Differentiating FFTW fftshift #1652
Labels
Comments
|
Can you paste the whole code to reproduce and error log? |
|
Certainly! Here is a simplified version of the full code that produces the error: using Enzyme, Pkg, UnPack, Plots, HDF5, LinearAlgebra, LineSearches, DSP, Optim, FFTW, SignalAnalysis, Statistics, Random, LaTeXStrings
Xd = ones(4,200000)
t = range(start=5,step=0.005,length=200000)
dt_data = 0.005
N_t_data = Int(500/dt_data)
X_1011 = transpose(Xd[:, 1:N_t_data])
ts = range(; start=0, step = dt_data, length=N_t_data)
ts = ts*2
# Y variables parameterization
function U_det(Xk, b0, b1, b2, b3)
U = b0 + b1*Xk + b2/10.0*Xk^2 + b3/100.0*Xk^3
return U
end
function U_p(Xk, b0, b1, b2, b3, e_tk)
U = U_det(Xk, b0, b1, b2, b3) + e_tk
return U
end
# dX rhs
function dXrhs(k, X, b0, b1, b2, b3, e_t)
dXk = - X[k - 1] * (X[k - 2] - X[k + 1]) - X[k] + F - U_p(X[k], b0, b1, b2, b3, e_t[k])
return dXk
end
# Euler step
function euler_step(model_struct)
@unpack N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2, e_t, e_tm1, sig_e_t, sig_e_tm1, X_array = model_struct
# Compute derivatives
dX[1] = - X[K] * (X[K - 1] - X[2]) - X[1] + F - U_p(X[1], b0, b1, b2, b3, e_t[1])
dX[2] = - X[1] * (X[K] - X[3]) - X[2] + F - U_p(X[2], b0, b1, b2, b3, e_t[2])
dX[K] = - X[K - 1] * (X[K - 2] - X[1]) - X[K] + F - U_p(X[K], b0, b1, b2, b3, e_t[K])
# Update X
X[1] += dt * dX[1]
X[2] += dt * dX[2]
X[K] += dt * dX[3]
for k = 3:K-1
dX[k] = dXrhs(k, X, b0, b1, b2, b3, e_t)
X[k] += dt * dX[k]
end
# Update dXs
dXm1[:] = dX[:]
dXm2[:] = dXm1[:]
@pack! model_struct = X, dX, dXm1, dXm2
return nothing
end
# AB-3 rhs
function adams_bashforth3(x, dt, dx, dxm1, dxm2)
rhs = x + dt * ((23/12) * dx
- (16/12) * dxm1
+ (5/12) * dxm2)
return rhs
end
# AB-3 step
function ab3_step(model_struct)
@unpack N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2, e_t, e_tm1, sig_e_t, sig_e_tm1, X_array = model_struct
# Compute derivatives
dX[1] = - X[K] * (X[K - 1] - X[2]) - X[1] + F - U_p(X[1], b0, b1, b2, b3, e_t[1])
dX[2] = - X[1] * (X[K] - X[3]) - X[2] + F - U_p(X[2], b0, b1, b2, b3, e_t[2])
dX[K] = - X[K - 1] * (X[K - 2] - X[1]) - X[K] + F - U_p(X[K], b0, b1, b2, b3, e_t[K])
# Update X
X[1] = adams_bashforth3(X[1], dt, dX[1], dXm1[1], dXm2[1])
X[2] = adams_bashforth3(X[2], dt, dX[2], dXm1[2], dXm2[2])
X[K] = adams_bashforth3(X[K], dt, dX[K], dXm1[K], dXm2[K])
for k = 3:K-1
dX[k] = dXrhs(k, X, b0, b1, b2, b3, e_t)
X[k] = adams_bashforth3(X[k], dt, dX[k], dXm1[k], dXm2[k])
end
# Update dXs
dXm1[:] = dX[:]
dXm2[:] = dXm1[:]
@pack! model_struct = X, dX, dXm1, dXm2
return nothing
end
# AR(1) process
function AR1(model_struct)
@unpack N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2, e_t, e_tm1, sig_e_t, sig_e_tm1, X_array = model_struct
for k = 1:K
sig_e_t[k] = sig1*abs(X[k]) + sig0
e_t[k] = sig_e_t[k]*(1-10^(-phi))*e_tm1[k]/sig_e_tm1[k] + sig_e_t[k]*sqrt(1 - (1-10^(-phi))^2)*randn()
end
# Update e_t, sig_e_t
e_tm1[:] = e_t[:]
sig_e_tm1[:] = sig_e_t[:]
@pack! model_struct = e_t, e_tm1, sig_e_t, sig_e_tm1
return nothing
end
mutable struct cost
# Model set-up parameters
N_t::Int64
dt::Float64
K::Int64
h::Float64
F::Float64
b::Float64
c::Float64
# Forecast control parameters
b0::Float64
b1::Float64
b2::Float64
b3::Float64
phi::Float64
sig0::Float64
sig1::Float64
# Variables
X::Vector{Float64}
dX::Vector{Float64}
dXm1::Vector{Float64}
dXm2::Vector{Float64}
# Random
e_t::Vector{Float64}
e_tm1::Vector{Float64}
sig_e_t::Vector{Float64}
sig_e_tm1::Vector{Float64}
# Data in time array
X_array::Matrix{Float64}
# Data
data::Matrix{Float64}
step_factor::Int64
# Objective function
J0::Float64
end
function init_cost(N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, data, step_factor, J0)
# Memory allocation
dX = zeros(Float64, K)
dXm1 = zeros(Float64, K)
dXm2 = zeros(Float64, K)
e_t = zeros(Float64, K)
e_tm1 = zeros(Float64, K)
sig_e_t = zeros(Float64, K) .+ sig0
sig_e_tm1 = zeros(Float64, K) .+ sig0
X_array = zeros(Float64, K, N_t+1)
# Construct initial L96 object
J0_L96 = cost(N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2,
e_t, e_tm1, sig_e_t, sig_e_tm1, X_array, data, step_factor, J0)
# Set intial conditions in X_array
J0_L96.X_array[:, 1] = J0_L96.X[:]
return J0_L96
end
function forward_model_cost(model_struct)
@unpack N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX,
dXm1, dXm2, e_t, e_tm1, sig_e_t, sig_e_tm1, X_array, data, step_factor, J0 = model_struct
js = 1:step_factor:N_t
for i in 1:2
AR1(model_struct)
euler_step(model_struct)
X_array[:, i+1] = X[:]
end
for i in 2:N_t
AR1(model_struct)
ab3_step(model_struct)
X_array[:, i+1] = X[:]
end
model_signal = X_array[2,1:5:N_t].-mean(X_array[2,1:5:N_t])
data_signal = data[2,:].-mean(data[2,:])
ff = fftshift(fft(model_signal))[1321:1681]
ffd = fftshift(fft(data_signal))[1321:1681]
f = log10.(abs2.(ff))
fd = log10.(abs2.(ffd))
misfit = f .- fd
J0 = J0 + dot(misfit, misfit)
@pack! model_struct = X_array, J0
return J0
end
# Data parameter
step_factor = 5
# Model parameters
N_t = 500000
dt = 0.001
K = 4
h = 5.
F = 20.
b = 25.
c = 25.
