", i32 4}
> - !1 = !{i32 2, i32 0}
> --!2 = !{!"clang version 8.0.0 "}
> -+!2 = !{!"clang version 9.0.0 (https://llvm.org/git/clang
> 04fb8964a801a5c5d7baa5a22272243a7d183896) (https://llvm.org/git/llvm
384f64397f6ad95a361b72d62c07d7bac9f24163)"}
> -diff --git a/test/transcoding/block_w_struct_return.ll
> b/test/transcoding/block_w_struct_return.ll
> -index a68820f..ebd2c5f 100644
> ---- a/test/transcoding/block_w_struct_return.ll
> -+++ b/test/transcoding/block_w_struct_return.ll
> -@@ -16,6 +16,8 @@
> - ; res[tid] = kernelBlock(aa).a - 6;
> - ; }
> -
> -+; clang -cc1 -triple spir -cl-std=cl2.0 -disable-llvm-passes
> -finclude-default-header block_w_struct_return.cl -emit-llvm -o
> test/transcoding/block_w_struct_return.ll
> -+
> - ; RUN: llvm-as %s -o %t.bc
> - ; RUN: llvm-spirv %t.bc -spirv-text -o %t.spv.txt
> - ; RUN: FileCheck < %t.spv.txt %s --check-prefix=CHECK-SPIRV
> -@@ -28,12 +30,14 @@
> - ; CHECK-SPIRV: Name [[BlockInv:[0-9]+]]
> "__block_ret_struct_block_invoke"
> -
> - ; CHECK-SPIRV: 4 TypeInt [[IntTy:[0-9]+]] 32
> -+; CHECK-SPIRV: 4 TypeInt [[Int8Ty:[0-9]+]] 8
> -+; CHECK-SPIRV: 4 TypePointer [[Int8Ptr:[0-9]+]] 8 [[Int8Ty]]
> - ; CHECK-SPIRV: 3 TypeStruct [[StructTy:[0-9]+]] [[IntTy]]
> - ; CHECK-SPIRV: 4 TypePointer [[StructPtrTy:[0-9]+]] 7 [[StructTy]]
> -
> - ; CHECK-SPIRV: 4 Variable [[StructPtrTy]] [[StructArg:[0-9]+]] 7
> - ; CHECK-SPIRV: 4 Variable [[StructPtrTy]] [[StructRet:[0-9]+]] 7
> --; CHECK-SPIRV: 4 PtrCastToGeneric {{[0-9]+}} [[BlockLit:[0-9]+]]
> {{[0-9]+}}
> -+; CHECK-SPIRV: 4 PtrCastToGeneric [[Int8Ptr]] [[BlockLit:[0-9]+]]
> {{[0-9]+}}
> - ; CHECK-SPIRV: 7 FunctionCall {{[0-9]+}} {{[0-9]+}} [[BlockInv]]
> [[StructRet]] [[BlockLit]] [[StructArg]]
> -
> - ; CHECK-LLVM: %[[StructA:.*]] = type { i32 }
> -@@ -42,20 +46,21 @@
> - target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-
> v192:256-v256:256-v512:512-v1024:1024"
> - target triple = "spir64-unknown-unknown"
> -
> -+%struct.__opencl_block_literal_generic = type { i32, i32, i8
> addrspace(4)* }
> - %struct.A = type { i32 }
> -
> -- at __block_literal_global = internal addrspace(1) constant { i32, i32
> } { i32 8, i32 4 }, align 4
> -+ at __block_literal_global = internal addrspace(1) constant { i32,
> i32, i8 addrspace(4)* } { i32 16, i32 8, i8 addrspace(4)*
> addrspacecast (i8* bitcast (void (%struct.A*, i8 addrspace(4)*,
> %struct.A*)* @__block_ret_struct_block_invoke to i8*) to i8
> addrspace(4)*) }, align 8
> -
> - ; Function Attrs: convergent noinline nounwind optnone
> --define spir_kernel void @block_ret_struct(i32 addrspace(1)* %res)
> #0 !kernel_arg_addr_space !4 !kernel_arg_access_qual !5
> !kernel_arg_type !6 !kernel_arg_base_type !6 !kernel_arg_type_qual !7
> !kernel_arg_host_accessible !8 !kernel_arg_pipe_depth !9
> !kernel_arg_pipe_io !7 !kernel_arg_buffer_location !7 {
> -+define spir_kernel void @block_ret_struct(i32 addrspace(1)* %res)
> #0 !kernel_arg_addr_space !3 !kernel_arg_access_qual !4
> !kernel_arg_type !5 !kernel_arg_base_type !5 !kernel_arg_type_qual !6
> {
> - entry:
> - %res.addr = alloca i32 addrspace(1)*, align 8
> -- %kernelBlock = alloca void (%struct.A*, %struct.A*)
> addrspace(4)*, align 8
> -+ %kernelBlock = alloca %struct.__opencl_block_literal_generic
> addrspace(4)*, align 8
> - %tid = alloca i64, align 8
> - %aa = alloca %struct.A, align 4
> - %tmp = alloca %struct.A, align 4
> - store i32 addrspace(1)* %res, i32 addrspace(1)** %res.addr, align
> 8
> -- store void (%struct.A*, %struct.A*) addrspace(4)* addrspacecast
> (void (%struct.A*, %struct.A*) addrspace(1)* bitcast ({ i32, i32 }
> addrspace(1)* @__block_literal_global to void (%struct.A*,
> %struct.A*) addrspace(1)*) to void (%struct.A*, %struct.A*)
> addrspace(4)*), void (%struct.A*, %struct.A*) addrspace(4)**
> %kernelBlock, align 8
> -+ store %struct.__opencl_block_literal_generic addrspace(4)*
> addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)*
> bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> @__block_literal_global to %struct.__opencl_block_literal_generic
> addrspace(1)*) to %struct.__opencl_block_literal_generic
> addrspace(4)*), %struct.__opencl_block_literal_generic addrspace(4)**
> %kernelBlock, align 8
> - %call = call spir_func i64 @_Z13get_global_idj(i32 0) #4
> - store i64 %call, i64* %tid, align 8
> - %0 = load i32 addrspace(1)*, i32 addrspace(1)** %res.addr, align
> 8
> -@@ -64,7 +69,7 @@ entry:
> - store i32 -1, i32 addrspace(1)* %arrayidx, align 4
> - %a = getelementptr inbounds %struct.A, %struct.A* %aa, i32 0, i32
> 0
> - store i32 5, i32* %a, align 4
> -- call spir_func void @__block_ret_struct_block_invoke(%struct.A*
> sret %tmp, i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast
> ({ i32, i32 } addrspace(1)* @__block_literal_global to i8
> addrspace(1)*) to i8 addrspace(4)*), %struct.A* byval align 4 %aa) #5
> -+ call spir_func void @__block_ret_struct_block_invoke(%struct.A*
> sret %tmp, i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast
> ({ i32, i32, i8 addrspace(4)* } addrspace(1)* @__block_literal_global
> to i8 addrspace(1)*) to i8 addrspace(4)*), %struct.A* byval align 4
> %aa) #5
> - %a1 = getelementptr inbounds %struct.A, %struct.A* %tmp, i32 0,
> i32 0
> - %2 = load i32, i32* %a1, align 4
> - %sub = sub nsw i32 %2, 6
> -@@ -79,10 +84,10 @@ entry:
> - define internal spir_func void
> @__block_ret_struct_block_invoke(%struct.A* noalias sret %agg.result,
> i8 addrspace(4)* %.block_descriptor, %struct.A* byval align 4 %a) #1
> {
> - entry:
> - %.block_descriptor.addr = alloca i8 addrspace(4)*, align 8
> -- %block.addr = alloca <{ i32, i32 }> addrspace(4)*, align 8
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)* }>
> addrspace(4)*, align 8
> - store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 8
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32 }> addrspace(4)*
> -- store <{ i32, i32 }> addrspace(4)* %block, <{ i32, i32 }>
> addrspace(4)** %block.addr, align 8
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)* }> addrspace(4)*
> -+ store <{ i32, i32, i8 addrspace(4)* }> addrspace(4)* %block, <{
> i32, i32, i8 addrspace(4)* }> addrspace(4)** %block.addr, align 8
> - %a1 = getelementptr inbounds %struct.A, %struct.A* %a, i32 0, i32
> 0
> - store i32 6, i32* %a1, align 4
> - %0 = bitcast %struct.A* %agg.result to i8*
> -@@ -97,30 +102,22 @@ declare void @llvm.memcpy.p0i8.p0i8.i64(i8*
> nocapture writeonly, i8* nocapture r
> - ; Function Attrs: convergent nounwind readnone
> - declare spir_func i64 @_Z13get_global_idj(i32) #3
> -
> --attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-
> pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-
> tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-
> math"="false" "no-trapping-math"="false" "stack-protector-buffer-
> size"="8" "uniform-work-group-size"="false" "unsafe-fp-math"="false"
> "use-soft-float"="false" }
> --attributes #1 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-
> pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-
> tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-
> math"="false" "no-trapping-math"="false" "stack-protector-buffer-
> size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
> -+attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "uniform-work-group-size"="false" "unsafe-
> fp-math"="false" "use-soft-float"="false" }
> -+attributes #1 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-
> float"="false" }
> - attributes #2 = { argmemonly nounwind }
> - attributes #3 = { convergent nounwind readnone "correctly-rounded-
> divide-sqrt-fp-math"="false" "denorms-are-zero"="false" "disable-
> tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-
> elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false"
> "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-
> float"="false" }
> - attributes #4 = { convergent nounwind readnone }
> - attributes #5 = { convergent }
> -
> - !llvm.module.flags = !{!0}
> --!opencl.enable.FP_CONTRACT = !{}
> - !opencl.ocl.version = !{!1}
> - !opencl.spir.version = !{!1}
> --!opencl.used.extensions = !{!2}
> --!opencl.used.optional.core.features = !{!2}
> --!opencl.compiler.options = !{!2}
> --!llvm.ident = !{!3}
> -+!llvm.ident = !{!2}
> -
> - !0 = !{i32 1, !"wchar_size", i32 4}
> - !1 = !{i32 2, i32 0}
> --!2 = !{}
> --!3 = !{!"clang version 7.0.0"}
> --!4 = !{i32 1}
> --!5 = !{!"none"}
> --!6 = !{!"int*"}
> --!7 = !{!""}
> --!8 = !{i1 false}
> --!9 = !{i32 0}
> --
> -+!2 = !{!"clang version 9.0.0 (https://llvm.org/git/clang
> 04fb8964a801a5c5d7baa5a22272243a7d183896) (https://llvm.org/git/llvm
384f64397f6ad95a361b72d62c07d7bac9f24163)"}
> -+!3 = !{i32 1}
> -+!4 = !{!"none"}
> -+!5 = !{!"int*"}
> -+!6 = !{!""}
> -diff --git a/test/transcoding/enqueue_kernel.ll
> b/test/transcoding/enqueue_kernel.ll
> -index 1f0b360..761043e 100644
> ---- a/test/transcoding/enqueue_kernel.ll
> -+++ b/test/transcoding/enqueue_kernel.ll
> -@@ -51,11 +51,12 @@
> - ; ModuleID = 'enqueue_kernel.cl'
> - source_filename = "enqueue_kernel.cl"
> - target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-
> v96:128-v192:256-v256:256-v512:512-v1024:1024"
> --target triple = "spir-unknown-unknown"
> -+target triple = "spir"
> -
> - %opencl.queue_t = type opaque
> - %struct.ndrange_t = type { i32 }
> - %opencl.clk_event_t = type opaque
> -+%struct.__opencl_block_literal_generic = type { i32, i32, i8
> addrspace(4)* }
> -
> - ; CHECK-SPIRV: EntryPoint {{[0-9]+}} [[BlockKer1:[0-9]+]]
> "__device_side_enqueue_block_invoke_kernel"
> - ; CHECK-SPIRV: EntryPoint {{[0-9]+}} [[BlockKer2:[0-9]+]]
> "__device_side_enqueue_block_invoke_2_kernel"
> -@@ -66,89 +67,123 @@ target triple = "spir-unknown-unknown"
> -
> - ; CHECK-SPIRV: TypeInt [[Int32Ty:[0-9]+]] 32
> - ; CHECK-SPIRV: TypeInt [[Int8Ty:[0-9]+]] 8
> --; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt8:[0-9]+]] 8
> - ; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt0:[0-9]+]] 0
> --; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt17:[0-9]+]] 17
> -+; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt17:[0-9]+]] 21
> - ; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt2:[0-9]+]] 2
> --; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt20:[0-9]+]] 20
> --; CHECK-SPIRV: TypeVoid [[VoidTy:[0-9]+]]
> -+; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt8:[0-9]+]] 8
> -+; CHECK-SPIRV: Constant [[Int32Ty]] [[ConstInt20:[0-9]+]] 24
> -
> - ; CHECK-SPIRV: TypePointer {{[0-9]+}} 7 {{[0-9]+}}
> -+; CHECK-SPIRV: TypePointer [[Int8PtrGenTy:[0-9]+]] 8 [[Int8Ty]]
