derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
237 lines
9.8 KiB
LLVM
237 lines
9.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefix=SDAG %s
|
|
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefix=GISEL %s
|
|
|
|
; Test for a caller/callee mismatch in SGPR assignment for inreg args.
|
|
;
|
|
; The following test case demonstrates the bug with a function that takes 32
|
|
; inreg i32 arguments and returns the first one. In the callee, arg 0 is read
|
|
; from s16 (v_mov_b32_e32 v0, s16). In the caller, arg 0 (value 42) is written
|
|
; to s0 (s_mov_b32 s0, 42). Since s0 != s16, the callee does not see the value
|
|
; the caller intended to pass.
|
|
;
|
|
; The root cause is that the scratch resource descriptor occupies SGPR0-3. On
|
|
; the callee side, LowerFormalArguments marks SGPR0-3 as allocated in CCState
|
|
; before running the CC analysis, so CC_AMDGPU_Func skips them. On the caller
|
|
; side, LowerCall adds the scratch resource to RegsToPass without marking
|
|
; SGPR0-3 in CCState, so CC_AMDGPU_Func treats them as available and assigns
|
|
; user inreg args there.
|
|
|
|
define i32 @callee_returns_arg0(
|
|
; CHECK-LABEL: callee_returns_arg0:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s16
|
|
; CHECK-NEXT: s_setpc_b64 s[30:31]
|
|
; SDAG-LABEL: callee_returns_arg0:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, s16
|
|
; SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GISEL-LABEL: callee_returns_arg0:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, s16
|
|
; GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
i32 inreg %a0, i32 inreg %a1, i32 inreg %a2, i32 inreg %a3,
|
|
i32 inreg %a4, i32 inreg %a5, i32 inreg %a6, i32 inreg %a7,
|
|
i32 inreg %a8, i32 inreg %a9, i32 inreg %a10, i32 inreg %a11,
|
|
i32 inreg %a12, i32 inreg %a13, i32 inreg %a14, i32 inreg %a15,
|
|
i32 inreg %a16, i32 inreg %a17, i32 inreg %a18, i32 inreg %a19,
|
|
i32 inreg %a20, i32 inreg %a21, i32 inreg %a22, i32 inreg %a23,
|
|
i32 inreg %a24, i32 inreg %a25, i32 inreg %a26, i32 inreg %a27,
|
|
i32 inreg %a28, i32 inreg %a29, i32 inreg %a30, i32 inreg %a31) #0 {
|
|
ret i32 %a0
|
|
}
|
|
|
|
define i32 @caller_passes_42() #0 {
|
|
; CHECK-LABEL: caller_passes_42:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; CHECK-NEXT: s_mov_b32 s42, s33
|
|
; CHECK-NEXT: s_mov_b32 s33, s32
|
|
; CHECK-NEXT: s_xor_saveexec_b64 s[16:17], -1
|
|
; CHECK-NEXT: buffer_store_dword v18, off, s[0:3], s33 ; 4-byte Folded Spill
|
|
; CHECK-NEXT: s_mov_b64 exec, s[16:17]
