Skip to content

AMDGPU: Stop handling legacy amdgpu-unsafe-fp-atomics attribute #101699

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 13, 2024
Merged

AMDGPU: Stop handling legacy amdgpu-unsafe-fp-atomics attribute #101699

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 1 addition & 8 deletions llvm/lib/Target/AMDGPU/SIISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16165,14 +16165,7 @@ static bool globalMemoryFPAtomicIsLegal(const GCNSubtarget &Subtarget,
} else if (Subtarget.supportsAgentScopeFineGrainedRemoteMemoryAtomics())
return true;

if (RMW->hasMetadata("amdgpu.no.fine.grained.memory"))
return true;

// TODO: Auto-upgrade this attribute to the metadata in function body and stop
// checking it.
return RMW->getFunction()
->getFnAttribute("amdgpu-unsafe-fp-atomics")
.getValueAsBool();
return RMW->hasMetadata("amdgpu.no.fine.grained.memory");
}

/// \return Action to perform on AtomicRMWInsts for integer operations.
Expand Down
10 changes: 5 additions & 5 deletions llvm/test/CodeGen/AMDGPU/GlobalISel/global-atomic-fadd.f32-no-rtn.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,7 +113,7 @@ define amdgpu_ps void @global_atomic_fadd_f32_saddr_no_rtn_flat_intrinsic(ptr ad
ret void
}

define amdgpu_ps void @global_atomic_fadd_f32_no_rtn_atomicrmw(ptr addrspace(1) %ptr, float %data) #0 {
define amdgpu_ps void @global_atomic_fadd_f32_no_rtn_atomicrmw(ptr addrspace(1) %ptr, float %data) {
; GFX908_GFX11-LABEL: name: global_atomic_fadd_f32_no_rtn_atomicrmw
; GFX908_GFX11: bb.1 (%ir-block.0):
; GFX908_GFX11-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2
Expand All @@ -135,11 +135,11 @@ define amdgpu_ps void @global_atomic_fadd_f32_no_rtn_atomicrmw(ptr addrspace(1)
; GFX90A_GFX940-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX90A_GFX940-NEXT: GLOBAL_ATOMIC_ADD_F32 [[REG_SEQUENCE]], [[COPY2]], 0, 0, implicit $exec :: (load store syncscope("wavefront") monotonic (s32) on %ir.ptr, addrspace 1)
; GFX90A_GFX940-NEXT: S_ENDPGM 0
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
ret void
}

define amdgpu_ps void @global_atomic_fadd_f32_saddr_no_rtn_atomicrmw(ptr addrspace(1) inreg %ptr, float %data) #0 {
define amdgpu_ps void @global_atomic_fadd_f32_saddr_no_rtn_atomicrmw(ptr addrspace(1) inreg %ptr, float %data) {
; GFX90A_GFX940-LABEL: name: global_atomic_fadd_f32_saddr_no_rtn_atomicrmw
; GFX90A_GFX940: bb.1 (%ir-block.0):
; GFX90A_GFX940-NEXT: successors: %bb.2(0x40000000), %bb.5(0x40000000)
Expand Down Expand Up @@ -213,11 +213,11 @@ define amdgpu_ps void @global_atomic_fadd_f32_saddr_no_rtn_atomicrmw(ptr addrspa
; GFX90A_GFX940-NEXT: bb.5 (%ir-block.33):
; GFX90A_GFX940-NEXT: SI_END_CF [[SI_IF]], implicit-def $exec, implicit-def $scc, implicit $exec
; GFX90A_GFX940-NEXT: S_ENDPGM 0
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
ret void
}

declare float @llvm.amdgcn.global.atomic.fadd.f32.p1.f32(ptr addrspace(1), float)
declare float @llvm.amdgcn.flat.atomic.fadd.f32.p1.f32(ptr addrspace(1), float)

attributes #0 = {"amdgpu-unsafe-fp-atomics"="true" }
!0 = !{}
10 changes: 5 additions & 5 deletions llvm/test/CodeGen/AMDGPU/GlobalISel/global-atomic-fadd.f32-rtn.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,7 +120,7 @@ define amdgpu_ps float @global_atomic_fadd_f32_saddr_rtn_flat_intrinsic(ptr addr
ret float %ret
}

