Matt ArsenaultandClaude
b719e2e27c
Mechanically migrate the command-line target spelling on llc/opt RUN lines in llvm/test/CodeGen/AMDGPU from -mtriple=amdgcn ... -mcpu=<gfx> to the folded subarch triple form (e.g. -mtriple=amdgpu9.00-amd-amdhsa), dropping the redundant -mcpu. Co-Authored-By: Claude <noreply@anthropic.com> (Claude-Opus-4.8)
234 lines
8.8 KiB
LLVM
234 lines
8.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc -mtriple=amdgpu7.00-amd-amdhsa < %s | FileCheck -check-prefixes=GCN,GFX7 %s
|
|
; RUN: llc -mtriple=amdgpu9.00-amd-amdhsa < %s | FileCheck -check-prefixes=GCN,GFX9 %s
|
|
; RUN: llc -mtriple=amdgpu11.00-amd-amdhsa -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
|
|
; RUN: llc -mtriple=amdgpu11.00-amd-amdhsa -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
|
|
|
|
; Demonstrate that the conversion of bitmasks affecting the sign bit on integers to srcmods
|
|
; does not apply to canonicalizing instructions.
|
|
|
|
define double @v_uitofp_i32_to_f64_abs(i32 %arg0) nounwind {
|
|
; GCN-LABEL: v_uitofp_i32_to_f64_abs:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: v_and_b32_e32 v0, 0x7fffffff, v0
|
|
; GCN-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_uitofp_i32_to_f64_abs:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0x7fffffff, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.abs = and i32 %arg0, u0x7fffffff
|
|
%cvt = uitofp i32 %arg0.abs to double
|
|
ret double %cvt
|
|
}
|
|
|
|
define double @v_uitofp_i32_to_f64_neg(i32 %arg0) nounwind {
|
|
; GCN-LABEL: v_uitofp_i32_to_f64_neg:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: v_and_b32_e32 v0, 0x80000000, v0
|
|
; GCN-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_uitofp_i32_to_f64_neg:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0x80000000, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.neg = and i32 %arg0, u0x80000000
|
|
%cvt = uitofp i32 %arg0.neg to double
|
|
ret double %cvt
|
|
}
|
|
|
|
define double @s_uitofp_i32_to_f64_abs(i32 inreg %arg0) nounwind {
|
|
; GCN-LABEL: s_uitofp_i32_to_f64_abs:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: s_bitset0_b32 s16, 31
|
|
; GCN-NEXT: v_cvt_f64_u32_e32 v[0:1], s16
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: s_uitofp_i32_to_f64_abs:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: s_bitset0_b32 s0, 31
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-NEXT: v_cvt_f64_u32_e32 v[0:1], s0
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.abs = and i32 %arg0, u0x7fffffff
|
|
%cvt = uitofp i32 %arg0.abs to double
|
|
ret double %cvt
|
|
}
|
|
|
|
define double @s_uitofp_i32_to_f64_neg(i32 inreg %arg0) nounwind {
|
|
; GCN-LABEL: s_uitofp_i32_to_f64_neg:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: s_and_b32 s4, s16, 0x80000000
|
|
; GCN-NEXT: v_cvt_f64_u32_e32 v[0:1], s4
|
|
; GCN-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: s_uitofp_i32_to_f64_neg:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: s_and_b32 s0, s0, 0x80000000
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-NEXT: v_cvt_f64_u32_e32 v[0:1], s0
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.neg = and i32 %arg0, u0x80000000
|
|
%cvt = uitofp i32 %arg0.neg to double
|
|
ret double %cvt
|
|
}
|
|
|
|
define half @v_uitofp_i16_to_f16_abs(i16 %arg0) nounwind {
|
|
; GFX7-LABEL: v_uitofp_i16_to_f16_abs:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0x7fff, v0
|
|
; GFX7-NEXT: v_cvt_f32_u32_e32 v0, v0
|
|
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
|
|
; GFX7-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_uitofp_i16_to_f16_abs:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0x7fff, v0
|
|
; GFX9-NEXT: v_cvt_f16_u16_e32 v0, v0
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-TRUE16-LABEL: v_uitofp_i16_to_f16_abs:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0x7fff, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_cvt_f16_u16_e32 v0.l, v0.l
