Accelerated short-read alignment, derived from bwa-mem2 — drop-in replacement with correctness fixes, performance improvements, single-binary methylation, and shared-memory index.
See the code
bwa-mem3 is a short-read aligner derived from bwa-mem2, carrying correctness fixes, performance improvements, and new features (methylation alignment, shared-memory index, mimalloc allocator) maintained by Fulcrum Genomics.
Wall-clock speedup of the current release (v0.14.0*) against bwa 0.7.19, bwa-mem2 v2.2.1, and minibwa, on the wgs-5M sample. Cells are stock / --fast.
| arch | wall_s | vs bwa | vs bwa-mem2 | vs minibwa |
|---|---|---|---|---|
| ARM | 42.23 / 18.08 | 5.69x / 13.30x | — | 0.96x / 2.24x |
| x86 | 24.15 / 13.07 | 8.07x / 14.91x | 3.52x / 6.51x | 1.26x / 2.32x |
* v0.14.0, like v0.13.0, is measured with a denser suffix-array index — stride 2 instead of the stock stride 8 — built with bwa-mem3 re-sa -u 1 (or staged with bwa-mem3 shm -u 1), both new in v0.13.0. Output is byte-identical by construction (each added suffix-array sample is computed from the on-disk ones), and measured so: a full-content fgumi compare bams of wgs-5M (10,030,558 records) aligned on the stride-2 index against the stock stride-8 index — v0.13.0 against the output-identical v0.12.0 — found 0 differences on both benchmark hosts, Graviton4 m8g (NEON) and AMD m8a (AVX-512), and the change that added it (#510) found 0 differences on 5M-pair hg38 WGS and WES slices. It costs ~12 GB more memory for the human genome. Both v0.13.0 and v0.14.0 are measured on it in the tables below, so v0.14.0's vs prev is a like-for-like comparison. Every older release and every comparator uses the stock index.
[!TIP] 📈 Full release-history table — every bwa-mem3 release since v0.2.1, full methodology, and version pins.
Click to expand
Graviton4 (m8g, arm64/NEON)
release wall_s vs bwa vs bwa-mem2 vs minibwa vs prev bwa 240.38 1.00x — 0.17x — bwa-mem2 — — — — — minibwa 40.59 5.92x — 1.00x — v0.2.1 120.54 1.99x — 0.34x — v0.2.2 122.73 1.96x — 0.33x 0.982x v0.3.0 102.11 2.35x — 0.40x 1.202x v0.4.0 104.14 2.31x — 0.39x 0.980x v0.5.0 104.47 / 37.23 2.30x / 6.46x — 0.39x / 1.09x 0.997x v0.6.0 97.69 / 36.59 2.46x / 6.57x — 0.42x / 1.11x 1.069x / 1.017x v0.7.0 92.06 / 39.78 2.61x / 6.04x — 0.44x / 1.02x 1.061x / 0.920x v0.8.0 76.34 / 28.33 3.15x / 8.48x — 0.53x / 1.43x 1.206x / 1.404x v0.9.0 76.65 / 28.34 3.14x / 8.48x — 0.53x / 1.43x 0.996x / 1.000x v0.10.0 71.11 / 27.80 3.38x / 8.65x — 0.57x / 1.46x 1.078x / 1.020x v0.11.0 64.32 / 26.30 3.74x / 9.14x — 0.63x / 1.54x 1.106x / 1.057x v0.12.0 58.68 / 19.22 4.10x / 12.51x — 0.69x / 2.11x 1.096x / 1.369x v0.13.0* 55.02 / 18.73 4.37x / 12.84x — 0.74x / 2.17x 1.066x / 1.026x v0.14.0* 42.23 / 18.08 5.69x / 13.30x — 0.96x / 2.24x 1.303x / 1.036x AMD (m8a, x86-64/AVX-512)
