r600g: rewrite FMASK allocation, fix FMASK texturing with 2 and 4 samples

This fixes and enables texturing with compressed MSAA colorbuffers
on Evergreen and Cayman. For the first time, multisample textures work
on Cayman.

This requires the libdrm flag RADEON_SURF_FMASK.

v2: require libdrm_radeon 2.4.45

Reviewed-by: Alex Deucher <[email protected]>

8 files changed, 43 insertions, 37 deletions

diff --git a/configure.ac b/configure.ac
index 5bc8c94a2ec..eef4327804c 100644
--- a/configure.ac
+++ b/configure.ac
@@ -31,7 +31,7 @@ AC_SUBST([OSMESA_VERSION])
 
 dnl Versions for external dependencies
 LIBDRM_REQUIRED=2.4.24
-LIBDRM_RADEON_REQUIRED=2.4.44
+LIBDRM_RADEON_REQUIRED=2.4.45
 LIBDRM_INTEL_REQUIRED=2.4.38
 LIBDRM_NVVIEUX_REQUIRED=2.4.33
 LIBDRM_NOUVEAU_REQUIRED="2.4.33 libdrm >= 2.4.41"
diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 6797b223742..4eb976849c5 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -1101,7 +1101,7 @@ evergreen_create_sampler_view_custom(struct pipe_context *ctx,
 	uint32_t word4 = 0, yuv_format = 0, pitch = 0;
 	unsigned char swizzle[4], array_mode = 0, non_disp_tiling = 0;
 	unsigned height, depth, width;
-	unsigned macro_aspect, tile_split, bankh, bankw, nbanks;
+	unsigned macro_aspect, tile_split, bankh, bankw, nbanks, fmask_bankh;
 	enum pipe_format pipe_format = state->format;
 	struct radeon_surface_level *surflevel;
 
@@ -1188,6 +1188,7 @@ evergreen_create_sampler_view_custom(struct pipe_context *ctx,
 	macro_aspect = eg_macro_tile_aspect(macro_aspect);
 	bankw = eg_bank_wh(bankw);
 	bankh = eg_bank_wh(bankh);
+	fmask_bankh = eg_bank_wh(tmp->fmask_bank_height);
 
 	/* 128 bit formats require tile type = 1 */
 	if (rscreen->chip_class == CAYMAN) {
@@ -1219,8 +1220,7 @@ evergreen_create_sampler_view_custom(struct pipe_context *ctx,
 
 	/* TEX_RESOURCE_WORD3.MIP_ADDRESS */
 	if (texture->nr_samples > 1 && rscreen->msaa_texture_support == MSAA_TEXTURE_COMPRESSED) {
-		/* XXX the 2x and 4x cases are broken. */
-		if (tmp->is_depth || tmp->resource.b.b.nr_samples != 8) {
+		if (tmp->is_depth) {
 			/* disable FMASK (0 = disabled) */
 			view->tex_resource_words[3] = 0;
 			view->skip_mip_address_reloc = true;
@@ -1239,6 +1239,8 @@ evergreen_create_sampler_view_custom(struct pipe_context *ctx,
 				       S_030010_ENDIAN_SWAP(endian));
 	view->tex_resource_words[5] = S_030014_BASE_ARRAY(state->u.tex.first_layer) |
 				      S_030014_LAST_ARRAY(state->u.tex.last_layer);
+	view->tex_resource_words[6] = S_030018_TILE_SPLIT(tile_split);
+
 	if (texture->nr_samples > 1) {
 		unsigned log_samples = util_logbase2(texture->nr_samples);
 		if (rscreen->chip_class == CAYMAN) {
@@ -1246,13 +1248,14 @@ evergreen_create_sampler_view_custom(struct pipe_context *ctx,
 		}
 		/* LAST_LEVEL holds log2(nr_samples) for multisample textures */
 		view->tex_resource_words[5] |= S_030014_LAST_LEVEL(log_samples);
+		view->tex_resource_words[6] |= S_030018_FMASK_BANK_HEIGHT(fmask_bankh);
 	} else {
 		view->tex_resource_words[4] |= S_030010_BASE_LEVEL(state->u.tex.first_level);
 		view->tex_resource_words[5] |= S_030014_LAST_LEVEL(state->u.tex.last_level);
+		/* aniso max 16 samples */
+		view->tex_resource_words[6] |= S_030018_MAX_ANISO(4);
 	}
-	/* aniso max 16 samples */
-	view->tex_resource_words[6] = (S_030018_MAX_ANISO(4)) |
-				      (S_030018_TILE_SPLIT(tile_split));
+
 	view->tex_resource_words[7] = S_03001C_DATA_FORMAT(format) |
 				      S_03001C_TYPE(V_03001C_SQ_TEX_VTX_VALID_TEXTURE) |
 				      S_03001C_BANK_WIDTH(bankw) |
@@ -1576,7 +1579,7 @@ void evergreen_init_color_surface(struct r600_context *rctx,
 		surf->cb_color_fmask = surf->cb_color_base;
 		surf->cb_color_cmask = surf->cb_color_base;
 	}
-	surf->cb_color_fmask_slice = S_028C88_TILE_MAX(slice);
+	surf->cb_color_fmask_slice = S_028C88_TILE_MAX(rtex->fmask_slice_tile_max);
 	surf->cb_color_cmask_slice = S_028C80_TILE_MAX(rtex->cmask_slice_tile_max);
 
