Skip to contents

Changelog

Source: NEWS.md

anvl (development version)

Breaking changes

  • Renamed dim/dims to axis/axes throughout the package (an axis is an index, a dimension is a size); ndims() is now naxes().
  • Renamed the primary array argument of prim_* / nv_* functions from operand to x.
  • Renamed the "xla" backend to "pjrt".
  • xla() has been removed; use jit() instead.
  • nv_rdunif() has been renamed to nv_sample_int(), mirroring R’s sample.int().
  • nv_runif()’s lower/upper arguments are now min/max, nv_rnorm()’s mu/sigma are now mean/sd, and nv_rbinom()’s n is now size, matching the corresponding R functions.

Bug fixes

  • nv_sample_int() (formerly nv_rdunif()) was off by one: the first integer was drawn twice as often as it should have been, and the last integer was never drawn at all.

Features

  • New nv_sample() samples from an arbitrary 1-D population, like R’s sample(). Unlike in R, the population argument is never overloaded with a count – sampling the integers 1 to n is nv_sample_int().
  • nv_rnorm()’s mean and sd are now arrayish, so they may vary across the sample instead of being scalars.
  • Dimension arguments (dim, dims, dimension, permutation) now accept negative values that count from the end, so -1 refers to the last dimension. This works at both layers: in the prim_* primitives (prim_transpose(), prim_concatenate(), prim_reduce_*(), prim_reduce(), prim_cumsum() / prim_cumprod() / prim_cummax() / prim_cummin(), prim_argmax(), prim_argmin(), prim_reverse(), prim_iota(), prim_sort()) and in the nv_* functions built on top of them, including nv_squeeze(), nv_unsqueeze(), nv_select(), nv_top_k(), nv_quantile() and nv_median() (#396).
  • nv_reshape() and prim_reshape() accept a single -1 in shape, whose extent is then inferred from the number of elements of the input, e.g. nv_reshape(x, c(2, -1)) (#396).
  • Reductions now reject dimensions that are out of range for the operand instead of silently ignoring them (#396).
  • New nv_lower_tri() and nv_upper_tri() (with nv_lower_tri_like() / nv_upper_tri_like()) return a boolean triangular mask for a given shape, mirroring base R’s lower.tri() / upper.tri(). As in base R, the main diagonal is excluded by default; pass diagonal = 0L to include it. Use nv_tril() / nv_triu() to zero out a triangle of an existing array.
  • trace_fn() gained an optimize argument controlling which graph optimization passes run on the traced graph. TRUE runs all passes, FALSE (default) runs none, and a character vector (e.g. c("inline_scalars", "remove_unused_constants")) selects a subset. jit() always traces with all passes enabled.
  • New nv_dnorm() computes the normal distribution’s probability density function (or, with log = TRUE, its log-density).
  • New nv_pnorm() computes the normal distribution’s cumulative distribution function (lower_tail = FALSE for the upper tail, log_p = TRUE for the log-probability, staying accurate far into either tail).
  • New nv_qnorm() computes the normal distribution’s quantile function, with the same lower_tail / log_p arguments. log_p = TRUE accepts log-probabilities below the smallest representable p, remaining stable far into tail probabilities.
  • nv_array(), nv_scalar(), as_array(), and the as.integer() / as.double() / as.logical() / as.vector() methods for AnvlArray gained a check argument that opts into scanning for NA values during host -> device and device -> host transfers. See the “Gotchas” vignette.
  • nv_var() and nv_sd() now default to dims = NULL, which reduces over all dimensions and returns a scalar, consistent with the other reductions.
  • Supports 1-3d convolutions.

Performance

  • Most nv_*() API functions are now JIT-compiled internally (via a new @jit roxygen roclet), speeding up eager-mode execution.
  • Tracing (trace_fn()) performance has been improved.
  • Tracing now accumulates primitive calls in a fastmap::fastqueue (amortised-O(1) append) instead of an R list grown with c() (copy-on-modify, O(n^2)). Tracing large unrolled graphs is substantially faster, e.g. ~1.36x for an 8000-op chain, with the gain growing with graph size.
  • StableHLO lowering forwards the trace-time output types to the hlo_* builders (via an output_types argument passed to the lowering rules), so stablehlo skips redundant type inference when lowering the graph.
  • Calling jit()ted functions is now significantly faster.

Bug fixes

  • NULL is now treated as an empty node when flattening and unflattening trees. It contributes no leaves but is preserved structurally, so functions with optional arguments (e.g. function(x, y = NULL)) round-trip correctly.

  • nv_argmax() / nv_argmin() and nv_cummax() / nv_cummin() now break ties order-independently, so they return the same result on GPU as on CPU (#368). nv_argmax() / nv_argmin() prefer the smallest index; nv_cummax() / nv_cummin() prefer the last occurrence.

  • nv_diag() now errors on non-1-D input instead of silently producing an incorrect result.

  • Errors raised while tracing are now phrased in anvl’s own vocabulary (#298). Messages originating in the stablehlo package used the StableHLO spec’s terminology; they now speak of arrays instead of tensors, and of x instead of operand. For example, `operand` must have dtype float became `x` must have dtype float.