define amdgpu_ps float @global_atomic_fadd_f32_rtn_atomicrmw(ptr addrspace(1) %ptr, float %data) #0 {
define amdgpu_ps float @global_atomic_fadd_f32_rtn_atomicrmw(ptr addrspace(1) %ptr, float %data) {
; GFX90A_GFX940-LABEL: name: global_atomic_fadd_f32_rtn_atomicrmw
; GFX90A_GFX940: bb.1 (%ir-block.0):
; GFX90A_GFX940-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2
Expand All @@ -144,11 +144,11 @@ define amdgpu_ps float @global_atomic_fadd_f32_rtn_atomicrmw(ptr addrspace(1) %p
; GFX11-NEXT: [[GLOBAL_ATOMIC_ADD_F32_RTN:%[0-9]+]]:vgpr_32 = GLOBAL_ATOMIC_ADD_F32_RTN [[REG_SEQUENCE]], [[COPY2]], 0, 1, implicit $exec :: (load store syncscope("wavefront") monotonic (s32) on %ir.ptr, addrspace 1)
; GFX11-NEXT: $vgpr0 = COPY [[GLOBAL_ATOMIC_ADD_F32_RTN]]
; GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
ret float %ret
}

define amdgpu_ps float @global_atomic_fadd_f32_saddr_rtn_atomicrmw(ptr addrspace(1) inreg %ptr, float %data) #0 {
define amdgpu_ps float @global_atomic_fadd_f32_saddr_rtn_atomicrmw(ptr addrspace(1) inreg %ptr, float %data) {
; GFX90A-LABEL: name: global_atomic_fadd_f32_saddr_rtn_atomicrmw
; GFX90A: bb.1 (%ir-block.0):
; GFX90A-NEXT: successors: %bb.2(0x40000000), %bb.4(0x40000000)
Expand Down Expand Up @@ -422,11 +422,11 @@ define amdgpu_ps float @global_atomic_fadd_f32_saddr_rtn_atomicrmw(ptr addrspace
; GFX11-NEXT: bb.6 (%ir-block.38):
; GFX11-NEXT: $vgpr0 = COPY [[PHI]]
; GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic
%ret = atomicrmw fadd ptr addrspace(1) %ptr, float %data syncscope("wavefront") monotonic, !amdgpu.no.fine.grained.memory !0, !amdgpu.ignore.denormal.mode !0
ret float %ret
}

declare float @llvm.amdgcn.global.atomic.fadd.f32.p1.f32(ptr addrspace(1), float)
declare float @llvm.amdgcn.flat.atomic.fadd.f32.p1.f32(ptr addrspace(1), float)

attributes #0 = {"amdgpu-unsafe-fp-atomics"="true" }
!0 = !{}
34 changes: 17 additions & 17 deletions llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -365,7 +365,7 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_agent_scope_
; GFX1132-DPP-NEXT: s_nop 0
; GFX1132-DPP-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4
%result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -1158,7 +1158,7 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX1132-DPP-NEXT: .LBB1_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value()
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -1602,7 +1602,7 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_one_as_scope
; GFX1132-DPP-NEXT: global_atomic_add_f32 v1, v0, s[0:1]
; GFX1132-DPP-NEXT: .LBB2_2:
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 syncscope("one-as") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -2396,7 +2396,7 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_one_as_scope
; GFX1132-DPP-NEXT: .LBB3_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value() strictfp
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("one-as") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -3694,7 +3694,7 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX1132-DPP-NEXT: .LBB5_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value()
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -4488,7 +4488,7 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX1132-DPP-NEXT: .LBB6_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value() strictfp
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -6739,7 +6739,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_agent
; GFX1132-DPP-NEXT: .LBB9_3:
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 syncscope("agent") monotonic, align 4
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -8066,7 +8066,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.float.value()
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic, align 4
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -8584,7 +8584,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_one_a
; GFX1132-DPP-NEXT: s_cbranch_execnz .LBB11_2
; GFX1132-DPP-NEXT: .LBB11_3:
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 syncscope("one-as") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -9525,7 +9525,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_one_a
; GFX1132-DPP-NEXT: .LBB12_3:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.double.value() strictfp
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("one-as") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -10043,7 +10043,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_agent
; GFX1132-DPP-NEXT: s_cbranch_execnz .LBB13_2
; GFX1132-DPP-NEXT: .LBB13_3:
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 syncscope("agent") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 syncscope("agent") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -10984,7 +10984,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX1132-DPP-NEXT: .LBB14_3:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.double.value()
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -11925,7 +11925,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX1132-DPP-NEXT: .LBB15_3:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.float.value() strictfp
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -12859,7 +12859,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_defau
; GFX1132-DPP-NEXT: .LBB16_3:
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 monotonic, align 4
%result = atomicrmw fadd ptr addrspace(1) %ptr, double 4.0 monotonic, align 4, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -14186,7 +14186,7 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_defau
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.float.value() strictfp
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue monotonic, align 4
%result = atomicrmw fadd ptr addrspace(1) %ptr, double %divValue monotonic, align 4, !amdgpu.no.fine.grained.memory !1, !amdgpu.ignore.denormal.mode !1
ret void
}

