🏡 index : cold-air-networks/mirror-webgpu-cts.git

export const description = `
Execution tests for the 'trunc' builtin function

S is abstract-float, f32, f16
T is S or vecN<S>
@const fn trunc(e: T ) -> T
Returns the nearest whole number whose absolute value is less than or equal to e.
Component-wise when T is a vector.
`;

import { makeTestGroup } from '../../../../../../common/framework/test_group.js';
import { AllFeaturesMaxLimitsGPUTest } from '../../../../../gpu_test.js';
import { kValue } from '../../../../../util/constants.js';
import { Type, f32, f16 } from '../../../../../util/conversion.js';
import { FP } from '../../../../../util/floating_point.js';
import {
  allInputSources,
  onlyConstInputSource,
  run,
  basicExpressionBuilder,
} from '../../expression.js';

import { abstractFloatBuiltin, builtin } from './builtin.js';
import { d } from './trunc.cache.js';

export const g = makeTestGroup(AllFeaturesMaxLimitsGPUTest);

g.test('abstract_float')
  .specURL('https://www.w3.org/TR/WGSL/#float-builtin-functions')
  .desc(`abstract float tests`)
  .params(u =>
    u
      .combine('inputSource', onlyConstInputSource)
      .combine('vectorize', [undefined, 2, 3, 4] as const)
  )
  .fn(async t => {
    const cases = await d.get('abstract');
    await run(
      t,
      abstractFloatBuiltin('trunc'),
      [Type.abstractFloat],
      Type.abstractFloat,
      t.params,
      cases
    );
  });

g.test('f32')
  .specURL('https://www.w3.org/TR/WGSL/#float-builtin-functions')
  .desc(`f32 tests`)
  .params(u =>
    u.combine('inputSource', allInputSources).combine('vectorize', [undefined, 2, 3, 4] as const)
  )
  .fn(async t => {
    const cases = await d.get('f32');
    await run(t, builtin('trunc'), [Type.f32], Type.f32, t.params, cases);
  });

g.test('f16')
  .specURL('https://www.w3.org/TR/WGSL/#float-builtin-functions')
  .desc(`f16 tests`)
  .params(u =>
    u.combine('inputSource', allInputSources).combine('vectorize', [undefined, 2, 3, 4] as const)
  )
  .fn(async t => {
    t.skipIfDeviceDoesNotHaveFeature('shader-f16');
    const cases = await d.get('f16');
    await run(t, builtin('trunc'), [Type.f16], Type.f16, t.params, cases);
  });

g.test('subnormal_division')
  .specURL('https://www.w3.org/TR/WGSL/#float-builtin-functions')
  .desc(`truncate subnormals made even smaller to zero`)
  .params(u => u.combine('inputSource', allInputSources).combine('type', ['f32', 'f16'] as const))
  .fn(async t => {
    const type = t.params.type;
    if (type === 'f16') {
      t.skipIfDeviceDoesNotHaveFeature('shader-f16');
    }

    const trait = FP[type as 'f32' | 'f16'];
    const scalarBuilder = type === 'f32' ? f32 : f16;
    const constant = kValue[type as 'f32' | 'f16'];
    const cases = [
      // Smallest subnormals (closest to zero)
      {
        input: [scalarBuilder(constant.negative.subnormal.max)],
        expected: trait.toInterval(constant.negative.zero),
      },
      {
        input: [scalarBuilder(constant.positive.subnormal.min)],
        expected: trait.toInterval(constant.positive.zero),
      },
    ];

    await run(
      t,
      // Divided by 10.0 to make subnormal even smaller and unrepresentable as subnormal, should truncate to zero
      basicExpressionBuilder(values => `trunc(${values[0]} / 10.0)`),
      [Type[type as 'f32' | 'f16']],
      Type[type as 'f32' | 'f16'],
      t.params,
      cases
    );
  });