release wall_s vs bwa vs bwa-mem2 vs minibwa vs prev bwa 194.94 1.00x 0.44x 0.16x — bwa-mem2 85.04 2.29x 1.00x 0.36x — minibwa 30.36 6.42x 2.80x 1.00x — v0.2.1 56.63 3.44x 1.50x 0.54x 1.502x v0.2.2 53.06 3.67x 1.60x 0.57x 1.067x v0.3.0 51.30 3.80x 1.66x 0.59x 1.034x v0.4.0 51.49 3.79x 1.65x 0.59x 0.996x v0.5.0 51.44 / 22.84 3.79x / 8.53x 1.65x / 3.72x 0.59x / 1.33x 1.001x v0.6.0 50.03 / 22.34 3.90x / 8.73x 1.70x / 3.81x 0.61x / 1.36x 1.028x / 1.023x v0.7.0 48.34 / 22.31 4.03x / 8.74x 1.76x / 3.81x 0.63x / 1.36x 1.035x / 1.001x v0.8.0 43.37 / 18.69 4.49x / 10.43x 1.96x / 4.55x 0.70x / 1.62x 1.115x / 1.194x v0.9.0 43.29 / 18.17 4.50x / 10.73x 1.96x / 4.68x 0.70x / 1.67x 1.002x / 1.028x v0.10.0 42.82 / 18.16 4.55x / 10.73x 1.99x / 4.68x 0.71x / 1.67x 1.011x / 1.001x v0.11.0 41.77 / 17.64 4.67x / 11.05x 2.04x / 4.82x 0.73x / 1.72x 1.025x / 1.030x v0.12.0 32.59 / 14.64 5.98x / 13.31x 2.61x / 5.81x 0.93x / 2.07x 1.282x / 1.205x v0.13.0* 29.83 / 13.65 6.54x / 14.28x 2.85x / 6.23x 1.02x / 2.22x 1.093x / 1.073x v0.14.0* 24.15 / 13.07 8.07x / 14.91x 3.52x / 6.51x 1.26x / 2.32x 1.235x / 1.045x
vs previs the release-over-release speedup (prev_wall / this_wall,>1= faster) vs the previous release on this same host,stock / --fast. The first release's predecessor is upstreambwa-mem2— bwa-mem3 is its successor — so v0.2.1'svs previs its speedup over bwa-mem2 (blank on ARM, where upstream has no build).Version pins:
bwa0.7.19 ·bwa-mem2v2.2.1 ·minibwacommitd6d9f87d(minibwa-0.7). "ARM" = Graviton4 m8g (arm64/NEON, no SMT); "x86" = AMD m8a (x86-64/AVX-512, no SMT). Both are the general-purpose siblings of the c8g/c8a hosts used through v0.12.0 — same CPU family, same core count, no SMT, but 4 GiB/vCPU instead of 2 so every historical arm fits in memory — so absolute times are not comparable with earlier versions of this table, only ratios within it. (The x86 arm replaced an earlier Intel c7i arm, which ran 16 vCPUs over 8 physical cores under 2-way SMT and so wasn't a real core-for-core match for Graviton's 16 real cores.) No ARMbwa-mem2build exists, hence the blank cells there. Every arm for a given arch ran interleaved on one fixed on-demand host — 3 reps each, median wall-clock shown — so these are same-host comparisons, not medians pooled across separate runs.—means the release predates the comparator or predates--fast. * v0.13.0 and v0.14.0 rows use the stride-2 suffix-array index described above; all other rows use the stock stride-8 index. Regenerated at each release; see Benchmarks.
[!WARNING]
--fastis not alignment-identical to the default preset — it trades some sensitivity/specificity at the extremes (repetitive/multi-mapping regions, low-MAPQreads) for the speedup above. See "Three ways to run it" below before switching a production pipeline to it.
