mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-15 20:29:48 +00:00
35cc03a1ea
Now that we sometimes know the address space, this can theoretically do a better job. This needs better test coverage, but this mostly depends on first updating the loop optimizatiosn to provide the address space. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239053 91177308-0d34-0410-b5e6-96231b3b80d8
243 lines
7.3 KiB
LLVM
243 lines
7.3 KiB
LLVM
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown < %s | FileCheck -check-prefix=OPT %s
|
|
; RUN: llc -march=amdgcn -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN %s
|
|
|
|
declare i32 @llvm.r600.read.tidig.x() #0
|
|
|
|
; OPT-LABEL: @test_sink_global_small_offset_i32(
|
|
; OPT-NOT: getelementptr i32, i32 addrspace(1)* %in
|
|
; OPT: br i1
|
|
; OPT: ptrtoint
|
|
|
|
; GCN-LABEL: {{^}}test_sink_global_small_offset_i32:
|
|
; GCN: {{^}}BB0_2:
|
|
define void @test_sink_global_small_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond) {
|
|
entry:
|
|
%out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
|
|
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i64 7
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
%tmp1 = load i32, i32 addrspace(1)* %in.gep
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; OPT-LABEL: @test_sink_global_small_max_i32_ds_offset(
|
|
; OPT: %in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 65535
|
|
; OPT: br i1
|
|
|
|
; GCN-LABEL: {{^}}test_sink_global_small_max_i32_ds_offset:
|
|
; GCN: s_and_saveexec_b64
|
|
; GCN: buffer_load_sbyte {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, s{{[0-9]+$}}
|
|
; GCN: {{^}}BB1_2:
|
|
; GCN: s_or_b64 exec
|
|
define void @test_sink_global_small_max_i32_ds_offset(i32 addrspace(1)* %out, i8 addrspace(1)* %in, i32 %cond) {
|
|
entry:
|
|
%out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 99999
|
|
%in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 65535
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
%tmp1 = load i8, i8 addrspace(1)* %in.gep
|
|
%tmp2 = sext i8 %tmp1 to i32
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp2, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}test_sink_global_small_max_mubuf_offset:
|
|
; GCN: s_and_saveexec_b64
|
|
; GCN: buffer_load_sbyte {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offset:4095{{$}}
|
|
; GCN: {{^}}BB2_2:
|
|
; GCN: s_or_b64 exec
|
|
define void @test_sink_global_small_max_mubuf_offset(i32 addrspace(1)* %out, i8 addrspace(1)* %in, i32 %cond) {
|
|
entry:
|
|
%out.gep = getelementptr i32, i32 addrspace(1)* %out, i32 1024
|
|
%in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 4095
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
%tmp1 = load i8, i8 addrspace(1)* %in.gep
|
|
%tmp2 = sext i8 %tmp1 to i32
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp2, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}test_sink_global_small_max_plus_1_mubuf_offset:
|
|
; GCN: s_and_saveexec_b64
|
|
; GCN: buffer_load_sbyte {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, s{{[0-9]+$}}
|
|
; GCN: {{^}}BB3_2:
|
|
; GCN: s_or_b64 exec
|
|
define void @test_sink_global_small_max_plus_1_mubuf_offset(i32 addrspace(1)* %out, i8 addrspace(1)* %in, i32 %cond) {
|
|
entry:
|
|
%out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 99999
|
|
%in.gep = getelementptr i8, i8 addrspace(1)* %in, i64 4096
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
%tmp1 = load i8, i8 addrspace(1)* %in.gep
|
|
%tmp2 = sext i8 %tmp1 to i32
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp2, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; OPT-LABEL: @test_no_sink_flat_small_offset_i32(
|
|
; OPT: getelementptr i32, i32 addrspace(4)* %in
|
|
; OPT: br i1
|
|
; OPT-NOT: ptrtoint
|
|
|
|
; GCN-LABEL: {{^}}test_no_sink_flat_small_offset_i32:
