CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
352 lines
10 KiB
LLVM
352 lines
10 KiB
LLVM
; RUN: llc -mtriple=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
|
|
|
|
declare void @llvm.memcpy.p3.p3.i32(ptr addrspace(3) nocapture, ptr addrspace(3) nocapture, i32, i1) nounwind
|
|
declare void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) nocapture, ptr addrspace(1) nocapture, i64, i1) nounwind
|
|
declare void @llvm.memcpy.p1.p2.i64(ptr addrspace(1) nocapture, ptr addrspace(4) nocapture, i64, i1) nounwind
|
|
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_lds_to_lds_align1:
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
; SI-DAG: ds_read_u8
|
|
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
; SI-DAG: ds_write_b8
|
|
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_lds_to_lds_align1(ptr addrspace(3) noalias %out, ptr addrspace(3) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p3.p3.i32(ptr addrspace(3) %out, ptr addrspace(3) %in, i32 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_lds_to_lds_align2:
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
; SI-DAG: ds_read_u16
|
|
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
; SI-DAG: ds_write_b16
|
|
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_lds_to_lds_align2(ptr addrspace(3) noalias %out, ptr addrspace(3) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p3.p3.i32(ptr addrspace(3) align 2 %out, ptr addrspace(3) align 2 %in, i32 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_lds_to_lds_align4:
|
|
; SI: ds_read2_b32
|
|
; SI: ds_read2_b32
|
|
; SI: ds_read2_b32
|
|
; SI: ds_read2_b32
|
|
|
|
; SI: ds_write2_b32
|
|
; SI: ds_write2_b32
|
|
; SI: ds_write2_b32
|
|
; SI: ds_write2_b32
|
|
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_lds_to_lds_align4(ptr addrspace(3) noalias %out, ptr addrspace(3) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p3.p3.i32(ptr addrspace(3) align 4 %out, ptr addrspace(3) align 4 %in, i32 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FIXME: Use 64-bit ops
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_lds_to_lds_align8:
|
|
|
|
; SI: ds_read2_b64
|
|
; SI: ds_read2_b64
|
|
|
|
; SI: ds_write2_b64
|
|
; SI: ds_write2_b64
|
|
|
|
; SI-DAG: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_lds_to_lds_align8(ptr addrspace(3) noalias %out, ptr addrspace(3) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p3.p3.i32(ptr addrspace(3) align 8 %out, ptr addrspace(3) align 8 %in, i32 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_global_to_global_align1:
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
; SI-DAG: buffer_load_ubyte
|
|
; SI-DAG: buffer_store_byte
|
|
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_global_to_global_align1(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) %out, ptr addrspace(1) %in, i64 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_global_to_global_align2:
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
; SI-DAG: buffer_load_ushort
|
|
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
; SI-DAG: buffer_store_short
|
|
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_global_to_global_align2(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) align 2 %out, ptr addrspace(1) align 2 %in, i64 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_global_to_global_align4:
|
|
; SI: buffer_load_dwordx4
|
|
; SI: buffer_load_dwordx4
|
|
; SI: buffer_store_dwordx4
|
|
; SI: buffer_store_dwordx4
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_global_to_global_align4(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) align 4 %out, ptr addrspace(1) align 4 %in, i64 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_global_to_global_align8:
|
|
; SI: buffer_load_dwordx4
|
|
; SI: buffer_load_dwordx4
|
|
; SI: buffer_store_dwordx4
|
|
; SI: buffer_store_dwordx4
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_global_to_global_align8(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) align 8 %out, ptr addrspace(1) align 8 %in, i64 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_small_memcpy_i64_global_to_global_align16:
|
|
; SI: buffer_load_dwordx4
|
|
; SI: buffer_load_dwordx4
|
|
; SI: buffer_store_dwordx4
|
|
; SI: buffer_store_dwordx4
|
|
; SI: s_endpgm
|
|
define amdgpu_kernel void @test_small_memcpy_i64_global_to_global_align16(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in) nounwind {
|
|
call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) align 16 %out, ptr addrspace(1) align 16 %in, i64 32, i1 false) nounwind
|
|
ret void
|
|
}
|
|
|
|
; Test shouldConvertConstantLoadToIntImm
|
|
@hello.align4 = private unnamed_addr addrspace(4) constant [16 x i8] c"constant string\00", align 4
|
|
@hello.align1 = private unnamed_addr addrspace(4) constant [16 x i8] c"constant string\00", align 1
|
|
|
|
; FUNC-LABEL: {{^}}test_memcpy_const_string_align4:
|
|
; SI: s_getpc_b64
|
|
; SI: s_add_u32 s{{[0-9]+}}, s{{[0-9]+}}, .Lhello.align4@rel32@lo+4
|
|
; SI: s_addc_u32
|
|
; SI-DAG: s_load_dwordx8
|
|
; SI-DAG: s_load_dwordx2
|
|
; SI-DAG: buffer_store_dwordx4
|
|
; SI-DAG: buffer_store_dwordx4
|
|
define amdgpu_kernel void @test_memcpy_const_string_align4(ptr addrspace(1) noalias %out) nounwind {
|
|
call void @llvm.memcpy.p1.p2.i64(ptr addrspace(1) align 4 %out, ptr addrspace(4) align 4 @hello.align4, i64 32, i1 false)
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}test_memcpy_const_string_align1:
|
|
; SI-NOT: buffer_load
|
|
; SI: v_mov_b32_e32 v{{[0-9]+}}, 0x
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
; SI: buffer_store_byte
|
|
define amdgpu_kernel void @test_memcpy_const_string_align1(ptr addrspace(1) noalias %out) nounwind {
|
|
call void @llvm.memcpy.p1.p2.i64(ptr addrspace(1) %out, ptr addrspace(4) @hello.align1, i64 32, i1 false)
|
|
ret void
|
|
}
|