--compat, --fastbwa-mem3 has three alignment modes that differ in what alignments come out, not just in speed:
| mode | where reads align | when to use |
|---|---|---|
| plain (default) | bwa-mem2's alignments plus bonafide correctness fixes, with two extra tags (MQ:i, HN:i) and an enriched header. On the cells re-measured for release 0.7.1, the complete alignment-record stream (tags stripped) is byte-identical to bwa-mem2 v2.2.1 on wgs-5M/wes-5M/hic-1M (x86 c6a AVX2, with a c6a/c8g cross-arch check confirming the Arm c8g NEON build matches) — differing only by those additive tags and the header. Separately, a 1.07M-record HG00096 WGS slice shows zero diverging primary alignments (x86, primary-only; not part of the cross-arch or complete-stream checks). | Migrating a pipeline, validating against bwa/bwa-mem2, or any new pipeline. |
--compat=bwa-mem2 / --compat=bwa-mem | Byte-for-byte identical alignment records to a specific upstream (bwa-mem2 v2.2.1 or bwa 0.7.19), @PG excluded and -t/-K matched. The two targets are not interchangeable. | Diff-clean validation against an existing bwa/bwa-mem2 golden. |
--fast | Faster, and not record-compatible with the default: it reshuffles the low-confidence tail (~85% of the reads it re-places had MAPQ 0; the confident MAPQ-60 core moves on ≤0.5%, 0.011% on wgs-5M) while staying accuracy-neutral against golden truth (≤0.02 pp across the WGS and methylation sims). Figures from the benchmark release-validation cells (wgs-5M/wes-5M/panel-twist-5M at 5 M reads, hic-1M/sbx-1M at 1 M) across every SIMD tier (AVX2 c6a, AVX-512 c7a/c7i, NEON c7g/c8g; meth on m7i), each a .4xlarge host at -t 16, -K 160000000. | High-throughput pipelines where you care about the confident, uniquely-mapped calls. |
--compat is mutually exclusive with --fast (and with --meth and --proper-pair-from-emitted). See Alignment modes for the full side-by-side and Equivalence with bwa-mem2 for the field-by-field audit.
By default bwa-mem3 keeps bwa-mem2's command-line defaults, so it drops into an existing pipeline unchanged. For the fastest configuration — and what each recommended deviation from the bwa defaults trades for speed — see Settings profiles: bwa drop-in vs recommended.
Full documentation: https://bwa-mem3.readthedocs.io
The recommended way to install bwa-mem3 is via bioconda:
mamba install -c bioconda bwa-mem3
bwa-mem3 version
Prebuilt packages are available for linux-64, linux-aarch64, and osx-arm64.
git clone --recursive https://github.com/fg-labs/bwa-mem3.git
cd bwa-mem3
make
./bwa-mem3 version
See the installation guide for prerequisites and architecture-specific notes.
MIT. See the License page in the docs.
Please cite the bwa-mem2 paper (Vasimuddin Md et al., IPDPS 2019). See the Citation page for BibTeX.
File issues and pull requests on fg-labs/bwa-mem3.
Accelerated short-read alignment, derived from bwa-mem2 — drop-in replacement with correctness fixes, performance improvements, single-binary methylation, and shared-memory index.
See the code
bwa-mem3 is a short-read aligner derived from bwa-mem2, carrying correctness fixes, performance improvements, and new features (methylation alignment, shared-memory index, mimalloc allocator) maintained by Fulcrum Genomics.
Wall-clock speedup of the current release (v0.14.0*) against bwa 0.7.19, bwa-mem2 v2.2.1, and minibwa, on the wgs-5M sample. Cells are stock / --fast.
| arch | wall_s | vs bwa | vs bwa-mem2 | vs minibwa |
|---|---|---|---|---|
| ARM | 42.23 / 18.08 | 5.69x / 13.30x | — | 0.96x / 2.24x |
| x86 | 24.15 / 13.07 | 8.07x / 14.91x | 3.52x / 6.51x | 1.26x / 2.32x |
* v0.14.0, like v0.13.0, is measured with a denser suffix-array index — stride 2 instead of the stock stride 8 — built with bwa-mem3 re-sa -u 1 (or staged with bwa-mem3 shm -u 1), both new in v0.13.0. Output is byte-identical by construction (each added suffix-array sample is computed from the on-disk ones), and measured so: a full-content fgumi compare bams of wgs-5M (10,030,558 records) aligned on the stride-2 index against the stock stride-8 index — v0.13.0 against the output-identical v0.12.0 — found 0 differences on both benchmark hosts, Graviton4 m8g (NEON) and AMD m8a (AVX-512), and the change that added it (#510) found 0 differences on 5M-pair hg38 WGS and WES slices. It costs ~12 GB more memory for the human genome. Both v0.13.0 and v0.14.0 are measured on it in the tables below, so v0.14.0's vs prev is a like-for-like comparison. Every older release and every comparator uses the stock index.
[!TIP] 📈 Full release-history table — every bwa-mem3 release since v0.2.1, full methodology, and version pins.