|
|
; GCN: flat_load_dword
|
|
; GCN: {{^}}BB4_2:
|
|
|
|
define void @test_no_sink_flat_small_offset_i32(i32 addrspace(4)* %out, i32 addrspace(4)* %in, i32 %cond) {
|
|
entry:
|
|
%out.gep = getelementptr i32, i32 addrspace(4)* %out, i64 999999
|
|
%in.gep = getelementptr i32, i32 addrspace(4)* %in, i64 7
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
%tmp1 = load i32, i32 addrspace(4)* %in.gep
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(4)* %out.gep
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; OPT-LABEL: @test_sink_scratch_small_offset_i32(
|
|
; OPT-NOT: getelementptr [512 x i32]
|
|
; OPT: br i1
|
|
; OPT: ptrtoint
|
|
|
|
; GCN-LABEL: {{^}}test_sink_scratch_small_offset_i32:
|
|
; GCN: s_and_saveexec_b64
|
|
; GCN: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:4092{{$}}
|
|
; GCN: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:4092{{$}}
|
|
; GCN: {{^}}BB5_2:
|
|
define void @test_sink_scratch_small_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond, i32 %arg) {
|
|
entry:
|
|
%alloca = alloca [512 x i32], align 4
|
|
%out.gep.0 = getelementptr i32, i32 addrspace(1)* %out, i64 999998
|
|
%out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i64 999999
|
|
%add.arg = add i32 %arg, 8
|
|
%alloca.gep = getelementptr [512 x i32], [512 x i32]* %alloca, i32 0, i32 1023
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
store volatile i32 123, i32* %alloca.gep
|
|
%tmp1 = load volatile i32, i32* %alloca.gep
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep.0
|
|
%load = load volatile i32, i32* %alloca.gep
|
|
store i32 %load, i32 addrspace(1)* %out.gep.1
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; OPT-LABEL: @test_no_sink_scratch_large_offset_i32(
|
|
; OPT: %alloca.gep = getelementptr [512 x i32], [512 x i32]* %alloca, i32 0, i32 1024
|
|
; OPT: br i1
|
|
; OPT-NOT: ptrtoint
|
|
|
|
; GCN-LABEL: {{^}}test_no_sink_scratch_large_offset_i32:
|
|
; GCN: s_and_saveexec_b64
|
|
; GCN: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
|
|
; GCN: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
|
|
; GCN: {{^}}BB6_2:
|
|
define void @test_no_sink_scratch_large_offset_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %cond, i32 %arg) {
|
|
entry:
|
|
%alloca = alloca [512 x i32], align 4
|
|
%out.gep.0 = getelementptr i32, i32 addrspace(1)* %out, i64 999998
|
|
%out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i64 999999
|
|
%add.arg = add i32 %arg, 8
|
|
%alloca.gep = getelementptr [512 x i32], [512 x i32]* %alloca, i32 0, i32 1024
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
store volatile i32 123, i32* %alloca.gep
|
|
%tmp1 = load volatile i32, i32* %alloca.gep
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep.0
|
|
%load = load volatile i32, i32* %alloca.gep
|
|
store i32 %load, i32 addrspace(1)* %out.gep.1
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}test_sink_global_vreg_sreg_i32:
|
|
; GCN: s_and_saveexec_b64
|
|
; GCN: buffer_load_dword {{v[0-9]+}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
|
|
; GCN: {{^}}BB7_2:
|
|
define void @test_sink_global_vreg_sreg_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in, i32 %offset, i32 %cond) {
|
|
entry:
|
|
%offset.ext = zext i32 %offset to i64
|
|
%out.gep = getelementptr i32, i32 addrspace(1)* %out, i64 999999
|
|
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i64 %offset.ext
|
|
%tmp0 = icmp eq i32 %cond, 0
|
|
br i1 %tmp0, label %endif, label %if
|
|
|
|
if:
|
|
%tmp1 = load i32, i32 addrspace(1)* %in.gep
|
|
br label %endif
|
|
|
|
endif:
|
|
%x = phi i32 [ %tmp1, %if ], [ 0, %entry ]
|
|
store i32 %x, i32 addrspace(1)* %out.gep
|
|
br label %done
|
|
|
|
done:
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind readnone }
|
|
attributes #1 = { nounwind }
|