shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/annotate-kernel-features-hsa.ll
Shilei Tian 578741b5e8
[AMDGPU][Attributor] Rework update of AAAMDWavesPerEU (#123995) 
Currently, we use `AAAMDWavesPerEU` to iteratively update values based
on attributes from the associated function, potentially propagating
user-annotated values, along with `AAAMDFlatWorkGroupSize`. Similarly,
we have `AAAMDFlatWorkGroupSize`. However, since the value calculated
through the flat workgroup size always dominates the user annotation
(i.e., the attribute), running `AAAMDWavesPerEU` iteratively is
unnecessary if no user-annotated value exists.

This PR completely rewrites how the `amdgpu-waves-per-eu` attribute is
handled in `AMDGPUAttributor`. The key changes are as follows:

- `AAAMDFlatWorkGroupSize` remains unchanged.
- `AAAMDWavesPerEU` now only propagates user-annotated values.
- A new function is added to check and update `amdgpu-waves-per-eu`
based on the following rules:
- No waves per eu, no flat workgroup size: Assume a flat workgroup size
of `1,1024` and compute waves per eu based on this.
- No waves per eu, flat workgroup size exists: Use the provided flat
workgroup size to compute waves-per-eu.
- Waves per eu exists, no flat workgroup size: This is a tricky case. In
this PR, we assume a flat workgroup size of `1,1024`, but this can be
adjusted if a different approach is preferred. Alternatively, we could
directly use the user-annotated value.
- Both waves per eu and flat workgroup size exist: If there’s a
conflict, the value derived from the flat workgroup size takes
precedence over waves per eu.

This PR also updates the logic for merging two waves per eu pairs. The
current implementation, which uses `clampStateAndIndicateChange` to
compute a union, might not be ideal. If we think from ensure proper
resource allocation perspective, for instance, if one pair specifies a
minimum of 2 waves per eu, and another specifies a minimum of 4, we
should guarantee that 4 waves per eu can be supported, as failing to do
so could result in excessive resource allocation per wave. A similar
principle applies to the upper bound. Thus, the PR uses the following
approach for merging two pairs, `lo_a,up_a` and `lo_b,up_b`: `max(lo_a,
lo_b), max(up_a, up_b)`. This ensures that resource allocation adheres
to the stricter constraints from both inputs.

Fix #123092.
2025-05-17 01:01:09 -04:00

491 lines
26 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-globals
; RUN: opt -mtriple=amdgcn-unknown-amdhsa -S -passes=amdgpu-attributor < %s | FileCheck -check-prefix=HSA %s
declare i32 @llvm.amdgcn.workgroup.id.x() #0
declare i32 @llvm.amdgcn.workgroup.id.y() #0
declare i32 @llvm.amdgcn.workgroup.id.z() #0
declare i32 @llvm.amdgcn.workitem.id.x() #0
declare i32 @llvm.amdgcn.workitem.id.y() #0
declare i32 @llvm.amdgcn.workitem.id.z() #0
declare ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() #0
declare ptr addrspace(4) @llvm.amdgcn.queue.ptr() #0
declare ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr() #0
declare ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr() #0
declare i1 @llvm.amdgcn.is.shared(ptr nocapture) #2
declare i1 @llvm.amdgcn.is.private(ptr nocapture) #2
define amdgpu_kernel void @use_tgid_x(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_x
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1:[0-9]+]] {
; HSA-NEXT: [[VAL:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val = call i32 @llvm.amdgcn.workgroup.id.x()
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tgid_y(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_y
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR2:[0-9]+]] {
; HSA-NEXT: [[VAL:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val = call i32 @llvm.amdgcn.workgroup.id.y()
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @multi_use_tgid_y(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@multi_use_tgid_y
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR2]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workgroup.id.y()
store volatile i32 %val0, ptr addrspace(1) %ptr
%val1 = call i32 @llvm.amdgcn.workgroup.id.y()
store volatile i32 %val1, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tgid_x_y(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_x_y
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR2]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workgroup.id.x()
%val1 = call i32 @llvm.amdgcn.workgroup.id.y()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tgid_z(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_z
