* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
- * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
- * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
*/
#include "util/u_format.h"
#include "nvc0_graph_macros.h"
+#ifndef NOUVEAU_GETPARAM_GRAPH_UNITS
+# define NOUVEAU_GETPARAM_GRAPH_UNITS 13
+#endif
+
static boolean
nvc0_screen_is_format_supported(struct pipe_screen *pscreen,
enum pipe_format format,
if (!util_format_is_supported(format, bindings))
return FALSE;
- switch (format) {
- case PIPE_FORMAT_R8G8B8A8_UNORM:
- case PIPE_FORMAT_R8G8B8X8_UNORM:
- /* HACK: GL requires equal formats for MS resolve and window is BGRA */
- if (bindings & PIPE_BIND_RENDER_TARGET)
- return FALSE;
- default:
- break;
- }
-
if ((bindings & PIPE_BIND_SAMPLER_VIEW) && (target != PIPE_BUFFER))
if (util_format_get_blocksizebits(format) == 3 * 32)
return FALSE;
case PIPE_CAP_SEAMLESS_CUBE_MAP:
case PIPE_CAP_CUBE_MAP_ARRAY:
case PIPE_CAP_TEXTURE_BUFFER_OBJECTS:
+ case PIPE_CAP_TEXTURE_MULTISAMPLE:
return 1;
+ case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE:
+ return 65536;
case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE:
return (class_3d >= NVE4_3D_CLASS) ? 1 : 0;
case PIPE_CAP_TWO_SIDED_STENCIL:
case PIPE_CAP_DEPTH_CLIP_DISABLE:
case PIPE_CAP_POINT_SPRITE:
+ case PIPE_CAP_TGSI_TEXCOORD:
return 1;
case PIPE_CAP_SM3:
return 1;
case PIPE_CAP_QUERY_TIME_ELAPSED:
case PIPE_CAP_OCCLUSION_QUERY:
case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME:
+ case PIPE_CAP_QUERY_PIPELINE_STATISTICS:
return 1;
case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS:
return 4;
case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY:
case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY:
case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY:
- case PIPE_CAP_TEXTURE_MULTISAMPLE:
return 0;
case PIPE_CAP_COMPUTE:
return (class_3d >= NVE4_3D_CLASS) ? 1 : 0;
+ case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER:
+ return 1;
+ case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK:
+ return PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_NV50;
default:
NOUVEAU_ERR("unknown PIPE_CAP %d\n", param);
return 0;
case PIPE_SHADER_CAP_MAX_INPUTS:
if (shader == PIPE_SHADER_VERTEX)
return 32;
+ /* NOTE: These only count our slots for GENERIC varyings.
+ * The address space may be larger, but the actual hard limit seems to be
+ * less than what the address space layout permits, so don't add TEXCOORD,
+ * COLOR, etc. here.
+ */
if (shader == PIPE_SHADER_FRAGMENT)
- return (0x200 + 0x20 + 0x80) / 16; /* generic + colors + TexCoords */
- return (0x200 + 0x40 + 0x80) / 16; /* without 0x60 for per-patch inputs */
+ return 0x1f0 / 16;
+ /* Actually this counts CLIPVERTEX, which occupies the last generic slot,
+ * and excludes 0x60 per-patch inputs.
