diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.ptr.buffer.load.format.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.ptr.buffer.load.format.ll | 280 |
1 files changed, 280 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.ptr.buffer.load.format.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.ptr.buffer.load.format.ll index b0bd4e428ef2..c5202b84fa1e 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.ptr.buffer.load.format.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.ptr.buffer.load.format.ll @@ -2,6 +2,7 @@ ;RUN: llc < %s -mtriple=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck --check-prefixes=GFX6 %s ;RUN: llc < %s -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck --check-prefixes=GFX8PLUS %s ;RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs | FileCheck --check-prefixes=GFX11 %s +;RUN: llc < %s -mtriple=amdgcn -mattr=-enable-prt-strict-null -mcpu=gfx1100 -verify-machineinstrs | FileCheck --check-prefixes=NOPRT %s define amdgpu_ps {<4 x float>, <4 x float>, <4 x float>} @buffer_load(ptr addrspace(8) inreg) { ; GFX6-LABEL: buffer_load: @@ -31,6 +32,16 @@ define amdgpu_ps {<4 x float>, <4 x float>, <4 x float>} @buffer_load(ptr addrsp ; GFX11-NEXT: buffer_load_format_xyzw v[8:11], v8, s[0:3], 0 idxen slc ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v8, 0 +; NOPRT-NEXT: s_clause 0x2 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v8, s[0:3], 0 idxen +; NOPRT-NEXT: buffer_load_format_xyzw v[4:7], v8, s[0:3], 0 idxen glc +; NOPRT-NEXT: buffer_load_format_xyzw v[8:11], v8, s[0:3], 0 idxen slc +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 0, i32 0, i32 0) %data_glc = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 0, i32 0, i32 1) @@ -62,6 +73,13 @@ define amdgpu_ps <4 x float> @buffer_load_immoffs(ptr addrspace(8) inreg) { ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen offset:42 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_immoffs: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v0, 0 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen offset:42 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 42, i32 0, i32 0) ret <4 x float> %data @@ -126,6 +144,25 @@ define amdgpu_ps <4 x float> @buffer_load_immoffs_large(ptr addrspace(8) inreg) ; GFX11-NEXT: v_dual_add_f32 v0, v8, v0 :: v_dual_add_f32 v3, v11, v3 ; GFX11-NEXT: v_add_f32_e32 v2, v10, v2 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_immoffs_large: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v8, 0 +; NOPRT-NEXT: s_movk_i32 s4, 0x7ffc +; NOPRT-NEXT: s_clause 0x1 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v8, s[0:3], 60 idxen offset:4092 +; NOPRT-NEXT: buffer_load_format_xyzw v[4:7], v8, s[0:3], s4 idxen offset:4092 +; NOPRT-NEXT: s_mov_b32 s4, 0x8ffc +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: v_add_f32_e32 v1, v1, v5 +; NOPRT-NEXT: buffer_load_format_xyzw v[8:11], v8, s[0:3], s4 idxen offset:4 +; NOPRT-NEXT: v_dual_add_f32 v0, v0, v4 :: v_dual_add_f32 v3, v3, v7 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: v_dual_add_f32 v2, v2, v6 :: v_dual_add_f32 v1, v9, v1 +; NOPRT-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) +; NOPRT-NEXT: v_dual_add_f32 v0, v8, v0 :: v_dual_add_f32 v3, v11, v3 +; NOPRT-NEXT: v_add_f32_e32 v2, v10, v2 +; NOPRT-NEXT: ; return to shader part epilog main_body: %d.0 = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 4092, i32 60, i32 0) %d.1 = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 4092, i32 32764, i32 0) @@ -156,6 +193,13 @@ define amdgpu_ps <4 x float> @buffer_load_voffset_large_12bit(ptr addrspace(8) i ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen offset:4092 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_voffset_large_12bit: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v0, 0 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen offset:4092 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 4092, i32 0, i32 0) ret <4 x float> %data @@ -188,6 +232,15 @@ define amdgpu_ps <4 x float> @buffer_load_voffset_large_13bit(ptr addrspace(8) i ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_voffset_large_13bit: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: s_mov_b32 s4, 0 +; NOPRT-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; NOPRT-NEXT: v_dual_mov_b32 v1, 0x1000 :: v_dual_mov_b32 v0, s4 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 8188, i32 0, i32 0) ret <4 x float> %data @@ -220,6 +273,15 @@ define amdgpu_ps <4 x float> @buffer_load_voffset_large_16bit(ptr addrspace(8) i ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_voffset_large_16bit: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: s_mov_b32 s4, 0 +; NOPRT-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; NOPRT-NEXT: v_dual_mov_b32 v1, 0xf000 :: v_dual_mov_b32 v0, s4 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 65532, i32 0, i32 0) ret <4 x float> %data @@ -252,6 +314,15 @@ define amdgpu_ps <4 x float> @buffer_load_voffset_large_23bit(ptr addrspace(8) i ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_voffset_large_23bit: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: s_mov_b32 s4, 0 +; NOPRT-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; NOPRT-NEXT: v_dual_mov_b32 v1, 0x7ff000 :: v_dual_mov_b32 v0, s4 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 8388604, i32 0, i32 0) ret <4 x float> %data @@ -284,6 +355,15 @@ define amdgpu_ps <4 x float> @buffer_load_voffset_large_24bit(ptr addrspace(8) i ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_voffset_large_24bit: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: s_mov_b32 s4, 0 +; NOPRT-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; NOPRT-NEXT: v_dual_mov_b32 v1, 0xfff000 :: v_dual_mov_b32 v0, s4 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:4092 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 16777212, i32 0, i32 0) ret <4 x float> %data @@ -307,6 +387,12 @@ define amdgpu_ps <4 x float> @buffer_load_idx(ptr addrspace(8) inreg, i32) { ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_idx: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v0, s[0:3], 0 idxen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 %1, i32 0, i32 0, i32 0) ret <4 x float> %data @@ -339,6 +425,15 @@ define amdgpu_ps <4 x float> @buffer_load_ofs(ptr addrspace(8) inreg, i32) { ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_ofs: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: s_mov_b32 s4, 0 +; NOPRT-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; NOPRT-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s4 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 %1, i32 0, i32 0) ret <4 x float> %data @@ -371,6 +466,15 @@ define amdgpu_ps <4 x float> @buffer_load_ofs_imm(ptr addrspace(8) inreg, i32) { ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:60 ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_ofs_imm: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: s_mov_b32 s4, 0 +; NOPRT-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; NOPRT-NEXT: v_dual_mov_b32 v1, v0 :: v_dual_mov_b32 v0, s4 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen offset:60 +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %ofs = add i32 %1, 60 %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 0, i32 %ofs, i32 0, i32 0) @@ -395,6 +499,12 @@ define amdgpu_ps <4 x float> @buffer_load_both(ptr addrspace(8) inreg, i32, i32) ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_both: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[0:1], s[0:3], 0 idxen offen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 %1, i32 %2, i32 0, i32 0) ret <4 x float> %data @@ -421,6 +531,13 @@ define amdgpu_ps <4 x float> @buffer_load_both_reversed(ptr addrspace(8) inreg, ; GFX11-NEXT: buffer_load_format_xyzw v[0:3], v[1:2], s[0:3], 0 idxen offen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_both_reversed: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v2, v0 +; NOPRT-NEXT: buffer_load_format_xyzw v[0:3], v[1:2], s[0:3], 0 idxen offen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <4 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v4f32(ptr addrspace(8) %0, i32 %2, i32 %1, i32 0, i32 0) ret <4 x float> %data @@ -447,6 +564,13 @@ define amdgpu_ps float @buffer_load_x(ptr addrspace(8) inreg %rsrc) { ; GFX11-NEXT: buffer_load_format_x v0, v0, s[0:3], 0 idxen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_x: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v0, 0 +; NOPRT-NEXT: buffer_load_format_x v0, v0, s[0:3], 0 idxen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call float @llvm.amdgcn.struct.ptr.buffer.load.format.f32(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) ret float %data @@ -473,6 +597,13 @@ define amdgpu_ps float @buffer_load_x_i32(ptr addrspace(8) inreg %rsrc) { ; GFX11-NEXT: buffer_load_format_x v0, v0, s[0:3], 0 idxen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_x_i32: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v0, 0 +; NOPRT-NEXT: buffer_load_format_x v0, v0, s[0:3], 0 idxen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call i32 @llvm.amdgcn.struct.ptr.buffer.load.format.i32(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %fdata = bitcast i32 %data to float @@ -500,6 +631,13 @@ define amdgpu_ps <2 x float> @buffer_load_xy(ptr addrspace(8) inreg %rsrc) { ; GFX11-NEXT: buffer_load_format_xy v[0:1], v0, s[0:3], 0 idxen ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_xy: +; NOPRT: ; %bb.0: ; %main_body +; NOPRT-NEXT: v_mov_b32_e32 v0, 0 +; NOPRT-NEXT: buffer_load_format_xy v[0:1], v0, s[0:3], 0 idxen +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: ; return to shader part epilog main_body: %data = call <2 x float> @llvm.amdgcn.struct.ptr.buffer.load.format.v2f32(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) ret <2 x float> %data @@ -509,6 +647,10 @@ define amdgpu_cs float @buffer_load_v4i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX6-LABEL: buffer_load_v4i32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 +; GFX6-NEXT: v_mov_b32_e32 v4, v2 +; GFX6-NEXT: v_mov_b32_e32 v5, v2 +; GFX6-NEXT: v_mov_b32_e32 v6, v2 ; GFX6-NEXT: buffer_load_format_xyzw v[2:6], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -523,6 +665,10 @@ define amdgpu_cs float @buffer_load_v4i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX8PLUS-LABEL: buffer_load_v4i32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v4, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v5, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v6, v2 ; GFX8PLUS-NEXT: buffer_load_format_xyzw v[2:6], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dwordx4 v[0:1], v[2:5] @@ -533,11 +679,25 @@ define amdgpu_cs float @buffer_load_v4i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX11-LABEL: buffer_load_v4i32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 +; GFX11-NEXT: v_mov_b32_e32 v4, v2 +; GFX11-NEXT: v_mov_b32_e32 v5, v2 +; GFX11-NEXT: v_mov_b32_e32 v6, v2 ; GFX11-NEXT: buffer_load_format_xyzw v[2:6], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off ; GFX11-NEXT: v_mov_b32_e32 v0, v6 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_v4i32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v6, 0 +; NOPRT-NEXT: buffer_load_format_xyzw v[2:6], v6, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b128 v[0:1], v[2:5], off +; NOPRT-NEXT: v_mov_b32_e32 v0, v6 +; NOPRT-NEXT: ; return to shader part epilog %load = call { <4 x i32>, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_v4i32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { <4 x i32>, i32 } %load, 0 store <4 x i32> %data, ptr addrspace(1) %out @@ -550,6 +710,10 @@ define amdgpu_cs float @buffer_load_v4f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX6-LABEL: buffer_load_v4f32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 +; GFX6-NEXT: v_mov_b32_e32 v4, v2 +; GFX6-NEXT: v_mov_b32_e32 v5, v2 +; GFX6-NEXT: v_mov_b32_e32 v6, v2 ; GFX6-NEXT: buffer_load_format_xyzw v[2:6], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -564,6 +728,10 @@ define amdgpu_cs float @buffer_load_v4f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX8PLUS-LABEL: buffer_load_v4f32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v4, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v5, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v6, v2 ; GFX8PLUS-NEXT: buffer_load_format_xyzw v[2:6], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dwordx4 v[0:1], v[2:5] @@ -574,11 +742,25 @@ define amdgpu_cs float @buffer_load_v4f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX11-LABEL: buffer_load_v4f32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 +; GFX11-NEXT: v_mov_b32_e32 v4, v2 +; GFX11-NEXT: v_mov_b32_e32 v5, v2 +; GFX11-NEXT: v_mov_b32_e32 v6, v2 ; GFX11-NEXT: buffer_load_format_xyzw v[2:6], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off ; GFX11-NEXT: v_mov_b32_e32 v0, v6 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_v4f32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v6, 0 +; NOPRT-NEXT: buffer_load_format_xyzw v[2:6], v6, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b128 v[0:1], v[2:5], off +; NOPRT-NEXT: v_mov_b32_e32 v0, v6 +; NOPRT-NEXT: ; return to shader part epilog %load = call { <4 x float>, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_v4f32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { <4 x float>, i32 } %load, 0 store <4 x float> %data, ptr addrspace(1) %out @@ -591,6 +773,9 @@ define amdgpu_cs float @buffer_load_v3i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX6-LABEL: buffer_load_v3i32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 +; GFX6-NEXT: v_mov_b32_e32 v4, v2 +; GFX6-NEXT: v_mov_b32_e32 v5, v2 ; GFX6-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -606,6 +791,9 @@ define