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR3:[0-9]+]] {
; HSA-NEXT: [[VAL:%.*]] = call i32 @llvm.amdgcn.workgroup.id.z()
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val = call i32 @llvm.amdgcn.workgroup.id.z()
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tgid_x_z(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_x_z
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR3]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.z()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workgroup.id.x()
%val1 = call i32 @llvm.amdgcn.workgroup.id.z()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tgid_y_z(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_y_z
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR4:[0-9]+]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.z()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workgroup.id.y()
%val1 = call i32 @llvm.amdgcn.workgroup.id.z()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tgid_x_y_z(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tgid_x_y_z
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR4]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: [[VAL2:%.*]] = call i32 @llvm.amdgcn.workgroup.id.z()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL2]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workgroup.id.x()
%val1 = call i32 @llvm.amdgcn.workgroup.id.y()
%val2 = call i32 @llvm.amdgcn.workgroup.id.z()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
store volatile i32 %val2, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tidig_x(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tidig_x
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[VAL:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val = call i32 @llvm.amdgcn.workitem.id.x()
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tidig_y(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tidig_y
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR5:[0-9]+]] {
; HSA-NEXT: [[VAL:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val = call i32 @llvm.amdgcn.workitem.id.y()
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tidig_z(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tidig_z
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR6:[0-9]+]] {
; HSA-NEXT: [[VAL:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val = call i32 @llvm.amdgcn.workitem.id.z()
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tidig_x_tgid_x(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tidig_x_tgid_x
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
%val1 = call i32 @llvm.amdgcn.workgroup.id.x()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tidig_y_tgid_y(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tidig_y_tgid_y
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR7:[0-9]+]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workitem.id.y()
%val1 = call i32 @llvm.amdgcn.workgroup.id.y()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_tidig_x_y_z(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_tidig_x_y_z
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR8:[0-9]+]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
; HSA-NEXT: [[VAL2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL2]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
%val1 = call i32 @llvm.amdgcn.workitem.id.y()
%val2 = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
store volatile i32 %val2, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_all_workitems(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_all_workitems
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR9:[0-9]+]] {
; HSA-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; HSA-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
; HSA-NEXT: [[VAL2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()
; HSA-NEXT: [[VAL3:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
; HSA-NEXT: [[VAL4:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
; HSA-NEXT: [[VAL5:%.*]] = call i32 @llvm.amdgcn.workgroup.id.z()
; HSA-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL2]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL3]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL4]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: store volatile i32 [[VAL5]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
%val1 = call i32 @llvm.amdgcn.workitem.id.y()
%val2 = call i32 @llvm.amdgcn.workitem.id.z()
%val3 = call i32 @llvm.amdgcn.workgroup.id.x()
%val4 = call i32 @llvm.amdgcn.workgroup.id.y()