|
|
; CHECK-NEXT: s_addk_i32 s32, 0x400
|
|
; CHECK-NEXT: s_getpc_b64 s[16:17]
|
|
; CHECK-NEXT: s_add_u32 s16, s16, callee_returns_arg0@gotpcrel32@lo+4
|
|
; CHECK-NEXT: s_addc_u32 s17, s17, callee_returns_arg0@gotpcrel32@hi+12
|
|
; CHECK-NEXT: s_load_dwordx2 s[40:41], s[16:17], 0x0
|
|
; CHECK-NEXT: v_writelane_b32 v18, s30, 0
|
|
; CHECK-NEXT: s_mov_b32 s16, 42
|
|
; CHECK-NEXT: s_mov_b32 s17, 1
|
|
; CHECK-NEXT: s_mov_b32 s18, 2
|
|
; CHECK-NEXT: s_mov_b32 s19, 3
|
|
; CHECK-NEXT: s_mov_b32 s20, 4
|
|
; CHECK-NEXT: s_mov_b32 s21, 5
|
|
; CHECK-NEXT: s_mov_b32 s22, 6
|
|
; CHECK-NEXT: s_mov_b32 s23, 7
|
|
; CHECK-NEXT: s_mov_b32 s24, 8
|
|
; CHECK-NEXT: s_mov_b32 s25, 9
|
|
; CHECK-NEXT: s_mov_b32 s26, 10
|
|
; CHECK-NEXT: s_mov_b32 s27, 11
|
|
; CHECK-NEXT: s_mov_b32 s28, 12
|
|
; CHECK-NEXT: s_mov_b32 s29, 13
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, 14
|
|
; CHECK-NEXT: v_mov_b32_e32 v1, 15
|
|
; CHECK-NEXT: v_mov_b32_e32 v2, 16
|
|
; CHECK-NEXT: v_mov_b32_e32 v3, 17
|
|
; CHECK-NEXT: v_mov_b32_e32 v4, 18
|
|
; CHECK-NEXT: v_mov_b32_e32 v5, 19
|
|
; CHECK-NEXT: v_mov_b32_e32 v6, 20
|
|
; CHECK-NEXT: v_mov_b32_e32 v7, 21
|
|
; CHECK-NEXT: v_mov_b32_e32 v8, 22
|
|
; CHECK-NEXT: v_mov_b32_e32 v9, 23
|
|
; CHECK-NEXT: v_mov_b32_e32 v10, 24
|
|
; CHECK-NEXT: v_mov_b32_e32 v11, 25
|
|
; CHECK-NEXT: v_mov_b32_e32 v12, 26
|
|
; CHECK-NEXT: v_mov_b32_e32 v13, 27
|
|
; CHECK-NEXT: v_mov_b32_e32 v14, 28
|
|
; CHECK-NEXT: v_mov_b32_e32 v15, 29
|
|
; CHECK-NEXT: v_mov_b32_e32 v16, 30
|
|
; CHECK-NEXT: v_mov_b32_e32 v17, 31
|
|
; CHECK-NEXT: v_writelane_b32 v18, s31, 1
|
|
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-NEXT: s_swappc_b64 s[30:31], s[40:41]
|
|
; CHECK-NEXT: v_readlane_b32 s31, v18, 1
|
|
; CHECK-NEXT: v_readlane_b32 s30, v18, 0
|
|
; CHECK-NEXT: s_mov_b32 s32, s33
|
|
; CHECK-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; CHECK-NEXT: buffer_load_dword v18, off, s[0:3], s33 ; 4-byte Folded Reload
|
|
; CHECK-NEXT: s_mov_b64 exec, s[4:5]
|
|
; CHECK-NEXT: s_mov_b32 s33, s42
|
|
; CHECK-NEXT: s_waitcnt vmcnt(0)
|
|
; CHECK-NEXT: s_setpc_b64 s[30:31]
|
|
; SDAG-LABEL: caller_passes_42:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; SDAG-NEXT: s_mov_b32 s42, s33
|
|
; SDAG-NEXT: s_mov_b32 s33, s32
|
|
; SDAG-NEXT: s_xor_saveexec_b64 s[16:17], -1
|
|
; SDAG-NEXT: buffer_store_dword v18, off, s[0:3], s33 ; 4-byte Folded Spill
|
|
; SDAG-NEXT: s_mov_b64 exec, s[16:17]
|
|
; SDAG-NEXT: v_writelane_b32 v18, s30, 0
|
|
; SDAG-NEXT: s_addk_i32 s32, 0x400
