libmath: add jets for Chebyshev variants of libmath#63

Open

sigilante wants to merge 28 commits into

mainfrom

sigilante/transcendental-jets

Commits on Jun 11, 2026

math: jet-anchor structure for @rd exp + vere jet mirror

sigilante
and
claude
committed
math: @rd log jet + shared-source harness

sigilante
and
claude
committed

Commits on Jun 12, 2026

math: complete all 15 @rd transcendental jets (bit-exact, verified on-ship)

sigilante
and
claude
committed
math: split vere jet mirror into vere/ (32-bit) + vere64/ (64-bit)

sigilante
and
claude
committed
math: @rs transcendental jets + honor door rounding mode in composites

sigilante
and
claude
committed
math: README — document @rs jets + composite rounding-mode handling

sigilante
and
claude
committed
math: @rh (half-precision) transcendental jets

sigilante
and
claude
committed
math: @rq foundation -- f128 exp core + rounding dispatch + tooling

sigilante
and
claude
committed
math: @rq composites honor the door rounding mode

sigilante
and
claude
committed
math: README -- rationale for why transcendental kernels hardcode round-nearest

sigilante
and
claude
committed
math: pow-n -- reject negative/non-integer; @rh native; @rq guarded

sigilante
and
claude
committed
math: native f16 @rh exp + log (no @rs delegation)

sigilante
and
claude
committed
math: native f16 @rh sin/cos + rh-trig sub-core

sigilante
and
claude
committed
math: native f16 @rh atan + rh-atan sub-core; fix leading-zero hex literals

sigilante
and
claude
committed
math: native f16 @rh asin/acos + rh-ainv sub-core

sigilante
and
claude
committed
math: native f16 @rh tan/atan2/pow/sqt/cbt/log-2/log-10 + lr

sigilante
and
claude
committed
math: native f16 @rh C jets (drop _rh_1/_rh_2 -> _rs_* delegation)

sigilante
and
claude
committed
math: propagate native f16 @rh jets to 32-bit libmath/vere copy

sigilante
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

libmath: add jets for Chebyshev variants of libmath#63

libmath: add jets for Chebyshev variants of libmath#63
sigilante wants to merge 28 commits into
mainfrom
sigilante/transcendental-jets

Commits on Jun 11, 2026

math: jet-anchor structure for @rd exp + vere jet mirror

math: @rd log jet + shared-source harness

Commits on Jun 12, 2026

math: complete all 15 @rd transcendental jets (bit-exact, verified on-ship)

math: split vere jet mirror into vere/ (32-bit) + vere64/ (64-bit)

math: @rs transcendental jets + honor door rounding mode in composites

math: README — document @rs jets + composite rounding-mode handling

math: @rh (half-precision) transcendental jets

math: @rq foundation -- f128 exp core + rounding dispatch + tooling

math: @rq composites honor the door rounding mode

math: README -- rationale for why transcendental kernels hardcode round-nearest

math: pow-n -- reject negative/non-integer; @rh native; @rq guarded

math: native f16 @rh exp + log (no @rs delegation)

math: native f16 @rh sin/cos + rh-trig sub-core

math: native f16 @rh atan + rh-atan sub-core; fix leading-zero hex literals

math: native f16 @rh asin/acos + rh-ainv sub-core

math: native f16 @rh tan/atan2/pow/sqt/cbt/log-2/log-10 + lr

math: native f16 @rh C jets (drop _rh_1/_rh_2 -> _rs_* delegation)

math: propagate native f16 @rh jets to 32-bit libmath/vere copy

Commits on Jun 13, 2026

math: @rh jet wrappers must set softfloat_roundingMode, not _math_rnd

math: faithful @rq exp via fdlibm rational reconstruction

tools: @rq atan/asin/acos faithfulness audit (rq_inv_check.py)

math: sync _rq_exp jet to the faithful fdlibm reconstruction

math: port the 13 remaining @rq C jet cores (algorithm complete)

math: @rq u3 ABI wrappers (15 fns) with softfloat_roundingMode handling

math: register the @rq door jets (135/tree.c, q.h, w.h)

math: propagate @rq jets + registration to 32-bit libmath/vere

docs: README -- all four doors (@rd/@rs/@rh/@rq) now jetted

math: complete @rq door jet hints + document jet-attachment finding