f7d03707d1
Attempting to pass a `ptr addrspace(7)` to functions that take `ptr` arguments produces undesirable `addrspacecast(addrspacecast(p8 x to p7) to p0) => addrspacecast(p8 x to p0)` folds. This results in illegal GEP operations on buffer resources, which can't be GEP'd. (However, note that, while unimplemneted, addressspacecast from ptr addrspace(7) to ptr is legal - it's just an effective address computation) To resolve this problem, and thus prevent illegal `getelementptr T, ptr addrspace(8) %x, ...` s from being produces, this commit extends amdgcn.make.buffer.rsrc to also be variadic in its result type, auto-upgrading old manglings. The logic for handling a make.buffer.rsrc in instruction selection remains untouched and expects the output type to be a ptr addrspace(8), as does the Clang lowering for its builtin (the pointer-to-pointer version might want a different name in clang). LowerBufferFatPointers has been updated to lower amdgcn.make.buffer.rsrc.p7.p* to amdgcn.make.buffer.rsrc.p8.p* . This'll also make exposing buffer fat pointers in Clang easier, since you don't have to cast between a `__amdgcn_rsrc_t` and a pointer.
48 lines
3.8 KiB
LLVM
48 lines
3.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes
|
|
; RUN: opt -passes=function-attrs -S < %s | FileCheck --check-prefixes=COMMON,FNATTRS %s
|
|
; RUN: opt -passes=attributor-light -S < %s | FileCheck --check-prefixes=COMMON,ATTRIBUTOR %s
|
|
|
|
;; target triple = "amdgcn-amd-amdhsa"
|
|
target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-p7:160:256:256:32-p8:128:128-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-ni:7:8"
|
|
|
|
define amdgpu_kernel void @test_make_buffer_rsrc(ptr %p, ptr %q) {
|
|
; FNATTRS: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite)
|
|
; FNATTRS-LABEL: define {{[^@]+}}@test_make_buffer_rsrc
|
|
; FNATTRS-SAME: (ptr readonly captures(none) [[P:%.*]], ptr writeonly captures(none) [[Q:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
; FNATTRS-NEXT: [[P_RSRC:%.*]] = call ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr [[P]], i16 0, i32 4, i32 822243328)
|
|
; FNATTRS-NEXT: [[Q_RSRC:%.*]] = call ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr [[Q]], i16 0, i32 4, i32 822243328)
|
|
; FNATTRS-NEXT: [[V:%.*]] = call i8 @llvm.amdgcn.raw.ptr.buffer.load.i8(ptr addrspace(8) [[P_RSRC]], i32 0, i32 0, i32 0)
|
|
; FNATTRS-NEXT: call void @llvm.amdgcn.raw.ptr.buffer.store.i8(i8 [[V]], ptr addrspace(8) [[Q_RSRC]], i32 0, i32 0, i32 0)
|
|
; FNATTRS-NEXT: ret void
|
|
;
|
|
; ATTRIBUTOR: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite)
|
|
; ATTRIBUTOR-LABEL: define {{[^@]+}}@test_make_buffer_rsrc
|
|
; ATTRIBUTOR-SAME: (ptr nofree readonly captures(none) [[P:%.*]], ptr nofree writeonly captures(none) [[Q:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
; ATTRIBUTOR-NEXT: [[P_RSRC:%.*]] = call ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr [[P]], i16 0, i32 4, i32 822243328) #[[ATTR4:[0-9]+]]
|
|
; ATTRIBUTOR-NEXT: [[Q_RSRC:%.*]] = call ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr [[Q]], i16 0, i32 4, i32 822243328) #[[ATTR4]]
|
|
; ATTRIBUTOR-NEXT: [[V:%.*]] = call i8 @llvm.amdgcn.raw.ptr.buffer.load.i8(ptr addrspace(8) readonly captures(none) [[P_RSRC]], i32 0, i32 0, i32 0) #[[ATTR5:[0-9]+]]
|
|
; ATTRIBUTOR-NEXT: call void @llvm.amdgcn.raw.ptr.buffer.store.i8(i8 [[V]], ptr addrspace(8) writeonly captures(none) [[Q_RSRC]], i32 0, i32 0, i32 0) #[[ATTR6:[0-9]+]]
|
|
; ATTRIBUTOR-NEXT: ret void
|
|
;
|
|
%p.rsrc = call ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr %p, i16 0, i32 4, i32 822243328)
|
|
%q.rsrc = call ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr %q, i16 0, i32 4, i32 822243328)
|
|
%v = call i8 @llvm.amdgcn.raw.ptr.buffer.load.i8(ptr addrspace(8) %p.rsrc, i32 0, i32 0, i32 0)
|
|
call void @llvm.amdgcn.raw.ptr.buffer.store.i8(i8 %v, ptr addrspace(8) %q.rsrc, i32 0, i32 0, i32 0)
|
|
ret void
|
|
}
|
|
|
|
; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)
|
|
declare ptr addrspace(8) @llvm.amdgcn.make.buffer.rsrc.p8.p0(ptr readnone, i16, i32, i32) #0
|
|
|
|
; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: read)
|
|
declare i8 @llvm.amdgcn.raw.ptr.buffer.load.i8(ptr addrspace(8) nocapture readonly, i32, i32, i32 immarg) #1
|
|
|
|
; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: write)
|
|
declare void @llvm.amdgcn.raw.ptr.buffer.store.i8(i8, ptr addrspace(8) nocapture writeonly, i32, i32, i32 immarg) #2
|
|
|
|
attributes #0 = { mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none) }
|
|
attributes #1 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: read) }
|
|
attributes #2 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: write) }
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; COMMON: {{.*}}
|