}
}
-static struct brw_address
+UNUSED static struct brw_address
rw_bo(struct brw_bo *bo, uint32_t offset)
{
return (struct brw_address) {
};
}
+static struct brw_address
+rw_32_bo(struct brw_bo *bo, uint32_t offset)
+{
+ return (struct brw_address) {
+ .bo = bo,
+ .offset = offset,
+ .reloc_flags = RELOC_WRITE | RELOC_32BIT,
+ };
+}
+
+static struct brw_address
+ro_32_bo(struct brw_bo *bo, uint32_t offset)
+{
+ return (struct brw_address) {
+ .bo = bo,
+ .offset = offset,
+ .reloc_flags = RELOC_32BIT,
+ };
+}
+
UNUSED static struct brw_address
ggtt_bo(struct brw_bo *bo, uint32_t offset)
{
struct GENX(VERTEX_BUFFER_STATE) buf_state = {
.VertexBufferIndex = buffer_nr,
.BufferPitch = stride,
- .BufferStartingAddress = ro_bo(bo, start_offset),
+
+ /* The VF cache designers apparently cut corners, and made the cache
+ * only consider the bottom 32 bits of memory addresses. If you happen
+ * to have two vertex buffers which get placed exactly 4 GiB apart and
+ * use them in back-to-back draw calls, you can get collisions. To work
+ * around this problem, we restrict vertex buffers to the low 32 bits of
+ * the address space.
+ */
+ .BufferStartingAddress = ro_32_bo(bo, start_offset),
#if GEN_GEN >= 8
.BufferSize = end_offset - start_offset,
#endif
#endif
#endif
-#if GEN_GEN == 10
+#if GEN_GEN == 11
+ .VertexBufferMOCS = ICL_MOCS_WB,
+#elif GEN_GEN == 10
.VertexBufferMOCS = CNL_MOCS_WB,
#elif GEN_GEN == 9
.VertexBufferMOCS = SKL_MOCS_WB,
ib.CutIndexEnable = brw->prim_restart.enable_cut_index;
#endif
ib.IndexFormat = brw_get_index_type(index_buffer->index_size);
- ib.BufferStartingAddress = ro_bo(brw->ib.bo, 0);
+
+ /* The VF cache designers apparently cut corners, and made the cache
+ * only consider the bottom 32 bits of memory addresses. If you happen
+ * to have two index buffers which get placed exactly 4 GiB apart and
+ * use them in back-to-back draw calls, you can get collisions. To work
+ * around this problem, we restrict index buffers to the low 32 bits of
+ * the address space.
+ */
+ ib.BufferStartingAddress = ro_32_bo(brw->ib.bo, 0);
#if GEN_GEN >= 8
ib.IndexBufferMOCS = GEN_GEN >= 9 ? SKL_MOCS_WB : BDW_MOCS_WB;
ib.BufferSize = brw->ib.size;
#endif
if (wm_prog_data->base.total_scratch) {
- wm.ScratchSpaceBasePointer = rw_bo(stage_state->scratch_bo, 0);
+ wm.ScratchSpaceBasePointer = rw_32_bo(stage_state->scratch_bo, 0);
wm.PerThreadScratchSpace =
ffs(stage_state->per_thread_scratch) - 11;
}
/* ---------------------------------------------------------------------- */
+/* We restrict scratch buffers to the bottom 32 bits of the address space
+ * by using rw_32_bo().
+ *
+ * General State Base Address is a bit broken. If the address + size as
+ * seen by STATE_BASE_ADDRESS overflows 48 bits, the GPU appears to treat
+ * all accesses to the buffer as being out of bounds and returns zero.
+ */
+
#define INIT_THREAD_DISPATCH_FIELDS(pkt, prefix) \
pkt.KernelStartPointer = KSP(brw, stage_state->prog_offset); \
pkt.SamplerCount = \
pkt.FloatingPointMode = stage_prog_data->use_alt_mode; \
\
if (stage_prog_data->total_scratch) { \
- pkt.ScratchSpaceBasePointer = rw_bo(stage_state->scratch_bo, 0); \
+ pkt.ScratchSpaceBasePointer = rw_32_bo(stage_state->scratch_bo, 0); \
pkt.PerThreadScratchSpace = \
ffs(stage_state->per_thread_scratch) - 11; \
} \
assert(vue_prog_data->dispatch_mode == DISPATCH_MODE_SIMD8 ||
vue_prog_data->dispatch_mode == DISPATCH_MODE_4X2_DUAL_OBJECT);
+ assert(GEN_GEN < 11 ||
+ vue_prog_data->dispatch_mode == DISPATCH_MODE_SIMD8);
#if GEN_GEN == 6
/* From the BSpec, 3D Pipeline > Geometry > Vertex Shader > State,
#elif GEN_GEN >= 8
/* _NEW_VIEWPORT | _NEW_BUFFERS: Screen Space Viewport
* The hardware will take the intersection of the drawing rectangle,
- * scissor rectangle, and the viewport extents. We don't need to be
- * smart, and can therefore just program the viewport extents.
