radeonsi: Fix compute copies for subsampled formats.

We cannot do image stores (or render) to subsampled formats.

Reinterpret as R32_UINT instead.

si_set_shader_image_desc already uses the blockwidth from
the view formats, so the image width adjustments are
already implemented.

This is still icky with mipmapping on GFX9+ though, but
since it is mostly a video format I don't think that will
be much of an issue and broken mipmapping is still better
than broken everything.

Fixes: e5167a9276d "radeonsi: disable SDMA on gfx8 to fix corruption on RX 580"
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2535
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3853>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3853>

diff --git a/src/gallium/drivers/radeonsi/si_compute_blit.c b/src/gallium/drivers/radeonsi/si_compute_blit.c
index 3e544d68500111c55d462db468a559a9aa7dc2a2..73020f9d3b410ed4be671b4685e7e9217391107b 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_compute_blit.c
+++ b/src/gallium/drivers/radeonsi/si_compute_blit.c
@@ -392,8 +392,23 @@ void si_compute_copy_image(struct si_context *sctx,
        unsigned width = src_box->width;
        unsigned height = src_box->height;
        unsigned depth = src_box->depth;
+       enum pipe_format src_format = util_format_linear(src->format);
+       enum pipe_format dst_format = util_format_linear(dst->format);
 
-       unsigned data[] = {src_box->x, src_box->y, src_box->z, 0, dstx, dsty, dstz, 0};
+       assert(util_format_is_subsampled_422(src_format) ==
+              util_format_is_subsampled_422(dst_format));
+
+       if (util_format_is_subsampled_422(src_format))
+               src_format = dst_format = PIPE_FORMAT_R32_UINT;
+
+       unsigned x_div = util_format_get_blockwidth(src->format) /
+                        util_format_get_blockwidth(src_format);
+       assert(src_box->x % x_div == 0);
+       assert(width % x_div == 0);
+
+       unsigned data[] = {src_box->x / x_div, src_box->y, src_box->z, 0,
+                          dstx / x_div, dsty, dstz, 0};
+       width /= x_div;
 
        if (width == 0 || height == 0)
                return;
@@ -430,7 +445,7 @@ void si_compute_copy_image(struct si_context *sctx,
        struct pipe_image_view image[2] = {0};
        image[0].resource = src;
        image[0].shader_access = image[0].access = PIPE_IMAGE_ACCESS_READ;
-       image[0].format = util_format_linear(src->format);
+       image[0].format = src_format;
        image[0].u.tex.level = src_level;
        image[0].u.tex.first_layer = 0;
        image[0].u.tex.last_layer =
@@ -438,7 +453,7 @@ void si_compute_copy_image(struct si_context *sctx,
                                                : (unsigned)(src->array_size - 1);
        image[1].resource = dst;
        image[1].shader_access = image[1].access = PIPE_IMAGE_ACCESS_WRITE;
-       image[1].format = util_format_linear(dst->format);
+       image[1].format = dst_format;
        image[1].u.tex.level = dst_level;
        image[1].u.tex.first_layer = 0;
        image[1].u.tex.last_layer =