> -+; CHECK-SPIRV: TypeVoid [[VoidTy:[0-9]+]]
> - ; CHECK-SPIRV: TypePointer [[Int32LocPtrTy:[0-9]+]] 7 [[Int32Ty]]
> - ; CHECK-SPIRV: TypeDeviceEvent [[EventTy:[0-9]+]]
> --; CHECK-SPIRV: TypePointer [[Int8PtrGenTy:[0-9]+]] 8 [[Int8Ty]]
> - ; CHECK-SPIRV: TypePointer [[EventPtrTy:[0-9]+]] 8 [[EventTy]]
> - ; CHECK-SPIRV: TypeFunction [[BlockTy1:[0-9]+]] [[VoidTy]]
> [[Int8PtrGenTy]]
> - ; CHECK-SPIRV: TypeFunction [[BlockTy2:[0-9]+]] [[VoidTy]]
> [[Int8PtrGenTy]]
> - ; CHECK-SPIRV: TypeFunction [[BlockTy3:[0-9]+]] [[VoidTy]]
> [[Int8PtrGenTy]]
> - ; CHECK-SPIRV: ConstantNull [[EventPtrTy]] [[EventNull:[0-9]+]]
> -
> --; CHECK-LLVM: [[BlockTy1:%[0-9a-z\.]+]] = type { i32, i32 }
> --; CHECK-LLVM: [[BlockTy2:%[0-9a-z\.]+]] = type <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>
> --; CHECK-LLVM: [[BlockTy3:%[0-9a-z\.]+]] = type <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>
> --; CHECK-LLVM: [[BlockTy4:%[0-9a-z\.]+]] = type <{ i32, i32 }>
> -+; CHECK-LLVM: [[BlockTy1:%[0-9a-z\.]+]] = type { i32, i32, i8
> addrspace(4)* }
> -+; CHECK-LLVM: [[BlockTy2:%[0-9a-z\.]+]] = type <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>
> -+; CHECK-LLVM: [[BlockTy3:%[0-9a-z\.]+]] = type <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> -+; CHECK-LLVM: [[BlockTy4:%[0-9a-z\.]+]] = type <{ i32, i32, i8
> addrspace(4)* }>
> -
> --; CHECK-LLVM: @__block_literal_global = internal addrspace(1)
> constant [[BlockTy1]] { i32 8, i32 4 }, align 4
> --; CHECK-LLVM: @__block_literal_global.1 = internal addrspace(1)
> constant [[BlockTy1]] { i32 8, i32 4 }, align 4
> -+; CHECK-LLVM: @__block_literal_global = internal addrspace(1)
> constant [[BlockTy1]] { i32 12, i32 4, i8 addrspace(4)* addrspacecast
> (i8* null to i8 addrspace(4)*) }, align 4
> -+; CHECK-LLVM: @__block_literal_global.1 = internal addrspace(1)
> constant [[BlockTy1]] { i32 12, i32 4, i8 addrspace(4)* addrspacecast
> (i8* null to i8 addrspace(4)*) }, align 4
> -
> -- at __block_literal_global = internal addrspace(1) constant { i32, i32
> } { i32 8, i32 4 }, align 4
> -- at __block_literal_global.1 = internal addrspace(1) constant { i32,
> i32 } { i32 8, i32 4 }, align 4
> -+ at __block_literal_global = internal addrspace(1) constant { i32,
> i32, i8 addrspace(4)* } { i32 12, i32 4, i8 addrspace(4)*
> addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8
> addrspace(3)*)* @__device_side_enqueue_block_invoke_3 to i8*) to i8
> addrspace(4)*) }, align 4
> -+ at __block_literal_global.1 = internal addrspace(1) constant { i32,
> i32, i8 addrspace(4)* } { i32 12, i32 4, i8 addrspace(4)*
> addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*,
> i8 addrspace(3)*, i8 addrspace(3)*)*
> @__device_side_enqueue_block_invoke_4 to i8*) to i8 addrspace(4)*) },
> align 4
> -
> - ; Function Attrs: convergent noinline nounwind optnone
> --define spir_kernel void @device_side_enqueue(i32 addrspace(1)* %a,
> i32 addrspace(1)* %b, i32 %i, i8 signext %c0) #0
> !kernel_arg_addr_space !4 !kernel_arg_access_qual !5 !kernel_arg_type
> !6 !kernel_arg_base_type !6 !kernel_arg_type_qual !7 {
> -+define spir_kernel void @device_side_enqueue(i32 addrspace(1)* %a,
> i32 addrspace(1)* %b, i32 %i, i8 signext %c0) #0
> !kernel_arg_addr_space !3 !kernel_arg_access_qual !4 !kernel_arg_type
> !5 !kernel_arg_base_type !5 !kernel_arg_type_qual !6 {
> - entry:
> -+ %a.addr = alloca i32 addrspace(1)*, align 4
> -+ %b.addr = alloca i32 addrspace(1)*, align 4
> -+ %i.addr = alloca i32, align 4
> -+ %c0.addr = alloca i8, align 1
> - %default_queue = alloca %opencl.queue_t*, align 4
> - %flags = alloca i32, align 4
> - %ndrange = alloca %struct.ndrange_t, align 4
> - %clk_event = alloca %opencl.clk_event_t*, align 4
> - %event_wait_list = alloca %opencl.clk_event_t*, align 4
> - %event_wait_list2 = alloca [1 x %opencl.clk_event_t*], align 4
> -- %block = alloca <{ i32, i32, i32 addrspace(1)*, i32, i8 }>, align
> 4
> -- %block3 = alloca <{ i32, i32, i32 addrspace(1)*, i32, i32
> addrspace(1)* }>, align 4
> -+ %block = alloca <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*,
> i32, i8 }>, align 4
> -+ %tmp = alloca %struct.ndrange_t, align 4
> -+ %block3 = alloca <{ i32, i32, i8 addrspace(4)*, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, align 4
> -+ %tmp4 = alloca %struct.ndrange_t, align 4
> - %c = alloca i8, align 1
> -+ %tmp11 = alloca %struct.ndrange_t, align 4
> -+ %block_sizes = alloca [1 x i32], align 4
> -+ %tmp12 = alloca %struct.ndrange_t, align 4
> -+ %block_sizes13 = alloca [3 x i32], align 4
> -+ store i32 addrspace(1)* %a, i32 addrspace(1)** %a.addr, align 4
> -+ store i32 addrspace(1)* %b, i32 addrspace(1)** %b.addr, align 4
> -+ store i32 %i, i32* %i.addr, align 4
> -+ store i8 %c0, i8* %c0.addr, align 1
> - store i32 0, i32* %flags, align 4
> - %arrayinit.begin = getelementptr inbounds [1 x
> %opencl.clk_event_t*], [1 x %opencl.clk_event_t*]* %event_wait_list2,
> i32 0, i32 0
> - %0 = load %opencl.clk_event_t*, %opencl.clk_event_t** %clk_event,
> align 4
> - store %opencl.clk_event_t* %0, %opencl.clk_event_t**
> %arrayinit.begin, align 4
> - %1 = load %opencl.queue_t*, %opencl.queue_t** %default_queue,
> align 4
> - %2 = load i32, i32* %flags, align 4
> -- %block.size = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8
> }>* %block, i32 0, i32 0
> -- store i32 17, i32* %block.size, align 4
> -- %block.align = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8
> }>* %block, i32 0, i32 1
> -+ %3 = bitcast %struct.ndrange_t* %tmp to i8*
> -+ %4 = bitcast %struct.ndrange_t* %ndrange to i8*
> -+ call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %3, i8* align 4
> %4, i32 4, i1 false)
> -+ %block.size = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>* %block, i32 0, i32 0
> -+ store i32 21, i32* %block.size, align 4
> -+ %block.align = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>* %block, i32 0, i32 1
> - store i32 4, i32* %block.align, align 4
> -- %block.captured = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8
> }>* %block, i32 0, i32 2
> -- store i32 addrspace(1)* %a, i32 addrspace(1)** %block.captured,
> align 4
> -- %block.captured1 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8
> }>* %block, i32 0, i32 3
> -- store i32 %i, i32* %block.captured1, align 4
> -- %block.captured2 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8
> }>* %block, i32 0, i32 4
> -- store i8 %c0, i8* %block.captured2, align 4
> -- %3 = bitcast <{ i32, i32, i32 addrspace(1)*, i32, i8 }>* %block
> to void ()*
> -- %4 = addrspacecast void ()* %3 to i8 addrspace(4)*
> -+ %block.invoke = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>* %block, i32 0, i32 2
> -+ store i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8
> addrspace(4)*)* @__device_side_enqueue_block_invoke to i8*) to i8
> addrspace(4)*), i8 addrspace(4)** %block.invoke, align 4
> -+ %block.captured = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>* %block, i32 0, i32 3
> -+ %5 = load i32 addrspace(1)*, i32 addrspace(1)** %a.addr, align 4
> -+ store i32 addrspace(1)* %5, i32 addrspace(1)** %block.captured,
> align 4
> -+ %block.captured1 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>* %block, i32 0, i32 4
> -+ %6 = load i32, i32* %i.addr, align 4
> -+ store i32 %6, i32* %block.captured1, align 4
> -+ %block.captured2 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>* %block, i32 0, i32 5
> -+ %7 = load i8, i8* %c0.addr, align 1
> -+ store i8 %7, i8* %block.captured2, align 4
> -+ %8 = bitcast <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*,
> i32, i8 }>* %block to %struct.__opencl_block_literal_generic*
> -+ %9 = addrspacecast %struct.__opencl_block_literal_generic* %8 to
> i8 addrspace(4)*
> -
> - ; CHECK-SPIRV: PtrCastToGeneric [[Int8PtrGenTy]] [[BlockLit1:[0-
> 9]+]]
> - ; CHECK-SPIRV: EnqueueKernel [[Int32Ty]] {{[0-9]+}} {{[0-9]+}}
> {{[0-9]+}} {{[0-9]+}}
> - ; [[ConstInt0]] [[EventNull]]
> [[EventNull]]
> - ; [[BlockKer1]] [[BlockLit1]]
> [[ConstInt17]] [[ConstInt8]]
> -
> --; CHECK-LLVM: [[Block2:%[0-9]+]] = addrspacecast [[BlockTy2]]*
> %block to i8 addrspace(4)*
> -+; CHECK-LLVM: [[Block2:%[0-9]+]] = bitcast [[BlockTy2]]* %block to
> %struct.__opencl_block_literal_generic*
> -+; CHECK-LLVM: [[Block2Ptr:%[0-9]+]] = addrspacecast
> %struct.__opencl_block_literal_generic* [[Block2]] to i8
> addrspace(4)*
> - ; CHECK-LLVM: [[BlockInv2:%[0-9]+]] = addrspacecast void (i8
> addrspace(4)*)* @__device_side_enqueue_block_invoke_kernel to i8
> addrspace(4)*
> --; CHECK-LLVM: call i32
> @__enqueue_kernel_basic_events(%opencl.queue_t* {{.*}}, i32 {{.*}},
> %struct.ndrange_t* {{.*}}, i32 0, %opencl.clk_event_t* addrspace(4)*
> null, %opencl.clk_event_t* addrspace(4)* null, i8 addrspace(4)*
> [[BlockInv2]], i8 addrspace(4)* [[Block2]])
> --
> -- %5 = call i32 @__enqueue_kernel_basic(%opencl.queue_t* %1, i32
> %2, %struct.ndrange_t* byval %ndrange, i8 addrspace(4)* addrspacecast
> (i8* bitcast (void (i8 addrspace(4)*)*
> @__device_side_enqueue_block_invoke_kernel to i8*) to i8
> addrspace(4)*), i8 addrspace(4)* %4)
> -- %6 = addrspacecast %opencl.clk_event_t** %event_wait_list to
> %opencl.clk_event_t* addrspace(4)*
> -- %7 = addrspacecast %opencl.clk_event_t** %clk_event to
> %opencl.clk_event_t* addrspace(4)*
> -- %block.size5 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>* %block3, i32 0, i32 0
> -- store i32 20, i32* %block.size5, align 4
> -- %block.align6 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>* %block3, i32 0, i32 1
> -+; CHECK-LLVM: call i32
> @__enqueue_kernel_basic_events(%opencl.queue_t* {{.*}}, i32 {{.*}},
> %struct.ndrange_t* {{.*}}, i32 0, %opencl.clk_event_t* addrspace(4)*
> null, %opencl.clk_event_t* addrspace(4)* null, i8 addrspace(4)*
> [[BlockInv2]], i8 addrspace(4)* [[Block2Ptr]])
> -+
> -+ %10 = call i32 @__enqueue_kernel_basic(%opencl.queue_t* %1, i32
> %2, %struct.ndrange_t* byval %tmp, i8 addrspace(4)* addrspacecast
> (i8* bitcast (void (i8 addrspace(4)*)*
> @__device_side_enqueue_block_invoke_kernel to i8*) to i8
> addrspace(4)*), i8 addrspace(4)* %9)
> -+ %11 = load %opencl.queue_t*, %opencl.queue_t** %default_queue,
> align 4
> -+ %12 = load i32, i32* %flags, align 4
> -+ %13 = bitcast %struct.ndrange_t* %tmp4 to i8*
> -+ %14 = bitcast %struct.ndrange_t* %ndrange to i8*
> -+ call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %13, i8* align 4
> %14, i32 4, i1 false)
> -+ %15 = addrspacecast %opencl.clk_event_t** %event_wait_list to
> %opencl.clk_event_t* addrspace(4)*
