v0.9.0

Indicatrix v0.9.0

Highlights

• Ray-Traced, a new third engine beside Solid and Path-Traced: fast, deterministic, with Fire. You can use it in the live view and for exports, render jobs, remote rendering, the web app and the Android app.
• The backdrop is a lit tray. The surface under the stone is now seen through its windows too, so a window shows what is really behind the stone. White is the new default.
• Optimize and Retarget accept designs from older GemCad files, and Retarget's limits now follow how far the facets turn.
• Web app on the GPU: every Ray-Traced material class and isotropic Path-Traced stones render on the GPU in Chrome and Edge.
• Android: Optimize and Retarget are included by default, and dragging in the Render view no longer stalls.

Rendering

New: Ray-Traced engine

• A Whitted-style RGB approximation. It uses one d-line refractive index, or three indices with Fire for dispersion. Anti-Aliasing offers 1, 4, 8 or 16 taps. The first tap shows at once, the picture finishes at the tap count, and no denoiser is used.
• It tints by the material's absorption color and renders frosted facets. It leaves out polarization, birefringent doubling, scattering, zoning and edge rounding, and drops paths beyond the bounce cap. Path-Traced keeps the full spectral model.
• Optical metrics, tilt curves and Optical Diagnostics keep their own analytic tracer, so the engine choice doesn't change them.
• It runs on the GPU with a small kernel (four storage buffers) and falls back to the CPU automatically.
• Exports: the export and tilt-video dialogs each have their own Renderer choice, with Anti-Aliasing and Fire in place of the sample count (16 taps by default). Render jobs freeze the engine and its settings: Ray-Traced jobs use job format 2, and existing Path-Traced jobs (format 1) are unaffected. Ray-Traced output never adds the diagnostics overlay.
• Catalogue: a Thumbnail Renderer setting picks the engine for library pictures (Path-Traced by default).

Concave stones on the GPU

• The GPU now traces concave stones with up to 128 tool placements, every placement a design can hold. It was 64 before, and stones with 65 to 128 placements fell back to the CPU. This applies to the desktop, remote workers, the web app and the Android app.

New: the backdrop is a lit tray

• The backdrop is a matte tray under the stone, lit by the lighting rig. Every ray that leaves the stone sees it, whether the ray missed the stone or came out through the pavilion. Before, only the camera saw the backdrop, and a window showed the lighting's dark ground.
• Four choices:
  • White: 90 % reflectance, the new default.
  • Grey: 18 %, the photographic grey card.
  • Black: 2 %, velvet.
  • As Lit: the lighting's own ground. For the Grading tray this is nothing below the girdle (the ISO convention), so windows stay dark.
• The tray is lit, not a light source, so it follows the lighting, the light direction and the exposure. In a side view the horizon appears where it physically is. HDR environment maps ignore the backdrop.
• Two built-in lightings bring their own tray: Studio Softbox uses Grey and Dramatic Spotlight uses Black. Every other lighting starts on White, and any tray can still be picked afterwards.
• Brilliance, extinction, windowing and the tilt curves don't change with the backdrop.

Optical Diagnostics

• Scintillation, a new fifth layer: violet diamonds mark the pixels that sparkle. Each pixel's center ray is traced at five viewing angles within ±3° of yaw. The summary adds Visible-Area Scintillation (Map), and hovering a pixel shows how many of the five poses returned light.
• Exports:
  • Still exports stay clean unless Include Optical Diagnostics (4 Samples/Pixel) is checked.
  • The chosen layers, opacity, palette, mode and Scintillation are frozen with the export, including preset fan-outs and queued jobs.
  • The diagnostics are traced at the full output size, with the same picture whatever the core count. Each PNG gets an .optical.txt legend.
• An export that can't produce a diagnostic pass is refused before anything renders. That covers ASET lighting without an HDR map, a stone without facets, and no layer selected.

Fixes

• Scattering stones (for example the built-in ruby): light sampling after a scattering event charged the shadow ray the wrong extinction. Renders with light sampling on came out about 0.5 % darker than with it off. Fixed on the CPU tracer and the GPU kernels.

