AMDGPU: Remove invalid testcase for enqueue kernel

[arsenm]

Summary

The call didn't have the right calling convention, but calls to
kernels are supposed to be illegal anyway.

[rampitec]

Unfortunately OpenCL allows calls to kernels. We even had a pass to lower such calls.

[arsenm]

In D140643#4023773, @rampitec wrote:

Unfortunately OpenCL allows calls to kernels. We even had a pass to lower such calls.

That’s hacked around separately

[rampitec]

It seems to remove the test when a caller is in a called function, because the called function is a kernel. But should it be an another function it shall be valid, yet I do not see such test. This is in fact what it would become after kernel call lowering. It seems worth to add such test with a non-kernel caller.

[arsenm]

In D140643#4023847, @rampitec wrote:

It seems to remove the test when a caller is in a called function, because the called function is a kernel. But should it be an another function it shall be valid, yet I do not see such test. This is in fact what it would become after kernel call lowering. It seems worth to add such test with a non-kernel caller.

We probably should have a separate phase ordering test for the current scheme, which is not what we have here. However, I'm working on a patch to move this into clang since the comment at the top of the file explaining why this is impossible is wrong

[rampitec]

In D140643#4028853, @arsenm wrote:

In D140643#4023847, @rampitec wrote:

It seems to remove the test when a caller is in a called function, because the called function is a kernel. But should it be an another function it shall be valid, yet I do not see such test. This is in fact what it would become after kernel call lowering. It seems worth to add such test with a non-kernel caller.

We probably should have a separate phase ordering test for the current scheme, which is not what we have here. However, I'm working on a patch to move this into clang since the comment at the top of the file explaining why this is impossible is wrong

But the test for a call for a non-inlined regular function is still needed regardless of pass ordering.

[arsenm]

In D140643#4029383, @rampitec wrote:

In D140643#4028853, @arsenm wrote:

In D140643#4023847, @rampitec wrote:

It seems to remove the test when a caller is in a called function, because the called function is a kernel. But should it be an another function it shall be valid, yet I do not see such test. This is in fact what it would become after kernel call lowering. It seems worth to add such test with a non-kernel caller.

We probably should have a separate phase ordering test for the current scheme, which is not what we have here. However, I'm working on a patch to move this into clang since the comment at the top of the file explaining why this is impossible is wrong

But the test for a call for a non-inlined regular function is still needed regardless of pass ordering.

It doesn't matter after D141701, this callgraph analysis is broken anyway

[arsenm]

ping, the test is simply invalid. calls to kernels would never be considered here

[yaxunl]

LGTM

[arsenm]

bbc7b30fbf52c0f0b5383b9b9f5ebdb45db98b55

llvm/test/CodeGen/AMDGPU/enqueue-kernel.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes --check-globals --include-generated-funcs
 ; RUN: opt -data-layout=A5 -amdgpu-lower-enqueued-block -S < %s | FileCheck %s
 
 %struct.ndrange_t = type { i32 }
 %opencl.queue_t = type opaque
 
 define amdgpu_kernel void @non_caller(ptr addrspace(1) %a, i8 %b, ptr addrspace(1) %c, i64 %d) {
   ret void
 }
 
-define amdgpu_kernel void @caller_indirect(ptr addrspace(1) %a, i8 %b, ptr addrspace(1) %c, i64 %d) {
-  call void @caller(ptr addrspace(1) %a, i8 %b, ptr addrspace(1) %c, i64 %d)
-  ret void
-}
-
 define amdgpu_kernel void @caller(ptr addrspace(1) %a, i8 %b, ptr addrspace(1) %c, i64 %d) {
 entry:
   %block = alloca <{ i32, i32, ptr addrspace(1), i8 }>, align 8, addrspace(5)
   %inst = alloca %struct.ndrange_t, align 4, addrspace(5)
   %block2 = alloca <{ i32, i32, ptr addrspace(1), ptr addrspace(1), i64, i8 }>, align 8, addrspace(5)
   %inst3 = alloca %struct.ndrange_t, align 4, addrspace(5)
   %block.size = getelementptr inbounds <{ i32, i32, ptr addrspace(1), i8 }>, ptr addrspace(5) %block, i32 0, i32 0
   store i32 25, ptr addrspace(5) %block.size, align 8

 ; CHECK: @[[__AMDGPU_ENQUEUED_KERNEL_RUNTIME_HANDLE:[a-zA-Z0-9_$"\\.-]+]] = addrspace(1) global [[BLOCK_RUNTIME_HANDLE_T:%.*]] zeroinitializer
 ; CHECK: @[[__AMDGPU_ENQUEUED_KERNEL_1_RUNTIME_HANDLE:[a-zA-Z0-9_$"\\.-]+]] = addrspace(1) global [[BLOCK_RUNTIME_HANDLE_T:%.*]] zeroinitializer
 ;.
 ; CHECK-LABEL: define {{[^@]+}}@non_caller
 ; CHECK-SAME: (ptr addrspace(1) [[A:%.*]], i8 [[B:%.*]], ptr addrspace(1) [[C:%.*]], i64 [[D:%.*]]) {
 ; CHECK-NEXT:    ret void
 ;
 ;
-; CHECK-LABEL: define {{[^@]+}}@caller_indirect
-; CHECK-SAME: (ptr addrspace(1) [[A:%.*]], i8 [[B:%.*]], ptr addrspace(1) [[C:%.*]], i64 [[D:%.*]]) {
-; CHECK-NEXT:    call void @caller(ptr addrspace(1) [[A]], i8 [[B]], ptr addrspace(1) [[C]], i64 [[D]])
-; CHECK-NEXT:    ret void
-;
-;
 ; CHECK-LABEL: define {{[^@]+}}@caller
 ; CHECK-SAME: (ptr addrspace(1) [[A:%.*]], i8 [[B:%.*]], ptr addrspace(1) [[C:%.*]], i64 [[D:%.*]]) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[BLOCK:%.*]] = alloca <{ i32, i32, ptr addrspace(1), i8 }>, align 8, addrspace(5)
 ; CHECK-NEXT:    [[INST:%.*]] = alloca [[STRUCT_NDRANGE_T:%.*]], align 4, addrspace(5)
 ; CHECK-NEXT:    [[BLOCK2:%.*]] = alloca <{ i32, i32, ptr addrspace(1), ptr addrspace(1), i64, i8 }>, align 8, addrspace(5)
 ; CHECK-NEXT:    [[INST3:%.*]] = alloca [[STRUCT_NDRANGE_T]], align 4, addrspace(5)
 ; CHECK-NEXT:    [[BLOCK_SIZE:%.*]] = getelementptr inbounds <{ i32, i32, ptr addrspace(1), i8 }>, ptr addrspace(5) [[BLOCK]], i32 0, i32 0