[gui] Add GGUI set_image support for non-Vector fields and numpy ndarrays.

[Hyiker]

Related issue = close #5633

Fixing the issue and adding additional support for flat fields like (400, 400, 3). Also, I implemented support for numpy.ndarray following the document.

[netlify[bot]]

✅ Deploy Preview for docsite-preview ready!

Name	Link
🔨 Latest commit	65e96dbf515145098b205d17f63c6598b5f343ac
🔍 Latest deploy log	https://app.netlify.com/sites/docsite-preview/deploys/62f77e18961e850009b07605
😎 Deploy Preview	https://deploy-preview-5654--docsite-preview.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site settings.

[Morcki]

Sorry, there have been some confilcts with others whose commitment include adding support for image texture type, I fix some confilcts, may you help to check it? Very sorry to you for I do not merge into master in time cause there have been some bugs on the github CI workflow which took some time for our team to fix on this week. Thanks a lot again for your contributions! Wish you best!

[Hyiker]

Sorry, there have been some confilcts with others whose commitment include adding support for image texture type, I fix some confilcts, may you help to check it? Very sorry to you for I do not merge into master in time cause there have been some bugs on the github CI workflow which took some time for our team to fix on this week. Thanks a lot again for your contributions! Wish you best!

Sure, I rewrite some lines to improve the readability. Additionally, I created a test for flat fields in test_ggui.py, but when I'm trying to create one for numpy field, the test failed with:

E           RuntimeError: [D:/a/taichi/taichi/taichi/codegen/spirv/spirv_ir_builder.cpp:register_value@1185] tmp1 is existed.

However, in case I actually use a numpy ndarray as input, it just worked fine, I'm really confused. My numpy test code was like:

@pytest.mark.skipif(not _ti_core.GGUI_AVAILABLE, reason="GGUI Not Available")
@test_utils.test(arch=supported_archs)
def test_set_image_flat_field():
    window = ti.ui.Window('test', (640, 480), show_window=False)
    canvas = window.get_canvas()

    img = ti.field(ti.f32, (512, 512, 4))

    @ti.kernel
    def init_img():
        for i, j in ti.ndrange(img.shape[0], img.shape[1]):
            img[i, j, 0] = i / 512
            img[i, j, 1] = j / 512
            img[i, j, 2] = 0
            img[i, j, 3] = 1.0

    init_img()

    def render():
        canvas.set_image(img.to_numpy())

    for _ in range(RENDER_REPEAT):
        render()
        write_temp_image(window)
    render()
    verify_image(window, 'test_set_image')
    window.destroy()

I'm not quite familiar with taichi kernel code, wondering if u could help me solve it.

[Morcki]

The problem is due to this line for i, j in ti.ndrange(src.shape[0], src.shape[1]):, you can replace it using ti.grouped() to solve this problem. It seems like using ti.ndrange() for ti.ndarray() is not supported under vulkan backend, I think this is a bug. You can issue this for us.

Here is a simple test:

import taichi as ti

ti.init(arch=ti.vulkan)

@ti.kernel
def test_loop_under_vulkan(test_arr : ti.types.ndarray()):
    for i, j in ti.ndrange(test_arr.shape[0], test_arr.shape[1]):
        pass

ti_arr = ti.ndarray(ti.f32, (3,3,3))
test_loop_under_vulkan(ti_arr)

[Hyiker]

Thx for reviewing, I've applied all your suggestion. But the numpy test problem stills exists even if I used ti.grouped, I think it might be another bug and may open an individual issue for it.

[Morcki]

So that's strange, cause I success with ti.grouped on my dev environment(Ubuntu 18). Yeah, it seems like an individual issue which should be talked about in another issue.