+ * scissor rectangle, and the viewport extents. However, emitting
+ * 3DSTATE_DRAWING_RECTANGLE is expensive since it requires a full
+ * pipeline stall so we're better off just being a little more clever
+ * with our viewport so we can emit it once at context creation time.
*/
+ const float viewport_Xmin = MAX2(ctx->ViewportArray[i].X, 0);
+ const float viewport_Ymin = MAX2(ctx->ViewportArray[i].Y, 0);
const float viewport_Xmax =
- ctx->ViewportArray[i].X + ctx->ViewportArray[i].Width;
+ MIN2(ctx->ViewportArray[i].X + ctx->ViewportArray[i].Width, fb_width);
const float viewport_Ymax =
- ctx->ViewportArray[i].Y + ctx->ViewportArray[i].Height;
+ MIN2(ctx->ViewportArray[i].Y + ctx->ViewportArray[i].Height, fb_height);
if (render_to_fbo) {
- sfv.XMinViewPort = ctx->ViewportArray[i].X;
+ sfv.XMinViewPort = viewport_Xmin;
sfv.XMaxViewPort = viewport_Xmax - 1;
- sfv.YMinViewPort = ctx->ViewportArray[i].Y;
+ sfv.YMinViewPort = viewport_Ymin;
sfv.YMaxViewPort = viewport_Ymax - 1;
} else {
- sfv.XMinViewPort = ctx->ViewportArray[i].X;
+ sfv.XMinViewPort = viewport_Xmin;
sfv.XMaxViewPort = viewport_Xmax - 1;
sfv.YMinViewPort = fb_height - viewport_Ymax;
- sfv.YMaxViewPort = fb_height - ctx->ViewportArray[i].Y - 1;
+ sfv.YMaxViewPort = fb_height - viewport_Ymin - 1;
}
#endif
#if GEN_GEN >= 9
/* prepare the active component dwords */
- const int num_inputs = urb_entry_read_length * 2;
- for (int input_index = 0; input_index < num_inputs; input_index++) {
- sbe.AttributeActiveComponentFormat[input_index] = ACTIVE_COMPONENT_XYZW;
- }
+ for (int i = 0; i < 32; i++)
+ sbe.AttributeActiveComponentFormat[i] = ACTIVE_COMPONENT_XYZW;
#endif
}
ps.SampleMask = genX(determine_sample_mask(brw));
#endif
- /* 3DSTATE_PS expects the number of threads per PSD, which is always 64;
- * it implicitly scales for different GT levels (which have some # of
- * PSDs).
+ /* 3DSTATE_PS expects the number of threads per PSD, which is always 64
+ * for pre Gen11 and 128 for gen11+; On gen11+ If a programmed value is
+ * k, it implies 2(k+1) threads. It implicitly scales for different GT
+ * levels (which have some # of PSDs).
*
- * In Gen8 the format is U8-2 whereas in Gen9 it is U8-1.
+ * In Gen8 the format is U8-2 whereas in Gen9+ it is U9-1.
*/
#if GEN_GEN >= 9
ps.MaximumNumberofThreadsPerPSD = 64 - 1;
if (prog_data->base.total_scratch) {
ps.ScratchSpaceBasePointer =
- rw_bo(stage_state->scratch_bo,
- ffs(stage_state->per_thread_scratch) - 11);
+ rw_32_bo(stage_state->scratch_bo,
+ ffs(stage_state->per_thread_scratch) - 11);
}
}
}
if (!tes_prog_data) {
brw_batch_emit(brw, GENX(3DSTATE_DS), ds);
} else {
+ assert(GEN_GEN < 11 ||
+ vue_prog_data->dispatch_mode == DISPATCH_MODE_SIMD8);
+
brw_batch_emit(brw, GENX(3DSTATE_DS), ds) {
INIT_THREAD_DISPATCH_FIELDS(ds, Patch);
*/
per_thread_scratch_value = stage_state->per_thread_scratch / 1024 - 1;
}
- vfe.ScratchSpaceBasePointer = rw_bo(stage_state->scratch_bo, 0);
+ vfe.ScratchSpaceBasePointer = rw_32_bo(stage_state->scratch_bo, 0);
vfe.PerThreadScratchSpace = per_thread_scratch_value;
}
const uint32_t subslices = MAX2(brw->screen->subslice_total, 1);
vfe.MaximumNumberofThreads = devinfo->max_cs_threads * subslices - 1;
vfe.NumberofURBEntries = GEN_GEN >= 8 ? 2 : 0;
+#if GEN_GEN < 11
vfe.ResetGatewayTimer =
Resettingrelativetimerandlatchingtheglobaltimestamp;
+#endif
#if GEN_GEN < 9
vfe.BypassGatewayControl = BypassingOpenGatewayCloseGatewayprotocol;
#endif