|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 |
| 2 | +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx700 < %s | FileCheck -check-prefixes=GCN,GFX7 %s |
| 3 | +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9 %s |
| 4 | +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s |
| 5 | +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s |
| 6 | + |
| 7 | +; Ensure that the conversion of bitmasks affecting the sign bit on integers to srcmods |
| 8 | +; does not apply to canonicalizing instructions. |
| 9 | + |
| 10 | +define double @v_uitofp_i32_to_f64_abs(i32 %arg0) nounwind { |
| 11 | +; GCN-LABEL: v_uitofp_i32_to_f64_abs: |
| 12 | +; GCN: ; %bb.0: |
| 13 | +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| 14 | +; GCN-NEXT: v_and_b32_e32 v0, 0x7fffffff, v0 |
| 15 | +; GCN-NEXT: v_cvt_f64_u32_e32 v[0:1], v0 |
| 16 | +; GCN-NEXT: s_setpc_b64 s[30:31] |
| 17 | +; |
| 18 | +; GFX11-LABEL: v_uitofp_i32_to_f64_abs: |
| 19 | +; GFX11: ; %bb.0: |
| 20 | +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| 21 | +; GFX11-NEXT: v_and_b32_e32 v0, 0x7fffffff, v0 |
| 22 | +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) |
| 23 | +; GFX11-NEXT: v_cvt_f64_u32_e32 v[0:1], v0 |
| 24 | +; GFX11-NEXT: s_setpc_b64 s[30:31] |
| 25 | + %arg0.abs = and i32 %arg0, u0x7fffffff |
| 26 | + %cvt = uitofp i32 %arg0.abs to double |
| 27 | + ret double %cvt |
| 28 | +} |
| 29 | + |
| 30 | +define double @v_uitofp_i32_to_f64_neg(i32 %arg0) nounwind { |
| 31 | +; GCN-LABEL: v_uitofp_i32_to_f64_neg: |
| 32 | +; GCN: ; %bb.0: |
| 33 | +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| 34 | +; GCN-NEXT: v_and_b32_e32 v0, 0x80000000, v0 |
| 35 | +; GCN-NEXT: v_cvt_f64_u32_e32 v[0:1], v0 |
| 36 | +; GCN-NEXT: s_setpc_b64 s[30:31] |
| 37 | +; |
| 38 | +; GFX11-LABEL: v_uitofp_i32_to_f64_neg: |
| 39 | +; GFX11: ; %bb.0: |
| 40 | +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| 41 | +; GFX11-NEXT: v_and_b32_e32 v0, 0x80000000, v0 |
| 42 | +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) |
| 43 | +; GFX11-NEXT: v_cvt_f64_u32_e32 v[0:1], v0 |
| 44 | +; GFX11-NEXT: s_setpc_b64 s[30:31] |
| 45 | + %arg0.neg = and i32 %arg0, u0x80000000 |
| 46 | + %cvt = uitofp i32 %arg0.neg to double |
| 47 | + ret double %cvt |
| 48 | +} |
| 49 | +;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: |
| 50 | +; GFX11-FAKE16: {{.*}} |
| 51 | +; GFX11-TRUE16: {{.*}} |
| 52 | +; GFX7: {{.*}} |
| 53 | +; GFX9: {{.*}} |
0 commit comments