Desktop studio and CAD

Optimize and Retarget

• Older GemCad files no longer get refused. Many GemCad 4.x files write facet depths with five decimals. Their meets therefore miss by a few millionths of the stone, and the meet-preservation check used to reject them. Meets are now recognised within 0.005 % of the stone, and each meet's existing miss is recorded: a candidate must keep it, not close it. Exact designs behave exactly as before, and a meet that really opens is still refused.
• Retarget's limits follow the turn. A change of refractive index turns the facets, about ten degrees from diamond to sapphire. The stone's depth, girdle height and meet movement may now change by as much as that turn implies, measured for each tier from how far it reaches in from the girdle outline. Meets, girdle thickness, table and aspect are not relaxed. Optimize keeps its fixed limits.
• The target RI follows the material. Choosing a material in the Retarget dialog now clears the RI override taken over from the design. Before, a design that carried its own RI kept it, and the shift came out as zero. The web app has the same fix.
• Concave designs are scored with their tools: the Optimize score, Retarget's before and after columns, the Sweep, History pictures, library views, and the preview and tilt batches. Flat designs are unchanged. To refresh cached pictures of concave designs, use Regenerate All.

Library

• Loading a design sets its material straight away, without a visit to the Stone tab. The material comes from the design file, else the entry's preview material, else the built-in material that matches the stone's RI. When it was inferred from the RI, an amber "from library" badge shows until you pick another one.
• The preview offer shows its quality settings first. At first start and in the import and Regenerate offers, Picture Quality has Preview Image Size and Preview Samples, and moving a slider counts the designs again.

Rough planner

• Stones (Up To) (0 to 99, where 0 means no limit) caps how many stones one layout may hold. It is saved with the plan, and plans saved earlier open with no limit.

Interface

• Export and tilt-video export: Include Optical Diagnostics (4 Samples/Pixel).
• Render & Output: the anti-aliasing choices sit under an Anti-Aliasing label.
• Ray-Traced hints no longer say "requires Path-Traced", on the desktop, the web and Android.
• Optimize candidates: the Tone column is wider, so "L* 100.0" fits.
• The manual is updated to match.

CLI (indicatrix-cli)

• New census command (read-only): it measures what the meet-preserving check does to the built-in templates, design files or a whole library (opened read-only).
  • It writes one CSV or JSON row per design, target material and mode, plus percentile tables.
  • --out-dir writes the retargeted .isdd of every valid row.
  • Columns include gemcad_version, mast_decimals, valence, split_ratio and max_slack_ppm.
• Batches: info, solve, metrics, validate, optimize, retarget, sweep and export take any number of design files. --out-dir DIR writes one result per design. A failing file doesn't stop the others.
• retarget --girdle-allowance is the dialog's "Allow a Thicker Girdle" (up to +10 %). The CLI applies only a valid retarget, as the desktop dialog does.
• Render jobs take --engine path|ray.

Web app

• GPU rendering in Chrome and Edge: every Ray-Traced material class and isotropic Path-Traced stones (such as diamond and spinel) render on the GPU beside the CPU Workers, including concave stones with up to 64 tools. Anisotropic and color-zoned Path-Traced stones, and stones with more tools, render on the CPU, and the page says why.
• GPU start-up test:
  • It runs at page load, and whenever GPU Rendering is switched on.
  • A progress card shows the adapter, the limits, the kernel compile and a test render, while the CPU Workers keep rendering.
  • The Render panel then says what the page renders with.
  • Test GPU runs the test by hand, and Forget GPU Test Result clears a remembered failure.
• GPU Rendering is on by default.
• The Render tab offers Path-Traced and Ray-Traced, the four tap choices and Fire. PNG export uses the selected engine.

Android app

• Optimize and Retarget are included by default. The validated-design-search feature is now a default feature.
• Fixed: dragging in the Render view. Orbiting, flipping the stone or dragging the light used to stall and flicker until the finger lifted. Now the Solid picture follows the finger, and one trace starts at the final pose after you release. With the Remote backend, a drag cancels the session once instead of restarting it over and over.
• Ray-Traced:
  • Draft, Normal and Fine use 1, 4 and 8 taps, and export uses 16.
  • Device-GPU pictures go up to 1,048,576 pixels.
  • Remote Live view, Send To Server and cancellation all work with it.