+ */
+ return 0x200 / 16;
case PIPE_SHADER_CAP_MAX_CONSTS:
return 65536 / 16;
case PIPE_SHADER_CAP_MAX_CONST_BUFFERS:
if (screen->blitter)
nvc0_blitter_destroy(screen);
+ if (screen->pm.prog) {
+ screen->pm.prog->code = NULL; /* hardcoded, don't FREE */
+ nvc0_program_destroy(NULL, screen->pm.prog);
+ }
nouveau_bo_ref(NULL, &screen->text);
nouveau_bo_ref(NULL, &screen->uniform_bo);
}
}
+boolean
+nvc0_screen_resize_tls_area(struct nvc0_screen *screen,
+ uint32_t lpos, uint32_t lneg, uint32_t cstack)
+{
+ struct nouveau_bo *bo = NULL;
+ int ret;
+ uint64_t size = (lpos + lneg) * 32 + cstack;
+
+ if (size >= (1 << 20)) {
+ NOUVEAU_ERR("requested TLS size too large: 0x%"PRIx64"\n", size);
+ return FALSE;
+ }
+
+ size *= (screen->base.device->chipset >= 0xe0) ? 64 : 48; /* max warps */
+ size *= screen->mp_count;
+
+ size = align(size, 1 << 17);
+
+ ret = nouveau_bo_new(screen->base.device, NOUVEAU_BO_VRAM, 1 << 17, size,
+ NULL, &bo);
+ if (ret) {
+ NOUVEAU_ERR("failed to allocate TLS area, size: 0x%"PRIx64"\n", size);
+ return FALSE;
+ }
+ nouveau_bo_ref(NULL, &screen->tls);
+ screen->tls = bo;
+ return TRUE;
+}
+
#define FAIL_SCREEN_INIT(str, err) \
do { \
NOUVEAU_ERR(str, err); \
struct pipe_screen *pscreen;
struct nouveau_object *chan;
struct nouveau_pushbuf *push;
+ uint64_t value;
uint32_t obj_class;
int ret;
unsigned i;
pscreen->get_param = nvc0_screen_get_param;
pscreen->get_shader_param = nvc0_screen_get_shader_param;
pscreen->get_paramf = nvc0_screen_get_paramf;
+ pscreen->get_driver_query_info = nvc0_screen_get_driver_query_info;
nvc0_screen_init_resource_functions(pscreen);
BEGIN_NVC0(push, SUBC_2D(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push, screen->eng2d->oclass);
+ BEGIN_NVC0(push, NVC0_2D(SINGLE_GPC), 1);
+ PUSH_DATA (push, 0);
BEGIN_NVC0(push, NVC0_2D(OPERATION), 1);
PUSH_DATA (push, NVC0_2D_OPERATION_SRCCOPY);
BEGIN_NVC0(push, NVC0_2D(CLIP_ENABLE), 1);
PUSH_DATA (push, 0x17);
}
+ IMMED_NVC0(push, NVC0_3D(ZETA_COMP_ENABLE), dev->drm_version >= 0x01000101);
+ BEGIN_NVC0(push, NVC0_3D(RT_COMP_ENABLE(0)), 8);
+ for (i = 0; i < 8; ++i)
+ PUSH_DATA(push, dev->drm_version >= 0x01000101);
+
BEGIN_NVC0(push, NVC0_3D(RT_CONTROL), 1);
PUSH_DATA (push, 1);
PUSH_DATAh(push, screen->uniform_bo->offset + (5 << 16) + (6 << 9));
PUSH_DATA (push, screen->uniform_bo->offset + (5 << 16) + (6 << 9));
- /* max MPs * max warps per MP (TODO: ask kernel) */
- if (screen->eng3d->oclass >= NVE4_3D_CLASS)
- screen->tls_size = 8 * 64 * 32;
- else
- screen->tls_size = 16 * 48 * 32;
- screen->tls_size *= NVC0_CAP_MAX_PROGRAM_TEMPS * 16;
- screen->tls_size = align(screen->tls_size, 1 << 17);
+ if (dev->drm_version >= 0x01000101) {
+ ret = nouveau_getparam(dev, NOUVEAU_GETPARAM_GRAPH_UNITS, &value);
+ if (ret) {
+ NOUVEAU_ERR("NOUVEAU_GETPARAM_GRAPH_UNITS failed.\n");
+ goto fail;
+ }
+ } else {
+ if (dev->chipset >= 0xe0 && dev->chipset < 0xf0)
+ value = (8 << 8) | 4;
+ else
+ value = (16 << 8) | 4;
+ }
+ screen->mp_count = value >> 8;
+ screen->mp_count_compute = screen->mp_count;
- ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 17,
- screen->tls_size, NULL, &screen->tls);
- if (ret)
- goto fail;
+ nvc0_screen_resize_tls_area(screen, 128 * 16, 0, 0x200);
BEGIN_NVC0(push, NVC0_3D(CODE_ADDRESS_HIGH), 2);
PUSH_DATAh(push, screen->text->offset);
BEGIN_NVC0(push, NVC0_3D(TEMP_ADDRESS_HIGH), 4);
PUSH_DATAh(push, screen->tls->offset);
PUSH_DATA (push, screen->tls->offset);
- PUSH_DATA (push, screen->tls_size >> 32);
- PUSH_DATA (push, screen->tls_size);
+ PUSH_DATA (push, screen->tls->size >> 32);
+ PUSH_DATA (push, screen->tls->size);
BEGIN_NVC0(push, NVC0_3D(WARP_TEMP_ALLOC), 1);
PUSH_DATA (push, 0);
BEGIN_NVC0(push, NVC0_3D(LOCAL_BASE), 1);
projects / mesa.git / blobdiff