b0 = 5.4
b1 = 4.4
b2 = -4.3
b3 = 1.2
phi = 0.19
sig0 = 1.5
sig1 = 1.1
X = copy(Xd[:,1])
const data = X_1011'
# Initial conditions
J0 = 0.0
function adjoint_J0(model_struct)
@unpack N_t, dt, K = model_struct
@unpack data, step_factor, J0 = model_struct
adjoint_vars = init_cost(
N_t, dt,
K, h, F, b, c,
b0, b1, b2, b3, phi, sig0, sig1,
zeros(Float64, K),
zeros(Float64, K, N_t+1), step_factor, 0.0)
autodiff(Reverse, forward_model_cost, Duplicated(model_struct, adjoint_vars))
return adjoint_vars
end
J0_model = init_cost(N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, data,
step_factor, J0)
J0_for_adjoint = init_cost(N_t, dt, K, phi, sig0, sig1, h, F, b, c, b0, b1, b2, b3, X, data,
step_factor, J0)
forward_model_cost(J0_model)
J0_adj = adjoint_J0(J0_for_adjoint)
print(J0_adj.b0)The error it produces is: The code runs fine (no errors) if the following block is removed: ff = fftshift(fft(model_signal))[1321:1681]
ffd = fftshift(fft(data_signal))[1321:1681]
f = log10.(abs2.(ff))
fd = log10.(abs2.(ffd))
misfit = f .- fd
J0 = J0 + dot(misfit, misfit)I'm not sure how to produce the error log you might be looking for. Apologies for the length of the code and my unfamiliarity with the error log -- I can probably come up with a simpler example to produce this error if needed. |
|
@trostelm can you share the "data.h5" file, or update the code to not need it [e.g. make it use zeros or an appropriate size or something] |
|
Of course, I've updated the code above. Thanks!! |
|
I believe the problem is not # ╔═╡ 964afc32-481e-11ef-0ce7-ddc5430a25d2
using Enzyme
# ╔═╡ 477e8d70-716c-4b7c-90d6-4e1f6fab6d5a
using FFTW
# ╔═╡ e87d8cd9-ea20-4727-8960-3c3aa7983c5b
using Zygote
# ╔═╡ 1eb7c19c-7ff4-402b-beb6-0df22e28f9de
fftest(x) = sum(abs2, x.^2 .- fftshift(x))
# ╔═╡ 3d8f7184-0f5f-4d19-9aac-81038ff269fa
ll = [1.0,2.0, 3.0, 4.0, 5.0]
# ╔═╡ 294b28b5-2cce-4bfb-8d1a-b9a3809bfc03
dll = make_zero(ll)
# ╔═╡ b0d7a3bb-c87d-43d0-b8ff-0d00ec0927af
Enzyme.autodiff(Enzyme.ReverseWithPrimal, fftest, Duplicated(ll, dll))
# ╔═╡ 16bea1a4-c6a5-441a-baa2-31b0127e1a85
dll
# ╔═╡ 218eb198-015b-4f35-bea5-1ccae095dda7
Zygote.gradient(fftest, ll)
|
using Enzyme, UnPack, LinearAlgebra, Random
Xd = ones(4,200000)
t = range(start=5,step=0.005,length=200000)
dt_data = 0.005
N_t_data = Int(500/dt_data)
X_1011 = transpose(Xd[:, 1:N_t_data])
ts = range(; start=0, step = dt_data, length=N_t_data)
ts = ts*2
# Y variables parameterization
function U_det(Xk, b0, b1, b2, b3)
U = b0 + b1*Xk + b2/10.0*Xk^2 + b3/100.0*Xk^3
return U
end
function U_p(Xk, b0, b1, b2, b3, e_tk)
U = U_det(Xk, b0, b1, b2, b3) + e_tk
return U
end
# dX rhs
function dXrhs(k, X, b0, b1, b2, b3, e_t)
dXk = - X[k - 1] * (X[k - 2] - X[k + 1]) - X[k] + F - U_p(X[k], b0, b1, b2, b3, e_t[k])
return dXk
end
# Euler step
function euler_step(model_struct)
@unpack N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2, e_t, e_tm1, sig_e_t, sig_e_tm1, X_array = model_struct
# Compute derivatives
dX[1] = - X[K] * (X[K - 1] - X[2]) - X[1] + F - U_p(X[1], b0, b1, b2, b3, e_t[1])
dX[2] = - X[1] * (X[K] - X[3]) - X[2] + F - U_p(X[2], b0, b1, b2, b3, e_t[2])
dX[K] = - X[K - 1] * (X[K - 2] - X[1]) - X[K] + F - U_p(X[K], b0, b1, b2, b3, e_t[K])
# Update X
X[1] += dt * dX[1]
X[2] += dt * dX[2]
X[K] += dt * dX[3]
for k = 3:K-1
dX[k] = dXrhs(k, X, b0, b1, b2, b3, e_t)
X[k] += dt * dX[k]
end
# Update dXs
dXm1[:] = dX[:]
dXm2[:] = dXm1[:]
return nothing
end
# AB-3 rhs
function adams_bashforth3(x, dt, dx, dxm1, dxm2)
rhs = x + dt * ((23/12) * dx
- (16/12) * dxm1
+ (5/12) * dxm2)
return rhs
end
# AB-3 step
function ab3_step(model_struct)
@unpack N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2, e_t, e_tm1, sig_e_t, sig_e_tm1, X_array = model_struct
# Compute derivatives
dX[1] = - X[K] * (X[K - 1] - X[2]) - X[1] + F - U_p(X[1], b0, b1, b2, b3, e_t[1])
# Update X
X[1] = dX[1]
return nothing
end
mutable struct cost
# Model set-up parameters
N_t::Int64
dt::Float64
K::Int64
h::Float64
F::Float64
b::Float64
c::Float64
# Forecast control parameters
b0::Float64
b1::Float64
b2::Float64
b3::Float64
phi::Float64
sig0::Float64
sig1::Float64
# Variables
X::Vector{Float64}
dX::Vector{Float64}
dXm1::Vector{Float64}
dXm2::Vector{Float64}
# Random
e_t::Vector{Float64}
e_tm1::Vector{Float64}
sig_e_t::Vector{Float64}
sig_e_tm1::Vector{Float64}
# Data in time array
X_array::Matrix{Float64}
# Data
data::Matrix{Float64}
step_factor::Int64
# Objective function
J0::Float64
end
function init_cost(N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, data, step_factor, J0)
# Memory allocation
dX = zeros(Float64, K)
dXm1 = zeros(Float64, K)
dXm2 = zeros(Float64, K)
e_t = zeros(Float64, K)
e_tm1 = zeros(Float64, K)
sig_e_t = zeros(Float64, K) .+ sig0
sig_e_tm1 = zeros(Float64, K) .+ sig0
X_array = zeros(Float64, K, N_t+1)
# Construct initial L96 object
J0_L96 = cost(N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, dX, dXm1, dXm2,
e_t, e_tm1, sig_e_t, sig_e_tm1, X_array, data, step_factor, J0)
# Set intial conditions in X_array
J0_L96.X_array[:, 1] = J0_L96.X[:]
return J0_L96
end
function forward_model_cost(model_struct)
for i in 1:2
euler_step(model_struct)
end
for i in 2:model_struct.N_t
ab3_step(model_struct)
end
ff = model_struct.X
return dot(ff, ff)
end
# Data parameter
step_factor = 5
# Model parameters
N_t = 500000
dt = 0.001
K = 4
h = 5.
F = 20.
b = 25.
c = 25.
b0 = 5.4
b1 = 4.4
b2 = -4.3
b3 = 1.2
phi = 0.19
sig0 = 1.5
sig1 = 1.1
X = copy(Xd[:,1])
const data = X_1011'
# Initial conditions
J0 = 0.0
function adjoint_J0(model_struct)
@unpack N_t, dt, K = model_struct
@unpack data, step_factor, J0 = model_struct
adjoint_vars = init_cost(
N_t, dt,
K, h, F, b, c,
b0, b1, b2, b3, phi, sig0, sig1,
zeros(Float64, K),
zeros(Float64, K, N_t+1), step_factor, 0.0)
autodiff(Reverse, forward_model_cost, Duplicated(model_struct, adjoint_vars))
return adjoint_vars
end
J0_model = init_cost(N_t, dt, K, h, F, b, c, b0, b1, b2, b3, phi, sig0, sig1, X, data,
step_factor, J0)
J0_for_adjoint = init_cost(N_t, dt, K, phi, sig0, sig1, h, F, b, c, b0, b1, b2, b3, X, data,
step_factor, J0)
forward_model_cost(J0_model)
J0_adj = adjoint_J0(J0_for_adjoint)
print(J0_adj.b0) |
using Enzyme, LinearAlgebra, Random
Enzyme.API.printall!(true)
Enzyme.Compiler.DumpPostOpt[] = true
# Y variables parameterization
function U_det(Xk, b0, b1, b2, b3)
U = b2 + Xk + b3 + Xk
return U
end
function U_p(Xk, b0, b1, b2, b3, e_tk)
U = U_det(Xk, b0, b1, b2, b3)
return U
end
function euler_step(model_struct)
X = model_struct.X
dX = model_struct.dX
dXm1 = model_struct.dXm1
@inbounds X[1] = 0.1 * @inbounds dX[1]
for k = 3:3
X[k] += X[k] + F
end
# Update dXs
@inbounds dXm1[:] = dX[:]
return nothing
end
# AB-3 step
function ab3_step(model_struct)
b2 = model_struct.b2
b3 = model_struct.b3
e_t = model_struct.e_t
dX = model_struct.dX
X = model_struct.X
# Compute derivatives
X1 = @inbounds X[1]
X2 = @inbounds X[2]
e1 = @inbounds e_t[1]
@inbounds dX[1] = X2 * X2 + F - U_p(X1, 1.0, 1.0, b2, b3, e1)
# Update X
@inbounds X[1] = @inbounds dX[1]
return nothing
end
mutable struct cost
# Forecast control parameters
b2::Float64
b3::Float64
# Variables
X::Vector{Float64}
dX::Vector{Float64}
dXm1::Vector{Float64}
# Random
e_t::Vector{Float64}
# Data in time array
X_array::Matrix{Float64}
end
function init_cost(N_t, K, b2, b3, X)
# Memory allocation
dX = zeros(Float64, K)
dXm1 = zeros(Float64, K)
e_t = zeros(Float64, K)
X_array = zeros(Float64, K, N_t+1)
# Construct initial L96 object
J0_L96 = cost(b2, b3, X, dX, dXm1,
e_t, X_array)
return J0_L96
end
function forward_model_cost(model_struct, N)
for i in 1:2
euler_step(model_struct)
end
for i in 2:N
ab3_step(model_struct)
end
ff = model_struct.X
return @inbounds ff[1]
end
# Model parameters
N_t = 500000
K = 4
F = 20.
b2 = -4.3
b3 = 1.2
X = ones(4)
J0_model = init_cost(N_t, K, b2, b3, X)
forward_model_cost(J0_model, N_t)
model_struct = init_cost(N_t, K, b2, b3, X)
adjoint_vars = init_cost(
N_t,
K,
b2, b3,
zeros(Float64, K))
autodiff(Reverse, forward_model_cost, Duplicated(model_struct, adjoint_vars), Const(N_t))resulting in |
using Enzyme, LinearAlgebra, Random
Enzyme.API.printall!(true)
Enzyme.Compiler.DumpPostOpt[] = true
# Y variables parameterization
function U_det(Xk, b2, b3)
U = b2 + Xk + b3 + Xk
return U
end
function U_p(Xk, b0, b1, b2, b3, e_tk)
U = U_det(Xk, b2, b3)
return U
end
function euler_step(X, dX, dXm1)
@inbounds X[1] = 0.1 * @inbounds dX[1]
for k = 3:3
Xk = @inbounds X[k]
X[k] += Xk + F
end
# Update dXs
@inbounds dXm1[:] = dX[:]
return nothing
end
# AB-3 step
function ab3_step(X, dX, b2, b3, e1)
# Compute derivatives
X1 = @inbounds X[1]
X2 = @inbounds X[2]
@inbounds dX[1] = X2 * X2 + F - U_p(X1, 1.0, 1.0, b2, b3, e1)
# Update X
@inbounds X[1] = @inbounds dX[1]
return nothing
end
function forward_model_cost(dXm1, N, b2, b3, e1, X, dX)
# dXm1 = model_struct.dXm1
for i in 1:2
euler_step(X, dX, dXm1)
end
for i in 2:N
ab3_step(X, dX, b2, b3, e1)
end
return @inbounds X[1]
end
# Model parameters
N_t = 500000
K = 4
F = 20.