|
|
; SDAG-NEXT: v_writelane_b32 v18, s31, 1
|
|
; SDAG-NEXT: s_getpc_b64 s[16:17]
|
|
; SDAG-NEXT: s_add_u32 s16, s16, callee_returns_arg0@gotpcrel32@lo+4
|
|
; SDAG-NEXT: s_addc_u32 s17, s17, callee_returns_arg0@gotpcrel32@hi+12
|
|
; SDAG-NEXT: s_load_dwordx2 s[40:41], s[16:17], 0x0
|
|
; SDAG-NEXT: s_mov_b32 s16, 42
|
|
; SDAG-NEXT: s_mov_b32 s17, 1
|
|
; SDAG-NEXT: s_mov_b32 s18, 2
|
|
; SDAG-NEXT: s_mov_b32 s19, 3
|
|
; SDAG-NEXT: s_mov_b32 s20, 4
|
|
; SDAG-NEXT: s_mov_b32 s21, 5
|
|
; SDAG-NEXT: s_mov_b32 s22, 6
|
|
; SDAG-NEXT: s_mov_b32 s23, 7
|
|
; SDAG-NEXT: s_mov_b32 s24, 8
|
|
; SDAG-NEXT: s_mov_b32 s25, 9
|
|
; SDAG-NEXT: s_mov_b32 s26, 10
|
|
; SDAG-NEXT: s_mov_b32 s27, 11
|
|
; SDAG-NEXT: s_mov_b32 s28, 12
|
|
; SDAG-NEXT: s_mov_b32 s29, 13
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, 14
|
|
; SDAG-NEXT: v_mov_b32_e32 v1, 15
|
|
; SDAG-NEXT: v_mov_b32_e32 v2, 16
|
|
; SDAG-NEXT: v_mov_b32_e32 v3, 17
|
|
; SDAG-NEXT: v_mov_b32_e32 v4, 18
|
|
; SDAG-NEXT: v_mov_b32_e32 v5, 19
|
|
; SDAG-NEXT: v_mov_b32_e32 v6, 20
|
|
; SDAG-NEXT: v_mov_b32_e32 v7, 21
|
|
; SDAG-NEXT: v_mov_b32_e32 v8, 22
|
|
; SDAG-NEXT: v_mov_b32_e32 v9, 23
|
|
; SDAG-NEXT: v_mov_b32_e32 v10, 24
|
|
; SDAG-NEXT: v_mov_b32_e32 v11, 25
|
|
; SDAG-NEXT: v_mov_b32_e32 v12, 26
|
|
; SDAG-NEXT: v_mov_b32_e32 v13, 27
|
|
; SDAG-NEXT: v_mov_b32_e32 v14, 28
|
|
; SDAG-NEXT: v_mov_b32_e32 v15, 29
|
|
; SDAG-NEXT: v_mov_b32_e32 v16, 30
|
|
; SDAG-NEXT: v_mov_b32_e32 v17, 31
|
|
; SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SDAG-NEXT: s_swappc_b64 s[30:31], s[40:41]
|
|
; SDAG-NEXT: v_readlane_b32 s30, v18, 0
|
|
; SDAG-NEXT: v_readlane_b32 s31, v18, 1
|
|
; SDAG-NEXT: s_mov_b32 s32, s33
|
|
; SDAG-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; SDAG-NEXT: buffer_load_dword v18, off, s[0:3], s33 ; 4-byte Folded Reload
|
|
; SDAG-NEXT: s_mov_b64 exec, s[4:5]
|
|
; SDAG-NEXT: s_mov_b32 s33, s42
|
|
; SDAG-NEXT: s_waitcnt vmcnt(0)
|
|
; SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GISEL-LABEL: caller_passes_42:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GISEL-NEXT: s_mov_b32 s42, s33
|
|
; GISEL-NEXT: s_mov_b32 s33, s32
|
|
; GISEL-NEXT: s_xor_saveexec_b64 s[16:17], -1
|
|
; GISEL-NEXT: buffer_store_dword v18, off, s[0:3], s33 ; 4-byte Folded Spill
|
|
; GISEL-NEXT: s_mov_b64 exec, s[16:17]
|
|
; GISEL-NEXT: v_writelane_b32 v18, s30, 0
|
|
; GISEL-NEXT: s_addk_i32 s32, 0x400
|
|
; GISEL-NEXT: v_writelane_b32 v18, s31, 1
|
|
; GISEL-NEXT: s_getpc_b64 s[16:17]
|
|
