Mesa (staging/21.1): radv: fix missing ITERATE_256 for D/S MSAA images that are TC-compat HTILE
GitLab Mirror
gitlab-mirror at kemper.freedesktop.org
Tue May 11 20:51:07 UTC 2021
Module: Mesa
Branch: staging/21.1
Commit: 420a2b20406dd5cb41033adb0e1c67de5540d8c3
URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=420a2b20406dd5cb41033adb0e1c67de5540d8c3
Author: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Date: Mon May 3 17:56:27 2021 +0200
radv: fix missing ITERATE_256 for D/S MSAA images that are TC-compat HTILE
To make them readable by shaders, only needed on GFX10+.
This also fixes corruption with Control and MSAA.
Cc: 21.1 mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10610>
(cherry picked from commit 33ede796d5b9149076738b031137b4389dab71a1)
---
.pick_status.json | 2 +-
src/amd/ci/deqp-radv-sienna_cichlid-aco-fails.txt | 33 -----------------------
src/amd/common/ac_gpu_info.c | 5 ++++
src/amd/common/ac_gpu_info.h | 1 +
src/amd/vulkan/radv_device.c | 12 +++++++++
src/amd/vulkan/radv_image.c | 4 +++
src/amd/vulkan/radv_private.h | 11 ++++++++
7 files changed, 34 insertions(+), 34 deletions(-)
diff --git a/.pick_status.json b/.pick_status.json
index f5ba516aa1d..387f6e302b9 100644
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -1921,7 +1921,7 @@
"description": "radv: fix missing ITERATE_256 for D/S MSAA images that are TC-compat HTILE",
"nominated": true,
"nomination_type": 0,
- "resolution": 0,
+ "resolution": 1,
"master_sha": null,
"because_sha": null
},
diff --git a/src/amd/ci/deqp-radv-sienna_cichlid-aco-fails.txt b/src/amd/ci/deqp-radv-sienna_cichlid-aco-fails.txt
index d21e84f713a..677f350943f 100644
--- a/src/amd/ci/deqp-radv-sienna_cichlid-aco-fails.txt
+++ b/src/amd/ci/deqp-radv-sienna_cichlid-aco-fails.txt
@@ -121,36 +121,3 @@ dEQP-VK.drm_format_modifiers.create_list_modifiers.r16g16b16_uscaled,Crash
dEQP-VK.image.mismatched_formats.sparse_image_read.d16_unorm_with_r16,Crash
dEQP-VK.image.mismatched_formats.sparse_image_read.d32_sfloat_with_r32f,Crash
dEQP-VK.image.mismatched_formats.sparse_image_read.s8_uint_with_r8ui,Crash
-
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.keep.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.keep.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.replace.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.replace.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.keep.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.keep.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.replace.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.replace.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.keep.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.keep.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.replace.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.noshaderrate.replace.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.keep.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.replace.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.keep.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.noattachment.shaderrate.replace.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.keep.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.keep.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.replace.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.replace.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.keep.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.keep.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.replace.keep.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.replace.keep.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.keep.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.replace.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.replace.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.keep.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.replace.replace.151x431.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.keep.replace.256x256.samples8.vs,Fail
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.noshaderrate.keep.replace.151x431.samples8.vsœ
-dEQP-VK.fragment_shading_rate.fragstencil.dynamic.attachment.shaderrate.replace.replace.256x256.samples8.vs
diff --git a/src/amd/common/ac_gpu_info.c b/src/amd/common/ac_gpu_info.c
index ec574bd854d..120dc10bec0 100644
--- a/src/amd/common/ac_gpu_info.c
+++ b/src/amd/common/ac_gpu_info.c
@@ -887,6 +887,11 @@ bool ac_query_gpu_info(int fd, void *dev_p, struct radeon_info *info,
info->has_image_load_dcc_bug = info->family == CHIP_DIMGREY_CAVEFISH ||
info->family == CHIP_VANGOGH;
+ /* DB has a bug when ITERATE_256 is set to 1 that can cause a hang. The
+ * workaround is to set DECOMPRESS_ON_Z_PLANES to 2 for 4X MSAA D/S images.
+ */
+ info->has_two_planes_iterate256_bug = info->chip_class == GFX10;
+
/* Support for GFX10.3 was added with F32_ME_FEATURE_VERSION_31 but the
* firmware version wasn't bumped.
*/
diff --git a/src/amd/common/ac_gpu_info.h b/src/amd/common/ac_gpu_info.h
index 546c1e9fb5d..7eb9c944d9a 100644
--- a/src/amd/common/ac_gpu_info.h
+++ b/src/amd/common/ac_gpu_info.h
@@ -77,6 +77,7 @@ struct radeon_info {
bool has_ls_vgpr_init_bug;
bool has_zero_index_buffer_bug;
bool has_image_load_dcc_bug;
+ bool has_two_planes_iterate256_bug;
bool has_32bit_predication;
bool has_3d_cube_border_color_mipmap;
diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
index ed11e27f53f..c4cf1d040fd 100644
--- a/src/amd/vulkan/radv_device.c
+++ b/src/amd/vulkan/radv_device.c
@@ -6579,6 +6579,14 @@ radv_calc_decompress_on_z_planes(struct radv_device *device, struct radv_image_v
if (iview->vk_format == VK_FORMAT_D16_UNORM && iview->image->info.samples > 1)
max_zplanes = 2;
+ /* Workaround for a DB hang when ITERATE_256 is set to 1. Only affects 4X MSAA D/S images. */
+ if (device->physical_device->rad_info.has_two_planes_iterate256_bug &&
+ radv_image_get_iterate256(device, iview->image) &&
+ !radv_image_tile_stencil_disabled(device, iview->image) &&
+ iview->image->info.samples == 4) {
+ max_zplanes = 1;
+ }
+
max_zplanes = max_zplanes + 1;
} else {
if (iview->vk_format == VK_FORMAT_D16_UNORM) {
@@ -6682,8 +6690,12 @@ radv_initialise_ds_surface(struct radv_device *device, struct radv_ds_buffer_inf
ds->db_z_info |= S_028038_DECOMPRESS_ON_N_ZPLANES(max_zplanes);
if (device->physical_device->rad_info.chip_class >= GFX10) {
+ bool iterate256 = radv_image_get_iterate256(device, iview->image);
+
ds->db_z_info |= S_028040_ITERATE_FLUSH(1);
ds->db_stencil_info |= S_028044_ITERATE_FLUSH(1);
+ ds->db_z_info |= S_028040_ITERATE_256(iterate256);
+ ds->db_stencil_info |= S_028044_ITERATE_256(iterate256);
} else {
ds->db_z_info |= S_028038_ITERATE_FLUSH(1);
ds->db_stencil_info |= S_02803C_ITERATE_FLUSH(1);
diff --git a/src/amd/vulkan/radv_image.c b/src/amd/vulkan/radv_image.c
index 51259b0cc0a..c7d102882a6 100644
--- a/src/amd/vulkan/radv_image.c
+++ b/src/amd/vulkan/radv_image.c
@@ -926,6 +926,10 @@ gfx10_make_texture_descriptor(struct radv_device *device, struct radv_image *ima
S_00A018_ALPHA_IS_ON_MSB(vi_alpha_is_on_msb(device, vk_format));
}
+ if (radv_image_get_iterate256(device, image)) {
+ state[6] |= S_00A018_ITERATE_256(1);
+ }
+
/* Initialize the sampler view for FMASK. */
if (fmask_state) {
if (radv_image_has_fmask(image)) {
diff --git a/src/amd/vulkan/radv_private.h b/src/amd/vulkan/radv_private.h
index 34d8d64a9f1..9c5104558d5 100644
--- a/src/amd/vulkan/radv_private.h
+++ b/src/amd/vulkan/radv_private.h
@@ -2076,6 +2076,17 @@ radv_get_htile_initial_value(const struct radv_device *device, const struct radv
return initial_value;
}
+static inline bool
+radv_image_get_iterate256(struct radv_device *device, struct radv_image *image)
+{
+ /* ITERATE_256 is required for depth or stencil MSAA images that are TC-compatible HTILE. */
+ return device->physical_device->rad_info.chip_class >= GFX10 &&
+ (image->usage & (VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT |
+ VK_IMAGE_USAGE_TRANSFER_DST_BIT)) &&
+ radv_image_is_tc_compat_htile(image) &&
+ image->info.samples > 1;
+}
+
unsigned radv_image_queue_family_mask(const struct radv_image *image, uint32_t family,
uint32_t queue_family);
More information about the mesa-commit
mailing list