> -+ %16 = addrspacecast %opencl.clk_event_t** %clk_event to
> %opencl.clk_event_t* addrspace(4)*
> -+ %block.size5 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>*
> %block3, i32 0, i32 0
> -+ store i32 24, i32* %block.size5, align 4
> -+ %block.align6 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>*
> %block3, i32 0, i32 1
> - store i32 4, i32* %block.align6, align 4
> -- %block.captured7 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>* %block3, i32 0, i32 2
> -- store i32 addrspace(1)* %a, i32 addrspace(1)** %block.captured7,
> align 4
> -- %block.captured8 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>* %block3, i32 0, i32 3
> -- store i32 %i, i32* %block.captured8, align 4
> -- %block.captured9 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>* %block3, i32 0, i32 4
> -- store i32 addrspace(1)* %b, i32 addrspace(1)** %block.captured9,
> align 4
> -- %8 = bitcast <{ i32, i32, i32 addrspace(1)*, i32, i32
> addrspace(1)* }>* %block3 to void ()*
> -- %9 = addrspacecast void ()* %8 to i8 addrspace(4)*
> -+ %block.invoke7 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>*
> %block3, i32 0, i32 2
> -+ store i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8
> addrspace(4)*)* @__device_side_enqueue_block_invoke_2 to i8*) to i8
> addrspace(4)*), i8 addrspace(4)** %block.invoke7, align 4
> -+ %block.captured8 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>*
> %block3, i32 0, i32 3
> -+ %17 = load i32 addrspace(1)*, i32 addrspace(1)** %a.addr, align 4
> -+ store i32 addrspace(1)* %17, i32 addrspace(1)** %block.captured8,
> align 4
> -+ %block.captured9 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>*
> %block3, i32 0, i32 4
> -+ %18 = load i32, i32* %i.addr, align 4
> -+ store i32 %18, i32* %block.captured9, align 4
> -+ %block.captured10 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>*
> %block3, i32 0, i32 5
> -+ %19 = load i32 addrspace(1)*, i32 addrspace(1)** %b.addr, align 4
> -+ store i32 addrspace(1)* %19, i32 addrspace(1)**
> %block.captured10, align 4
> -+ %20 = bitcast <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*,
> i32, i32 addrspace(1)* }>* %block3 to
> %struct.__opencl_block_literal_generic*
> -+ %21 = addrspacecast %struct.__opencl_block_literal_generic* %20
> to i8 addrspace(4)*
> -+
> -
> - ; CHECK-SPIRV: PtrCastToGeneric [[EventPtrTy]] [[Event1:[0-9]+]]
> - ; CHECK-SPIRV: PtrCastToGeneric [[EventPtrTy]] [[Event2:[0-9]+]]
> -@@ -158,16 +193,24 @@ entry:
> - ; [[ConstInt2]] [[Event1]] [[Event2]]
> - ; [[BlockKer2]] [[BlockLit2]]
> [[ConstInt20]] [[ConstInt8]]
> -
> --; CHECK-LLVM: [[Block3:%[0-9]+]] = addrspacecast [[BlockTy3]]*
> %block3 to i8 addrspace(4)*
> -+; CHECK-LLVM: [[Block3:%[0-9]+]] = bitcast [[BlockTy3]]* %block3 to
> %struct.__opencl_block_literal_generic*
> -+; CHECK-LLVM: [[Block3Ptr:%[0-9]+]] = addrspacecast
> %struct.__opencl_block_literal_generic* [[Block3]] to i8 addrspace(4)
> - ; CHECK-LLVM: [[BlockInv3:%[0-9]+]] = addrspacecast void (i8
> addrspace(4)*)* @__device_side_enqueue_block_invoke_2_kernel to i8
> addrspace(4)*
> --; CHECK-LLVM: call i32
> @__enqueue_kernel_basic_events(%opencl.queue_t* {{.*}}, i32 {{.*}},
> %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t* addrspace(4)*
> {{.*}}, %opencl.clk_event_t* addrspace(4)* {{.*}}, i8 addrspace(4)*
> [[BlockInv3]], i8 addrspace(4)* [[Block3]])
> --
> -- %10 = call i32 @__enqueue_kernel_basic_events(%opencl.queue_t*
> %1, i32 %2, %struct.ndrange_t* %ndrange, i32 2, %opencl.clk_event_t*
> addrspace(4)* %6, %opencl.clk_event_t* addrspace(4)* %7, i8
> addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)*
> @__device_side_enqueue_block_invoke_2_kernel to i8*) to i8
> addrspace(4)*), i8 addrspace(4)* %9)
> -- %11 = alloca [1 x i32]
> -- %12 = getelementptr [1 x i32], [1 x i32]* %11, i32 0, i32 0
> -- %13 = load i8, i8* %c, align 1
> -- %14 = zext i8 %13 to i32
> -- store i32 %14, i32* %12, align 4
> -+; CHECK-LLVM: call i32
> @__enqueue_kernel_basic_events(%opencl.queue_t* {{.*}}, i32 {{.*}},
> %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t* addrspace(4)*
> {{.*}}, %opencl.clk_event_t* addrspace(4)* {{.*}}, i8 addrspace(4)*
> [[BlockInv3]], i8 addrspace(4)* [[Block3Ptr]])
> -+
> -+ %22 = call i32 @__enqueue_kernel_basic_events(%opencl.queue_t*
> %11, i32 %12, %struct.ndrange_t* %tmp4, i32 2, %opencl.clk_event_t*
> addrspace(4)* %15, %opencl.clk_event_t* addrspace(4)* %16, i8
> addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)*
> @__device_side_enqueue_block_invoke_2_kernel to i8*) to i8
> addrspace(4)*), i8 addrspace(4)* %21)
> -+ %23 = load %opencl.queue_t*, %opencl.queue_t** %default_queue,
> align 4
> -+ %24 = load i32, i32* %flags, align 4
> -+ %25 = bitcast %struct.ndrange_t* %tmp11 to i8*
> -+ %26 = bitcast %struct.ndrange_t* %ndrange to i8*
> -+ call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %25, i8* align 4
> %26, i32 4, i1 false)
> -+ %arraydecay = getelementptr inbounds [1 x %opencl.clk_event_t*],
> [1 x %opencl.clk_event_t*]* %event_wait_list2, i32 0, i32 0
> -+ %27 = addrspacecast %opencl.clk_event_t** %arraydecay to
> %opencl.clk_event_t* addrspace(4)*
> -+ %28 = addrspacecast %opencl.clk_event_t** %clk_event to
> %opencl.clk_event_t* addrspace(4)*
> -+ %29 = getelementptr [1 x i32], [1 x i32]* %block_sizes, i32 0,
> i32 0
> -+ %30 = load i8, i8* %c, align 1
> -+ %31 = zext i8 %30 to i32
> -+ store i32 %31, i32* %29, align 4
> -
> - ; CHECK-SPIRV: PtrAccessChain [[Int32LocPtrTy]] [[LocalBuf31:[0-
> 9]+]]
> - ; CHECK-SPIRV: Bitcast {{[0-9]+}} [[BlockLit3Tmp:[0-9]+]]
> [[BlockGlb1:[0-9]+]]
> -@@ -182,14 +225,18 @@ entry:
> - ; CHECK-LLVM: [[BlockInv0:%[0-9]+]] = addrspacecast void (i8
> addrspace(4)*, i8 addrspace(3)*)*
> @__device_side_enqueue_block_invoke_3_kernel to i8 addrspace(4)*
> - ; CHECK-LLVM: call i32
> @__enqueue_kernel_events_varargs(%opencl.queue_t* {{.*}}, i32 {{.*}},
> %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t* addrspace(4)*
> {{.*}}, %opencl.clk_event_t* addrspace(4)* {{.*}}, i8 addrspace(4)*
> [[BlockInv0]], i8 addrspace(4)* [[Block0]], i32 1, i32* {{.*}})
> -
> -- %15 = call i32 @__enqueue_kernel_events_varargs(%opencl.queue_t*
> %1, i32 %2, %struct.ndrange_t* %ndrange, i32 2, %opencl.clk_event_t*
> addrspace(4)* %6, %opencl.clk_event_t* addrspace(4)* %7, i8
> addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8
> addrspace(3)*)* @__device_side_enqueue_block_invoke_3_kernel to i8*)
> to i8 addrspace(4)*), i8 addrspace(4)* addrspacecast (i8
> addrspace(1)* bitcast ({ i32, i32 } addrspace(1)*
> @__block_literal_global to i8 addrspace(1)*) to i8 addrspace(4)*),
> i32 1, i32* %12)
> -- %16 = alloca [3 x i32]
> -- %17 = getelementptr [3 x i32], [3 x i32]* %16, i32 0, i32 0
> -- store i32 1, i32* %17, align 4
> -- %18 = getelementptr [3 x i32], [3 x i32]* %16, i32 0, i32 1
> -- store i32 2, i32* %18, align 4
> -- %19 = getelementptr [3 x i32], [3 x i32]* %16, i32 0, i32 2
> -- store i32 4, i32* %19, align 4
> -+ %32 = call i32 @__enqueue_kernel_events_varargs(%opencl.queue_t*
> %23, i32 %24, %struct.ndrange_t* %tmp11, i32 2, %opencl.clk_event_t*
> addrspace(4)* %27, %opencl.clk_event_t* addrspace(4)* %28, i8
> addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8
> addrspace(3)*)* @__device_side_enqueue_block_invoke_3_kernel to i8*)
> to i8 addrspace(4)*), i8 addrspace(4)* addrspacecast (i8
> addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> @__block_literal_global to i8 addrspace(1)*) to i8 addrspace(4)*),
> i32 1, i32* %29)
> -+ %33 = load %opencl.queue_t*, %opencl.queue_t** %default_queue,
> align 4
> -+ %34 = load i32, i32* %flags, align 4
> -+ %35 = bitcast %struct.ndrange_t* %tmp12 to i8*
> -+ %36 = bitcast %struct.ndrange_t* %ndrange to i8*
> -+ call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %35, i8* align 4
> %36, i32 4, i1 false)
> -+ %37 = getelementptr [3 x i32], [3 x i32]* %block_sizes13, i32 0,
> i32 0
> -+ store i32 1, i32* %37, align 4
> -+ %38 = getelementptr [3 x i32], [3 x i32]* %block_sizes13, i32 0,
> i32 1
> -+ store i32 2, i32* %38, align 4
> -+ %39 = getelementptr [3 x i32], [3 x i32]* %block_sizes13, i32 0,
> i32 2
> -+ store i32 4, i32* %39, align 4
> -
> - ; CHECK-SPIRV: PtrAccessChain [[Int32LocPtrTy]] [[LocalBuf41:[0-
> 9]+]]
> - ; CHECK-SPIRV: PtrAccessChain [[Int32LocPtrTy]] [[LocalBuf42:[0-
> 9]+]]
> -@@ -206,24 +253,27 @@ entry:
> - ; CHECK-LLVM: [[BlockInv1:%[0-9]+]] = addrspacecast void (i8
> addrspace(4)*, i8 addrspace(3)*, i8 addrspace(3)*, i8 addrspace(3)*)*
> @__device_side_enqueue_block_invoke_4_kernel to i8 addrspace(4)*
> - ; CHECK-LLVM: call i32
> @__enqueue_kernel_events_varargs(%opencl.queue_t* {{.*}}, i32 {{.*}},
> %struct.ndrange_t* {{.*}}, i32 0, %opencl.clk_event_t* addrspace(4)*
> null, %opencl.clk_event_t* addrspace(4)* null, i8 addrspace(4)*
> [[BlockInv1]], i8 addrspace(4)* [[Block1]], i32 3, i32* {{.*}})
> -
> -- %20 = call i32 @__enqueue_kernel_varargs(%opencl.queue_t* %1, i32
> %2, %struct.ndrange_t* %ndrange, i8 addrspace(4)* addrspacecast (i8*
> bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*, i8 addrspace(3)*,
> i8 addrspace(3)*)* @__device_side_enqueue_block_invoke_4_kernel to
> i8*) to i8 addrspace(4)*), i8 addrspace(4)* addrspacecast (i8
> addrspace(1)* bitcast ({ i32, i32 } addrspace(1)*
> @__block_literal_global.1 to i8 addrspace(1)*) to i8 addrspace(4)*),
> i32 3, i32* %17)
> -+ %40 = call i32 @__enqueue_kernel_varargs(%opencl.queue_t* %33,
> i32 %34, %struct.ndrange_t* %tmp12, i8 addrspace(4)* addrspacecast
> (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*, i8
> addrspace(3)*, i8 addrspace(3)*)*
> @__device_side_enqueue_block_invoke_4_kernel to i8*) to i8
> addrspace(4)*), i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)*
> @__block_literal_global.1 to i8 addrspace(1)*) to i8 addrspace(4)*),
> i32 3, i32* %37)
> - ret void
> - }
> -
> -+; Function Attrs: argmemonly nounwind
> -+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture writeonly,
> i8* nocapture readonly, i32, i1) #1
> -+
> - ; Function Attrs: convergent noinline nounwind optnone
> - define internal spir_func void
> @__device_side_enqueue_block_invoke(i8 addrspace(4)*
> %.block_descriptor) #2 {
> - entry:
> - %.block_descriptor.addr = alloca i8 addrspace(4)*, align 4
> -- %block.addr = alloca <{ i32, i32, i32 addrspace(1)*, i32, i8 }>
> addrspace(4)*, align 4
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)*, i32
> addrspace(1)*, i32, i8 }> addrspace(4)*, align 4