Click to expand
Graviton4 (m8g, arm64/NEON)
release wall_s vs bwa vs bwa-mem2 vs minibwa vs prev bwa 240.38 1.00x — 0.17x — bwa-mem2 — — — — — minibwa 40.59 5.92x — 1.00x — v0.2.1 120.54 1.99x — 0.34x — v0.2.2 122.73 1.96x — 0.33x 0.982x v0.3.0 102.11 2.35x — 0.40x 1.202x v0.4.0 104.14 2.31x — 0.39x 0.980x v0.5.0 104.47 / 37.23 2.30x / 6.46x — 0.39x / 1.09x 0.997x v0.6.0 97.69 / 36.59 2.46x / 6.57x — 0.42x / 1.11x 1.069x / 1.017x v0.7.0 92.06 / 39.78 2.61x / 6.04x — 0.44x / 1.02x 1.061x / 0.920x v0.8.0 76.34 / 28.33 3.15x / 8.48x — 0.53x / 1.43x 1.206x / 1.404x v0.9.0 76.65 / 28.34 3.14x / 8.48x — 0.53x / 1.43x 0.996x / 1.000x v0.10.0 71.11 / 27.80 3.38x / 8.65x — 0.57x / 1.46x 1.078x / 1.020x v0.11.0 64.32 / 26.30 3.74x / 9.14x — 0.63x / 1.54x 1.106x / 1.057x v0.12.0 58.68 / 19.22 4.10x / 12.51x — 0.69x / 2.11x 1.096x / 1.369x v0.13.0* 55.02 / 18.73 4.37x / 12.84x — 0.74x / 2.17x 1.066x / 1.026x v0.14.0* 42.23 / 18.08 5.69x / 13.30x — 0.96x / 2.24x 1.303x / 1.036x AMD (m8a, x86-64/AVX-512)
release wall_s vs bwa vs bwa-mem2 vs minibwa vs prev bwa 194.94 1.00x 0.44x 0.16x — bwa-mem2 85.04 2.29x 1.00x 0.36x — minibwa 30.36 6.42x 2.80x 1.00x — v0.2.1 56.63 3.44x 1.50x 0.54x 1.502x v0.2.2 53.06 3.67x 1.60x 0.57x 1.067x v0.3.0 51.30 3.80x 1.66x 0.59x 1.034x v0.4.0 51.49 3.79x 1.65x 0.59x 0.996x v0.5.0 51.44 / 22.84 3.79x / 8.53x 1.65x / 3.72x 0.59x / 1.33x 1.001x v0.6.0 50.03 / 22.34 3.90x / 8.73x 1.70x / 3.81x 0.61x / 1.36x 1.028x / 1.023x v0.7.0 48.34 / 22.31 4.03x / 8.74x 1.76x / 3.81x 0.63x / 1.36x 1.035x / 1.001x v0.8.0 43.37 / 18.69 4.49x / 10.43x 1.96x / 4.55x 0.70x / 1.62x 1.115x / 1.194x v0.9.0 43.29 / 18.17 4.50x / 10.73x 1.96x / 4.68x 0.70x / 1.67x 1.002x / 1.028x v0.10.0 42.82 / 18.16 4.55x / 10.73x 1.99x / 4.68x 0.71x / 1.67x 1.011x / 1.001x v0.11.0 41.77 / 17.64 4.67x / 11.05x 2.04x / 4.82x 0.73x / 1.72x 1.025x / 1.030x v0.12.0 32.59 / 14.64 5.98x / 13.31x 2.61x / 5.81x 0.93x / 2.07x 1.282x / 1.205x v0.13.0* 29.83 / 13.65 6.54x / 14.28x 2.85x / 6.23x 1.02x / 2.22x 1.093x / 1.073x v0.14.0* 24.15 / 13.07 8.07x / 14.91x 3.52x / 6.51x 1.26x / 2.32x 1.235x / 1.045x
vs previs the release-over-release speedup (prev_wall / this_wall,>1= faster) vs the previous release on this same host,stock / --fast. The first release's predecessor is upstreambwa-mem2— bwa-mem3 is its successor — so v0.2.1'svs previs its speedup over bwa-mem2 (blank on ARM, where upstream has no build).Version pins:
bwa0.7.19 ·bwa-mem2v2.2.1 ·minibwacommitd6d9f87d(minibwa-0.7). "ARM" = Graviton4 m8g (arm64/NEON, no SMT); "x86" = AMD m8a (x86-64/AVX-512, no SMT). Both are the general-purpose siblings of the c8g/c8a hosts used through v0.12.0 — same CPU family, same core count, no SMT, but 4 GiB/vCPU instead of 2 so every historical arm fits in memory — so absolute times are not comparable with earlier versions of this table, only ratios within it. (The x86 arm replaced an earlier Intel c7i arm, which ran 16 vCPUs over 8 physical cores under 2-way SMT and so wasn't a real core-for-core match for Graviton's 16 real cores.) No ARMbwa-mem2build exists, hence the blank cells there. Every arm for a given arch ran interleaved on one fixed on-demand host — 3 reps each, median wall-clock shown — so these are same-host comparisons, not medians pooled across separate runs.—means the release predates the comparator or predates--fast. * v0.13.0 and v0.14.0 rows use the stride-2 suffix-array index described above; all other rows use the stock stride-8 index. Regenerated at each release; see Benchmarks.