X = ones(4)
dXm1 = zeros(Float64, K)
dX = zeros(Float64, K)
forward_model_cost(dXm1, N_t, 1.2, -4.3, 1.2, X, dX)
autodiff(Reverse,
forward_model_cost,
Duplicated(dXm1, zero(dXm1)),
Const(N_t), Const(1.2), Const(-4.3), Const(1.2),
Duplicated(X, zero(X)),
Duplicated(dX, zero(dX))
) |
cc @gbaraldi |
after simplification :
; Function Attrs: mustprogress nofree readonly willreturn
define "enzyme_type"="{[-1]:Float@double}" double @preprocess_julia___2607(double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1) local_unnamed_addr #4 !dbg !23 {
top:
%2 = call {}*** @julia.get_pgcstack() #5
call void @llvm.dbg.value(metadata double %0, metadata !26, metadata !DIExpression()) #5, !dbg !28
call void @llvm.dbg.value(metadata double %1, metadata !27, metadata !DIExpression()) #5, !dbg !28
%ptls_field3 = getelementptr inbounds {}**, {}*** %2, i64 2
%3 = bitcast {}*** %ptls_field3 to i64***
%ptls_load45 = load i64**, i64*** %3, align 8, !tbaa !16
%4 = getelementptr inbounds i64*, i64** %ptls_load45, i64 2
%safepoint = load i64*, i64** %4, align 8, !tbaa !20
fence syncscope("singlethread") seq_cst
call void @julia.safepoint(i64* %safepoint) #5, !dbg !29
fence syncscope("singlethread") seq_cst
%5 = fadd double %0, %1, !dbg !29
ret double %5, !dbg !29
}
; Function Attrs: mustprogress nofree readonly willreturn
define internal double @augmented_julia___2607(double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1) local_unnamed_addr #4 !dbg !30 {
top:
%2 = alloca double, align 8
%3 = call {}*** @julia.get_pgcstack() #5
%ptls_field3 = getelementptr inbounds {}**, {}*** %3, i64 2
%4 = bitcast {}*** %ptls_field3 to i64***
%ptls_load45 = load i64**, i64*** %4, align 8, !tbaa !16, !alias.scope !35, !noalias !38
%5 = getelementptr inbounds i64*, i64** %ptls_load45, i64 2
%safepoint = load i64*, i64** %5, align 8, !tbaa !20, !alias.scope !40, !noalias !43
fence syncscope("singlethread") seq_cst
call void @julia.safepoint(i64* %safepoint) #5, !dbg !45
fence syncscope("singlethread") seq_cst
%6 = fadd double %0, %1, !dbg !45
store double %6, double* %2, align 8, !dbg !45
%7 = load double, double* %2, align 8, !dbg !45
ret double %7, !dbg !45
}
; Function Attrs: mustprogress nofree readonly willreturn
define internal { double } @diffejulia___2607(double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1, double %differeturn) local_unnamed_addr #4 !dbg !49 {
top:
%"'de" = alloca double, align 8
%2 = getelementptr double, double* %"'de", i64 0
store double 0.000000e+00, double* %2, align 8
%"'de1" = alloca double, align 8
%3 = getelementptr double, double* %"'de1", i64 0
store double 0.000000e+00, double* %3, align 8
%4 = call {}*** @julia.get_pgcstack() #7
br label %inverttop, !dbg !54
inverttop: ; preds = %top
store double %differeturn, double* %"'de", align 8
%5 = load double, double* %"'de", align 8, !dbg !54
store double 0.000000e+00, double* %"'de", align 8, !dbg !54
%6 = load double, double* %"'de1", align 8, !dbg !54
%7 = fadd fast double %6, %5, !dbg !54
store double %7, double* %"'de1", align 8, !dbg !54
fence syncscope("singlethread") seq_cst
fence syncscope("singlethread") seq_cst
%8 = load double, double* %"'de1", align 8
%9 = insertvalue { double } undef, double %8, 0
ret { double } %9
}
; ModuleID = 'start'
source_filename = "start"
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-linux-gnu"
; Function Attrs: inaccessiblemem_or_argmemonly nofree noinline optnone
declare void @julia.safepoint(i64*) local_unnamed_addr #0
; Function Attrs: nofree noinline optnone readonly
define { double } @augmented_julia___2607wrap(double %0, double %1) #1 !dbg !4 {
entry:
%thread_ptr = call i8* asm "movq %fs:0, $0", "=r"() #7
%tls_ppgcstack = getelementptr i8, i8* %thread_ptr, i64 -8
%2 = bitcast i8* %tls_ppgcstack to {}****
%tls_pgcstack = load {}***, {}**** %2, align 8
%ptls_field3.i = getelementptr inbounds {}**, {}*** %tls_pgcstack, i64 2
%3 = bitcast {}*** %ptls_field3.i to i64***
%ptls_load45.i = load i64**, i64*** %3, align 8, !tbaa !8, !alias.scope !12, !noalias !15
%4 = getelementptr inbounds i64*, i64** %ptls_load45.i, i64 2
%safepoint.i = load i64*, i64** %4, align 8, !tbaa !17, !alias.scope !19, !noalias !22
fence syncscope("singlethread") seq_cst
%5 = load volatile i64, i64* %safepoint.i, align 8, !dbg !24
fence syncscope("singlethread") seq_cst
%6 = fadd double %0, %1, !dbg !24
%7 = insertvalue { double } zeroinitializer, double %6, 0
ret { double } %7
}
; Function Attrs: nofree noinline nosync optnone readonly
define [1 x { double }] @diffejulia___2607wrap(double %0, double %1, double %2) #2 !dbg !35 {
entry:
fence syncscope("singlethread") seq_cst
%.unpack1 = insertvalue { double } zeroinitializer, double %2, 0
%3 = insertvalue [1 x { double }] zeroinitializer, { double } %.unpack1, 0
ret [1 x { double }] %3
}
; Function Attrs: argmemonly nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #3
; Function Attrs: argmemonly nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture) #3
; Function Attrs: inaccessiblemem_or_argmemonly noinline optnone
declare void @ijl_gc_queue_root({} addrspace(10)*) #4
; Function Attrs: noinline optnone allocsize(2)
declare noalias nonnull {} addrspace(10)* @ijl_gc_pool_alloc(i8*, i32, i32) #5
; Function Attrs: noinline optnone allocsize(1)
declare noalias nonnull {} addrspace(10)* @ijl_gc_big_alloc(i8*, i64) #6
; Function Attrs: noinline optnone allocsize(1)
declare noalias nonnull {} addrspace(10)* @ijl_gc_alloc_typed(i8*, i64, i8*) #6
attributes #0 = { inaccessiblemem_or_argmemonly nofree noinline optnone "enzyme_inactive" "enzyme_no_escaping_allocation" "enzymejl_world"="31477" }
attributes #1 = { nofree noinline optnone readonly "enzymejl_world"="31477" }
attributes #2 = { nofree noinline nosync optnone readonly "enzymejl_world"="31477" }
attributes #3 = { argmemonly nocallback nofree noinline nosync nounwind optnone willreturn }
attributes #4 = { inaccessiblemem_or_argmemonly noinline optnone }
attributes #5 = { noinline optnone allocsize(2) }
attributes #6 = { noinline optnone allocsize(1) }
attributes #7 = { nounwind }
!llvm.module.flags = !{!0, !1}
!llvm.dbg.cu = !{!2}
!0 = !{i32 2, !"Dwarf Version", i32 4}
!1 = !{i32 2, !"Debug Info Version", i32 3}
!2 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, nameTableKind: None)
!3 = !DIFile(filename: "julia", directory: ".")
!4 = distinct !DISubprogram(name: "augmented_julia___2607wrap", linkageName: "augmented_julia___2607wrap", scope: null, file: !5, type: !6, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !7)
!5 = !DIFile(filename: "float.jl", directory: ".")