%val5 = call i32 @llvm.amdgcn.workgroup.id.z()
store volatile i32 %val0, ptr addrspace(1) %ptr
store volatile i32 %val1, ptr addrspace(1) %ptr
store volatile i32 %val2, ptr addrspace(1) %ptr
store volatile i32 %val3, ptr addrspace(1) %ptr
store volatile i32 %val4, ptr addrspace(1) %ptr
store volatile i32 %val5, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_dispatch_ptr(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_dispatch_ptr
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR10:[0-9]+]] {
; HSA-NEXT: [[DISPATCH_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
; HSA-NEXT: [[VAL:%.*]] = load i32, ptr addrspace(4) [[DISPATCH_PTR]], align 4
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%dispatch.ptr = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%val = load i32, ptr addrspace(4) %dispatch.ptr
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_queue_ptr(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_queue_ptr
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR11:[0-9]+]] {
; HSA-NEXT: [[DISPATCH_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.queue.ptr()
; HSA-NEXT: [[VAL:%.*]] = load i32, ptr addrspace(4) [[DISPATCH_PTR]], align 4
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%dispatch.ptr = call ptr addrspace(4) @llvm.amdgcn.queue.ptr()
%val = load i32, ptr addrspace(4) %dispatch.ptr
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_kernarg_segment_ptr(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_kernarg_segment_ptr
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[DISPATCH_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr()
; HSA-NEXT: [[VAL:%.*]] = load i32, ptr addrspace(4) [[DISPATCH_PTR]], align 4
; HSA-NEXT: store i32 [[VAL]], ptr addrspace(1) [[PTR]], align 4
; HSA-NEXT: ret void
;
%dispatch.ptr = call ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr()
%val = load i32, ptr addrspace(4) %dispatch.ptr
store i32 %val, ptr addrspace(1) %ptr
ret void
}
define amdgpu_kernel void @use_group_to_flat_addrspacecast(ptr addrspace(3) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_group_to_flat_addrspacecast
; HSA-SAME: (ptr addrspace(3) [[PTR:%.*]]) #[[ATTR12:[0-9]+]] {
; HSA-NEXT: [[STOF:%.*]] = addrspacecast ptr addrspace(3) [[PTR]] to ptr
; HSA-NEXT: store volatile i32 0, ptr [[STOF]], align 4
; HSA-NEXT: ret void
;
%stof = addrspacecast ptr addrspace(3) %ptr to ptr
store volatile i32 0, ptr %stof
ret void
}
define amdgpu_kernel void @use_private_to_flat_addrspacecast(ptr addrspace(5) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_private_to_flat_addrspacecast
; HSA-SAME: (ptr addrspace(5) [[PTR:%.*]]) #[[ATTR13:[0-9]+]] {
; HSA-NEXT: [[STOF:%.*]] = addrspacecast ptr addrspace(5) [[PTR]] to ptr
; HSA-NEXT: store volatile i32 0, ptr [[STOF]], align 4
; HSA-NEXT: ret void
;
%stof = addrspacecast ptr addrspace(5) %ptr to ptr
store volatile i32 0, ptr %stof
ret void
}
define amdgpu_kernel void @use_flat_to_group_addrspacecast(ptr %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_flat_to_group_addrspacecast
; HSA-SAME: (ptr [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[FTOS:%.*]] = addrspacecast ptr [[PTR]] to ptr addrspace(3)
; HSA-NEXT: store volatile i32 0, ptr addrspace(3) [[FTOS]], align 4
; HSA-NEXT: ret void
;
%ftos = addrspacecast ptr %ptr to ptr addrspace(3)
store volatile i32 0, ptr addrspace(3) %ftos
ret void
}
define amdgpu_kernel void @use_flat_to_private_addrspacecast(ptr %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_flat_to_private_addrspacecast
; HSA-SAME: (ptr [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[FTOS:%.*]] = addrspacecast ptr [[PTR]] to ptr addrspace(5)
; HSA-NEXT: store volatile i32 0, ptr addrspace(5) [[FTOS]], align 4
; HSA-NEXT: ret void
;
%ftos = addrspacecast ptr %ptr to ptr addrspace(5)
store volatile i32 0, ptr addrspace(5) %ftos
ret void
}
; No-op addrspacecast should not use queue ptr
define amdgpu_kernel void @use_global_to_flat_addrspacecast(ptr addrspace(1) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_global_to_flat_addrspacecast
; HSA-SAME: (ptr addrspace(1) [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[STOF:%.*]] = addrspacecast ptr addrspace(1) [[PTR]] to ptr
; HSA-NEXT: store volatile i32 0, ptr [[STOF]], align 4
; HSA-NEXT: ret void
;
%stof = addrspacecast ptr addrspace(1) %ptr to ptr
store volatile i32 0, ptr %stof
ret void
}
define amdgpu_kernel void @use_constant_to_flat_addrspacecast(ptr addrspace(4) %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_constant_to_flat_addrspacecast