> - store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 4
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i32 addrspace(1)*, i32, i8 }> addrspace(4)*
> -- store <{ i32, i32, i32 addrspace(1)*, i32, i8 }> addrspace(4)*
> %block, <{ i32, i32, i32 addrspace(1)*, i32, i8 }> addrspace(4)**
> %block.addr, align 4
> -- %block.capture.addr = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8 }>
> addrspace(4)* %block, i32 0, i32 4
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i8 }> addrspace(4)*
> -+ store <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i8
> }> addrspace(4)* %block, <{ i32, i32, i8 addrspace(4)*, i32
> addrspace(1)*, i32, i8 }> addrspace(4)** %block.addr, align 4
> -+ %block.capture.addr = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }> addrspace(4)* %block,
> i32 0, i32 5
> - %0 = load i8, i8 addrspace(4)* %block.capture.addr, align 4
> - %conv = sext i8 %0 to i32
> -- %block.capture.addr1 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8 }>
> addrspace(4)* %block, i32 0, i32 2
> -+ %block.capture.addr1 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }> addrspace(4)* %block,
> i32 0, i32 3
> - %1 = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)*
> %block.capture.addr1, align 4
> -- %block.capture.addr2 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i8 }>, <{ i32, i32, i32 addrspace(1)*, i32, i8 }>
> addrspace(4)* %block, i32 0, i32 3
> -+ %block.capture.addr2 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }>, <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i8 }> addrspace(4)* %block,
> i32 0, i32 4
> - %2 = load i32, i32 addrspace(4)* %block.capture.addr2, align 4
> - %arrayidx = getelementptr inbounds i32, i32 addrspace(1)* %1, i32
> %2
> - store i32 %conv, i32 addrspace(1)* %arrayidx, align 4
> -@@ -243,19 +293,19 @@ declare i32
> @__enqueue_kernel_basic(%opencl.queue_t*, i32, %struct.ndrange_t*, i
> - define internal spir_func void
> @__device_side_enqueue_block_invoke_2(i8 addrspace(4)*
> %.block_descriptor) #2 {
> - entry:
> - %.block_descriptor.addr = alloca i8 addrspace(4)*, align 4
> -- %block.addr = alloca <{ i32, i32, i32 addrspace(1)*, i32, i32
> addrspace(1)* }> addrspace(4)*, align 4
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)*, i32
> addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)*, align 4
> - store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 4
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i32 addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)*
> -- store <{ i32, i32, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> addrspace(4)* %block, <{ i32, i32, i32 addrspace(1)*, i32, i32
> addrspace(1)* }> addrspace(4)** %block.addr, align 4
> -- %block.capture.addr = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)* %block, i32 0,
> i32 4
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> addrspace(4)*
> -+ store <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32
> addrspace(1)* }> addrspace(4)* %block, <{ i32, i32, i8 addrspace(4)*,
> i32 addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)**
> %block.addr, align 4
> -+ %block.capture.addr = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> addrspace(4)* %block, i32 0, i32 5
> - %0 = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)*
> %block.capture.addr, align 4
> -- %block.capture.addr1 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)* %block, i32 0,
> i32 3
> -+ %block.capture.addr1 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> addrspace(4)* %block, i32 0, i32 4
> - %1 = load i32, i32 addrspace(4)* %block.capture.addr1, align 4
> - %arrayidx = getelementptr inbounds i32, i32 addrspace(1)* %0, i32
> %1
> - %2 = load i32, i32 addrspace(1)* %arrayidx, align 4
> -- %block.capture.addr2 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)* %block, i32 0,
> i32 2
> -+ %block.capture.addr2 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> addrspace(4)* %block, i32 0, i32 3
> - %3 = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)*
> %block.capture.addr2, align 4
> -- %block.capture.addr3 = getelementptr inbounds <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32, i32, i32
> addrspace(1)*, i32, i32 addrspace(1)* }> addrspace(4)* %block, i32 0,
> i32 3
> -+ %block.capture.addr3 = getelementptr inbounds <{ i32, i32, i8
> addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>, <{ i32,
> i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>
> addrspace(4)* %block, i32 0, i32 4
> - %4 = load i32, i32 addrspace(4)* %block.capture.addr3, align 4
> - %arrayidx4 = getelementptr inbounds i32, i32 addrspace(1)* %3,
> i32 %4
> - store i32 %2, i32 addrspace(1)* %arrayidx4, align 4
> -@@ -276,11 +326,11 @@ define internal spir_func void
> @__device_side_enqueue_block_invoke_3(i8 addrspac
> - entry:
> - %.block_descriptor.addr = alloca i8 addrspace(4)*, align 4
> - %p.addr = alloca i8 addrspace(3)*, align 4
> -- %block.addr = alloca <{ i32, i32 }> addrspace(4)*, align 4
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)* }>
> addrspace(4)*, align 4
> - store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 4
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32 }> addrspace(4)*
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)* }> addrspace(4)*
> - store i8 addrspace(3)* %p, i8 addrspace(3)** %p.addr, align 4
> -- store <{ i32, i32 }> addrspace(4)* %block, <{ i32, i32 }>
> addrspace(4)** %block.addr, align 4
> -+ store <{ i32, i32, i8 addrspace(4)* }> addrspace(4)* %block, <{
> i32, i32, i8 addrspace(4)* }> addrspace(4)** %block.addr, align 4
> - ret void
> - }
> -
> -@@ -300,13 +350,13 @@ entry:
> - %p1.addr = alloca i8 addrspace(3)*, align 4
> - %p2.addr = alloca i8 addrspace(3)*, align 4
> - %p3.addr = alloca i8 addrspace(3)*, align 4
> -- %block.addr = alloca <{ i32, i32 }> addrspace(4)*, align 4
> -+ %block.addr = alloca <{ i32, i32, i8 addrspace(4)* }>
> addrspace(4)*, align 4
> - store i8 addrspace(4)* %.block_descriptor, i8 addrspace(4)**
> %.block_descriptor.addr, align 4
> -- %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32 }> addrspace(4)*
> -+ %block = bitcast i8 addrspace(4)* %.block_descriptor to <{ i32,
> i32, i8 addrspace(4)* }> addrspace(4)*
> - store i8 addrspace(3)* %p1, i8 addrspace(3)** %p1.addr, align 4
> - store i8 addrspace(3)* %p2, i8 addrspace(3)** %p2.addr, align 4
> - store i8 addrspace(3)* %p3, i8 addrspace(3)** %p3.addr, align 4
> -- store <{ i32, i32 }> addrspace(4)* %block, <{ i32, i32 }>
> addrspace(4)** %block.addr, align 4
> -+ store <{ i32, i32, i8 addrspace(4)* }> addrspace(4)* %block, <{
> i32, i32, i8 addrspace(4)* }> addrspace(4)** %block.addr, align 4
> - ret void
> - }
> -
> -@@ -329,27 +379,20 @@ declare i32
> @__enqueue_kernel_varargs(%opencl.queue_t*, i32, %struct.ndrange_t*,
> - ; CHECK-LLVM-DAG: define spir_kernel void
> @__device_side_enqueue_block_invoke_3_kernel(i8 addrspace(4)*, i8
> addrspace(3)*)
> - ; CHECK-LLVM-DAG: define spir_kernel void
> @__device_side_enqueue_block_invoke_4_kernel(i8 addrspace(4)*, i8
> addrspace(3)*, i8 addrspace(3)*, i8 addrspace(3)*)
> -
> --attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false"
> "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-
> infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-
> math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-
> math"="false" "stack-protector-buffer-size"="8" "uniform-work-group-
> size"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
> -+attributes #0 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "uniform-work-group-size"="false" "unsafe-
> fp-math"="false" "use-soft-float"="false" }
> - attributes #1 = { argmemonly nounwind }
> --attributes #2 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false"
> "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-
> infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-
> math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-
> math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-
> math"="false" "use-soft-float"="false" }
> -+attributes #2 = { convergent noinline nounwind optnone "correctly-
> rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false"
> "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-
> vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-
> math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-
> signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-
> protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-
> float"="false" }
> - attributes #3 = { nounwind }
> -
> - !llvm.module.flags = !{!0}
> --!opencl.enable.FP_CONTRACT = !{}
> - !opencl.ocl.version = !{!1}
> - !opencl.spir.version = !{!1}
> --!opencl.used.extensions = !{!2}
> --!opencl.used.optional.core.features = !{!2}
> --!opencl.compiler.options = !{!2}
> --!llvm.ident = !{!3}
> -+!llvm.ident = !{!2}
> -
> - !0 = !{i32 1, !"wchar_size", i32 4}
> - !1 = !{i32 2, i32 0}
> --!2 = !{}
> --!3 = !{!"clang version 7.0.0"}
> --!4 = !{i32 1, i32 1, i32 0, i32 0}
> --!5 = !{!"none", !"none", !"none", !"none"}
> --!6 = !{!"int*", !"int*", !"int", !"char"}
> --!7 = !{!"", !"", !"", !""}
> --!8 = !{i1 false, i1 false, i1 false, i1 false}
> --!9 = !{i32 0, i32 0, i32 0, i32 0}
> -+!2 = !{!"clang version 9.0.0 (https://llvm.org/git/clang
> 04fb8964a801a5c5d7baa5a22272243a7d183896) (https://llvm.org/git/llvm
384f64397f6ad95a361b72d62c07d7bac9f24163)"}
> -+!3 = !{i32 1, i32 1, i32 0, i32 0}
> -+!4 = !{!"none", !"none", !"none", !"none"}
> -+!5 = !{!"int*", !"int*", !"int", !"char"}
> -+!6 = !{!"", !"", !"", !""}
> ---
> -2.7.4
> -
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-dont-export-targets-for-binaries.patch
> b/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-dont-
> export-targets-for-binaries.patch
> index 9d25bba..da2475f 100644
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> dont-export-targets-for-binaries.patch
> +++ b/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> dont-export-targets-for-binaries.patch
> @@ -1,38 +1,39 @@
> -From 7bbd0058362ac3bb5edd7a82d43e1785810776b3 Mon Sep 17 00:00:00
> 2001
> +From 559fb8f82295ec4dc64a132b6566939b85c1b6fe Mon Sep 17 00:00:00
> 2001
> From: Anuj Mittal <anuj.mittal at intel.com>
> -Date: Fri, 29 Mar 2019 08:56:53 +0800
> +Date: Thu, 15 Aug 2019 22:34:31 +0800
> Subject: [PATCH] dont export targets for binaries
>
> The projects using LLVM cmake modules look for target binaries in
> sysroot as a result which isn't desirable in this case and isn't
> needed
> either.