!6 = !DISubroutineType(types: !7)
!7 = !{}
!8 = !{!9, !9, i64 0}
!9 = !{!"jtbaa_gcframe", !10, i64 0}
!10 = !{!"jtbaa", !11, i64 0}
!11 = !{!"jtbaa"}
!12 = !{!13}
!13 = distinct !{!13, !14, !"primal"}
!14 = distinct !{!14, !" diff: %"}
!15 = !{!16}
!16 = distinct !{!16, !14, !"shadow_0"}
!17 = !{!18, !18, i64 0, i64 0}
!18 = !{!"jtbaa_const", !10, i64 0}
!19 = !{!20}
!20 = distinct !{!20, !21, !"primal"}
!21 = distinct !{!21, !" diff: %ptls_load45"}
!22 = !{!23}
!23 = distinct !{!23, !21, !"shadow_0"}
!24 = !DILocation(line: 409, scope: !25, inlinedAt: !34)
!25 = distinct !DISubprogram(name: "+", linkageName: "julia_+_2607", scope: null, file: !5, line: 409, type: !26, scopeLine: 409, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !30)
!26 = !DISubroutineType(types: !27)
!27 = !{!28, !29, !28, !28}
!28 = !DIBasicType(name: "Float64", size: 64, encoding: DW_ATE_unsigned)
!29 = !DICompositeType(tag: DW_TAG_structure_type, name: "#+", align: 8, elements: !7, runtimeLang: DW_LANG_Julia, identifier: "127923598536992")
!30 = !{!31, !32, !33}
!31 = !DILocalVariable(name: "#self#", arg: 1, scope: !25, file: !5, line: 409, type: !29)
!32 = !DILocalVariable(name: "x", arg: 2, scope: !25, file: !5, line: 409, type: !28)
!33 = !DILocalVariable(name: "y", arg: 3, scope: !25, file: !5, line: 409, type: !28)
!34 = distinct !DILocation(line: 0, scope: !4)
!35 = distinct !DISubprogram(name: "diffejulia___2607wrap", linkageName: "diffejulia___2607wrap", scope: null, file: !5, type: !6, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !7)
after simplification :
; Function Attrs: mustprogress nofree readonly willreturn
define "enzyme_type"="{[-1]:Float@double}" double @preprocess_julia___2738(double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1) local_unnamed_addr #4 !dbg !23 {
top:
%2 = call {}*** @julia.get_pgcstack() #5
call void @llvm.dbg.value(metadata double %0, metadata !26, metadata !DIExpression()) #5, !dbg !28
call void @llvm.dbg.value(metadata double %1, metadata !27, metadata !DIExpression()) #5, !dbg !28
%ptls_field3 = getelementptr inbounds {}**, {}*** %2, i64 2
%3 = bitcast {}*** %ptls_field3 to i64***
%ptls_load45 = load i64**, i64*** %3, align 8, !tbaa !16
%4 = getelementptr inbounds i64*, i64** %ptls_load45, i64 2
%safepoint = load i64*, i64** %4, align 8, !tbaa !20
fence syncscope("singlethread") seq_cst
call void @julia.safepoint(i64* %safepoint) #5, !dbg !29
fence syncscope("singlethread") seq_cst
%5 = fadd double %0, %1, !dbg !29
ret double %5, !dbg !29
}
; Function Attrs: mustprogress nofree readonly willreturn
define internal double @augmented_julia___2738(double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1) local_unnamed_addr #4 !dbg !30 {
top:
%2 = alloca double, align 8
%3 = call {}*** @julia.get_pgcstack() #5
%ptls_field3 = getelementptr inbounds {}**, {}*** %3, i64 2
%4 = bitcast {}*** %ptls_field3 to i64***
%ptls_load45 = load i64**, i64*** %4, align 8, !tbaa !16, !alias.scope !35, !noalias !38
%5 = getelementptr inbounds i64*, i64** %ptls_load45, i64 2
%safepoint = load i64*, i64** %5, align 8, !tbaa !20, !alias.scope !40, !noalias !43
fence syncscope("singlethread") seq_cst
call void @julia.safepoint(i64* %safepoint) #5, !dbg !45
fence syncscope("singlethread") seq_cst
%6 = fadd double %0, %1, !dbg !45
store double %6, double* %2, align 8, !dbg !45
%7 = load double, double* %2, align 8, !dbg !45
ret double %7, !dbg !45
}
; Function Attrs: mustprogress nofree readonly willreturn
define internal { double, double } @diffejulia___2738(double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1, double %differeturn) local_unnamed_addr #4 !dbg !49 {
top:
%"'de" = alloca double, align 8
%2 = getelementptr double, double* %"'de", i64 0
store double 0.000000e+00, double* %2, align 8
%"'de1" = alloca double, align 8
%3 = getelementptr double, double* %"'de1", i64 0
store double 0.000000e+00, double* %3, align 8
%"'de2" = alloca double, align 8
%4 = getelementptr double, double* %"'de2", i64 0
store double 0.000000e+00, double* %4, align 8
%5 = call {}*** @julia.get_pgcstack() #7
br label %inverttop, !dbg !54
inverttop: ; preds = %top
store double %differeturn, double* %"'de", align 8
%6 = load double, double* %"'de", align 8, !dbg !54
store double 0.000000e+00, double* %"'de", align 8, !dbg !54
%7 = load double, double* %"'de1", align 8, !dbg !54
%8 = fadd fast double %7, %6, !dbg !54
store double %8, double* %"'de1", align 8, !dbg !54
%9 = load double, double* %"'de2", align 8, !dbg !54
%10 = fadd fast double %9, %6, !dbg !54
store double %10, double* %"'de2", align 8, !dbg !54
fence syncscope("singlethread") seq_cst
fence syncscope("singlethread") seq_cst
%11 = load double, double* %"'de1", align 8
%12 = load double, double* %"'de2", align 8
%13 = insertvalue { double, double } undef, double %11, 0
%14 = insertvalue { double, double } %13, double %12, 1
ret { double, double } %14
}
; ModuleID = 'start'
source_filename = "start"
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-linux-gnu"
; Function Attrs: inaccessiblemem_or_argmemonly nofree noinline optnone
declare void @julia.safepoint(i64*) local_unnamed_addr #0
; Function Attrs: nofree noinline optnone readonly
define { double } @augmented_julia___2738wrap(double %0, double %1) #1 !dbg !4 {
entry:
%thread_ptr = call i8* asm "movq %fs:0, $0", "=r"() #7
%tls_ppgcstack = getelementptr i8, i8* %thread_ptr, i64 -8
%2 = bitcast i8* %tls_ppgcstack to {}****
%tls_pgcstack = load {}***, {}**** %2, align 8
%ptls_field3.i = getelementptr inbounds {}**, {}*** %tls_pgcstack, i64 2
%3 = bitcast {}*** %ptls_field3.i to i64***
%ptls_load45.i = load i64**, i64*** %3, align 8, !tbaa !8, !alias.scope !12, !noalias !15
%4 = getelementptr inbounds i64*, i64** %ptls_load45.i, i64 2
%safepoint.i = load i64*, i64** %4, align 8, !tbaa !17, !alias.scope !19, !noalias !22
fence syncscope("singlethread") seq_cst
%5 = load volatile i64, i64* %safepoint.i, align 8, !dbg !24
fence syncscope("singlethread") seq_cst
%6 = fadd double %0, %1, !dbg !24
%7 = insertvalue { double } zeroinitializer, double %6, 0
ret { double } %7
}
; Function Attrs: nofree noinline nosync optnone readonly
define [1 x [2 x double]] @diffejulia___2738wrap(double %0, double %1, double %2) #2 !dbg !35 {
entry:
fence syncscope("singlethread") seq_cst
%3 = insertvalue [2 x double] zeroinitializer, double %2, 0
%.unpack3 = insertvalue [2 x double] %3, double %2, 1
%4 = insertvalue [1 x [2 x double]] zeroinitializer, [2 x double] %.unpack3, 0
ret [1 x [2 x double]] %4
}
; Function Attrs: argmemonly nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #3
; Function Attrs: argmemonly nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture) #3
; Function Attrs: inaccessiblemem_or_argmemonly noinline optnone
declare void @ijl_gc_queue_root({} addrspace(10)*) #4
; Function Attrs: noinline optnone allocsize(2)
declare noalias nonnull {} addrspace(10)* @ijl_gc_pool_alloc(i8*, i32, i32) #5
; Function Attrs: noinline optnone allocsize(1)
declare noalias nonnull {} addrspace(10)* @ijl_gc_big_alloc(i8*, i64) #6
; Function Attrs: noinline optnone allocsize(1)
declare noalias nonnull {} addrspace(10)* @ijl_gc_alloc_typed(i8*, i64, i8*) #6
attributes #0 = { inaccessiblemem_or_argmemonly nofree noinline optnone "enzyme_inactive" "enzyme_no_escaping_allocation" "enzymejl_world"="31477" }
attributes #1 = { nofree noinline optnone readonly "enzymejl_world"="31477" }
attributes #2 = { nofree noinline nosync optnone readonly "enzymejl_world"="31477" }
attributes #3 = { argmemonly nocallback nofree noinline nosync nounwind optnone willreturn }
attributes #4 = { inaccessiblemem_or_argmemonly noinline optnone }
attributes #5 = { noinline optnone allocsize(2) }
attributes #6 = { noinline optnone allocsize(1) }
attributes #7 = { nounwind }
!llvm.module.flags = !{!0, !1}
!llvm.dbg.cu = !{!2}
!0 = !{i32 2, !"Dwarf Version", i32 4}
!1 = !{i32 2, !"Debug Info Version", i32 3}
!2 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, nameTableKind: None)
!3 = !DIFile(filename: "julia", directory: ".")
!4 = distinct !DISubprogram(name: "augmented_julia___2738wrap", linkageName: "augmented_julia___2738wrap", scope: null, file: !5, type: !6, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !7)
!5 = !DIFile(filename: "float.jl", directory: ".")