[!WARNING]
--fastis not alignment-identical to the default preset — it trades some sensitivity/specificity at the extremes (repetitive/multi-mapping regions, low-MAPQreads) for the speedup above. See "Three ways to run it" below before switching a production pipeline to it.
--compat, --fastbwa-mem3 has three alignment modes that differ in what alignments come out, not just in speed:
| mode | where reads align | when to use |
|---|---|---|
| plain (default) | bwa-mem2's alignments plus bonafide correctness fixes, with two extra tags (MQ:i, HN:i) and an enriched header. On the cells re-measured for release 0.7.1, the complete alignment-record stream (tags stripped) is byte-identical to bwa-mem2 v2.2.1 on wgs-5M/wes-5M/hic-1M (x86 c6a AVX2, with a c6a/c8g cross-arch check confirming the Arm c8g NEON build matches) — differing only by those additive tags and the header. Separately, a 1.07M-record HG00096 WGS slice shows zero diverging primary alignments (x86, primary-only; not part of the cross-arch or complete-stream checks). | Migrating a pipeline, validating against bwa/bwa-mem2, or any new pipeline. |
--compat=bwa-mem2 / --compat=bwa-mem | Byte-for-byte identical alignment records to a specific upstream (bwa-mem2 v2.2.1 or bwa 0.7.19), @PG excluded and -t/-K matched. The two targets are not interchangeable. | Diff-clean validation against an existing bwa/bwa-mem2 golden. |
--fast | Faster, and not record-compatible with the default: it reshuffles the low-confidence tail (~85% of the reads it re-places had MAPQ 0; the confident MAPQ-60 core moves on ≤0.5%, 0.011% on wgs-5M) while staying accuracy-neutral against golden truth (≤0.02 pp across the WGS and methylation sims). Figures from the benchmark release-validation cells (wgs-5M/wes-5M/panel-twist-5M at 5 M reads, hic-1M/sbx-1M at 1 M) across every SIMD tier (AVX2 c6a, AVX-512 c7a/c7i, NEON c7g/c8g; meth on m7i), each a .4xlarge host at -t 16, -K 160000000. | High-throughput pipelines where you care about the confident, uniquely-mapped calls. |
--compat is mutually exclusive with --fast (and with --meth and --proper-pair-from-emitted). See Alignment modes for the full side-by-side and Equivalence with bwa-mem2 for the field-by-field audit.
By default bwa-mem3 keeps bwa-mem2's command-line defaults, so it drops into an existing pipeline unchanged. For the fastest configuration — and what each recommended deviation from the bwa defaults trades for speed — see Settings profiles: bwa drop-in vs recommended.
Full documentation: https://bwa-mem3.readthedocs.io
The recommended way to install bwa-mem3 is via bioconda:
mamba install -c bioconda bwa-mem3
bwa-mem3 version
Prebuilt packages are available for linux-64, linux-aarch64, and osx-arm64.
git clone --recursive https://github.com/fg-labs/bwa-mem3.git
cd bwa-mem3
make
./bwa-mem3 version
See the installation guide for prerequisites and architecture-specific notes.
MIT. See the License page in the docs.
Please cite the bwa-mem2 paper (Vasimuddin Md et al., IPDPS 2019). See the Citation page for BibTeX.
File issues and pull requests on fg-labs/bwa-mem3.