; HSA-SAME: (ptr addrspace(4) [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[STOF:%.*]] = addrspacecast ptr addrspace(4) [[PTR]] to ptr
; HSA-NEXT: [[LD:%.*]] = load volatile i32, ptr [[STOF]], align 4
; HSA-NEXT: ret void
;
%stof = addrspacecast ptr addrspace(4) %ptr to ptr
%ld = load volatile i32, ptr %stof
ret void
}
define amdgpu_kernel void @use_flat_to_global_addrspacecast(ptr %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_flat_to_global_addrspacecast
; HSA-SAME: (ptr [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[FTOS:%.*]] = addrspacecast ptr [[PTR]] to ptr addrspace(1)
; HSA-NEXT: store volatile i32 0, ptr addrspace(1) [[FTOS]], align 4
; HSA-NEXT: ret void
;
%ftos = addrspacecast ptr %ptr to ptr addrspace(1)
store volatile i32 0, ptr addrspace(1) %ftos
ret void
}
define amdgpu_kernel void @use_flat_to_constant_addrspacecast(ptr %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_flat_to_constant_addrspacecast
; HSA-SAME: (ptr [[PTR:%.*]]) #[[ATTR1]] {
; HSA-NEXT: [[FTOS:%.*]] = addrspacecast ptr [[PTR]] to ptr addrspace(4)
; HSA-NEXT: [[LD:%.*]] = load volatile i32, ptr addrspace(4) [[FTOS]], align 4
; HSA-NEXT: ret void
;
%ftos = addrspacecast ptr %ptr to ptr addrspace(4)
%ld = load volatile i32, ptr addrspace(4) %ftos
ret void
}
define amdgpu_kernel void @use_is_shared(ptr %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_is_shared
; HSA-SAME: (ptr [[PTR:%.*]]) #[[ATTR12]] {
; HSA-NEXT: [[IS_SHARED:%.*]] = call i1 @llvm.amdgcn.is.shared(ptr [[PTR]])
; HSA-NEXT: [[EXT:%.*]] = zext i1 [[IS_SHARED]] to i32
; HSA-NEXT: store i32 [[EXT]], ptr addrspace(1) poison, align 4
; HSA-NEXT: ret void
;
%is.shared = call i1 @llvm.amdgcn.is.shared(ptr %ptr)
%ext = zext i1 %is.shared to i32
store i32 %ext, ptr addrspace(1) poison
ret void
}
define amdgpu_kernel void @use_is_private(ptr %ptr) #1 {
; HSA-LABEL: define {{[^@]+}}@use_is_private
; HSA-SAME: (ptr [[PTR:%.*]]) #[[ATTR12]] {
; HSA-NEXT: [[IS_PRIVATE:%.*]] = call i1 @llvm.amdgcn.is.private(ptr [[PTR]])
; HSA-NEXT: [[EXT:%.*]] = zext i1 [[IS_PRIVATE]] to i32
; HSA-NEXT: store i32 [[EXT]], ptr addrspace(1) poison, align 4
; HSA-NEXT: ret void
;
%is.private = call i1 @llvm.amdgcn.is.private(ptr %ptr)
%ext = zext i1 %is.private to i32
store i32 %ext, ptr addrspace(1) poison
ret void
}
define amdgpu_kernel void @use_alloca() #1 {
; HSA-LABEL: define {{[^@]+}}@use_alloca
; HSA-SAME: () #[[ATTR1]] {
; HSA-NEXT: [[ALLOCA:%.*]] = alloca i32, align 4, addrspace(5)
; HSA-NEXT: store i32 0, ptr addrspace(5) [[ALLOCA]], align 4
; HSA-NEXT: ret void
;
%alloca = alloca i32, addrspace(5)
store i32 0, ptr addrspace(5) %alloca
ret void
}
define amdgpu_kernel void @use_alloca_non_entry_block() #1 {
; HSA-LABEL: define {{[^@]+}}@use_alloca_non_entry_block
; HSA-SAME: () #[[ATTR1]] {
; HSA-NEXT: entry:
; HSA-NEXT: br label [[BB:%.*]]
; HSA: bb:
; HSA-NEXT: [[ALLOCA:%.*]] = alloca i32, align 4, addrspace(5)
; HSA-NEXT: store i32 0, ptr addrspace(5) [[ALLOCA]], align 4
; HSA-NEXT: ret void
;
entry:
br label %bb
bb:
%alloca = alloca i32, addrspace(5)
store i32 0, ptr addrspace(5) %alloca
ret void
}
define void @use_alloca_func() #1 {
; HSA-LABEL: define {{[^@]+}}@use_alloca_func
; HSA-SAME: () #[[ATTR1]] {
; HSA-NEXT: [[ALLOCA:%.*]] = alloca i32, align 4, addrspace(5)
; HSA-NEXT: store i32 0, ptr addrspace(5) [[ALLOCA]], align 4
; HSA-NEXT: ret void
;
%alloca = alloca i32, addrspace(5)
store i32 0, ptr addrspace(5) %alloca
ret void
}
attributes #0 = { nounwind readnone speculatable }
attributes #1 = { nounwind }
; AKF_HSA: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 500}
;.
; HSA: attributes #[[ATTR0:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
; HSA: attributes #[[ATTR1]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR2]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR3]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR4]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR5]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR6]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR7]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR8]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR9]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workitem-id-x" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR10]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR11]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR12]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; HSA: attributes #[[ATTR13]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
;.
Reference in New Issue View Git Blame Copy Permalink