; GISEL-NEXT: s_add_u32 s16, s16, callee_returns_arg0@gotpcrel32@lo+4
|
|
; GISEL-NEXT: s_addc_u32 s17, s17, callee_returns_arg0@gotpcrel32@hi+12
|
|
; GISEL-NEXT: s_load_dwordx2 s[40:41], s[16:17], 0x0
|
|
; GISEL-NEXT: s_mov_b32 s16, 42
|
|
; GISEL-NEXT: s_mov_b32 s17, 1
|
|
; GISEL-NEXT: s_mov_b32 s18, 2
|
|
; GISEL-NEXT: s_mov_b32 s19, 3
|
|
; GISEL-NEXT: s_mov_b32 s20, 4
|
|
; GISEL-NEXT: s_mov_b32 s21, 5
|
|
; GISEL-NEXT: s_mov_b32 s22, 6
|
|
; GISEL-NEXT: s_mov_b32 s23, 7
|
|
; GISEL-NEXT: s_mov_b32 s24, 8
|
|
; GISEL-NEXT: s_mov_b32 s25, 9
|
|
; GISEL-NEXT: s_mov_b32 s26, 10
|
|
; GISEL-NEXT: s_mov_b32 s27, 11
|
|
; GISEL-NEXT: s_mov_b32 s28, 12
|
|
; GISEL-NEXT: s_mov_b32 s29, 13
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 14
|
|
; GISEL-NEXT: v_mov_b32_e32 v1, 15
|
|
; GISEL-NEXT: v_mov_b32_e32 v2, 16
|
|
; GISEL-NEXT: v_mov_b32_e32 v3, 17
|
|
; GISEL-NEXT: v_mov_b32_e32 v4, 18
|
|
; GISEL-NEXT: v_mov_b32_e32 v5, 19
|
|
; GISEL-NEXT: v_mov_b32_e32 v6, 20
|
|
; GISEL-NEXT: v_mov_b32_e32 v7, 21
|
|
; GISEL-NEXT: v_mov_b32_e32 v8, 22
|
|
; GISEL-NEXT: v_mov_b32_e32 v9, 23
|
|
; GISEL-NEXT: v_mov_b32_e32 v10, 24
|
|
; GISEL-NEXT: v_mov_b32_e32 v11, 25
|
|
; GISEL-NEXT: v_mov_b32_e32 v12, 26
|
|
; GISEL-NEXT: v_mov_b32_e32 v13, 27
|
|
; GISEL-NEXT: v_mov_b32_e32 v14, 28
|
|
; GISEL-NEXT: v_mov_b32_e32 v15, 29
|
|
; GISEL-NEXT: v_mov_b32_e32 v16, 30
|
|
; GISEL-NEXT: v_mov_b32_e32 v17, 31
|
|
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GISEL-NEXT: s_swappc_b64 s[30:31], s[40:41]
|
|
; GISEL-NEXT: v_readlane_b32 s30, v18, 0
|
|
; GISEL-NEXT: v_readlane_b32 s31, v18, 1
|
|
; GISEL-NEXT: s_mov_b32 s32, s33
|
|
; GISEL-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GISEL-NEXT: buffer_load_dword v18, off, s[0:3], s33 ; 4-byte Folded Reload
|
|
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GISEL-NEXT: s_mov_b32 s33, s42
|
|
; GISEL-NEXT: s_waitcnt vmcnt(0)
|
|
; GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
%r = call i32 @callee_returns_arg0(
|
|
i32 inreg 42, i32 inreg 1, i32 inreg 2, i32 inreg 3,
|
|
i32 inreg 4, i32 inreg 5, i32 inreg 6, i32 inreg 7,
|
|
i32 inreg 8, i32 inreg 9, i32 inreg 10, i32 inreg 11,
|
|
i32 inreg 12, i32 inreg 13, i32 inreg 14, i32 inreg 15,
|
|
i32 inreg 16, i32 inreg 17, i32 inreg 18, i32 inreg 19,
|
|
i32 inreg 20, i32 inreg 21, i32 inreg 22, i32 inreg 23,
|
|
i32 inreg 24, i32 inreg 25, i32 inreg 26, i32 inreg 27,
|
|
i32 inreg 28, i32 inreg 29, i32 inreg 30, i32 inreg 31)
|
|
ret i32 %r
|
|
}
|
|
|
|
attributes #0 = { nounwind }
|