Worker and remote rendering

• Live views, exports, preview batches and server jobs carry the chosen engine. A coordinator splits Ray-Traced work by tap ranges across its own lane and its joined workers, and a lost worker's range is retried on another lane.
• indicatrix-worker render --engine ray|path sets the e...

v0.8.1

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

v0.8.0

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

v0.7.2

Indicatrix Cut 0.7.2

Tutorials

• 68 short tutorials, one for each tool. They cover tiers, meets, concave tiers, solving, the optimizer, Compare, the solid view, rendering, export, the library and the app itself. Open them from Help → Tutorials..., from Ctrl+K, or from the empty start screen. They can be searched, and each shows a Done mark once finished.
• Every design can become a tutorial. Choose Build this design on any library design, local or remote. Indicatrix writes a step-by-step lesson that cuts that design from scratch, checks the lesson works before showing it, and starts it.
• A welcome tour for new users, plus the long guided walkthroughs.

Lighting

• New lighting presets: ASET (green/red/blue zones), Shop lights, Window daylight, White tray, Illuminant A and Daylight sun (sky plus a real 0.27° sun). Daylight dome is now sky only.
• The Light tent is the default. The new Grading tray preset (D65 hemisphere with an observer head shadow) is the reference light for scoring.
• The head shadow size is adjustable (0–30°, default 16°), and the metrics follow it.
• New default view: near face-up, with the key light at 72° elevation. New view presets: Grading tray and Window daylight.

Optimizer

• Multi-start search: it tries several starting designs (default 8, up to 32) and keeps the best result. The default budget is now 800 evaluations. CLI: indicatrix-cli optimize --starts.
• Tone goals: new presets to make a stone lighter, darker or more saturated, with a tone column and color swatches. A result never moves the tone against your goal.
• Optimize keeps the design's look: a shape term, refitting, and a pavilion floor.
• Every design is scored under the Grading tray light, so desktop, CLI and web results match.
• New "Compare against" picker in the Optimize comparison: compare with the current design or the original design.

Retarget

• The crown follows the pavilion's stretch.
• When a girdle check fails, Retarget may make the girdle up to 10 % thicker (can be turned off).
• Retarget scores designs under the Grading tray light.

Color

• Custom color with an LCh editor (tone, saturation, hue). It accounts for the light path through the stone and shows swatches at 3, 5 and 10 mm. It is available in Design settings, the material editor and Live Render. Custom materials keep these colors.

Render jobs

• Render queue: collect still exports and tilt videos, then start, pause or resume them. A paused video resumes at its first unfinished frame.
• New CLI commands indicatrix-cli render and indicatrix-cli tilt-video. The queue can be exported as a .sh or .ps1 script.

Cutting instructions

• The cutting sheet follows the standard fantasy-cut layout:
  • facet codes in cutting order;
  • positive angles only;
  • pavilion first, then crown, then table;
  • Facet, Size and Design Data blocks;
  • four views.

Rough planner

• Import STL and PLY as well as OBJ, and choose the units.
• Smooth scans are no longer refused. They are simplified to a 64-plane outline while every stone is still checked against the real scan.
• Broken meshes are repaired automatically, and self-intersecting meshes are detected.
• Inclusions can be added as extra shells, or located from photos. Stones are planned around them.
• Concave designs are fitted by their carved outline. Mesh roughs plan faster, and their time limit can be set.

Concave tiers

• Clicking a tool facet selects its concave tier.
• Compare draws, traces and measures concave stones.

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.i

v0.7.1

Indicatrix Cut 0.7.1

Tutorials

• 68 short tutorials, one for each tool. They cover tiers, meets, concave tiers, solving, the optimizer, Compare, the solid view, rendering, export, the library and the app itself. Open them from Help → Tutorials..., from Ctrl+K, or from the empty start screen. They can be searched, and each shows a Done mark once finished.
• Every design can become a tutorial. Choose Build this design on any library design, local or remote. Indicatrix writes a step-by-step lesson that cuts that design from scratch, checks the lesson works before showing it, and starts it.
• A welcome tour for new users, plus the long guided walkthroughs.