!6 = !DISubroutineType(types: !7)
!7 = !{}
!8 = !{!9, !9, i64 0}
!9 = !{!"jtbaa_gcframe", !10, i64 0}
!10 = !{!"jtbaa", !11, i64 0}
!11 = !{!"jtbaa"}
!12 = !{!13}
!13 = distinct !{!13, !14, !"primal"}
!14 = distinct !{!14, !" diff: %"}
!15 = !{!16}
!16 = distinct !{!16, !14, !"shadow_0"}
!17 = !{!18, !18, i64 0, i64 0}
!18 = !{!"jtbaa_const", !10, i64 0}
!19 = !{!20}
!20 = distinct !{!20, !21, !"primal"}
!21 = distinct !{!21, !" diff: %ptls_load45"}
!22 = !{!23}
!23 = distinct !{!23, !21, !"shadow_0"}
!24 = !DILocation(line: 409, scope: !25, inlinedAt: !34)
!25 = distinct !DISubprogram(name: "+", linkageName: "julia_+_2738", scope: null, file: !5, line: 409, type: !26, scopeLine: 409, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !30)
!26 = !DISubroutineType(types: !27)
!27 = !{!28, !29, !28, !28}
!28 = !DIBasicType(name: "Float64", size: 64, encoding: DW_ATE_unsigned)
!29 = !DICompositeType(tag: DW_TAG_structure_type, name: "#+", align: 8, elements: !7, runtimeLang: DW_LANG_Julia, identifier: "127923598536992")
!30 = !{!31, !32, !33}
!31 = !DILocalVariable(name: "#self#", arg: 1, scope: !25, file: !5, line: 409, type: !29)
!32 = !DILocalVariable(name: "x", arg: 2, scope: !25, file: !5, line: 409, type: !28)
!33 = !DILocalVariable(name: "y", arg: 3, scope: !25, file: !5, line: 409, type: !28)
!34 = distinct !DILocation(line: 0, scope: !4)
!35 = distinct !DISubprogram(name: "diffejulia___2738wrap", linkageName: "diffejulia___2738wrap", scope: null, file: !5, type: !6, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !7)
after simplification :
; Function Attrs: mustprogress willreturn
define noundef nonnull align 16 dereferenceable(40) "enzyme_type"="{[-1]:Pointer, [-1,0]:Pointer, [-1,0,-1]:Float@double, [-1,8]:Integer, [-1,9]:Integer, [-1,10]:Integer, [-1,11]:Integer, [-1,12]:Integer, [-1,13]:Integer, [-1,14]:Integer, [-1,15]:Integer, [-1,16]:Integer, [-1,17]:Integer, [-1,18]:Integer, [-1,19]:Integer, [-1,20]:Integer, [-1,21]:Integer, [-1,22]:Integer, [-1,23]:Integer, [-1,24]:Integer, [-1,25]:Integer, [-1,26]:Integer, [-1,27]:Integer, [-1,28]:Integer, [-1,29]:Integer, [-1,30]:Integer, [-1,31]:Integer, [-1,32]:Integer, [-1,33]:Integer, [-1,34]:Integer, [-1,35]:Integer, [-1,36]:Integer, [-1,37]:Integer, [-1,38]:Integer, [-1,39]:Integer}" {} addrspace(10)* @preprocess_julia_setindex__2765({} addrspace(10)* noundef nonnull returned align 16 dereferenceable(40) "enzyme_type"="{[-1]:Pointer, [-1,0]:Pointer, [-1,0,-1]:Float@double, [-1,8]:Integer, [-1,9]:Integer, [-1,10]:Integer, [-1,11]:Integer, [-1,12]:Integer, [-1,13]:Integer, [-1,14]:Integer, [-1,15]:Integer, [-1,16]:Integer, [-1,17]:Integer, [-1,18]:Integer, [-1,19]:Integer, [-1,20]:Integer, [-1,21]:Integer, [-1,22]:Integer, [-1,23]:Integer, [-1,24]:Integer, [-1,25]:Integer, [-1,26]:Integer, [-1,27]:Integer, [-1,28]:Integer, [-1,29]:Integer, [-1,30]:Integer, [-1,31]:Integer, [-1,32]:Integer, [-1,33]:Integer, [-1,34]:Integer, [-1,35]:Integer, [-1,36]:Integer, [-1,37]:Integer, [-1,38]:Integer, [-1,39]:Integer}" "enzymejl_parmtype"="127923554507456" "enzymejl_parmtype_ref"="2" %0, double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1, i64 signext "enzyme_inactive" "enzyme_type"="{[-1]:Integer}" "enzymejl_parmtype"="127923625262592" "enzymejl_parmtype_ref"="0" %2) local_unnamed_addr #5 !dbg !53 {
top:
%3 = call {}*** @julia.get_pgcstack() #6
call void @llvm.dbg.value(metadata {} addrspace(10)* null, metadata !56, metadata !DIExpression(DW_OP_deref)) #6, !dbg !59
call void @llvm.dbg.value(metadata {} addrspace(10)* %0, metadata !56, metadata !DIExpression(DW_OP_deref)) #6, !dbg !59
call void @llvm.dbg.value(metadata double %1, metadata !57, metadata !DIExpression()) #6, !dbg !59
call void @llvm.dbg.value(metadata i64 %2, metadata !58, metadata !DIExpression()) #6, !dbg !59
%ptls_field3 = getelementptr inbounds {}**, {}*** %3, i64 2
%4 = bitcast {}*** %ptls_field3 to i64***
%ptls_load45 = load i64**, i64*** %4, align 8, !tbaa !23
%5 = getelementptr inbounds i64*, i64** %ptls_load45, i64 2
%safepoint = load i64*, i64** %5, align 8, !tbaa !27
fence syncscope("singlethread") seq_cst
call void @julia.safepoint(i64* %safepoint) #6, !dbg !60
fence syncscope("singlethread") seq_cst
%6 = add i64 %2, -1, !dbg !60
%7 = addrspacecast {} addrspace(10)* %0 to { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(11)*, !dbg !60
%arraylen_ptr = getelementptr inbounds { i8 addrspace(13)*, i64, i16, i16, i32 }, { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(11)* %7, i64 0, i32 1, !dbg !60
%arraylen = load i64, i64 addrspace(11)* %arraylen_ptr, align 8, !dbg !60, !tbaa !30, !range !33, !alias.scope !34, !noalias !37
%inbounds = icmp ult i64 %6, %arraylen, !dbg !60
br i1 %inbounds, label %idxend, label %oob, !dbg !60
oob: ; preds = %top
%errorbox = alloca i64, align 8, !dbg !60
store i64 %2, i64* %errorbox, align 8, !dbg !60, !noalias !61
%8 = addrspacecast {} addrspace(10)* %0 to {} addrspace(12)*, !dbg !60
call void @ijl_bounds_error_ints({} addrspace(12)* noundef %8, i64* noundef nonnull align 8 %errorbox, i64 noundef 1) #7, !dbg !60
unreachable, !dbg !60
idxend: ; preds = %top
%9 = addrspacecast {} addrspace(10)* %0 to double addrspace(13)* addrspace(11)*, !dbg !60
%arrayptr6 = load double addrspace(13)*, double addrspace(13)* addrspace(11)* %9, align 16, !dbg !60, !tbaa !45, !alias.scope !64, !noalias !37, !nonnull !14
%10 = getelementptr inbounds double, double addrspace(13)* %arrayptr6, i64 %6, !dbg !60
store double %1, double addrspace(13)* %10, align 8, !dbg !60, !tbaa !48, !alias.scope !51, !noalias !65
ret {} addrspace(10)* %0, !dbg !60
}
; Function Attrs: mustprogress willreturn
define internal { {} addrspace(10)*, {} addrspace(10)* } @augmented_julia_setindex__2765({} addrspace(10)* noundef nonnull align 16 dereferenceable(40) "enzyme_type"="{[-1]:Pointer, [-1,0]:Pointer, [-1,0,-1]:Float@double, [-1,8]:Integer, [-1,9]:Integer, [-1,10]:Integer, [-1,11]:Integer, [-1,12]:Integer, [-1,13]:Integer, [-1,14]:Integer, [-1,15]:Integer, [-1,16]:Integer, [-1,17]:Integer, [-1,18]:Integer, [-1,19]:Integer, [-1,20]:Integer, [-1,21]:Integer, [-1,22]:Integer, [-1,23]:Integer, [-1,24]:Integer, [-1,25]:Integer, [-1,26]:Integer, [-1,27]:Integer, [-1,28]:Integer, [-1,29]:Integer, [-1,30]:Integer, [-1,31]:Integer, [-1,32]:Integer, [-1,33]:Integer, [-1,34]:Integer, [-1,35]:Integer, [-1,36]:Integer, [-1,37]:Integer, [-1,38]:Integer, [-1,39]:Integer}" "enzymejl_parmtype"="127923554507456" "enzymejl_parmtype_ref"="2" %0, {} addrspace(10)* align 16 "enzyme_type"="{[-1]:Pointer, [-1,0]:Pointer, [-1,0,-1]:Float@double, [-1,8]:Integer, [-1,9]:Integer, [-1,10]:Integer, [-1,11]:Integer, [-1,12]:Integer, [-1,13]:Integer, [-1,14]:Integer, [-1,15]:Integer, [-1,16]:Integer, [-1,17]:Integer, [-1,18]:Integer, [-1,19]:Integer, [-1,20]:Integer, [-1,21]:Integer, [-1,22]:Integer, [-1,23]:Integer, [-1,24]:Integer, [-1,25]:Integer, [-1,26]:Integer, [-1,27]:Integer, [-1,28]:Integer, [-1,29]:Integer, [-1,30]:Integer, [-1,31]:Integer, [-1,32]:Integer, [-1,33]:Integer, [-1,34]:Integer, [-1,35]:Integer, [-1,36]:Integer, [-1,37]:Integer, [-1,38]:Integer, [-1,39]:Integer}" "enzymejl_parmtype"="127923554507456" "enzymejl_parmtype_ref"="2" %"'", double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1, i64 signext "enzyme_inactive" "enzyme_type"="{[-1]:Integer}" "enzymejl_parmtype"="127923625262592" "enzymejl_parmtype_ref"="0" %2) local_unnamed_addr #5 !dbg !66 {