Expand Down Expand Up @@ -14898,8 +14898,8 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_system_scope
ret void
}

attributes #0 = { "denormal-fp-math-f32"="preserve-sign,preserve-sign" "amdgpu-unsafe-fp-atomics"="true" }
attributes #1 = { strictfp "denormal-fp-math-f32"="preserve-sign,preserve-sign" "amdgpu-unsafe-fp-atomics"="true" }
attributes #0 = { "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { strictfp "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #2 = { strictfp }

!llvm.module.flags = !{!0}
Expand Down
26 changes: 13 additions & 13 deletions llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -273,7 +273,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_agent_scope_
; GFX1132-DPP-NEXT: s_nop 0
; GFX1132-DPP-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -1075,7 +1075,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_agent_scope_
; GFX1132-DPP-NEXT: .LBB1_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value()
%result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -1337,7 +1337,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_one_as_scope
; GFX1132-DPP-NEXT: s_nop 0
; GFX1132-DPP-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 syncscope("one-as") monotonic
%result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -2140,7 +2140,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_one_as_scope
; GFX1132-DPP-NEXT: .LBB3_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value()
%result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue syncscope("one-as") monotonic
%result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -2403,7 +2403,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_default_scop
; GFX1132-DPP-NEXT: s_nop 0
; GFX1132-DPP-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -3205,7 +3205,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_default_scop
; GFX1132-DPP-NEXT: .LBB5_2:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call float @div.float.value()
%result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -4069,7 +4069,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_uni_value_agent
; GFX1132-DPP-NEXT: .LBB6_3:
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fmax ptr addrspace(1) %ptr, double 4.0 syncscope("agent") monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, double 4.0 syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -5474,7 +5474,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_agent
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.double.value()
%result = atomicrmw fmax ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, double %divValue syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -5810,7 +5810,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_uni_value_one_a
; GFX1132-DPP-NEXT: s_cbranch_execnz .LBB8_2
; GFX1132-DPP-NEXT: .LBB8_3:
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fmax ptr addrspace(1) %ptr, double 4.0 syncscope("one-as") monotonic
%result = atomicrmw fmax ptr addrspace(1) %ptr, double 4.0 syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -6772,7 +6772,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_one_a
; GFX1132-DPP-NEXT: .LBB9_3:
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.double.value()
%result = atomicrmw fmax ptr addrspace(1) %ptr, double %divValue syncscope("one-as") monotonic
%result = atomicrmw fmax ptr addrspace(1) %ptr, double %divValue syncscope("one-as") monotonic, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -7636,7 +7636,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_uni_value_defau
; GFX1132-DPP-NEXT: .LBB10_3:
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%result = atomicrmw fmax ptr addrspace(1) %ptr, double 4.0 monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, double 4.0 monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -9041,7 +9041,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_defau
; GFX1132-DPP-NEXT: s_set_inst_prefetch_distance 0x2
; GFX1132-DPP-NEXT: s_endpgm
%divValue = call double @div.double.value()
%result = atomicrmw fmax ptr addrspace(1) %ptr, double %divValue monotonic, align 4
%result = atomicrmw fmax ptr addrspace(1) %ptr, double %divValue monotonic, align 4, !amdgpu.no.fine.grained.memory !1
ret void
}

Expand Down Expand Up @@ -9569,7 +9569,7 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_system_scope
ret void
}

attributes #0 = { "denormal-fp-math-f32"="preserve-sign,preserve-sign" "amdgpu-unsafe-fp-atomics"="true" }
attributes #0 = { "denormal-fp-math-f32"="preserve-sign,preserve-sign" }

!llvm.module.flags = !{!0}
!0 = !{i32 1, !"amdhsa_code_object_version", i32 500}
Expand Down
Loading
Loading