Lighting

• New lighting presets: ASET (green/red/blue zones), Shop lights, Window daylight, White tray, Illuminant A and Daylight sun (sky plus a real 0.27° sun). Daylight dome is now sky only.
• The Light tent is the default. The new Grading tray preset (D65 hemisphere with an observer head shadow) is the reference light for scoring.
• The head shadow size is adjustable (0–30°, default 16°), and the metrics follow it.
• New default view: near face-up, with the key light at 72° elevation. New view presets: Grading tray and Window daylight.

Optimizer

• Multi-start search: it tries several starting designs (default 8, up to 32) and keeps the best result. The default budget is now 800 evaluations. CLI: indicatrix-cli optimize --starts.
• Tone goals: new presets to make a stone lighter, darker or more saturated, with a tone column and color swatches. A result never moves the tone against your goal.
• Optimize keeps the design's look: a shape term, refitting, and a pavilion floor.
• Every design is scored under the Grading tray light, so desktop, CLI and web results match.
• New "Compare against" picker in the Optimize comparison: compare with the current design or the original design.

Retarget

• The crown follows the pavilion's stretch.
• When a girdle check fails, Retarget may make the girdle up to 10 % thicker (can be turned off).
• Retarget scores designs under the Grading tray light.

Color

• Custom color with an LCh editor (tone, saturation, hue). It accounts for the light path through the stone and shows swatches at 3, 5 and 10 mm. It is available in Design settings, the material editor and Live Render. Custom materials keep these colors.

Render jobs

• Render queue: collect still exports and tilt videos, then start, pause or resume them. A paused video resumes at its first unfinished frame.
• New CLI commands indicatrix-cli render and indicatrix-cli tilt-video. The queue can be exported as a .sh or .ps1 script.

Cutting instructions

• The cutting sheet follows the standard fantasy-cut layout:
  • facet codes in cutting order;
  • positive angles only;
  • pavilion first, then crown, then table;
  • Facet, Size and Design Data blocks;
  • four views.

Rough planner

• Import STL and PLY as well as OBJ, and choose the units.
• Smooth scans are no longer refused. They are simplified to a 64-plane outline while every stone is still checked against the real scan.
• Broken meshes are repaired automatically, and self-intersecting meshes are detected.
• Inclusions can be added as extra shells, or located from photos. Stones are planned around them.
• Concave designs are fitted by their carved outline. Mesh roughs plan faster, and their time limit can be set.

Concave tiers

• Clicking a tool facet selects its concave tier.
• Compare draws, traces and measures concave stones.

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

v0.7.0

Indicatrix Cut 0.7.0

Tutorials

• 68 short tutorials, one for each tool. They cover tiers, meets, concave tiers, solving, the optimizer, Compare, the solid view, rendering, export, the library and the app itself. Open them from Help → Tutorials..., from Ctrl+K, or from the empty start screen. They can be searched, and each shows a Done mark once finished.
• Every design can become a tutorial. Choose Build this design on any library design, local or remote. Indicatrix writes a step-by-step lesson that cuts that design from scratch, checks the lesson works before showing it, and starts it.
• A welcome tour for new users, plus the long guided walkthroughs.

Lighting

• New lighting presets: ASET (green/red/blue zones), Shop lights, Window daylight, White tray, Illuminant A and Daylight sun (sky plus a real 0.27° sun). Daylight dome is now sky only.
• The Light tent is the default. The new Grading tray preset (D65 hemisphere with an observer head shadow) is the reference light for scoring.
• The head shadow size is adjustable (0–30°, default 16°), and the metrics follow it.
• New default view: near face-up, with the key light at 72° elevation. New view presets: Grading tray and Window daylight.