>
> -Upstream-Status: Inappropriate [cross-compile specific]
> +Upstream-Status: Inappropriate [cross-compile specific]
>
> Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> +Signed-off-by: Naveen Saini <naveen.kumar.saini at intel.com>
> ---
> llvm/cmake/modules/AddLLVM.cmake | 9 ---------
> llvm/cmake/modules/TableGen.cmake | 6 ------
> 2 files changed, 15 deletions(-)
>
> diff --git a/llvm/cmake/modules/AddLLVM.cmake
> b/llvm/cmake/modules/AddLLVM.cmake
> -index 0df6845..b79f4fa 100644
> +index 619e986b8aa..d2bc1a25dd9 100644
> --- a/llvm/cmake/modules/AddLLVM.cmake
> +++ b/llvm/cmake/modules/AddLLVM.cmake
> -@@ -866,12 +866,6 @@ macro(add_llvm_tool name)
> +@@ -898,12 +898,6 @@ macro(add_llvm_tool name)
>
> if ( ${name} IN_LIST LLVM_TOOLCHAIN_TOOLS OR NOT
> LLVM_INSTALL_TOOLCHAIN_ONLY)
> if( LLVM_BUILD_TOOLS )
> +- set(export_to_llvmexports)
> - if(${name} IN_LIST LLVM_DISTRIBUTION_COMPONENTS OR
> - NOT LLVM_DISTRIBUTION_COMPONENTS)
> - set(export_to_llvmexports EXPORT LLVMExports)
> - set_property(GLOBAL PROPERTY LLVM_HAS_EXPORTS True)
> - endif()
> --
> +
> install(TARGETS ${name}
> ${export_to_llvmexports}
> - RUNTIME DESTINATION ${LLVM_TOOLS_INSTALL_DIR}
> -@@ -884,9 +878,6 @@ macro(add_llvm_tool name)
> +@@ -917,9 +911,6 @@ macro(add_llvm_tool name)
> endif()
> endif()
> endif()
> @@ -43,18 +44,19 @@ index 0df6845..b79f4fa 100644
> endmacro(add_llvm_tool name)
>
> diff --git a/llvm/cmake/modules/TableGen.cmake
> b/llvm/cmake/modules/TableGen.cmake
> -index 3c84ae7..141219f 100644
> +index 36c026b5c0f..537acd696d8 100644
> --- a/llvm/cmake/modules/TableGen.cmake
> +++ b/llvm/cmake/modules/TableGen.cmake
> -@@ -164,14 +164,8 @@ macro(add_tablegen target project)
> +@@ -148,15 +148,9 @@ macro(add_tablegen target project)
> endif()
>
> - if (${project} STREQUAL LLVM AND NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
> + if (${project} STREQUAL LLVM AND NOT LLVM_INSTALL_TOOLCHAIN_ONLY
> AND LLVM_BUILD_UTILS)
> +- set(export_to_llvmexports)
> - if(${target} IN_LIST LLVM_DISTRIBUTION_COMPONENTS OR
> - NOT LLVM_DISTRIBUTION_COMPONENTS)
> - set(export_to_llvmexports EXPORT LLVMExports)
> - endif()
> --
> +
> install(TARGETS ${target}
> ${export_to_llvmexports}
> RUNTIME DESTINATION ${LLVM_TOOLS_INSTALL_DIR})
> @@ -62,5 +64,5 @@ index 3c84ae7..141219f 100644
> - set_property(GLOBAL APPEND PROPERTY LLVM_EXPORTS ${target})
> endmacro()
> --
> -2.7.4
> +2.17.1
>
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0001-skip-building-tests.patch b/dynamic-
> layers/clang-layer/recipes-devtools/clang/files/0001-skip-building-
> tests.patch
> index c58e7d9..9a7b844 100644
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> skip-building-tests.patch
> +++ b/dynamic-layers/clang-layer/recipes-devtools/clang/files/0001-
> skip-building-tests.patch
> @@ -1,9 +1,9 @@
> -From 91db4c3cf7f290a3cab5caa316fc25a60dd409f1 Mon Sep 17 00:00:00
> 2001
> -From: Anuj Mittal <anuj.mittal at intel.com>
> -Date: Fri, 16 Aug 2019 20:25:16 +0800
> -Subject: [PATCH] llvm-spirv: skip including tests
> +From 48e50f06b1bbed94cdf5207587161d4bfce7366e Mon Sep 17 00:00:00
> 2001
> +From: Naveen Saini <naveen.kumar.saini at intel.com>
> +Date: Wed, 21 Aug 2019 14:35:31 +0800
> +Subject: [PATCH] llvm-spirv: skip building tests
>
> -Some of these need clang to be built and since we're building this
> in-tree,
> +Some of these need clang to be built and since we're building this
> in-tree,
> that leads to problems when compiling libcxx, compiler-rt which
> aren't built
> in-tree.
>
> @@ -13,12 +13,13 @@ all components, disable the building of tests
> altogether.
> Upstream-Status: Inappropriate
>
> Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> +Signed-off-by: Naveen Saini <naveen.kumar.saini at intel.com>
> ---
> CMakeLists.txt | 10 ----------
> 1 file changed, 10 deletions(-)
>
> diff --git a/CMakeLists.txt b/CMakeLists.txt
> -index d632a50..81ddf62 100644
> +index 1208741..20ca3e6 100644
> --- a/CMakeLists.txt
> +++ b/CMakeLists.txt
> @@ -15,13 +15,6 @@ if(CMAKE_SOURCE_DIR STREQUAL
> CMAKE_CURRENT_SOURCE_DIR)
> @@ -32,7 +33,7 @@ index d632a50..81ddf62 100644
> - )
> - endif(LLVM_INCLUDE_TESTS)
> -
> - find_package(LLVM 8.0.0 REQUIRED
> + find_package(LLVM 9.0.0 REQUIRED
> COMPONENTS
> Analysis
> @@ -56,9 +49,6 @@ set(LLVM_SPIRV_INCLUDE_DIRS
> ${CMAKE_CURRENT_SOURCE_DIR}/include)
> @@ -46,5 +47,5 @@ index d632a50..81ddf62 100644
> install(
> FILES
> --
> -2.7.4
> +2.17.1
>
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0002-OpenCL-Simplify-LLVM-IR-generated-for-
> OpenCL-blocks.patch b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0002-OpenCL-Simplify-LLVM-IR-generated-for-
> OpenCL-blocks.patch
> deleted file mode 100644
> index 2e935a1..0000000
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0002-
> OpenCL-Simplify-LLVM-IR-generated-for-OpenCL-blocks.patch
> +++ /dev/null
> @@ -1,294 +0,0 @@
> -From c94ec28600255098ffb9d73d1b386a7c8a535590 Mon Sep 17 00:00:00
> 2001
> -From: Andrew Savonichev <andrew.savonichev at intel.com>
> -Date: Thu, 21 Feb 2019 11:02:10 +0000
> -Subject: [PATCH 2/2] [OpenCL] Simplify LLVM IR generated for OpenCL
> blocks
> -
> -Summary:
> -Emit direct call of block invoke functions when possible, i.e. in
> case the
> -block is not passed as a function argument.
> -Also doing some refactoring of
> `CodeGenFunction::EmitBlockCallExpr()`
> -
> -Reviewers: Anastasia, yaxunl, svenvh
> -
> -Reviewed By: Anastasia
> -
> -Subscribers: cfe-commits
> -
> -Tags: #clang
> -
> -Differential Revision: https://reviews.llvm.org/D58388
> -
> -git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@354568
> 91177308-0d34-0410-b5e6-96231b3b80d8
> -
> -Upstream-Status: Backport
> -[
> https://github.com/llvm-mirror/clang/commit/eae71f8d05ce550c4e2595c9b7082cc2c7882c58
> ]
> -Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> ----
> - lib/CodeGen/CGBlocks.cpp | 77 +++++++++++++
> -------------
> - lib/CodeGen/CGOpenCLRuntime.cpp | 30 +++++++---
> - lib/CodeGen/CGOpenCLRuntime.h | 4 ++
> - test/CodeGenOpenCL/blocks.cl | 10 +---
> - test/CodeGenOpenCL/cl20-device-side-enqueue.cl | 34 +++++++++---
> - 5 files changed, 91 insertions(+), 64 deletions(-)
> -
> -diff --git a/lib/CodeGen/CGBlocks.cpp b/lib/CodeGen/CGBlocks.cpp
> -index fa3c3ee..10a0238 100644
> ---- a/lib/CodeGen/CGBlocks.cpp
> -+++ b/lib/CodeGen/CGBlocks.cpp
> -@@ -1261,52 +1261,49 @@ RValue
> CodeGenFunction::EmitBlockCallExpr(const CallExpr *E,
> - ReturnValueSlot
> ReturnValue) {
> - const BlockPointerType *BPT =
> - E->getCallee()->getType()->getAs<BlockPointerType>();
> --
> - llvm::Value *BlockPtr = EmitScalarExpr(E->getCallee());
> --
> -- // Get a pointer to the generic block literal.
> -- // For OpenCL we generate generic AS void ptr to be able to reuse
> the same
> -- // block definition for blocks with captures generated as private
> AS local
> -- // variables and without captures generated as global AS program
> scope
> -- // variables.
> -- unsigned AddrSpace = 0;
> -- if (getLangOpts().OpenCL)
> -- AddrSpace =
> getContext().getTargetAddressSpace(LangAS::opencl_generic);
> --
> -- llvm::Type *BlockLiteralTy =
> -- llvm::PointerType::get(CGM.getGenericBlockLiteralType(),
> AddrSpace);
> --
> -- // Bitcast the callee to a block literal.
> -- BlockPtr =
> -- Builder.CreatePointerCast(BlockPtr, BlockLiteralTy,
> "block.literal");
> --
> -- // Get the function pointer from the literal.
> -- llvm::Value *FuncPtr =
> -- Builder.CreateStructGEP(CGM.getGenericBlockLiteralType(),
> BlockPtr,
> -- CGM.getLangOpts().OpenCL ? 2 : 3);
> --
> -- // Add the block literal.
> -+ llvm::Type *GenBlockTy = CGM.getGenericBlockLiteralType();
> -+ llvm::Value *Func = nullptr;
> -+ QualType FnType = BPT->getPointeeType();
> -+ ASTContext &Ctx = getContext();
> - CallArgList Args;
> -
> -- QualType VoidPtrQualTy = getContext().VoidPtrTy;
> -- llvm::Type *GenericVoidPtrTy = VoidPtrTy;
> - if (getLangOpts().OpenCL) {
> -- GenericVoidPtrTy =
> CGM.getOpenCLRuntime().getGenericVoidPointerType();
> -- VoidPtrQualTy =
> -- getContext().getPointerType(getContext().getAddrSpaceQualTy
> pe(
> -- getContext().VoidTy, LangAS::opencl_generic));
> -- }
> --
> -- BlockPtr = Builder.CreatePointerCast(BlockPtr, GenericVoidPtrTy);
> -- Args.add(RValue::get(BlockPtr), VoidPtrQualTy);
> --
> -- QualType FnType = BPT->getPointeeType();
> -+ // For OpenCL, BlockPtr is already casted to generic block
> literal.
> -+
> -+ // First argument of a block call is a generic block literal
> casted to
> -+ // generic void pointer, i.e. i8 addrspace(4)*
> -+ llvm::Value *BlockDescriptor = Builder.CreatePointerCast(
> -+ BlockPtr,
> CGM.getOpenCLRuntime().getGenericVoidPointerType());
> -+ QualType VoidPtrQualTy = Ctx.getPointerType(
> -+ Ctx.getAddrSpaceQualType(Ctx.VoidTy,
> LangAS::opencl_generic));
> -+ Args.add(RValue::get(BlockDescriptor), VoidPtrQualTy);
> -+ // And the rest of the arguments.
> -+ EmitCallArgs(Args, FnType->getAs<FunctionProtoType>(), E-
> >arguments());
> -+
> -+ // We *can* call the block directly unless it is a function
> argument.
> -+ if (!isa<ParmVarDecl>(E->getCalleeDecl()))
> -+ Func = CGM.getOpenCLRuntime().getInvokeFunction(E-
> >getCallee());
> -+ else {
> -+ llvm::Value *FuncPtr = Builder.CreateStructGEP(GenBlockTy,
> BlockPtr, 2);
> -+ Func = Builder.CreateAlignedLoad(FuncPtr, getPointerAlign());
> -+ }
> -+ } else {
> -+ // Bitcast the block literal to a generic block literal.
> -+ BlockPtr = Builder.CreatePointerCast(
> -+ BlockPtr, llvm::PointerType::get(GenBlockTy, 0),
> "block.literal");
> -+ // Get pointer to the block invoke function
> -+ llvm::Value *FuncPtr = Builder.CreateStructGEP(GenBlockTy,
> BlockPtr, 3);
> -
> -- // And the rest of the arguments.
> -- EmitCallArgs(Args, FnType->getAs<FunctionProtoType>(), E-
> >arguments());
> -+ // First argument is a block literal casted to a void pointer
> -+ BlockPtr = Builder.CreatePointerCast(BlockPtr, VoidPtrTy);
> -+ Args.add(RValue::get(BlockPtr), Ctx.VoidPtrTy);
> -+ // And the rest of the arguments.
> -+ EmitCallArgs(Args, FnType->getAs<FunctionProtoType>(), E-
> >arguments());
> -
> -- // Load the function.
> -- llvm::Value *Func = Builder.CreateAlignedLoad(FuncPtr,
> getPointerAlign());
> -+ // Load the function.