amdgpu_cs float @buffer_load_v3i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX8PLUS-LABEL: buffer_load_v3i32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v4, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v5, v2 ; GFX8PLUS-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dwordx3 v[0:1], v[2:4] @@ -616,11 +804,24 @@ define amdgpu_cs float @buffer_load_v3i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX11-LABEL: buffer_load_v3i32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 +; GFX11-NEXT: v_mov_b32_e32 v4, v2 +; GFX11-NEXT: v_mov_b32_e32 v5, v2 ; GFX11-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b96 v[0:1], v[2:4], off ; GFX11-NEXT: v_mov_b32_e32 v0, v5 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_v3i32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v5, 0 +; NOPRT-NEXT: buffer_load_format_xyz v[2:5], v5, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b96 v[0:1], v[2:4], off +; NOPRT-NEXT: v_mov_b32_e32 v0, v5 +; NOPRT-NEXT: ; return to shader part epilog %load = call { <3 x i32>, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_v3i32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { <3 x i32>, i32 } %load, 0 store <3 x i32> %data, ptr addrspace(1) %out @@ -633,6 +834,9 @@ define amdgpu_cs float @buffer_load_v3f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX6-LABEL: buffer_load_v3f32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 +; GFX6-NEXT: v_mov_b32_e32 v4, v2 +; GFX6-NEXT: v_mov_b32_e32 v5, v2 ; GFX6-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -648,6 +852,9 @@ define amdgpu_cs float @buffer_load_v3f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX8PLUS-LABEL: buffer_load_v3f32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v4, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v5, v2 ; GFX8PLUS-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dwordx3 v[0:1], v[2:4] @@ -658,11 +865,24 @@ define amdgpu_cs float @buffer_load_v3f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX11-LABEL: buffer_load_v3f32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 +; GFX11-NEXT: v_mov_b32_e32 v4, v2 +; GFX11-NEXT: v_mov_b32_e32 v5, v2 ; GFX11-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b96 v[0:1], v[2:4], off ; GFX11-NEXT: v_mov_b32_e32 v0, v5 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_v3f32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v5, 0 +; NOPRT-NEXT: buffer_load_format_xyz v[2:5], v5, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b96 v[0:1], v[2:4], off +; NOPRT-NEXT: v_mov_b32_e32 v0, v5 +; NOPRT-NEXT: ; return to shader part epilog %load = call { <3 x float>, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_v3f32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { <3 x float>, i32 } %load, 0 store <3 x float> %data, ptr addrspace(1) %out @@ -675,6 +895,9 @@ define amdgpu_cs float @buffer_load_v2i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX6-LABEL: buffer_load_v2i32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 +; GFX6-NEXT: v_mov_b32_e32 v4, v2 +; GFX6-NEXT: v_mov_b32_e32 v5, v2 ; GFX6-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -689,6 +912,8 @@ define amdgpu_cs float @buffer_load_v2i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX8PLUS-LABEL: buffer_load_v2i32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v4, v2 ; GFX8PLUS-NEXT: buffer_load_format_xy v[2:4], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dwordx2 v[0:1], v[2:3] @@ -699,11 +924,23 @@ define amdgpu_cs float @buffer_load_v2i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX11-LABEL: buffer_load_v2i32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 +; GFX11-NEXT: v_mov_b32_e32 v4, v2 ; GFX11-NEXT: buffer_load_format_xy v[2:4], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off ; GFX11-NEXT: v_mov_b32_e32 v0, v4 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_v2i32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v4, 0 +; NOPRT-NEXT: buffer_load_format_xy v[2:4], v4, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b64 v[0:1], v[2:3], off +; NOPRT-NEXT: v_mov_b32_e32 v0, v4 +; NOPRT-NEXT: ; return to shader part epilog %load = call { <2 x i32>, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_v2i32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { <2 x i32>, i32 } %load, 0 store <2 x i32> %data, ptr addrspace(1) %out @@ -716,6 +953,9 @@ define amdgpu_cs float @buffer_load_v2f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX6-LABEL: buffer_load_v2f32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 +; GFX6-NEXT: v_mov_b32_e32 v4, v2 +; GFX6-NEXT: v_mov_b32_e32 v5, v2 ; GFX6-NEXT: buffer_load_format_xyz v[2:5], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -730,6 +970,8 @@ define amdgpu_cs float @buffer_load_v2f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX8PLUS-LABEL: buffer_load_v2f32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 +; GFX8PLUS-NEXT: v_mov_b32_e32 v4, v2 ; GFX8PLUS-NEXT: buffer_load_format_xy v[2:4], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dwordx2 v[0:1], v[2:3] @@ -740,11 +982,23 @@ define amdgpu_cs float @buffer_load_v2f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ; GFX11-LABEL: buffer_load_v2f32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 +; GFX11-NEXT: v_mov_b32_e32 v4, v2 ; GFX11-NEXT: buffer_load_format_xy v[2:4], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off ; GFX11-NEXT: v_mov_b32_e32 v0, v4 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_v2f32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v4, 0 +; NOPRT-NEXT: buffer_load_format_xy v[2:4], v4, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b64 v[0:1], v[2:3], off +; NOPRT-NEXT: v_mov_b32_e32 v0, v4 +; NOPRT-NEXT: ; return to shader part epilog %load = call { <2 x float>, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_v2f32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { <2 x float>, i32 } %load, 0 store <2 x float> %data, ptr addrspace(1) %out @@ -757,6 +1011,7 @@ define amdgpu_cs float @buffer_load_i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ad ; GFX6-LABEL: buffer_load_i32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 ; GFX6-NEXT: buffer_load_format_x v[2:3], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -771,6 +1026,7 @@ define amdgpu_cs float @buffer_load_i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ad ; GFX8PLUS-LABEL: buffer_load_i32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 ; GFX8PLUS-NEXT: buffer_load_format_x v[2:3], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dword v[0:1], v2 @@ -781,11 +1037,22 @@ define amdgpu_cs float @buffer_load_i32_tfe(ptr addrspace(8) inreg %rsrc, ptr ad ; GFX11-LABEL: buffer_load_i32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 ; GFX11-NEXT: buffer_load_format_x v[2:3], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b32 v[0:1], v2, off ; GFX11-NEXT: v_mov_b32_e32 v0, v3 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_i32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v3, 0 +; NOPRT-NEXT: buffer_load_format_x v[2:3], v3, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b32 v[0:1], v2, off +; NOPRT-NEXT: v_mov_b32_e32 v0, v3 +; NOPRT-NEXT: ; return to shader part epilog %load = call { i32, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_i32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { i32, i32 } %load, 0 store i32 %data, ptr addrspace(1) %out @@ -798,6 +1065,7 @@ define amdgpu_cs float @buffer_load_f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ad ; GFX6-LABEL: buffer_load_f32_tfe: ; GFX6: ; %bb.0: ; GFX6-NEXT: v_mov_b32_e32 v2, 0 +; GFX6-NEXT: v_mov_b32_e32 v3, v2 ; GFX6-NEXT: buffer_load_format_x v[2:3], v2, s[0:3], 0 idxen tfe ; GFX6-NEXT: s_mov_b32 s2, 0 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 @@ -812,6 +1080,7 @@ define amdgpu_cs float @buffer_load_f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ad ; GFX8PLUS-LABEL: buffer_load_f32_tfe: ; GFX8PLUS: ; %bb.0: ; GFX8PLUS-NEXT: v_mov_b32_e32 v2, 0 +; GFX8PLUS-NEXT: v_mov_b32_e32 v3, v2 ; GFX8PLUS-NEXT: buffer_load_format_x v[2:3], v2, s[0:3], 0 idxen tfe ; GFX8PLUS-NEXT: s_waitcnt vmcnt(0) ; GFX8PLUS-NEXT: flat_store_dword v[0:1], v2 @@ -822,11 +1091,22 @@ define amdgpu_cs float @buffer_load_f32_tfe(ptr addrspace(8) inreg %rsrc, ptr ad ; GFX11-LABEL: buffer_load_f32_tfe: ; GFX11: ; %bb.0: ; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_mov_b32_e32 v3, v2 ; GFX11-NEXT: buffer_load_format_x v[2:3], v2, s[0:3], 0 idxen tfe ; GFX11-NEXT: s_waitcnt vmcnt(0) ; GFX11-NEXT: global_store_b32 v[0:1], v2, off ; GFX11-NEXT: v_mov_b32_e32 v0, v3 ; GFX11-NEXT: ; return to shader part epilog +; +; NOPRT-LABEL: buffer_load_f32_tfe: +; NOPRT: ; %bb.0: +; NOPRT-NEXT: v_mov_b32_e32 v3, 0 +; NOPRT-NEXT: buffer_load_format_x v[2:3], v3, s[0:3], 0 idxen tfe +; NOPRT-NEXT: s_waitcnt vmcnt(0) +; NOPRT-NEXT: global_store_b32 v[0:1], v2, off +; NOPRT-NEXT: v_mov_b32_e32 v0, v3 +; NOPRT-NEXT: ; return to shader part epilog %load = call { float, i32 } @llvm.amdgcn.struct.ptr.buffer.load.format.sl_f32i32s(ptr addrspace(8) %rsrc, i32 0, i32 0, i32 0, i32 0) %data = extractvalue { float, i32 } %load, 0 store float %data, ptr addrspace(1) %out |