top:
%3 = alloca { {} addrspace(10)*, {} addrspace(10)* }, align 8
%4 = call {}*** @julia.get_pgcstack() #6
%ptls_field3 = getelementptr inbounds {}**, {}*** %4, i64 2
%5 = bitcast {}*** %ptls_field3 to i64***
%ptls_load45 = load i64**, i64*** %5, align 8, !tbaa !23, !alias.scope !72, !noalias !75
%6 = getelementptr inbounds i64*, i64** %ptls_load45, i64 2
%safepoint = load i64*, i64** %6, align 8, !tbaa !27, !alias.scope !77, !noalias !80
fence syncscope("singlethread") seq_cst
call void @julia.safepoint(i64* %safepoint) #6, !dbg !82
fence syncscope("singlethread") seq_cst
%7 = add i64 %2, -1, !dbg !82
%8 = addrspacecast {} addrspace(10)* %0 to { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(11)*, !dbg !82
%arraylen_ptr = getelementptr inbounds { i8 addrspace(13)*, i64, i16, i16, i32 }, { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(11)* %8, i64 0, i32 1, !dbg !82
%arraylen = load i64, i64 addrspace(11)* %arraylen_ptr, align 8, !dbg !82, !tbaa !30, !range !33, !alias.scope !83, !noalias !86
%inbounds = icmp ult i64 %7, %arraylen, !dbg !82
br i1 %inbounds, label %idxend, label %oob, !dbg !82
oob: ; preds = %top
%errorbox = alloca i64, align 8, !dbg !82
store i64 %2, i64* %errorbox, align 8, !dbg !82, !noalias !88
%9 = addrspacecast {} addrspace(10)* %0 to {} addrspace(12)*, !dbg !82
call void @ijl_bounds_error_ints({} addrspace(12)* noundef %9, i64* noundef nonnull align 8 %errorbox, i64 noundef 1) #7, !dbg !82
unreachable, !dbg !82
idxend: ; preds = %top
%"'ipc" = addrspacecast {} addrspace(10)* %"'" to double addrspace(13)* addrspace(11)*, !dbg !82
%10 = addrspacecast {} addrspace(10)* %0 to double addrspace(13)* addrspace(11)*, !dbg !82
%"arrayptr6'ipl" = load double addrspace(13)*, double addrspace(13)* addrspace(11)* %"'ipc", align 16, !dbg !82, !tbaa !45, !alias.scope !91, !noalias !94, !nonnull !14
%arrayptr6 = load double addrspace(13)*, double addrspace(13)* addrspace(11)* %10, align 16, !dbg !82, !tbaa !45, !alias.scope !95, !noalias !86, !nonnull !14
%11 = getelementptr inbounds double, double addrspace(13)* %arrayptr6, i64 %7, !dbg !82
store double %1, double addrspace(13)* %11, align 8, !dbg !82, !tbaa !48, !alias.scope !96, !noalias !99
%12 = getelementptr inbounds { {} addrspace(10)*, {} addrspace(10)* }, { {} addrspace(10)*, {} addrspace(10)* }* %3, i32 0, i32 0, !dbg !82
store {} addrspace(10)* %0, {} addrspace(10)** %12, align 8, !dbg !82
%13 = getelementptr inbounds { {} addrspace(10)*, {} addrspace(10)* }, { {} addrspace(10)*, {} addrspace(10)* }* %3, i32 0, i32 1, !dbg !82
store {} addrspace(10)* %"'", {} addrspace(10)** %13, align 8, !dbg !82
%14 = load { {} addrspace(10)*, {} addrspace(10)* }, { {} addrspace(10)*, {} addrspace(10)* }* %3, align 8, !dbg !82
ret { {} addrspace(10)*, {} addrspace(10)* } %14, !dbg !82
}
; Function Attrs: mustprogress willreturn
define internal { double } @diffejulia_setindex__2765({} addrspace(10)* align 16 dereferenceable(40) "enzyme_type"="{[-1]:Pointer, [-1,0]:Pointer, [-1,0,-1]:Float@double, [-1,8]:Integer, [-1,9]:Integer, [-1,10]:Integer, [-1,11]:Integer, [-1,12]:Integer, [-1,13]:Integer, [-1,14]:Integer, [-1,15]:Integer, [-1,16]:Integer, [-1,17]:Integer, [-1,18]:Integer, [-1,19]:Integer, [-1,20]:Integer, [-1,21]:Integer, [-1,22]:Integer, [-1,23]:Integer, [-1,24]:Integer, [-1,25]:Integer, [-1,26]:Integer, [-1,27]:Integer, [-1,28]:Integer, [-1,29]:Integer, [-1,30]:Integer, [-1,31]:Integer, [-1,32]:Integer, [-1,33]:Integer, [-1,34]:Integer, [-1,35]:Integer, [-1,36]:Integer, [-1,37]:Integer, [-1,38]:Integer, [-1,39]:Integer}" "enzymejl_parmtype"="127923554507456" "enzymejl_parmtype_ref"="2" %0, {} addrspace(10)* align 16 "enzyme_type"="{[-1]:Pointer, [-1,0]:Pointer, [-1,0,-1]:Float@double, [-1,8]:Integer, [-1,9]:Integer, [-1,10]:Integer, [-1,11]:Integer, [-1,12]:Integer, [-1,13]:Integer, [-1,14]:Integer, [-1,15]:Integer, [-1,16]:Integer, [-1,17]:Integer, [-1,18]:Integer, [-1,19]:Integer, [-1,20]:Integer, [-1,21]:Integer, [-1,22]:Integer, [-1,23]:Integer, [-1,24]:Integer, [-1,25]:Integer, [-1,26]:Integer, [-1,27]:Integer, [-1,28]:Integer, [-1,29]:Integer, [-1,30]:Integer, [-1,31]:Integer, [-1,32]:Integer, [-1,33]:Integer, [-1,34]:Integer, [-1,35]:Integer, [-1,36]:Integer, [-1,37]:Integer, [-1,38]:Integer, [-1,39]:Integer}" "enzymejl_parmtype"="127923554507456" "enzymejl_parmtype_ref"="2" %"'", double "enzyme_type"="{[-1]:Float@double}" "enzymejl_parmtype"="127923625261792" "enzymejl_parmtype_ref"="0" %1, i64 signext "enzyme_inactive" "enzyme_type"="{[-1]:Integer}" "enzymejl_parmtype"="127923625262592" "enzymejl_parmtype_ref"="0" %2) local_unnamed_addr #5 !dbg !104 {
top:
%"'de" = alloca double, align 8
%3 = getelementptr double, double* %"'de", i64 0
store double 0.000000e+00, double* %3, align 8
%4 = call {}*** @julia.get_pgcstack() #8
%5 = add i64 %2, -1, !dbg !110
br i1 true, label %idxend, label %oob, !dbg !110
oob: ; preds = %top
unreachable
idxend: ; preds = %top
%"'ipc" = addrspacecast {} addrspace(10)* %"'" to double addrspace(13)* addrspace(11)*, !dbg !110
%"arrayptr6'ipl" = load double addrspace(13)*, double addrspace(13)* addrspace(11)* %"'ipc", align 16, !dbg !110, !tbaa !45, !alias.scope !111, !noalias !114, !nonnull !14
%"'ipg" = getelementptr inbounds double, double addrspace(13)* %"arrayptr6'ipl", i64 %5, !dbg !110
br label %invertidxend, !dbg !110
inverttop: ; preds = %invertidxend
fence syncscope("singlethread") seq_cst
fence syncscope("singlethread") seq_cst
%6 = load double, double* %"'de", align 8
%7 = insertvalue { double } undef, double %6, 0
ret { double } %7
invertidxend: ; preds = %idxend
%8 = load double, double addrspace(13)* %"'ipg", align 8, !dbg !110, !tbaa !48, !alias.scope !116, !noalias !119
store double 0.000000e+00, double addrspace(13)* %"'ipg", align 8, !dbg !110, !tbaa !48, !alias.scope !116, !noalias !119
%9 = load double, double* %"'de", align 8, !dbg !110
%10 = fadd fast double %9, %8, !dbg !110
