; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=amdgcn -mcpu=hawaii < %s | FileCheck --check-prefix=GFX7 %s
; RUN: llc -mtriple=amdgcn -mcpu=tonga < %s | FileCheck --check-prefix=GFX8 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck --check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck --check-prefix=GFX10 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck --check-prefixes=GFX11 %s
define bfloat @raw_ptr_buffer_load_bf16(ptr addrspace(8) inreg %rsrc) {
; GFX7-LABEL: raw_ptr_buffer_load_bf16:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s11, s17
; GFX7-NEXT: s_mov_b32 s10, s16
; GFX7-NEXT: s_mov_b32 s9, s7
; GFX7-NEXT: s_mov_b32 s8, s6
; GFX7-NEXT: buffer_load_ushort v0, off, s[8:11], 0
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: raw_ptr_buffer_load_bf16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_mov_b32 s11, s17
; GFX8-NEXT: s_mov_b32 s10, s16
; GFX8-NEXT: s_mov_b32 s9, s7
; GFX8-NEXT: s_mov_b32 s8, s6
; GFX8-NEXT: buffer_load_ushort v0, off, s[8:11], 0
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: raw_ptr_buffer_load_bf16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s11, s17
; GFX9-NEXT: s_mov_b32 s10, s16
; GFX9-NEXT: s_mov_b32 s9, s7
; GFX9-NEXT: s_mov_b32 s8, s6
; GFX9-NEXT: buffer_load_ushort v0, off, s[8:11], 0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: raw_ptr_buffer_load_bf16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_mov_b32 s11, s17
; GFX10-NEXT: s_mov_b32 s10, s16
; GFX10-NEXT: s_mov_b32 s9, s7
; GFX10-NEXT: s_mov_b32 s8, s6
; GFX10-NEXT: buffer_load_ushort v0, off, s[8:11], 0
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: raw_ptr_buffer_load_bf16:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: buffer_load_u16 v0, off, s[0:3], 0
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%val = call bfloat @llvm.amdgcn.raw.ptr.buffer.load.v2bf16(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0)
ret bfloat %val
}
define <2 x bfloat> @raw_ptr_buffer_load_v2bf16(ptr addrspace(8) inreg %rsrc) {
; GFX7-LABEL: raw_ptr_buffer_load_v2bf16:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s11, s17
; GFX7-NEXT: s_mov_b32 s10, s16
; GFX7-NEXT: s_mov_b32 s9, s7
; GFX7-NEXT: s_mov_b32 s8, s6
; GFX7-NEXT: buffer_load_dword v1, off, s[8:11], 0
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v1
; GFX7-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: raw_ptr_buffer_load_v2bf16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_mov_b32 s11, s17
; GFX8-NEXT: s_mov_b32 s10, s16
; GFX8-NEXT: s_mov_b32 s9, s7
; GFX8-NEXT: s_mov_b32 s8, s6
; GFX8-NEXT: buffer_load_dword v0, off, s[8:11], 0
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: raw_ptr_buffer_load_v2bf16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s11, s17
; GFX9-NEXT: s_mov_b32 s10, s16
; GFX9-NEXT: s_mov_b32 s9, s7
; GFX9-NEXT: s_mov_b32 s8, s6
; GFX9-NEXT: buffer_load_dword v0, off, s[8:11], 0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: raw_ptr_buffer_load_v2bf16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_mov_b32 s11, s17
; GFX10-NEXT: s_mov_b32 s10, s16
; GFX10-NEXT: s_mov_b32 s9, s7
; GFX10-NEXT: s_mov_b32 s8, s6
; GFX10-NEXT: buffer_load_dword v0, off, s[8:11], 0
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: raw_ptr_buffer_load_v2bf16:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: buffer_load_b32 v0, off, s[0:3], 0
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x bfloat> @llvm.amdgcn.raw.ptr.buffer.load.v2bf16(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0)
ret <2 x bfloat> %val
}
define <4 x bfloat> @raw_ptr_buffer_load_v4bf16(ptr addrspace(8) inreg %rsrc) {
; GFX7-LABEL: raw_ptr_buffer_load_v4bf16:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s11, s17
; GFX7-NEXT: s_mov_b32 s10, s16
; GFX7-NEXT: s_mov_b32 s9, s7
; GFX7-NEXT: s_mov_b32 s8, s6
; GFX7-NEXT: buffer_load_dwordx2 v[2:3], off, s[8:11], 0
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v2
; GFX7-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
; GFX7-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: raw_ptr_buffer_load_v4bf16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_mov_b32 s11, s17
; GFX8-NEXT: s_mov_b32 s10, s16
; GFX8-NEXT: s_mov_b32 s9, s7
; GFX8-NEXT: s_mov_b32 s8, s6
; GFX8-NEXT: buffer_load_dwordx2 v[0:1], off, s[8:11], 0
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: raw_ptr_buffer_load_v4bf16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s11, s17
; GFX9-NEXT: s_mov_b32 s10, s16
; GFX9-NEXT: s_mov_b32 s9, s7
; GFX9-NEXT: s_mov_b32 s8, s6
; GFX9-NEXT: buffer_load_dwordx2 v[0:1], off, s[8:11], 0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: raw_ptr_buffer_load_v4bf16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_mov_b32 s11, s17
; GFX10-NEXT: s_mov_b32 s10, s16
; GFX10-NEXT: s_mov_b32 s9, s7
; GFX10-NEXT: s_mov_b32 s8, s6
; GFX10-NEXT: buffer_load_dwordx2 v[0:1], off, s[8:11], 0
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: raw_ptr_buffer_load_v4bf16:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: buffer_load_b64 v[0:1], off, s[0:3], 0
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%val = call <4 x bfloat> @llvm.amdgcn.raw.ptr.buffer.load.v4bf16(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0)
ret <4 x bfloat> %val
}
; FIXME
; define <6 x bfloat> @raw_ptr_buffer_load_v6bf16(ptr addrspace(8) inreg %rsrc) {
; %val = call <6 x bfloat> @llvm.amdgcn.raw.ptr.buffer.load.v6bf16(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0)
; ret <6 x bfloat> %val
; }
define <8 x bfloat> @raw_ptr_buffer_load_v8bf16(ptr addrspace(8) inreg %rsrc) {
; GFX7-LABEL: raw_ptr_buffer_load_v8bf16:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_mov_b32 s11, s17
; GFX7-NEXT: s_mov_b32 s10, s16
; GFX7-NEXT: s_mov_b32 s9, s7
; GFX7-NEXT: s_mov_b32 s8, s6
; GFX7-NEXT: buffer_load_dwordx4 v[4:7], off, s[8:11], 0
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v4
; GFX7-NEXT: v_and_b32_e32 v1, 0xffff0000, v4
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v5
; GFX7-NEXT: v_and_b32_e32 v3, 0xffff0000, v5
; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v6
; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v6
; GFX7-NEXT: v_lshlrev_b32_e32 v6, 16, v7
; GFX7-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: raw_ptr_buffer_load_v8bf16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_mov_b32 s11, s17
; GFX8-NEXT: s_mov_b32 s10, s16
; GFX8-NEXT: s_mov_b32 s9, s7
; GFX8-NEXT: s_mov_b32 s8, s6
; GFX8-NEXT: buffer_load_dwordx4 v[0:3], off, s[8:11], 0
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: raw_ptr_buffer_load_v8bf16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s11, s17
; GFX9-NEXT: s_mov_b32 s10, s16
; GFX9-NEXT: s_mov_b32 s9, s7
; GFX9-NEXT: s_mov_b32 s8, s6
; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[8:11], 0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: raw_ptr_buffer_load_v8bf16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_mov_b32 s11, s17
; GFX10-NEXT: s_mov_b32 s10, s16
; GFX10-NEXT: s_mov_b32 s9, s7
; GFX10-NEXT: s_mov_b32 s8, s6
; GFX10-NEXT: buffer_load_dwordx4 v[0:3], off, s[8:11], 0
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: raw_ptr_buffer_load_v8bf16:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: buffer_load_b128 v[0:3], off, s[0:3], 0
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%val = call <8 x bfloat> @llvm.amdgcn.raw.ptr.buffer.load.v8bf16(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0)
ret <8 x bfloat> %val
}
Codebase Browser
llvm