119 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GCN-SDAG %s3; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GCN-GISEL %s4 5; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN,GFX10PLUS,GFX10 %s6; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN,GFX10PLUS,GFX10 %s7 8; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GCN,GFX10PLUS,GFX11 %s9; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GCN,GFX10PLUS,GFX11 %s10 11define double @v_constained_fadd_f64_fpexcept_strict(double %x, double %y) #0 {12; GCN-LABEL: v_constained_fadd_f64_fpexcept_strict:13; GCN: ; %bb.0:14; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)15; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[2:3]16; GCN-NEXT: s_setpc_b64 s[30:31]17 %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.strict")18 ret double %val19}20 21define double @v_constained_fadd_f64_fpexcept_ignore(double %x, double %y) #0 {22; GCN-LABEL: v_constained_fadd_f64_fpexcept_ignore:23; GCN: ; %bb.0:24; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)25; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[2:3]26; GCN-NEXT: s_setpc_b64 s[30:31]27 %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")28 ret double %val29}30 31define double @v_constained_fadd_f64_fpexcept_maytrap(double %x, double %y) #0 {32; GCN-LABEL: v_constained_fadd_f64_fpexcept_maytrap:33; GCN: ; %bb.0:34; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)35; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[2:3]36; GCN-NEXT: s_setpc_b64 s[30:31]37 %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")38 ret double %val39}40 41define <2 x double> @v_constained_fadd_v2f64_fpexcept_strict(<2 x double> %x, <2 x double> %y) #0 {42; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_strict:43; GCN: ; %bb.0:44; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)45; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]46; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]47; GCN-NEXT: s_setpc_b64 s[30:31]48 %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")49 ret <2 x double> %val50}51 52define <2 x double> @v_constained_fadd_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 {53; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_ignore:54; GCN: ; %bb.0:55; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)56; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]57; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]58; GCN-NEXT: s_setpc_b64 s[30:31]59 %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")60 ret <2 x double> %val61}62 63define <2 x double> @v_constained_fadd_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 {64; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap:65; GCN: ; %bb.0:66; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)67; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]68; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]69; GCN-NEXT: s_setpc_b64 s[30:31]70 %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")71 ret <2 x double> %val72}73 74define <3 x double> @v_constained_fadd_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 {75; GCN-LABEL: v_constained_fadd_v3f64_fpexcept_strict:76; GCN: ; %bb.0:77; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)78; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[6:7]79; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[8:9]80; GCN-NEXT: v_add_f64 v[4:5], v[4:5], v[10:11]81; GCN-NEXT: s_setpc_b64 s[30:31]82 %val = call <3 x double> @llvm.experimental.constrained.fadd.v3f64(<3 x double> %x, <3 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")83 ret <3 x double> %val84}85 86define amdgpu_ps <2 x float> @s_constained_fadd_f64_fpexcept_strict(double inreg %x, double inreg %y) #0 {87; GCN-SDAG-LABEL: s_constained_fadd_f64_fpexcept_strict:88; GCN-SDAG: ; %bb.0:89; GCN-SDAG-NEXT: v_mov_b32_e32 v0, s490; GCN-SDAG-NEXT: v_mov_b32_e32 v1, s591; GCN-SDAG-NEXT: v_add_f64 v[0:1], s[2:3], v[0:1]92; GCN-SDAG-NEXT: ; return to shader part epilog93;94; GCN-GISEL-LABEL: s_constained_fadd_f64_fpexcept_strict:95; GCN-GISEL: ; %bb.0:96; GCN-GISEL-NEXT: v_mov_b32_e32 v0, s497; GCN-GISEL-NEXT: v_mov_b32_e32 v1, s598; GCN-GISEL-NEXT: v_add_f64 v[0:1], s[2:3], v[0:1]99; GCN-GISEL-NEXT: ; return to shader part epilog100;101; GFX10PLUS-LABEL: s_constained_fadd_f64_fpexcept_strict:102; GFX10PLUS: ; %bb.0:103; GFX10PLUS-NEXT: v_add_f64 v[0:1], s[2:3], s[4:5]104; GFX10PLUS-NEXT: ; return to shader part epilog105 %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.strict")106 %cast = bitcast double %val to <2 x float>107 ret <2 x float> %cast108}109 110declare double @llvm.experimental.constrained.fadd.f64(double, double, metadata, metadata) #1111declare <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double>, <2 x double>, metadata, metadata) #1112declare <3 x double> @llvm.experimental.constrained.fadd.v3f64(<3 x double>, <3 x double>, metadata, metadata) #1113 114attributes #0 = { strictfp }115attributes #1 = { inaccessiblememonly nounwind willreturn }116;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:117; GFX10: {{.*}}118; GFX11: {{.*}}119