  • jit() now rejects static arguments with reference semantics – an environment (and therefore also an R6 or reference class object) or an external pointer, including one nested inside a static list (#17). Such a value can be mutated in place while the compilation cache key stays equal, which silently reused a program compiled from its old contents. Functions and device objects remain valid static values.

  • Errors raised while tracing now speak of arrays instead of tensors (#298). Previously, messages that originated in the stablehlo package used the StableHLO spec’s terminology, e.g. `lhs` and `rhs` must have the same tensor type. x Got tensor<4xi32> and tensor<4xf32>.

anvl 0.3.0

Breaking Changes

New Features

  • On CPU, jitted XLA functions now back every non-aliased output with an R-owned RAWSXP. anvl appends a phantom donated input per unaliased output during lowering, allocates pjrt::pjrt_empty() buffers at execute time, and pjrt migrates the keepalive onto the output XPtr. The output’s host bytes are then managed by R’s GC.
  • Renamed user-facing API functions to match base R names: nv_sine() -> nv_sin(), nv_cosine() -> nv_cos(), nv_ceil() -> nv_ceiling(), nv_cholesky() -> nv_chol(). The corresponding primitives were renamed in step: prim_sine() -> prim_sin(), prim_cosine() -> prim_cos(), prim_cholesky() -> prim_chol().
  • nv_reduce_mean() was renamed to nv_mean().
  • nv_solve() no longer requires a to be symmetric positive-definite as it uses LU instead of Cholesky decomposition. Because of this, it is no longer differentiable, as the reverse rule for LU is not implemented yet.
  • nv_chol() / prim_chol() now default to lower = FALSE (upper-triangular factor), matching base R’s chol(). Previously defaulted to lower = TRUE.

New Features

Linear algebra

Element-wise math

  • New unary primitives and corresponding nv_*() functions: acos, acosh, asin, asinh, atan, atanh, cosh, sinh, digamma, lgamma, polygamma, erf, erf_inv, erfc.
  • New API functions nv_mod() (flooring remainder) and nv_trunc() (truncation toward zero).

Cumulative reductions

  • New primitives and corresponding nv_*() functions: cumsum, cumprod, cummax, cummin. prim_cumprod() does not yet have a reverse rule.

Sorting and searching

Array construction / shape

  • nv_array() gained a byrow argument that fills the array from an R object in row-major order, mirroring matrix(byrow = TRUE) (#165).
  • New nv_matrix(data, nrow, ncol, ...) which works like R’s matrix().
  • New API functions nv_rbind() and nv_cbind() and corresponding rbind() / cbind() generics.
  • New API function nv_flatten() for flattening to 1-D.

Misc

Other

Bug Fixes

  • The overloaded %% operator now calls the new nv_mod() to be consistent with base R.
  • The reverse rule for prim_reduce_prod() no longer produces NaN / Inf gradients when the input contains zeros.
  • The CI now actually runs the torch-comparison tests.
  • nv_runif() not properly respects the lower argument.

anvl 0.2.0

Breaking Changes

  • The package was renamed from anvil to anvl to avoid a conflict with the Bioconductor package AnVIL.
  • AnvilTensor/nv_tensor were renamed to AnvlArray and nv_array to be more in line with R’s array(). Also, nv_aten() was renamed to nv_aval().
  • Subsetting with list() (e.g. x[list(1, 3)]) is no longer supported. Use array() to wrap the indices instead, e.g. x[array(c(1L, 3L))]. This mirrors the input convention used everywhere else in the package.
  • Removed debug mode.
  • Remove NSE support for nvl_if. It now requires passing 0-argument closures as true and false arguments.
  • Primitives renamed from nvl_* to prim_*. The underlying primitive object containing the rules and metadata is now part of the JitPrimitive function via the primitive attribute.

New Features

  • Better composability: jit()ted functions can now be used in other jit()-calls. This is the mechanism underlying the new eager mode.
  • Eager mode was added: This means, you can now do nv_add(1, nv_array(1:2)) and it will actually perform the computation and not only do type inference.
  • An experimental {quickr} backend was added It only runs on CPU for now and supports a subset of available operations. You can enable it via the backend argument in jit() and nv_array() or via the anvl.default_backend option.
  • New primitives:
    • nvl_cholesky() to compute the Cholesky decomposition of a matrix.
    • nvl_triangular_solve() to solve a system of linear equations with a triangular matrix.
  • New API functions (+ corresponding R generic implementations):
  • jit() now accepts integer positions for the static argument.
  • New S3 methods dim(), nrow(), ncol(), and length() for anvl arrays.
  • Printing tensors via nv_print() now also works on GPUs.
  • R vectors of length 1 and arrays are now auto-converted when being passed to jitted functions.
  • Improved device handling in jit()

Performance

  • Many operations are now done asynchronously, which improves performance, especially on GPUs.

Bug Fixes

Documentation

  • New vignette on implementing Gaussian Processes.
  • New vignette on implementing Metropolis-Hastings sampling.

Platform support and installation

  • An installation guide was added.
  • Linux on ARM is now supported (CPU only).
  • To use the CUDA backend, it is now possible to install the cuda12.8 package (see installation guide), which only requires a compatible CUDA driver.

anvl 0.1.0

Initial release