|
|
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-FAKE16-LABEL: v_uitofp_i16_to_f16_abs:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0x7fff, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-FAKE16-NEXT: v_cvt_f16_u16_e32 v0, v0
|
|
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.abs = and i16 %arg0, u0x7fff
|
|
%cvt = uitofp i16 %arg0.abs to half
|
|
ret half %cvt
|
|
}
|
|
|
|
define half @v_uitofp_i16_to_f16_neg(i16 %arg0) nounwind {
|
|
; GFX7-LABEL: v_uitofp_i16_to_f16_neg:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0x8000, v0
|
|
; GFX7-NEXT: v_cvt_f32_u32_e32 v0, v0
|
|
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
|
|
; GFX7-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_uitofp_i16_to_f16_neg:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff8000, v0
|
|
; GFX9-NEXT: v_cvt_f16_u16_e32 v0, v0
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-TRUE16-LABEL: v_uitofp_i16_to_f16_neg:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0x8000, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_cvt_f16_u16_e32 v0.l, v0.l
|
|
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-FAKE16-LABEL: v_uitofp_i16_to_f16_neg:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff8000, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-FAKE16-NEXT: v_cvt_f16_u16_e32 v0, v0
|
|
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.neg = and i16 %arg0, u0x8000
|
|
%cvt = uitofp i16 %arg0.neg to half
|
|
ret half %cvt
|
|
}
|
|
|
|
define half @s_uitofp_i16_to_f16_abs(i16 inreg %arg0) nounwind {
|
|
; GFX7-LABEL: s_uitofp_i16_to_f16_abs:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s4, s16, 0x7fff
|
|
; GFX7-NEXT: v_cvt_f32_u32_e32 v0, s4
|
|
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
|
|
; GFX7-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: s_uitofp_i16_to_f16_abs:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_and_b32 s4, s16, 0x7fff
|
|
; GFX9-NEXT: v_cvt_f16_u16_e32 v0, s4
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-TRUE16-LABEL: s_uitofp_i16_to_f16_abs:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0x7fff
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-TRUE16-NEXT: v_cvt_f16_u16_e32 v0.l, s0
|
|
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-FAKE16-LABEL: s_uitofp_i16_to_f16_abs:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0x7fff
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-FAKE16-NEXT: v_cvt_f16_u16_e32 v0, s0
|
|
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.abs = and i16 %arg0, u0x7fff
|
|
%cvt = uitofp i16 %arg0.abs to half
|
|
ret half %cvt
|
|
}
|
|
|
|
define half @s_uitofp_i16_to_f16_neg(i16 inreg %arg0) nounwind {
|
|
; GFX7-LABEL: s_uitofp_i16_to_f16_neg:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s4, s16, 0x8000
|
|
; GFX7-NEXT: v_cvt_f32_u32_e32 v0, s4
|
|
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
|
|
; GFX7-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: s_uitofp_i16_to_f16_neg:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_and_b32 s4, s16, 0x8000
|
|
; GFX9-NEXT: v_cvt_f16_u16_e32 v0, s4
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-TRUE16-LABEL: s_uitofp_i16_to_f16_neg:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0x8000
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-TRUE16-NEXT: v_cvt_f16_u16_e32 v0.l, s0
|
|
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-FAKE16-LABEL: s_uitofp_i16_to_f16_neg:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0x8000
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-FAKE16-NEXT: v_cvt_f16_u16_e32 v0, s0
|
|
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
|
|
%arg0.neg = and i16 %arg0, u0x8000
|
|
%cvt = uitofp i16 %arg0.neg to half
|
|
ret half %cvt
|
|
}
|
|
|