store double %10, double* %"'de", align 8, !dbg !110
br label %inverttop
}
; ModuleID = 'start'
source_filename = "start"
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-linux-gnu"
; Function Attrs: inaccessiblemem_or_argmemonly nofree noinline optnone
declare void @julia.safepoint(i64*) local_unnamed_addr #0
; Function Attrs: noinline noreturn optnone
declare void @ijl_bounds_error_ints({} addrspace(12)*, i64*, i64) local_unnamed_addr #1
; Function Attrs: noinline optnone
define { {} addrspace(10)*, {} addrspace(10)* } @augmented_julia_setindex__2765wrap({} addrspace(10)* %0, {} addrspace(10)* %1, double %2, i64 %3) #2 !dbg !4 {
entry:
%thread_ptr = call i8* asm "movq %fs:0, $0", "=r"() #9
%tls_ppgcstack = getelementptr i8, i8* %thread_ptr, i64 -8
%4 = bitcast i8* %tls_ppgcstack to {}****
%tls_pgcstack = load {}***, {}**** %4, align 8
%ptls_field3.i = getelementptr inbounds {}**, {}*** %tls_pgcstack, i64 2
%5 = bitcast {}*** %ptls_field3.i to i64***
%ptls_load45.i = load i64**, i64*** %5, align 8, !tbaa !8, !alias.scope !12, !noalias !15
%6 = getelementptr inbounds i64*, i64** %ptls_load45.i, i64 2
%safepoint.i = load i64*, i64** %6, align 8, !tbaa !17, !alias.scope !19, !noalias !22
fence syncscope("singlethread") seq_cst
%7 = load volatile i64, i64* %safepoint.i, align 8, !dbg !24
fence syncscope("singlethread") seq_cst
%8 = add i64 %3, -1, !dbg !24
%9 = bitcast {} addrspace(10)* %0 to { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(10)*, !dbg !24
%10 = addrspacecast { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(10)* %9 to { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(11)*, !dbg !24
%arraylen_ptr.i = getelementptr inbounds { i8 addrspace(13)*, i64, i16, i16, i32 }, { i8 addrspace(13)*, i64, i16, i16, i32 } addrspace(11)* %10, i64 0, i32 1, !dbg !24
%arraylen.i = load i64, i64 addrspace(11)* %arraylen_ptr.i, align 8, !dbg !24, !tbaa !42, !range !45, !alias.scope !46, !noalias !51
%inbounds.i = icmp ult i64 %8, %arraylen.i, !dbg !24
br i1 %inbounds.i, label %augmented_julia_setindex__2765.exit, label %oob.i, !dbg !24
oob.i: ; preds = %entry
%errorbox.i = alloca i64, align 8, !dbg !24
store i64 %3, i64* %errorbox.i, align 8, !dbg !24, !noalias !57
%11 = addrspacecast {} addrspace(10)* %0 to {} addrspace(12)*, !dbg !24
call void @ijl_bounds_error_ints({} addrspace(12)* noundef %11, i64* noundef nonnull align 8 %errorbox.i, i64 noundef 1) #10, !dbg !24
unreachable, !dbg !24
augmented_julia_setindex__2765.exit: ; preds = %entry
%12 = bitcast {} addrspace(10)* %0 to double addrspace(13)* addrspace(10)*, !dbg !24
%13 = addrspacecast double addrspace(13)* addrspace(10)* %12 to double addrspace(13)* addrspace(11)*, !dbg !24
%arrayptr6.i = load double addrspace(13)*, double addrspace(13)* addrspace(11)* %13, align 16, !dbg !24, !tbaa !62, !alias.scope !64, !noalias !51, !nonnull !7
%14 = getelementptr inbounds double, double addrspace(13)* %arrayptr6.i, i64 %8, !dbg !24
store double %2, double addrspace(13)* %14, align 8, !dbg !24, !tbaa !67, !alias.scope !70, !noalias !73
%15 = insertvalue { {} addrspace(10)*, {} addrspace(10)* } zeroinitializer, {} addrspace(10)* %0, 0
%16 = insertvalue { {} addrspace(10)*, {} addrspace(10)* } %15, {} addrspace(10)* %1, 1
ret { {} addrspace(10)*, {} addrspace(10)* } %16
}
; Function Attrs: nofree noinline nosync optnone
define [1 x { double }] @diffejulia_setindex__2765wrap({} addrspace(10)* nocapture nofree readnone %0, {} addrspace(10)* nocapture nofree readonly %1, double %2, i64 %3) #3 !dbg !75 {
entry:
%4 = add i64 %3, -1, !dbg !76
%5 = bitcast {} addrspace(10)* %1 to double addrspace(13)* addrspace(10)*, !dbg !76
%"'ipc.i" = addrspacecast double addrspace(13)* addrspace(10)* %5 to double addrspace(13)* addrspace(11)*, !dbg !76
%"arrayptr6'ipl.i" = load double addrspace(13)*, double addrspace(13)* addrspace(11)* %"'ipc.i", align 16, !dbg !76, !tbaa !62, !alias.scope !84, !noalias !89, !nonnull !7
%"'ipg.i" = getelementptr inbounds double, double addrspace(13)* %"arrayptr6'ipl.i", i64 %4, !dbg !76
%6 = load double, double addrspace(13)* %"'ipg.i", align 8, !dbg !76, !tbaa !67, !alias.scope !91, !noalias !94
store double 0.000000e+00, double addrspace(13)* %"'ipg.i", align 8, !dbg !76, !tbaa !67, !alias.scope !91, !noalias !96
fence syncscope("singlethread") seq_cst
%.unpack1 = insertvalue { double } zeroinitializer, double %6, 0
%7 = insertvalue [1 x { double }] zeroinitializer, { double } %.unpack1, 0
ret [1 x { double }] %7
}
; Function Attrs: argmemonly nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #4
; Function Attrs: argmemonly nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture) #4
; Function Attrs: nocallback nofree noinline nosync nounwind optnone willreturn
declare i8* @llvm.stacksave() #5
; Function Attrs: nocallback nofree noinline nosync nounwind optnone willreturn
declare void @llvm.stackrestore(i8*) #5
; Function Attrs: inaccessiblemem_or_argmemonly noinline optnone
declare void @ijl_gc_queue_root({} addrspace(10)*) #6
; Function Attrs: noinline optnone allocsize(2)
declare noalias nonnull {} addrspace(10)* @ijl_gc_pool_alloc(i8*, i32, i32) #7
; Function Attrs: noinline optnone allocsize(1)
declare noalias nonnull {} addrspace(10)* @ijl_gc_big_alloc(i8*, i64) #8
; Function Attrs: noinline optnone allocsize(1)
declare noalias nonnull {} addrspace(10)* @ijl_gc_alloc_typed(i8*, i64, i8*) #8
attributes #0 = { inaccessiblemem_or_argmemonly nofree noinline optnone "enzyme_inactive" "enzyme_no_escaping_allocation" "enzymejl_world"="31477" }
attributes #1 = { noinline noreturn optnone "enzymejl_world"="31477" }
attributes #2 = { noinline optnone "enzymejl_world"="31477" }
attributes #3 = { nofree noinline nosync optnone "enzymejl_world"="31477" }
attributes #4 = { argmemonly nocallback nofree noinline nosync nounwind optnone willreturn }
attributes #5 = { nocallback nofree noinline nosync nounwind optnone willreturn }
attributes #6 = { inaccessiblemem_or_argmemonly noinline optnone }
attributes #7 = { noinline optnone allocsize(2) }
attributes #8 = { noinline optnone allocsize(1) }
attributes #9 = { nounwind }
attributes #10 = { mustprogress noreturn willreturn }
!llvm.module.flags = !{!0, !1}
!llvm.dbg.cu = !{!2}
!0 = !{i32 2, !"Dwarf Version", i32 4}
!1 = !{i32 2, !"Debug Info Version", i32 3}
!2 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, nameTableKind: None)
!3 = !DIFile(filename: "julia", directory: ".")
!4 = distinct !DISubprogram(name: "augmented_julia_setindex__2765wrap", linkageName: "augmented_julia_setindex__2765wrap", scope: null, file: !5, type: !6, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !7)
!5 = !DIFile(filename: "array.jl", directory: ".")