> -+ Func = Builder.CreateAlignedLoad(FuncPtr, getPointerAlign());
> -+ }
> -
> - const FunctionType *FuncTy = FnType->castAs<FunctionType>();
> - const CGFunctionInfo &FnInfo =
> -diff --git a/lib/CodeGen/CGOpenCLRuntime.cpp
> b/lib/CodeGen/CGOpenCLRuntime.cpp
> -index 7f6f595..75003e5 100644
> ---- a/lib/CodeGen/CGOpenCLRuntime.cpp
> -+++ b/lib/CodeGen/CGOpenCLRuntime.cpp
> -@@ -123,6 +123,23 @@ llvm::PointerType
> *CGOpenCLRuntime::getGenericVoidPointerType() {
> - CGM.getContext().getTargetAddressSpace(LangAS::opencl_generic
> ));
> - }
> -
> -+// Get the block literal from an expression derived from the block
> expression.
> -+// OpenCL v2.0 s6.12.5:
> -+// Block variable declarations are implicitly qualified with const.
> Therefore
> -+// all block variables must be initialized at declaration time and
> may not be
> -+// reassigned.
> -+static const BlockExpr *getBlockExpr(const Expr *E) {
> -+ const Expr *Prev = nullptr; // to make sure we do not stuck in
> infinite loop.
> -+ while(!isa<BlockExpr>(E) && E != Prev) {
> -+ Prev = E;
> -+ E = E->IgnoreCasts();
> -+ if (auto DR = dyn_cast<DeclRefExpr>(E)) {
> -+ E = cast<VarDecl>(DR->getDecl())->getInit();
> -+ }
> -+ }
> -+ return cast<BlockExpr>(E);
> -+}
> -+
> - /// Record emitted llvm invoke function and llvm block literal for
> the
> - /// corresponding block expression.
> - void CGOpenCLRuntime::recordBlockInfo(const BlockExpr *E,
> -@@ -137,20 +154,17 @@ void CGOpenCLRuntime::recordBlockInfo(const
> BlockExpr *E,
> - EnqueuedBlockMap[E].Kernel = nullptr;
> - }
> -
> -+llvm::Function *CGOpenCLRuntime::getInvokeFunction(const Expr *E) {
> -+ return EnqueuedBlockMap[getBlockExpr(E)].InvokeFunc;
> -+}
> -+
> - CGOpenCLRuntime::EnqueuedBlockInfo
> - CGOpenCLRuntime::emitOpenCLEnqueuedBlock(CodeGenFunction &CGF,
> const Expr *E) {
> - CGF.EmitScalarExpr(E);
> -
> - // The block literal may be assigned to a const variable. Chasing
> down
> - // to get the block literal.
> -- if (auto DR = dyn_cast<DeclRefExpr>(E)) {
> -- E = cast<VarDecl>(DR->getDecl())->getInit();
> -- }
> -- E = E->IgnoreImplicit();
> -- if (auto Cast = dyn_cast<CastExpr>(E)) {
> -- E = Cast->getSubExpr();
> -- }
> -- auto *Block = cast<BlockExpr>(E);
> -+ const BlockExpr *Block = getBlockExpr(E);
> -
> - assert(EnqueuedBlockMap.find(Block) != EnqueuedBlockMap.end() &&
> - "Block expression not emitted");
> -diff --git a/lib/CodeGen/CGOpenCLRuntime.h
> b/lib/CodeGen/CGOpenCLRuntime.h
> -index 750721f..4effc7e 100644
> ---- a/lib/CodeGen/CGOpenCLRuntime.h
> -+++ b/lib/CodeGen/CGOpenCLRuntime.h
> -@@ -92,6 +92,10 @@ public:
> - /// \param Block block literal emitted for the block expression.
> - void recordBlockInfo(const BlockExpr *E, llvm::Function *InvokeF,
> - llvm::Value *Block);
> -+
> -+ /// \return LLVM block invoke function emitted for an expression
> derived from
> -+ /// the block expression.
> -+ llvm::Function *getInvokeFunction(const Expr *E);
> - };
> -
> - }
> -diff --git a/test/CodeGenOpenCL/blocks.cl
> b/test/CodeGenOpenCL/blocks.cl
> -index 19aacc3..ab5a2c6 100644
> ---- a/test/CodeGenOpenCL/blocks.cl
> -+++ b/test/CodeGenOpenCL/blocks.cl
> -@@ -39,11 +39,8 @@ void foo(){
> - // SPIR: %[[blk_gen_ptr:.*]] = addrspacecast
> %struct.__opencl_block_literal_generic* %[[blk_ptr]] to
> %struct.__opencl_block_literal_generic addrspace(4)*
> - // SPIR: store %struct.__opencl_block_literal_generic
> addrspace(4)* %[[blk_gen_ptr]],
> %struct.__opencl_block_literal_generic addrspace(4)**
> %[[block_B:.*]],
> - // SPIR: %[[block_literal:.*]] = load
> %struct.__opencl_block_literal_generic addrspace(4)*,
> %struct.__opencl_block_literal_generic addrspace(4)** %[[block_B]]
> -- // SPIR: %[[invoke_addr:.*]] = getelementptr inbounds
> %struct.__opencl_block_literal_generic,
> %struct.__opencl_block_literal_generic addrspace(4)*
> %[[block_literal]], i32 0, i32 2
> - // SPIR: %[[blk_gen_ptr:.*]] = bitcast
> %struct.__opencl_block_literal_generic addrspace(4)*
> %[[block_literal]] to i8 addrspace(4)*
> -- // SPIR: %[[invoke_func_ptr:.*]] = load i8 addrspace(4)*, i8
> addrspace(4)* addrspace(4)* %[[invoke_addr]]
> -- // SPIR: %[[invoke_func:.*]] = addrspacecast i8 addrspace(4)*
> %[[invoke_func_ptr]] to i32 (i8 addrspace(4)*)*
> -- // SPIR: call {{.*}}i32 %[[invoke_func]](i8 addrspace(4)*
> %[[blk_gen_ptr]])
> -+ // SPIR: call {{.*}}i32 @__foo_block_invoke(i8 addrspace(4)*
> %[[blk_gen_ptr]])
> - // AMDGCN: %[[block_invoke:.*]] = getelementptr inbounds <{ i32,
> i32, i8*, i32 }>, <{ i32, i32, i8*, i32 }> addrspace(5)*
> %[[block:.*]], i32 0, i32 2
> - // AMDGCN: store i8* bitcast (i32 (i8*)* @__foo_block_invoke to
> i8*), i8* addrspace(5)* %[[block_invoke]]
> - // AMDGCN: %[[block_captured:.*]] = getelementptr inbounds <{
> i32, i32, i8*, i32 }>, <{ i32, i32, i8*, i32 }> addrspace(5)*
> %[[block]], i32 0, i32 3
> -@@ -53,11 +50,8 @@ void foo(){
> - // AMDGCN: %[[blk_gen_ptr:.*]] = addrspacecast
> %struct.__opencl_block_literal_generic addrspace(5)* %[[blk_ptr]] to
> %struct.__opencl_block_literal_generic*
> - // AMDGCN: store %struct.__opencl_block_literal_generic*
> %[[blk_gen_ptr]], %struct.__opencl_block_literal_generic*
> addrspace(5)* %[[block_B:.*]],
> - // AMDGCN: %[[block_literal:.*]] = load
> %struct.__opencl_block_literal_generic*,
> %struct.__opencl_block_literal_generic* addrspace(5)* %[[block_B]]
> -- // AMDGCN: %[[invoke_addr:.*]] = getelementptr inbounds
> %struct.__opencl_block_literal_generic,
> %struct.__opencl_block_literal_generic* %[[block_literal]], i32 0,
> i32 2
> - // AMDGCN: %[[blk_gen_ptr:.*]] = bitcast
> %struct.__opencl_block_literal_generic* %[[block_literal]] to i8*
> -- // AMDGCN: %[[invoke_func_ptr:.*]] = load i8*, i8**
> %[[invoke_addr]]
> -- // AMDGCN: %[[invoke_func:.*]] = bitcast i8* %[[invoke_func_ptr]]
> to i32 (i8*)*
> -- // AMDGCN: call {{.*}}i32 %[[invoke_func]](i8* %[[blk_gen_ptr]])
> -+ // AMDGCN: call {{.*}}i32 @__foo_block_invoke(i8*
> %[[blk_gen_ptr]])
> -
> - int (^ block_B)(void) = ^{
> - return i;
> -diff --git a/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> b/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> -index 8445016..1566912 100644
> ---- a/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> -+++ b/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> -@@ -312,9 +312,7 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - };
> -
> - // Uses global block literal [[BLG8]] and invoke function
> [[INVG8]].
> -- // COMMON: [[r1:%.*]] = load i8 addrspace(4)*, i8 addrspace(4)*
> addrspace(4)* getelementptr inbounds
> (%struct.__opencl_block_literal_generic,
> %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast
> (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32,
> i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to
> %struct.__opencl_block_literal_generic addrspace(1)*) to
> %struct.__opencl_block_literal_generic addrspace(4)*), i32 0, i32 2)
> -- // COMMON: [[r2:%.*]] = addrspacecast i8 addrspace(4)* [[r1]] to
> void (i8 addrspace(4)*)*
> -- // COMMON: call spir_func void [[r2]](i8 addrspace(4)*
> addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)*
> } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*))
> -+ // COMMON: call spir_func void
> @__device_side_enqueue_block_invoke_11(i8 addrspace(4)* addrspacecast
> (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* }
> addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*))
> - block_A();
> -
> - // Emits global block literal [[BLG8]] and block kernel
> [[INVGK8]]. [[INVGK8]] calls [[INVG8]].
> -@@ -333,15 +331,35 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - unsigned size = get_kernel_work_group_size(block_A);
> -
> - // Uses global block literal [[BLG8]] and invoke function
> [[INVG8]]. Make sure no redundant block literal and invoke functions
> are emitted.
> -- // COMMON: [[r1:%.*]] = load i8 addrspace(4)*, i8 addrspace(4)*
> addrspace(4)* getelementptr inbounds
> (%struct.__opencl_block_literal_generic,
> %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast
> (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32,
> i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to
> %struct.__opencl_block_literal_generic addrspace(1)*) to
> %struct.__opencl_block_literal_generic addrspace(4)*), i32 0, i32 2)
> -- // COMMON: [[r2:%.*]] = addrspacecast i8 addrspace(4)* [[r1]] to
> void (i8 addrspace(4)*)*
> -- // COMMON: call spir_func void [[r2]](i8 addrspace(4)*
> addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)*
> } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*))
> -+ // COMMON: call spir_func void
> @__device_side_enqueue_block_invoke_11(i8 addrspace(4)* addrspacecast
> (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* }
> addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*))
> - block_A();
> -
> -+ // Make sure that block invoke function is resolved correctly
> after sequence of assignements.
> -+ // COMMON: store %struct.__opencl_block_literal_generic
> addrspace(4)*
> -+ // COMMON-SAME: addrspacecast
> (%struct.__opencl_block_literal_generic addrspace(1)*
> -+ // COMMON-SAME: bitcast ({ i32, i32, i8 addrspace(4)* }
> addrspace(1)* [[BL_GLOBAL]] to %struct.__opencl_block_literal_generic
> addrspace(1)*)
> -+ // COMMON-SAME: to %struct.__opencl_block_literal_generic
> addrspace(4)*),
> -+ // COMMON-SAME: %struct.__opencl_block_literal_generic
> addrspace(4)** %b1,
> -+ bl_t b1 = block_G;
> -+ // COMMON: store %struct.__opencl_block_literal_generic
> addrspace(4)*
> -+ // COMMON-SAME: addrspacecast
> (%struct.__opencl_block_literal_generic addrspace(1)*
> -+ // COMMON-SAME: bitcast ({ i32, i32, i8 addrspace(4)* }
> addrspace(1)* [[BL_GLOBAL]] to %struct.__opencl_block_literal_generic
> addrspace(1)*)
> -+ // COMMON-SAME: to %struct.__opencl_block_literal_generic
> addrspace(4)*),
> -+ // COMMON-SAME: %struct.__opencl_block_literal_generic
> addrspace(4)** %b2,
> -+ bl_t b2 = b1;
> -+ // COMMON: call spir_func void @block_G_block_invoke(i8
> addrspace(4)* addrspacecast (i8 addrspace(1)*
> -+ // COMMON-SAME: bitcast ({ i32, i32, i8 addrspace(4)* }
> addrspace(1)* [[BL_GLOBAL]] to i8 addrspace(1)*)
> -+ // COOMON-SAME: to i8 addrspace(4)*), i8 addrspace(3)* null)
> -+ b2(0);
> -+ // Uses global block literal [[BL_GLOBAL]] and block kernel
> [[INV_G_K]]. [[INV_G_K]] calls [[INV_G]].
