| Changpeng Fang | 29fcf88 | 2018-02-01 18:41:33 +0000 | [diff] [blame] | 1 | ; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs -show-mc-encoding | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=UNPACKED %s |
| Changpeng Fang | 44dfa1d | 2018-01-12 21:12:19 +0000 | [diff] [blame] | 2 | ; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX81 %s |
| 3 | ; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX9 %s |
| 4 | |
| 5 | ; GCN-LABEL: {{^}}buffer_store_format_d16_x: |
| 6 | ; GCN: v_trunc_f16_e32 v[[LO:[0-9]+]], s{{[0-9]+}} |
| 7 | ; GCN: buffer_store_format_d16_x v[[LO]], v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 idxen |
| 8 | define amdgpu_kernel void @buffer_store_format_d16_x(<4 x i32> %rsrc, half %data, i32 %index) { |
| 9 | main_body: |
| 10 | call void @llvm.amdgcn.buffer.store.format.f16(half %data, <4 x i32> %rsrc, i32 %index, i32 0, i1 0, i1 0) |
| 11 | ret void |
| 12 | } |
| 13 | |
| 14 | ; GCN-LABEL: {{^}}buffer_store_format_d16_xy: |
| 15 | |
| Stanislav Mekhanoshin | c078ca9 | 2018-02-14 18:05:14 +0000 | [diff] [blame] | 16 | ; UNPACKED: flat_load_ushort v[[HI:[0-9]+]], v[{{[0-9]+:[0-9]+}}] ; |
| 17 | ; UNPACKED: flat_load_ushort v[[LO:[0-9]+]], v[{{[0-9]+:[0-9]+}}] ; |
| Changpeng Fang | 44dfa1d | 2018-01-12 21:12:19 +0000 | [diff] [blame] | 18 | ; UNPACKED: buffer_store_format_d16_xy v{{\[}}[[LO]]:[[HI]]{{\]}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 idxen |
| 19 | |
| 20 | ; PACKED: buffer_store_format_d16_xy v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 idxen |
| 21 | define amdgpu_kernel void @buffer_store_format_d16_xy(<4 x i32> %rsrc, <2 x half> %data, i32 %index) { |
| 22 | main_body: |
| 23 | call void @llvm.amdgcn.buffer.store.format.v2f16(<2 x half> %data, <4 x i32> %rsrc, i32 %index, i32 0, i1 0, i1 0) |
| 24 | ret void |
| 25 | } |
| 26 | |
| 27 | ; GCN-LABEL: {{^}}buffer_store_format_d16_xyzw: |
| 28 | |
| Stanislav Mekhanoshin | c078ca9 | 2018-02-14 18:05:14 +0000 | [diff] [blame] | 29 | ; UNPACKED: flat_load_ushort v[[HI:[0-9]+]], v[{{[0-9]+:[0-9]+}}] ; |
| 30 | ; UNPACKED: flat_load_ushort v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}] ; |
| 31 | ; UNPACKED: flat_load_ushort v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}] ; |
| 32 | ; UNPACKED: flat_load_ushort v[[LO:[0-9]+]], v[{{[0-9]+:[0-9]+}}] ; |
| Changpeng Fang | 44dfa1d | 2018-01-12 21:12:19 +0000 | [diff] [blame] | 33 | ; UNPACKED: buffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 idxen |
| 34 | |
| 35 | ; GFX81: v_or_b32_e32 v[[HI:[0-9]+]] |
| 36 | ; GFX81: v_or_b32_e32 v[[LO:[0-9]+]] |
| 37 | |
| 38 | ; GFX9: v_mov_b32_e32 v[[LO:[0-9]+]] |
| 39 | ; GFX9: v_mov_b32_e32 v[[HI:[0-9]+]] |
| 40 | |
| 41 | ; PACKED: buffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 idxen |
| 42 | define amdgpu_kernel void @buffer_store_format_d16_xyzw(<4 x i32> %rsrc, <4 x half> %data, i32 %index) { |
| 43 | main_body: |
| 44 | call void @llvm.amdgcn.buffer.store.format.v4f16(<4 x half> %data, <4 x i32> %rsrc, i32 %index, i32 0, i1 0, i1 0) |
| 45 | ret void |
| 46 | } |
| 47 | |
| 48 | declare void @llvm.amdgcn.buffer.store.format.f16(half, <4 x i32>, i32, i32, i1, i1) |
| 49 | declare void @llvm.amdgcn.buffer.store.format.v2f16(<2 x half>, <4 x i32>, i32, i32, i1, i1) |
| 50 | declare void @llvm.amdgcn.buffer.store.format.v4f16(<4 x half>, <4 x i32>, i32, i32, i1, i1) |