Optimizer

• Multi-start search: it tries several starting designs (default 8, up to 32) and keeps the best result. The default budget is now 800 evaluations. CLI: indicatrix-cli optimize --starts.
• Tone goals: new presets to make a stone lighter, darker or more saturated, with a tone column and color swatches. A result never moves the tone against your goal.
• Optimize keeps the design's look: a shape term, refitting, and a pavilion floor.
• Every design is scored under the Grading tray light, so desktop, CLI and web results match.
• New "Compare against" picker in the Optimize comparison: compare with the current design or the original design.

Retarget

• The crown follows the pavilion's stretch.
• When a girdle check fails, Retarget may make the girdle up to 10 % thicker (can be turned off).
• Retarget scores designs under the Grading tray light.

Color

• Custom color with an LCh editor (tone, saturation, hue). It accounts for the light path through the stone and shows swatches at 3, 5 and 10 mm. It is available in Design settings, the material editor and Live Render. Custom materials keep these colors.

Render jobs

• Render queue: collect still exports and tilt videos, then start, pause or resume them. A paused video resumes at its first unfinished frame.
• New CLI commands indicatrix-cli render and indicatrix-cli tilt-video. The queue can be exported as a .sh or .ps1 script.

Cutting instructions

• The cutting sheet follows the standard fantasy-cut layout:
  • facet codes in cutting order;
  • positive angles only;
  • pavilion first, then crown, then table;
  • Facet, Size and Design Data blocks;
  • four views.

Rough planner

• Import STL and PLY as well as OBJ, and choose the units.
• Smooth scans are no longer refused. They are simplified to a 64-plane outline while every stone is still checked against the real scan.
• Broken meshes are repaired automatically, and self-intersecting meshes are detected.
• Inclusions can be added as extra shells, or located from photos. Stones are planned around them.
• Concave designs are fitted by their carved outline. Mesh roughs plan faster, and their time limit can be set.

Concave tiers

• Clicking a tool facet selects its concave tier.
• Compare draws, traces and measures concave stones.

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

v0.6.2

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

v0.6.1

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

v0.6.0

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

Edit:
Added Windows binaries with icons + Linux binaries

v0.5.0

Which Download Should I Pick?

These binaries are built with Profile-Guided Optimization (PGO) for maximum ray-tracing performance. They are provided in three CPU vectorization tiers: AVX-512, AVX2, and Scalar.

If an executable is built for an instruction set your processor does not support, it will crash immediately on launch (with an Illegal Instruction error or crash code 0xc000001d).

👉 Recommended Order to Try:

1. Start with AVX-512 (Fastest — Try this first!)

   • Delivers the highest ray-tracing throughput. Supported on AMD Ryzen 7000/9000 series (Zen 4 / Zen 5) and supported Intel CPUs (11th Gen, Xeon, Skylake-X).
   • If it launches, keep using it! You get the maximum possible performance.
   • If it crashes immediately on launch (with an Illegal Instruction error or code 0xc000001d), your CPU does not support AVX-512 — simply step down to AVX2.

2. Step down to AVX2 (Broad compatibility)

   • Works on virtually any PC manufactured in the last decade (Intel 4th Gen Haswell 2013+ / AMD Ryzen 2017+).
   • Delivers excellent vectorized ray-tracing speed. If AVX-512 didn't work for you, this is the version to use.

3. Fall back to scalar (Universal compatibility)

   • Use this only if the avx2 build also refuses to start.
   • Runs on any 64-bit x86 processor (older legacy hardware, virtual machines without AVX passthrough, or low-power chips), but trades away SIMD vectorization speed.

---

GPU vs. CPU Builds

• *-gpu (Recommended): Uses your dedicated or integrated GPU via WGPU (Vulkan / DirectX 12 / Metal) for hardware-accelerated spectral compute. Automatically falls back to multi-threaded CPU if needed.
• *-cpu: Pure multi-threaded CPU tracer without graphics compute dependencies. Ideal for headless servers or systems without a compatible GPU driver.

---

Binaries in this Release:

• indicatrix-cut: The desktop 3D faceting studio & cutting schedule editor (Slint GUI).
• indicatrix-worker: The headless CLI renderer and remote mTLS compute daemon.

Edit:
Added Windows binaries with icons + Linux binaries