> -+ // COMMON: call i32
> @__get_kernel_preferred_work_group_size_multiple_impl(
> -+ // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> ({{.*}} [[INV_G_K:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> -+ // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]]
> to i8 addrspace(1)*) to i8 addrspace(4)*))
> -+ size = get_kernel_preferred_work_group_size_multiple(b2);
> -+
> - void (^block_C)(void) = ^{
> - callee(i, a);
> - };
> --
> - // Emits block literal on stack and block kernel [[INVLK3]].
> - // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast
> (void (i8 addrspace(4)*)*
> [[INVL3:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8
> addrspace(4)*), i8 addrspace(4)** %block.invoke
> - // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> %opencl.queue_t{{.*}}** %default_queue
> -@@ -404,8 +422,8 @@ kernel void device_side_enqueue(global int *a,
> global int *b, int i) {
> - // COMMON: define internal spir_func void [[INVG8]](i8
> addrspace(4)*{{.*}})
> - // COMMON: define internal spir_func void [[INVG9]](i8
> addrspace(4)*{{.*}}, i8 addrspace(3)* %{{.*}})
> - // COMMON: define internal spir_kernel void [[INVGK8]](i8
> addrspace(4)*{{.*}})
> -+// COMMON: define internal spir_kernel void [[INV_G_K]](i8
> addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}})
> - // COMMON: define internal spir_kernel void [[INVLK3]](i8
> addrspace(4)*{{.*}})
> - // COMMON: define internal spir_kernel void [[INVGK9]](i8
> addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}})
> --// COMMON: define internal spir_kernel void [[INV_G_K]](i8
> addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}})
> - // COMMON: define internal spir_kernel void [[INVGK10]](i8
> addrspace(4)*{{.*}})
> - // COMMON: define internal spir_kernel void [[INVGK11]](i8
> addrspace(4)*{{.*}})
> ---
> -1.8.3.1
> -
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0002-Remove-extra-semicolon.patch b/dynamic-
> layers/clang-layer/recipes-devtools/clang/files/0002-Remove-extra-
> semicolon.patch
> deleted file mode 100644
> index 5f5a957..0000000
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0002-
> Remove-extra-semicolon.patch
> +++ /dev/null
> @@ -1,27 +0,0 @@
> -From a2c093c8db7b4e3a5612d0fcce9e4fd1756d6e4b Mon Sep 17 00:00:00
> 2001
> -From: Alexey Sotkin <alexey.sotkin at intel.com>
> -Date: Mon, 5 Aug 2019 18:18:01 +0300
> -Subject: [PATCH] Remove extra semicolon
> -
> -Upstream-Status: Backport
> -Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> ----
> - lib/SPIRV/libSPIRV/SPIRVEnum.h | 2 +-
> - 1 file changed, 1 insertion(+), 1 deletion(-)
> -
> -diff --git a/lib/SPIRV/libSPIRV/SPIRVEnum.h
> b/lib/SPIRV/libSPIRV/SPIRVEnum.h
> -index c93a484..3a071e7 100644
> ---- a/lib/SPIRV/libSPIRV/SPIRVEnum.h
> -+++ b/lib/SPIRV/libSPIRV/SPIRVEnum.h
> -@@ -124,7 +124,7 @@ template <> inline void
> SPIRVMap<SPIRVExtensionKind, std::string>::init() {
> - add(SPV_INTEL_device_side_avc_motion_estimation,
> - "SPV_INTEL_device_side_avc_motion_estimation");
> - add(SPV_KHR_no_integer_wrap_decoration,
> "SPV_KHR_no_integer_wrap_decoration");
> --};
> -+}
> -
> - template <> inline void SPIRVMap<SPIRVExtInstSetKind,
> std::string>::init() {
> - add(SPIRVEIS_OpenCL, "OpenCL.std");
> ---
> -2.7.4
> -
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/0003-OpenCL-Fix-assertion-due-to-blocks.patch
> b/dynamic-layers/clang-layer/recipes-devtools/clang/files/0003-
> OpenCL-Fix-assertion-due-to-blocks.patch
> deleted file mode 100644
> index 510c7c6..0000000
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/files/0003-
> OpenCL-Fix-assertion-due-to-blocks.patch
> +++ /dev/null
> @@ -1,61 +0,0 @@
> -From 29e2813a2ab7d5569860bb07892dfef7b5374d96 Mon Sep 17 00:00:00
> 2001
> -From: Yaxun Liu <Yaxun.Liu at amd.com>
> -Date: Tue, 26 Feb 2019 16:20:41 +0000
> -Subject: [PATCH] [OpenCL] Fix assertion due to blocks
> -
> -A recent change caused assertion in
> CodeGenFunction::EmitBlockCallExpr when a block is called.
> -
> -There is code
> -
> - Func = CGM.getOpenCLRuntime().getInvokeFunction(E->getCallee());
> -getCalleeDecl calls Expr::getReferencedDeclOfCallee, which does not
> handle
> -BlockExpr and returns nullptr, which causes isa to assert.
> -
> -This patch fixes that.
> -
> -Differential Revision: https://reviews.llvm.org/D58658
> -
> -
> -git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@354893
> 91177308-0d34-0410-b5e6-96231b3b80d8
> -
> -Upstream-Status: Backport
> -[
> https://github.com/llvm-mirror/clang/commit/29e2813a2ab7d5569860bb07892dfef7b5374d96
> ]
> -Signed-off-by: Anuj Mittal <anuj.mittal at intel.com>
> ----
> - lib/AST/Expr.cpp | 2 ++
> - test/CodeGenOpenCL/blocks.cl | 6 ++++++
> - 2 files changed, 8 insertions(+)
> -
> -diff --git a/lib/AST/Expr.cpp b/lib/AST/Expr.cpp
> -index aef1eab..85690c7 100644
> ---- a/lib/AST/Expr.cpp
> -+++ b/lib/AST/Expr.cpp
> -@@ -1358,6 +1358,8 @@ Decl *Expr::getReferencedDeclOfCallee() {
> - return DRE->getDecl();
> - if (MemberExpr *ME = dyn_cast<MemberExpr>(CEE))
> - return ME->getMemberDecl();
> -+ if (auto *BE = dyn_cast<BlockExpr>(CEE))
> -+ return BE->getBlockDecl();
> -
> - return nullptr;
> - }
> -diff --git a/test/CodeGenOpenCL/blocks.cl
> b/test/CodeGenOpenCL/blocks.cl
> -index ab5a2c6..c3e2685 100644
> ---- a/test/CodeGenOpenCL/blocks.cl
> -+++ b/test/CodeGenOpenCL/blocks.cl
> -@@ -90,6 +90,12 @@ int get42() {
> - return blockArgFunc(^{return 42;});
> - }
> -
> -+// COMMON-LABEL: define {{.*}}@call_block
> -+// call {{.*}}@__call_block_block_invoke
> -+int call_block() {
> -+ return ^int(int num) { return num; } (11);
> -+}
> -+
> - // CHECK-DEBUG: !DIDerivedType(tag: DW_TAG_member, name: "__size"
> - // CHECK-DEBUG: !DIDerivedType(tag: DW_TAG_member, name: "__align"
> -
> ---
> -1.8.3.1
> -
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/BasicBlockUtils-Add-metadata-fixing-in-
> SplitBlockPre.patch b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/BasicBlockUtils-Add-metadata-fixing-in-
> SplitBlockPre.patch
> new file mode 100644
> index 0000000..17d7018
> --- /dev/null
> +++ b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/BasicBlockUtils-Add-metadata-fixing-in-
> SplitBlockPre.patch
> @@ -0,0 +1,108 @@
> +From eeb816d95f0910bd246e37bb2bb3923acf0edf6b Mon Sep 17 00:00:00
> 2001
> +From: Aleksander Us <aleksander.us at intel.com>
> +Date: Mon, 26 Aug 2019 15:47:41 +0300
> +Subject: [PATCH] [BasicBlockUtils] Add metadata fixing in
> + SplitBlockPredecessors.
> +
> +In case when BB is header of some loop and predecessor is latch of
> +this loop, metadata was not attached to newly created basic block.
> +This led to loss of loop metadata for other passes.
> +
> +Upstream-Status: Backport [
> https://github.com/intel/llvm-patches/commit/8af4449e2d201707f7f2f832b473a0439e255f32
> ]
> +Signed-off-by: Naveen Saini <naveen.kumar.saini at intel.com>
> +---
> + lib/Transforms/Utils/BasicBlockUtils.cpp | 23 ++++++++----
> + test/Transforms/LoopSimplify/loop_metadata.ll | 36
> +++++++++++++++++++
> + 2 files changed, 52 insertions(+), 7 deletions(-)
> + create mode 100644 test/Transforms/LoopSimplify/loop_metadata.ll
> +
> +diff --git a/lib/Transforms/Utils/BasicBlockUtils.cpp
> b/lib/Transforms/Utils/BasicBlockUtils.cpp
> +index 5fa371377c8..3a90ae061fb 100644
> +--- a/lib/Transforms/Utils/BasicBlockUtils.cpp
> ++++ b/lib/Transforms/Utils/BasicBlockUtils.cpp
> +@@ -579,24 +579,33 @@ BasicBlock
> *llvm::SplitBlockPredecessors(BasicBlock *BB,
> +
> + // The new block unconditionally branches to the old block.
> + BranchInst *BI = BranchInst::Create(BB, NewBB);
> ++ bool IsBBHeader = LI && LI->isLoopHeader(BB);
> ++ Loop *BBLoop = LI ? LI->getLoopFor(BB) : nullptr;
> + // Splitting the predecessors of a loop header creates a
> preheader block.
> +- if (LI && LI->isLoopHeader(BB))
> ++ if (IsBBHeader)
> + // Using the loop start line number prevents debuggers stepping
> into the
> + // loop body for this instruction.
> +- BI->setDebugLoc(LI->getLoopFor(BB)->getStartLoc());
> ++ BI->setDebugLoc(BBLoop->getStartLoc());
> + else
> + BI->setDebugLoc(BB->getFirstNonPHIOrDbg()->getDebugLoc());
> +
> + // Move the edges from Preds to point to NewBB instead of BB.
> +- for (unsigned i = 0, e = Preds.size(); i != e; ++i) {
> ++ for (BasicBlock *Pred : Preds) {
> ++ Instruction *PI = Pred->getTerminator();
> + // This is slightly more strict than necessary; the minimum
> requirement
> + // is that there be no more than one indirectbr branching to
> BB. And
> + // all BlockAddress uses would need to be updated.
> +- assert(!isa<IndirectBrInst>(Preds[i]->getTerminator()) &&
> ++ assert(!isa<IndirectBrInst>(PI) &&
> + "Cannot split an edge from an IndirectBrInst");
> +- assert(!isa<CallBrInst>(Preds[i]->getTerminator()) &&
> +- "Cannot split an edge from a CallBrInst");
> +- Preds[i]->getTerminator()->replaceUsesOfWith(BB, NewBB);
> ++ assert(!isa<CallBrInst>(PI) && "Cannot split an edge from a
> CallBrInst");
> ++ if (IsBBHeader && BBLoop->contains(Pred) && BBLoop-
> >isLoopLatch(Pred)) {
> ++ // Update loop metadata if it exists.
> ++ if (MDNode *LoopMD = PI->getMetadata(LLVMContext::MD_loop)) {
> ++ BI->setMetadata(LLVMContext::MD_loop, LoopMD);
> ++ PI->setMetadata(LLVMContext::MD_loop, nullptr);
> ++ }
> ++ }
> ++ PI->replaceUsesOfWith(BB, NewBB);
> + }
> +
> + // Insert a new PHI node into NewBB for every PHI node in BB and
> that new PHI
> +diff --git a/test/Transforms/LoopSimplify/loop_metadata.ll
> b/test/Transforms/LoopSimplify/loop_metadata.ll
> +new file mode 100644
> +index 00000000000..c15c92fe3ae
> +--- /dev/null
> ++++ b/test/Transforms/LoopSimplify/loop_metadata.ll
> +@@ -0,0 +1,36 @@
> ++; RUN: opt -S -loop-simplify < %s | FileCheck %s
> ++
> ++; CHECK: for.cond.loopexit:
> ++; CHECK: br label %for.cond, !llvm.loop !0
> ++; CHECK: br i1 %cmp1, label %for.body1, label %for.cond.loopexit
> ++
> ++define void @foo() {
> ++entry:
> ++ br label %for.cond
> ++
> ++for.cond: ; preds =
> %for.cond1, %entry
> ++ %j = phi i32 [ 0, %entry ], [ %add, %for.cond1 ]
> ++ %cmp = icmp ult i32 %j, 8
> ++ br i1 %cmp, label %for.body, label %for.end
> ++
> ++for.body: ; preds =
> %for.cond
> ++ %dummy1 = add i32 1, 1
> ++ %add = add nuw nsw i32 %j, 1
> ++ br label %for.cond1
> ++
> ++for.cond1: ; preds =
> %for.body1, %for.body
> ++ %i.0 = phi i32 [ 1, %for.body ], [ %inc, %for.body1 ]
> ++ %cmp1 = icmp ult i32 %i.0, 8
> ++ br i1 %cmp1, label %for.body1, label %for.cond, !llvm.loop !0
> ++
> ++for.body1: ; preds =
> %for.cond1
> ++ %dummy2 = add i32 1, 1
> ++ %inc = add nuw nsw i32 %i.0, 1
> ++ br label %for.cond1
> ++
> ++for.end: ; preds = %for.cond
> ++ ret void
> ++}
> ++
> ++!0 = distinct !{!0, !1}
> ++!1 = !{!"llvm.loop.unroll.full"}
> +--
> +2.18.0
> +
> diff --git a/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/IndVarSimplify-Do-not-use-SCEV-expander-for-
> IVCount-.patch b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/IndVarSimplify-Do-not-use-SCEV-expander-for-
> IVCount-.patch
> new file mode 100644
> index 0000000..ba95099
> --- /dev/null
> +++ b/dynamic-layers/clang-layer/recipes-
> devtools/clang/files/IndVarSimplify-Do-not-use-SCEV-expander-for-
> IVCount-.patch
> @@ -0,0 +1,143 @@
> +From 35e218a886f4c066eabd18685240d55270bd5a6d Mon Sep 17 00:00:00
> 2001
> +From: Aleksander Us <aleksander.us at intel.com>
> +Date: Mon, 26 Aug 2019 15:45:47 +0300
> +Subject: [PATCH] [IndVarSimplify] Do not use SCEV expander for
> IVCount in
> + LFTR when possible.