!6 = !DISubroutineType(types: !7)
!7 = !{}
!8 = !{!9, !9, i64 0}
!9 = !{!"jtbaa_gcframe", !10, i64 0}
!10 = !{!"jtbaa", !11, i64 0}
!11 = !{!"jtbaa"}
!12 = !{!13}
!13 = distinct !{!13, !14, !"primal"}
!14 = distinct !{!14, !" diff: %"}
!15 = !{!16}
!16 = distinct !{!16, !14, !"shadow_0"}
!17 = !{!18, !18, i64 0, i64 0}
!18 = !{!"jtbaa_const", !10, i64 0}
!19 = !{!20}
!20 = distinct !{!20, !21, !"primal"}
!21 = distinct !{!21, !" diff: %ptls_load45"}
!22 = !{!23}
!23 = distinct !{!23, !21, !"shadow_0"}
!24 = !DILocation(line: 1021, scope: !25, inlinedAt: !41)
!25 = distinct !DISubprogram(name: "setindex!", linkageName: "julia_setindex!_2765", scope: null, file: !5, line: 1021, type: !26, scopeLine: 1021, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !36)
!26 = !DISubroutineType(types: !27)
!27 = !{!28, !33, !28, !34, !35}
!28 = !DIDerivedType(tag: DW_TAG_typedef, name: "Array", baseType: !29)
!29 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !30, size: 64, align: 64)
!30 = !DICompositeType(tag: DW_TAG_structure_type, name: "jl_value_t", file: !31, line: 71, align: 64, elements: !32)
!31 = !DIFile(filename: "julia.h", directory: "")
!32 = !{!29}
!33 = !DICompositeType(tag: DW_TAG_structure_type, name: "#setindex!", align: 8, elements: !7, runtimeLang: DW_LANG_Julia, identifier: "127923603541552")
!34 = !DIBasicType(name: "Float64", size: 64, encoding: DW_ATE_unsigned)
!35 = !DIBasicType(name: "Int64", size: 64, encoding: DW_ATE_unsigned)
!36 = !{!37, !38, !39, !40}
!37 = !DILocalVariable(name: "#self#", arg: 1, scope: !25, file: !5, line: 1021, type: !33)
!38 = !DILocalVariable(name: "A", arg: 2, scope: !25, file: !5, line: 1021, type: !28)
!39 = !DILocalVariable(name: "x", arg: 3, scope: !25, file: !5, line: 1021, type: !34)
!40 = !DILocalVariable(name: "i1", arg: 4, scope: !25, file: !5, line: 1021, type: !35)
!41 = distinct !DILocation(line: 0, scope: !4)
!42 = !{!43, !43, i64 0}
!43 = !{!"jtbaa_arraylen", !44, i64 0}
!44 = !{!"jtbaa_array", !10, i64 0}
!45 = !{i64 0, i64 9223372036854775807}
!46 = !{!47, !49}
!47 = distinct !{!47, !48, !"primal"}
!48 = distinct !{!48, !" diff: %"}
!49 = !{!"jnoalias_typemd", !50}
!50 = !{!"jnoalias"}
!51 = !{!52, !53, !54, !55, !56}
!52 = distinct !{!52, !48, !"shadow_0"}
!53 = !{!"jnoalias_gcframe", !50}
!54 = !{!"jnoalias_stack", !50}
!55 = !{!"jnoalias_data", !50}
!56 = !{!"jnoalias_const", !50}
!57 = !{!58, !60}
!58 = distinct !{!58, !59, !"na_addr13"}
!59 = distinct !{!59, !"addr13"}
!60 = distinct !{!60, !61, !"na_addr13"}
!61 = distinct !{!61, !"addr13"}
!62 = !{!63, !63, i64 0}
!63 = !{!"jtbaa_arrayptr", !44, i64 0}
!64 = !{!58, !47, !65, !49}
!65 = distinct !{!65, !66, !"na_addr13"}
!66 = distinct !{!66, !"addr13"}
!67 = !{!68, !68, i64 0}
!68 = !{!"jtbaa_arraybuf", !69, i64 0}
!69 = !{!"jtbaa_data", !10, i64 0}
!70 = !{!71, !55}
!71 = distinct !{!71, !72, !"primal"}
!72 = distinct !{!72, !" diff: %arrayptr6"}
!73 = !{!58, !74, !65, !53, !54, !49, !56}
!74 = distinct !{!74, !72, !"shadow_0"}
!75 = distinct !DISubprogram(name: "diffejulia_setindex__2765wrap", linkageName: "diffejulia_setindex__2765wrap", scope: null, file: !5, type: !6, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !7)
!76 = !DILocation(line: 1021, scope: !77, inlinedAt: !83)
!77 = distinct !DISubprogram(name: "setindex!", linkageName: "julia_setindex!_2765", scope: null, file: !5, line: 1021, type: !26, scopeLine: 1021, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !78)
!78 = !{!79, !80, !81, !82}
!79 = !DILocalVariable(name: "#self#", arg: 1, scope: !77, file: !5, line: 1021, type: !33)
!80 = !DILocalVariable(name: "A", arg: 2, scope: !77, file: !5, line: 1021, type: !28)
!81 = !DILocalVariable(name: "x", arg: 3, scope: !77, file: !5, line: 1021, type: !34)
!82 = !DILocalVariable(name: "i1", arg: 4, scope: !77, file: !5, line: 1021, type: !35)
!83 = distinct !DILocation(line: 0, scope: !75)
!84 = !{!58, !85, !87, !49}
!85 = distinct !{!85, !86, !"shadow_0"}
!86 = distinct !{!86, !" diff: %"}
!87 = distinct !{!87, !88, !"na_addr13"}
!88 = distinct !{!88, !"addr13"}
!89 = !{!90, !53, !54, !55, !56}
!90 = distinct !{!90, !86, !"primal"}
!91 = !{!92, !55}
!92 = distinct !{!92, !93, !"shadow_0"}
!93 = distinct !{!93, !" diff: %arrayptr6"}
!94 = !{!95, !87, !53, !54, !49, !56}
!95 = distinct !{!95, !93, !"primal"}
!96 = !{!58, !95, !87, !53, !54, !49, !56}
GC error (probable corruption)
Allocations: 27089595 (Pool: 27035466; Big: 54129); GC: 39
Base.getindex
thread 0 ptr queue:
~~~~~~~~~~ ptr queue top ~~~~~~~~~~
<?#0x745891cbf8e0::(nil)>
==========
!!! ERROR in jl_ -- ABORTING !!!
==========
Array{Float64, (4,)}[0, 0, 0, 0]
==========
Array{Float64, (4,)}[inf, 0, 0, 0]
==========
<?#0x745891d17a00::(nil)>
==========
~~~~~~~~~~ ptr queue bottom ~~~~~~~~~~
[774624] signal (6.-6): Aborted
in expression starting at /home/wmoses/git/Enzyme.jl/fft.jl:76
pthread_kill at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
raise at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
abort at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
gc_dump_queue_and_abort at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:1822
gc_mark_outrefs at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:2528 [inlined]
gc_mark_loop_serial_ at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:2697
gc_mark_loop_serial at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:2720
gc_mark_loop at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:2901
_jl_gc_collect at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:3234
ijl_gc_collect at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:3531
maybe_collect at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:937
jl_gc_pool_alloc_inner at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:1300
ijl_gc_pool_alloc at /home/wmoses/git/Enzyme.jl/julia10/src/gc.c:1348
ab3_step at /home/wmoses/git/Enzyme.jl/fft.jl:41
forward_model_cost at /home/wmoses/git/Enzyme.jl/fft.jl:57 [inlined]
diffejulia_forward_model_cost_827wrap at /home/wmoses/git/Enzyme.jl/fft.jl:0
macro expansion at /home/wmoses/git/Enzyme.jl/src/compiler.jl:6819 [inlined]
enzyme_call at /home/wmoses/git/Enzyme.jl/src/compiler.jl:6419 [inlined]
CombinedAdjointThunk at /home/wmoses/git/Enzyme.jl/src/compiler.jl:6296 [inlined]
autodiff at /home/wmoses/git/Enzyme.jl/src/Enzyme.jl:314 [inlined]
autodiff at /home/wmoses/git/Enzyme.jl/src/Enzyme.jl:338 [inlined]
autodiff at /home/wmoses/git/Enzyme.jl/src/Enzyme.jl:323
unknown function (ip: 0x745883f1da8a)
_jl_invoke at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:2895
ijl_apply_generic at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:3077
jl_apply at /home/wmoses/git/Enzyme.jl/julia10/src/julia.h:1982
do_call at /home/wmoses/git/Enzyme.jl/julia10/src/interpreter.c:126
eval_value at /home/wmoses/git/Enzyme.jl/julia10/src/interpreter.c:223
eval_stmt_value at /home/wmoses/git/Enzyme.jl/julia10/src/interpreter.c:174
eval_body at /home/wmoses/git/Enzyme.jl/julia10/src/interpreter.c:617
jl_interpret_toplevel_thunk at /home/wmoses/git/Enzyme.jl/julia10/src/interpreter.c:775
jl_toplevel_eval_flex at /home/wmoses/git/Enzyme.jl/julia10/src/toplevel.c:934
jl_toplevel_eval_flex at /home/wmoses/git/Enzyme.jl/julia10/src/toplevel.c:877
ijl_toplevel_eval at /home/wmoses/git/Enzyme.jl/julia10/src/toplevel.c:943
ijl_toplevel_eval_in at /home/wmoses/git/Enzyme.jl/julia10/src/toplevel.c:985
eval at ./boot.jl:385 [inlined]
include_string at ./loading.jl:2076
jl_fptr_args at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:2537
_jl_invoke at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:2876
ijl_apply_generic at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:3077
_include at ./loading.jl:2136
include at ./Base.jl:495
jfptr_include_46707 at /home/wmoses/git/Enzyme.jl/julia10/usr/lib/julia/sys-debug.so (unknown line)
_jl_invoke at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:2876
ijl_apply_generic at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:3077
exec_options at ./client.jl:318
_start at ./client.jl:552
jfptr__start_83017 at /home/wmoses/git/Enzyme.jl/julia10/usr/lib/julia/sys-debug.so (unknown line)
_jl_invoke at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:2876
ijl_apply_generic at /home/wmoses/git/Enzyme.jl/julia10/src/gf.c:3077
jl_apply at /home/wmoses/git/Enzyme.jl/julia10/src/julia.h:1982
true_main at /home/wmoses/git/Enzyme.jl/julia10/src/jlapi.c:582
jl_repl_entrypoint at /home/wmoses/git/Enzyme.jl/julia10/src/jlapi.c:731
jl_load_repl at /home/wmoses/git/Enzyme.jl/julia10/cli/loader_lib.c:568
main at /home/wmoses/git/Enzyme.jl/julia10/cli/loader_exe.c:58
unknown function (ip: 0x74589c829d8f)
__libc_start_main at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
_start at ./julia10/julia (unknown line)
Allocations: 27089595 (Pool: 27035466; Big: 54129); GC: 39
wmoses@beast:~/git/Enzyme.jl (fcc2) $ |
|
Idk why LLC hates us but the module of relevance: |
|
from https://godbolt.org/z/EacKPahbf estimated problem child in 735: |
|
Running Julia alloc opt on https://godbolt.org/z/77f9K43bj breaks |
|
FYI for those following here. This turns out to be a bug in the julia compiler itself (fixed here JuliaLang/julia#55306) |
|
Closing as a relevant issue is open in julia proper -- and unfortauntely there's nothing we can do in this package. |
|
Thanks for looking into this @wsmoses ! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Not sure how to understand this error. This occurs when I use FFTW's fftshift to compute the cost function to be differentiated.
`Enzyme execution failed.
Enzyme: The original primal code hits this error condition, thus differentiating it does not make sense
Stacktrace:
[1] multiple call sites
@ unknown:0
Stacktrace:
[1] throwerr(cstr::Cstring)
@ Enzyme.Compiler ~/.julia/packages/Enzyme/Pljwm/src/compiler.jl:1678
`
The text was updated successfully, but these errors were encountered: