2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics to
4 develop this 3D driver.
6 Permission is hereby granted, free of charge, to any person obtaining
7 a copy of this software and associated documentation files (the
8 "Software"), to deal in the Software without restriction, including
9 without limitation the rights to use, copy, modify, merge, publish,
10 distribute, sublicense, and/or sell copies of the Software, and to
11 permit persons to whom the Software is furnished to do so, subject to
12 the following conditions:
14 The above copyright notice and this permission notice (including the
15 next paragraph) shall be included in all copies or substantial
16 portions of the Software.
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **********************************************************************/
29 * Keith Whitwell <keithw@vmware.com>
34 #include "intel_fbo.h"
35 #include "brw_context.h"
36 #include "brw_state.h"
37 #include "brw_defines.h"
40 /***********************************************************************
41 * WM unit - fragment programs and rasterization
45 brw_color_buffer_write_enabled(struct brw_context
*brw
)
47 struct gl_context
*ctx
= &brw
->ctx
;
48 /* BRW_NEW_FRAGMENT_PROGRAM */
49 const struct gl_fragment_program
*fp
= brw
->fragment_program
;
53 for (i
= 0; i
< ctx
->DrawBuffer
->_NumColorDrawBuffers
; i
++) {
54 struct gl_renderbuffer
*rb
= ctx
->DrawBuffer
->_ColorDrawBuffers
[i
];
58 (fp
->Base
.OutputsWritten
& BITFIELD64_BIT(FRAG_RESULT_COLOR
) ||
59 fp
->Base
.OutputsWritten
& BITFIELD64_BIT(FRAG_RESULT_DATA0
+ i
)) &&
60 (ctx
->Color
.ColorMask
[i
][0] ||
61 ctx
->Color
.ColorMask
[i
][1] ||
62 ctx
->Color
.ColorMask
[i
][2] ||
63 ctx
->Color
.ColorMask
[i
][3])) {
72 * Setup wm hardware state. See page 225 of Volume 2
75 brw_upload_wm_unit(struct brw_context
*brw
)
77 struct gl_context
*ctx
= &brw
->ctx
;
78 const struct gl_fragment_program
*fp
= brw
->fragment_program
;
79 struct brw_wm_unit_state
*wm
;
81 wm
= brw_state_batch(brw
, AUB_TRACE_WM_STATE
,
82 sizeof(*wm
), 32, &brw
->wm
.base
.state_offset
);
83 memset(wm
, 0, sizeof(*wm
));
85 if (brw
->wm
.prog_data
->prog_offset_16
) {
86 /* These two fields should be the same pre-gen6, which is why we
87 * only have one hardware field to program for both dispatch
90 assert(brw
->wm
.prog_data
->first_curbe_grf
==
91 brw
->wm
.prog_data
->first_curbe_grf_16
);
94 /* BRW_NEW_PROGRAM_CACHE | CACHE_NEW_WM_PROG */
95 wm
->thread0
.grf_reg_count
= brw
->wm
.prog_data
->reg_blocks
;
96 wm
->wm9
.grf_reg_count_2
= brw
->wm
.prog_data
->reg_blocks_16
;
98 wm
->thread0
.kernel_start_pointer
=
99 brw_program_reloc(brw
,
100 brw
->wm
.base
.state_offset
+
101 offsetof(struct brw_wm_unit_state
, thread0
),
102 brw
->wm
.base
.prog_offset
+
103 (wm
->thread0
.grf_reg_count
<< 1)) >> 6;
105 wm
->wm9
.kernel_start_pointer_2
=
106 brw_program_reloc(brw
,
107 brw
->wm
.base
.state_offset
+
108 offsetof(struct brw_wm_unit_state
, wm9
),
109 brw
->wm
.base
.prog_offset
+
110 brw
->wm
.prog_data
->prog_offset_16
+
111 (wm
->wm9
.grf_reg_count_2
<< 1)) >> 6;
113 wm
->thread1
.depth_coef_urb_read_offset
= 1;
114 /* Use ALT floating point mode for ARB fragment programs, because they
115 * require 0^0 == 1. Even though _CurrentFragmentProgram is used for
116 * rendering, CurrentProgram[MESA_SHADER_FRAGMENT] is used for this check
117 * to differentiate between the GLSL and non-GLSL cases.
119 if (ctx
->Shader
.CurrentProgram
[MESA_SHADER_FRAGMENT
] == NULL
)
120 wm
->thread1
.floating_point_mode
= BRW_FLOATING_POINT_NON_IEEE_754
;
122 wm
->thread1
.floating_point_mode
= BRW_FLOATING_POINT_IEEE_754
;
124 wm
->thread1
.binding_table_entry_count
=
125 brw
->wm
.prog_data
->base
.binding_table
.size_bytes
/ 4;
127 if (brw
->wm
.prog_data
->total_scratch
!= 0) {
128 wm
->thread2
.scratch_space_base_pointer
=
129 brw
->wm
.base
.scratch_bo
->offset64
>> 10; /* reloc */
130 wm
->thread2
.per_thread_scratch_space
=
131 ffs(brw
->wm
.prog_data
->total_scratch
) - 11;
133 wm
->thread2
.scratch_space_base_pointer
= 0;
134 wm
->thread2
.per_thread_scratch_space
= 0;
137 wm
->thread3
.dispatch_grf_start_reg
= brw
->wm
.prog_data
->first_curbe_grf
;
138 wm
->thread3
.urb_entry_read_length
=
139 brw
->wm
.prog_data
->num_varying_inputs
* 2;
140 wm
->thread3
.urb_entry_read_offset
= 0;
141 wm
->thread3
.const_urb_entry_read_length
=
142 brw
->wm
.prog_data
->curb_read_length
;
143 /* BRW_NEW_CURBE_OFFSETS */
144 wm
->thread3
.const_urb_entry_read_offset
= brw
->curbe
.wm_start
* 2;
147 wm
->wm4
.sampler_count
= 0; /* hardware requirement */
149 /* CACHE_NEW_SAMPLER */
150 wm
->wm4
.sampler_count
= (brw
->wm
.base
.sampler_count
+ 1) / 4;
153 if (brw
->wm
.base
.sampler_count
) {
155 wm
->wm4
.sampler_state_pointer
= (brw
->batch
.bo
->offset64
+
156 brw
->wm
.base
.sampler_offset
) >> 5;
158 wm
->wm4
.sampler_state_pointer
= 0;
161 /* BRW_NEW_FRAGMENT_PROGRAM */
162 wm
->wm5
.program_uses_depth
= (fp
->Base
.InputsRead
&
163 (1 << VARYING_SLOT_POS
)) != 0;
164 wm
->wm5
.program_computes_depth
= (fp
->Base
.OutputsWritten
&
165 BITFIELD64_BIT(FRAG_RESULT_DEPTH
)) != 0;
167 * Override for NULL depthbuffer case, required by the Pixel Shader Computed
170 if (!intel_get_renderbuffer(ctx
->DrawBuffer
, BUFFER_DEPTH
))
171 wm
->wm5
.program_computes_depth
= 0;
174 wm
->wm5
.program_uses_killpixel
= fp
->UsesKill
|| ctx
->Color
.AlphaEnabled
;
176 wm
->wm5
.enable_8_pix
= 1;
177 if (brw
->wm
.prog_data
->prog_offset_16
)
178 wm
->wm5
.enable_16_pix
= 1;
180 wm
->wm5
.max_threads
= brw
->max_wm_threads
- 1;
182 /* _NEW_BUFFERS | _NEW_COLOR */
183 if (brw_color_buffer_write_enabled(brw
) ||
184 wm
->wm5
.program_uses_killpixel
||
185 wm
->wm5
.program_computes_depth
) {
186 wm
->wm5
.thread_dispatch_enable
= 1;
189 wm
->wm5
.legacy_line_rast
= 0;
190 wm
->wm5
.legacy_global_depth_bias
= 0;
191 wm
->wm5
.early_depth_test
= 1; /* never need to disable */
192 wm
->wm5
.line_aa_region_width
= 0;
193 wm
->wm5
.line_endcap_aa_region_width
= 1;
195 /* _NEW_POLYGONSTIPPLE */
196 wm
->wm5
.polygon_stipple
= ctx
->Polygon
.StippleFlag
;
199 if (ctx
->Polygon
.OffsetFill
) {
200 wm
->wm5
.depth_offset
= 1;
201 /* Something wierd going on with legacy_global_depth_bias,
202 * offset_constant, scaling and MRD. This value passes glean
203 * but gives some odd results elsewere (eg. the
204 * quad-offset-units test).
206 wm
->global_depth_offset_constant
= ctx
->Polygon
.OffsetUnits
* 2;
208 /* This is the only value that passes glean:
210 wm
->global_depth_offset_scale
= ctx
->Polygon
.OffsetFactor
;
214 wm
->wm5
.line_stipple
= ctx
->Line
.StippleFlag
;
216 /* BRW_NEW_STATS_WM */
217 if (unlikely(INTEL_DEBUG
& DEBUG_STATS
) || brw
->stats_wm
)
218 wm
->wm4
.stats_enable
= 1;
220 /* Emit scratch space relocation */
221 if (brw
->wm
.prog_data
->total_scratch
!= 0) {
222 drm_intel_bo_emit_reloc(brw
->batch
.bo
,
223 brw
->wm
.base
.state_offset
+
224 offsetof(struct brw_wm_unit_state
, thread2
),
225 brw
->wm
.base
.scratch_bo
,
226 wm
->thread2
.per_thread_scratch_space
,
227 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
);
230 /* Emit sampler state relocation */
231 if (brw
->wm
.base
.sampler_count
!= 0) {
232 drm_intel_bo_emit_reloc(brw
->batch
.bo
,
233 brw
->wm
.base
.state_offset
+
234 offsetof(struct brw_wm_unit_state
, wm4
),
235 brw
->batch
.bo
, (brw
->wm
.base
.sampler_offset
|
236 wm
->wm4
.stats_enable
|
237 (wm
->wm4
.sampler_count
<< 2)),
238 I915_GEM_DOMAIN_INSTRUCTION
, 0);
241 brw
->state
.dirty
.cache
|= CACHE_NEW_WM_UNIT
;
244 const struct brw_tracked_state brw_wm_unit
= {
246 .mesa
= (_NEW_POLYGON
|
247 _NEW_POLYGONSTIPPLE
|
252 .brw
= (BRW_NEW_BATCH
|
253 BRW_NEW_PROGRAM_CACHE
|
254 BRW_NEW_FRAGMENT_PROGRAM
|
255 BRW_NEW_CURBE_OFFSETS
|
258 .cache
= (CACHE_NEW_WM_PROG
|
261 .emit
= brw_upload_wm_unit
,