> +
> +SCEV analysis cannot properly cache instruction with poison flags
> +(for example, add nsw outside of loop will not be reused by
> expander).
> +This can lead to generating of additional instructions by SCEV
> expander.
> +
> +Example IR:
> +
> + ...
> + %maxval = add nuw nsw i32 %a1, %a2
> + ...
> +for.body:
> + ...
> + %cmp22 = icmp ult i32 %ivadd, %maxval
> + br i1 %cmp22, label %for.body, label %for.end
> + ...
> +
> +SCEV expander will generate copy of %maxval in preheader but without
> +nuw/nsw flags. This can be avoided by explicit check that iv count
> +value gives the same SCEV expressions as calculated by LFTR.
> +
> +Upstream-Status: Backport [
> https://github.com/intel/llvm-patches/commit/fd6a6c97341a56fd21bc32bc940afea751312e8f
> ]
> +Signed-off-by: Naveen Saini <naveen.kumar.saini at intel.com>
> +---
> + lib/Transforms/Scalar/IndVarSimplify.cpp | 12 +++++++++-
> + test/Transforms/IndVarSimplify/add_nsw.ll | 23
> ++++++++++++++++++++
> + test/Transforms/IndVarSimplify/lftr-reuse.ll | 9 +++-----
> + test/Transforms/IndVarSimplify/udiv.ll | 1 +
> + 4 files changed, 38 insertions(+), 7 deletions(-)
> + create mode 100644 test/Transforms/IndVarSimplify/add_nsw.ll
> +
> +diff --git a/lib/Transforms/Scalar/IndVarSimplify.cpp
> b/lib/Transforms/Scalar/IndVarSimplify.cpp
> +index f9fc698a4a9..5e04dac8aa6 100644
> +--- a/lib/Transforms/Scalar/IndVarSimplify.cpp
> ++++ b/lib/Transforms/Scalar/IndVarSimplify.cpp
> +@@ -2375,6 +2375,17 @@ static Value *genLoopLimit(PHINode *IndVar,
> BasicBlock *ExitingBB,
> + if (UsePostInc)
> + IVLimit = SE->getAddExpr(IVLimit, SE->getOne(IVLimit-
> >getType()));
> +
> ++ // If computed limit is equal to old limit then do not use SCEV
> expander
> ++ // because it can lost NUW/NSW flags and create extra
> instructions.
> ++ BranchInst *BI = cast<BranchInst>(ExitingBB->getTerminator());
> ++ if (ICmpInst *Cmp = dyn_cast<ICmpInst>(BI->getOperand(0))) {
> ++ Value *Limit = Cmp->getOperand(0);
> ++ if (!L->isLoopInvariant(Limit))
> ++ Limit = Cmp->getOperand(1);
> ++ if (SE->getSCEV(Limit) == IVLimit)
> ++ return Limit;
> ++ }
> ++
> + // Expand the code for the iteration count.
> + assert(SE->isLoopInvariant(IVLimit, L) &&
> + "Computed iteration count is not loop invariant!");
> +@@ -2383,7 +2394,6 @@ static Value *genLoopLimit(PHINode *IndVar,
> BasicBlock *ExitingBB,
> + // SCEV expression (IVInit) for a pointer type IV value
> (IndVar).
> + Type *LimitTy = ExitCount->getType()->isPointerTy() ?
> + IndVar->getType() : ExitCount->getType();
> +- BranchInst *BI = cast<BranchInst>(ExitingBB->getTerminator());
> + return Rewriter.expandCodeFor(IVLimit, LimitTy, BI);
> + }
> + }
> +diff --git a/test/Transforms/IndVarSimplify/add_nsw.ll
> b/test/Transforms/IndVarSimplify/add_nsw.ll
> +new file mode 100644
> +index 00000000000..abd1cbb6c51
> +--- /dev/null
> ++++ b/test/Transforms/IndVarSimplify/add_nsw.ll
> +@@ -0,0 +1,23 @@
> ++; RUN: opt -indvars -S %s | FileCheck %s
> ++
> ++target datalayout = "e-p:32:32-i64:64-n8:16:32"
> ++
> ++; CHECK: for.body.preheader:
> ++; CHECK-NOT: add
> ++; CHECK: for.body:
> ++
> ++define void @foo(i32 %a1, i32 %a2) {
> ++entry:
> ++ %maxval = add nuw nsw i32 %a1, %a2
> ++ %cmp = icmp slt i32 %maxval, 1
> ++ br i1 %cmp, label %for.end, label %for.body
> ++
> ++for.body: ; preds = %entry,
> %for.body
> ++ %j.02 = phi i32 [ 0, %entry ], [ %add31, %for.body ]
> ++ %add31 = add nuw nsw i32 %j.02, 1
> ++ %cmp22 = icmp slt i32 %add31, %maxval
> ++ br i1 %cmp22, label %for.body, label %for.end
> ++
> ++for.end: ; preds =
> %for.body
> ++ ret void
> ++}
> +diff --git a/test/Transforms/IndVarSimplify/lftr-reuse.ll
> b/test/Transforms/IndVarSimplify/lftr-reuse.ll
> +index 14ae9738696..509d662b767 100644
> +--- a/test/Transforms/IndVarSimplify/lftr-reuse.ll
> ++++ b/test/Transforms/IndVarSimplify/lftr-reuse.ll
> +@@ -67,11 +67,9 @@ define void @expandOuterRecurrence(i32 %arg)
> nounwind {
> + ; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 0, [[SUB1]]
> + ; CHECK-NEXT: br i1 [[CMP1]], label [[OUTER_PREHEADER:%.*]],
> label [[EXIT:%.*]]
> + ; CHECK: outer.preheader:
> +-; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[ARG]], -1
> + ; CHECK-NEXT: br label [[OUTER:%.*]]
> + ; CHECK: outer:
> +-; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[TMP0]],
> [[OUTER_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[OUTER_INC:%.*]] ]
> +-; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_INC:%.*]], [[OUTER_INC]]
> ], [ 0, [[OUTER_PREHEADER]] ]
> ++; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_INC:%.*]],
> [[OUTER_INC:%.*]] ], [ 0, [[OUTER_PREHEADER]] ]
> + ; CHECK-NEXT: [[SUB2:%.*]] = sub nsw i32 [[ARG]], [[I]]
> + ; CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 [[SUB2]], 1
> + ; CHECK-NEXT: [[CMP2:%.*]] = icmp slt i32 0, [[SUB3]]
> +@@ -81,14 +79,13 @@ define void @expandOuterRecurrence(i32 %arg)
> nounwind {
> + ; CHECK: inner:
> + ; CHECK-NEXT: [[J:%.*]] = phi i32 [ 0, [[INNER_PH]] ], [
> [[J_INC:%.*]], [[INNER]] ]
> + ; CHECK-NEXT: [[J_INC]] = add nuw nsw i32 [[J]], 1
> +-; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[J_INC]],
> [[INDVARS_IV]]
> ++; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[J_INC]], [[SUB3]]
> + ; CHECK-NEXT: br i1 [[EXITCOND]], label [[INNER]], label
> [[OUTER_INC_LOOPEXIT:%.*]]
> + ; CHECK: outer.inc.loopexit:
> + ; CHECK-NEXT: br label [[OUTER_INC]]
> + ; CHECK: outer.inc:
> + ; CHECK-NEXT: [[I_INC]] = add nuw nsw i32 [[I]], 1
> +-; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i32 [[INDVARS_IV]], -1
> +-; CHECK-NEXT: [[EXITCOND1:%.*]] = icmp ne i32 [[I_INC]],
> [[TMP0]]
> ++; CHECK-NEXT: [[EXITCOND1:%.*]] = icmp ne i32 [[I_INC]],
> [[SUB1]]
> + ; CHECK-NEXT: br i1 [[EXITCOND1]], label [[OUTER]], label
> [[EXIT_LOOPEXIT:%.*]]
> + ; CHECK: exit.loopexit:
> + ; CHECK-NEXT: br label [[EXIT]]
> +diff --git a/test/Transforms/IndVarSimplify/udiv.ll
> b/test/Transforms/IndVarSimplify/udiv.ll
> +index b3f2c2a6a66..3530343ef4a 100644
> +--- a/test/Transforms/IndVarSimplify/udiv.ll
> ++++ b/test/Transforms/IndVarSimplify/udiv.ll
> +@@ -133,6 +133,7 @@ declare i32 @printf(i8* nocapture, ...) nounwind
> + ; CHECK-LABEL: @foo(
> + ; CHECK: for.body.preheader:
> + ; CHECK-NOT: udiv
> ++; CHECK: for.body:
> +
> + define void @foo(double* %p, i64 %n) nounwind {
> + entry:
> +--
> +2.18.0
> +
> diff --git a/dynamic-layers/clang-layer/recipes-devtools/clang/llvm-
> project-source.bbappend b/dynamic-layers/clang-layer/recipes-
> devtools/clang/llvm-project-source.bbappend
> index 50ab25e..23cc3fc 100644
> --- a/dynamic-layers/clang-layer/recipes-devtools/clang/llvm-project-
> source.bbappend
> +++ b/dynamic-layers/clang-layer/recipes-devtools/clang/llvm-project-
> source.bbappend
> @@ -1,14 +1,12 @@
> FILESEXTRAPATHS_prepend_intel-x86-common := "${THISDIR}/files:"
>
> SRC_URI_append_intel-x86-common = " \
> - file://0001-OpenCL-Change-type-
> of-block-pointer-for-OpenCL.patch;patchdir=clang \
> - file://0002-OpenCL-Simplify-
> LLVM-IR-generated-for-OpenCL-blocks.patch;patchdir=clang \
> - file://0003-OpenCL-Fix-
> assertion-due-to-blocks.patch;patchdir=clang \
> file://0001-dont-export-targets-
> for-binaries.patch \
> - git://github.com/KhronosGroup/SP
> IRV-LLVM-
> Translator.git;protocol=https;branch=llvm_release_80;destsuffix=git/l
> lvm/projects/llvm-spirv;name=spirv \
> - file://0001-Update-LowerOpenCL-
> pass-to-handle-new-blocks-represn.patch;patchdir=llvm/projects/llvm-
> spirv \
> - file://0002-Remove-extra-
> semicolon.patch;patchdir=llvm/projects/llvm-spirv \
> + file://0001-LLDBConfig.cmake-
> skip-python-version-checks.patch \
> + file://BasicBlockUtils-Add-
> metadata-fixing-in-SplitBlockPre.patch;patchdir=llvm \
> + file://IndVarSimplify-Do-not-
> use-SCEV-expander-for-IVCount-.patch;patchdir=llvm \
> + git://github.com/KhronosGroup/SP
> IRV-LLVM-
> Translator.git;protocol=https;branch=llvm_release_90;destsuffix=git/l
> lvm/projects/llvm-spirv;name=spirv \
> file://0001-skip-building-
> tests.patch;patchdir=llvm/projects/llvm-spirv \
> "
>
> -SRCREV_spirv = "1d48cd84d04a2f60b43ea3f66eb7c86f4e5973a9"
> +SRCREV_spirv = "70420631144a6a25613ae37178f2cc1d3607b65b"
> --
> 2.7.4
>
More information about the meta-intel
mailing list