 	surf->color_initialized = true;
diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h
index 757c9512e5f..8990d6c3fad 100644
--- a/src/gallium/drivers/r600/evergreend.h
+++ b/src/gallium/drivers/r600/evergreend.h
@@ -1140,9 +1140,12 @@
 #define   G_030014_LAST_ARRAY(x)                       (((x) >> 17) & 0x1FFF)
 #define   C_030014_LAST_ARRAY                          0xC001FFFF
 #define R_030018_SQ_TEX_RESOURCE_WORD6_0             0x030018
+/* FMASK_BANK_HEIGHT and MAX_ANISO share the first two bits.
+ * The former is only used with MSAA textures. */
 #define   S_030018_MAX_ANISO(x)                        (((x) & 0x7) << 0)
 #define   G_030018_MAX_ANISO(x)                        (((x) >> 0) & 0x7)
 #define   C_030018_MAX_ANISO                           0xFFFFFFF8
+#define   S_030018_FMASK_BANK_HEIGHT(x)                (((x) & 0x3) << 0)
 #define   S_030018_PERF_MODULATION(x)                  (((x) & 0x7) << 3)
 #define   G_030018_PERF_MODULATION(x)                  (((x) >> 3) & 0x7)
 #define   C_030018_PERF_MODULATION                     0xFFFFFFC7
diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c
index afe43896568..7d32eef89ee 100644
--- a/src/gallium/drivers/r600/r600_blit.c
+++ b/src/gallium/drivers/r600/r600_blit.c
@@ -300,11 +300,7 @@ static void r600_blit_decompress_color(struct pipe_context *ctx,
 		blend_decompress = rctx->custom_blend_decompress;
 		break;
 	case MSAA_TEXTURE_COMPRESSED:
-		/* XXX the 2x and 4x cases are broken. */
-		if (rtex->resource.b.b.nr_samples == 8)
-			blend_decompress = rctx->custom_blend_fmask_decompress;
-		else
-			blend_decompress = rctx->custom_blend_decompress;
+		blend_decompress = rctx->custom_blend_fmask_decompress;
 		break;
 	case MSAA_TEXTURE_SAMPLE_ZERO:
 	default:
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index 47cdb9e4926..efb45010a19 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -1271,8 +1271,7 @@ struct pipe_screen *r600_screen_create(struct radeon_winsys *ws)
 		break;
 	case CAYMAN:
 		rscreen->has_msaa = rscreen->info.drm_minor >= 19;
-		/* We should be able to read compressed MSAA textures, but it doesn't work. */
-		rscreen->msaa_texture_support = MSAA_TEXTURE_SAMPLE_ZERO;
+		rscreen->msaa_texture_support = MSAA_TEXTURE_COMPRESSED;
 		break;
 	default:
 		rscreen->has_msaa = FALSE;
diff --git a/src/gallium/drivers/r600/r600_resource.h b/src/gallium/drivers/r600/r600_resource.h
index e20e9421205..d5df63396cc 100644
--- a/src/gallium/drivers/r600/r600_resource.h
+++ b/src/gallium/drivers/r600/r600_resource.h
@@ -84,7 +84,9 @@ struct r600_texture {
 	/* FMASK and CMASK can only be used with MSAA textures for now.
 	 * MSAA textures cannot have mipmaps. */
 	unsigned			fmask_offset, fmask_size, fmask_bank_height;
-	unsigned			cmask_offset, cmask_size, cmask_slice_tile_max;
+	unsigned			fmask_slice_tile_max;
+	unsigned			cmask_offset, cmask_size;
+	unsigned			cmask_slice_tile_max;
 
 	struct r600_resource		*htile;
 	/* use htile only for first level */
@@ -97,6 +99,7 @@ struct r600_fmask_info {
 	unsigned size;
 	unsigned alignment;
 	unsigned bank_height;
+	unsigned slice_tile_max;
 };
 
 struct r600_cmask_info {
diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c
index 4e0e4a6de69..f0e36753d2b 100644
--- a/src/gallium/drivers/r600/r600_state.c
+++ b/src/gallium/drivers/r600/r600_state.c
@@ -1404,7 +1404,7 @@ static void r600_init_color_surface(struct r600_context *rctx,
 		if (rtex->fmask_size) {
 			color_info |= S_0280A0_TILE_MODE(V_0280A0_FRAG_ENABLE);
 			surf->cb_color_fmask = rtex->fmask_offset >> 8;
-			surf->cb_color_mask |= S_028100_FMASK_TILE_MAX(slice);
+			surf->cb_color_mask |= S_028100_FMASK_TILE_MAX(rtex->fmask_slice_tile_max);
 		} else { /* cmask only */
 			color_info |= S_0280A0_TILE_MODE(V_0280A0_CLEAR_ENABLE);
 		}
@@ -1455,7 +1455,7 @@ static void r600_init_color_surface(struct r600_context *rctx,
 		surf->cb_color_cmask = 0;
 		surf->cb_color_fmask = 0;
 		surf->cb_color_mask = S_028100_CMASK_BLOCK_MAX(cmask.slice_tile_max) |
-				      S_028100_FMASK_TILE_MAX(slice);
+				      S_028100_FMASK_TILE_MAX(fmask.slice_tile_max);
 	}
 
 	surf->cb_color_info = color_info;
diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c
index 3a1f2fea831..19afd65e2c8 100644
--- a/src/gallium/drivers/r600/r600_texture.c
+++ b/src/gallium/drivers/r600/r600_texture.c
@@ -284,50 +284,51 @@ void r600_texture_get_fmask_info(struct r600_screen *rscreen,
 				 unsigned nr_samples,
 				 struct r600_fmask_info *out)
 {
-	/* FMASK is allocated pretty much like an ordinary texture.
-	 * Here we use bpe in the units of bits, not bytes. */
+	/* FMASK is allocated like an ordinary texture. */
 	struct radeon_surface fmask = rtex->surface;
 
+	memset(out, 0, sizeof(*out));
+
+	fmask.bo_alignment = 0;
+	fmask.bo_size = 0;
+	fmask.nsamples = 1;
+	fmask.flags |= RADEON_SURF_FMASK;
+
 	switch (nr_samples) {
 	case 2:
-		/* This should be 8,1, but we should set nsamples > 1
-		 * for the allocator to treat it as a multisample surface.
-		 * Let's set 4,2 then. */
 	case 4:
-		fmask.bpe = 4;
-		fmask.nsamples = 2;
+		fmask.bpe = 1;
+		fmask.bankh = 4;
 		break;
 	case 8:
-		fmask.bpe = 8;
-		fmask.nsamples = 4;
-		break;
-	case 16:
-		fmask.bpe = 16;
-		fmask.nsamples = 4;
+		fmask.bpe = 4;
 		break;
 	default:
 		R600_ERR("Invalid sample count for FMASK allocation.\n");
 		return;
 	}
 
-	/* R600-R700 errata? Anyway, this fixes colorbuffer corruption. */
+	/* Overallocate FMASK on R600-R700 to fix colorbuffer corruption.
+	 * This can be fixed by writing a separate FMASK allocator specifically
+	 * for R600-R700 asics. */
 	if (rscreen->chip_class <= R700) {
 		fmask.bpe *= 2;
 	}
 
-	if (rscreen->chip_class >= EVERGREEN) {
-		fmask.bankh = nr_samples <= 4 ? 4 : 1;
-	}
-
 	if (rscreen->ws->surface_init(rscreen->ws, &fmask)) {
 		R600_ERR("Got error in surface_init while allocating FMASK.\n");
 		return;
 	}
+
 	assert(fmask.level[0].mode == RADEON_SURF_MODE_2D);
 
+	out->slice_tile_max = (fmask.level[0].nblk_x * fmask.level[0].nblk_y) / 64;
+	if (out->slice_tile_max)
+		out->slice_tile_max -= 1;
+
 	out->bank_height = fmask.bankh;
 	out->alignment = MAX2(256, fmask.bo_alignment);
-	out->size = (fmask.bo_size + 7) / 8;
+	out->size = fmask.bo_size;
 }
 
 static void r600_texture_allocate_fmask(struct r600_screen *rscreen,
@@ -339,6 +340,7 @@ static void r600_texture_allocate_fmask(struct r600_screen *rscreen,
 				    rtex->resource.b.b.nr_samples, &fmask);
 
 	rtex->fmask_bank_height = fmask.bank_height;
+	rtex->fmask_slice_tile_max = fmask.slice_tile_max;
 	rtex->fmask_offset = align(rtex->size, fmask.alignment);
 	rtex->fmask_size = fmask.size;
 	rtex->size = rtex->fmask_offset + rtex->fmask_size;