1 /* 2 * Copyright 2015 Intel Corporation 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 21 * IN THE SOFTWARE. 22 */ 23 24 #include "brw_context.h" 25 #include "brw_state.h" 26 #include "brw_defines.h" 27 #include "intel_batchbuffer.h" 28 #include "program/prog_parameter.h" 29 30 void 31 gen7_upload_constant_state(struct brw_context *brw, 32 const struct brw_stage_state *stage_state, 33 bool active, unsigned opcode) 34 { 35 uint32_t mocs = brw->gen < 8 ? GEN7_MOCS_L3 : 0; 36 37 /* Disable if the shader stage is inactive or there are no push constants. */ 38 active = active && stage_state->push_const_size != 0; 39 40 int dwords = brw->gen >= 8 ? 11 : 7; 41 BEGIN_BATCH(dwords); 42 OUT_BATCH(opcode << 16 | (dwords - 2)); 43 44 /* Workaround for SKL+ (we use option #2 until we have a need for more 45 * constant buffers). This comes from the documentation for 3DSTATE_CONSTANT_* 46 * 47 * The driver must ensure The following case does not occur without a flush 48 * to the 3D engine: 3DSTATE_CONSTANT_* with buffer 3 read length equal to 49 * zero committed followed by a 3DSTATE_CONSTANT_* with buffer 0 read length 50 * not equal to zero committed. Possible ways to avoid this condition 51 * include: 52 * 1. always force buffer 3 to have a non zero read length 53 * 2. always force buffer 0 to a zero read length 54 */ 55 if (brw->gen >= 9 && active) { 56 OUT_BATCH(0); 57 OUT_BATCH(stage_state->push_const_size); 58 } else { 59 OUT_BATCH(active ? stage_state->push_const_size : 0); 60 OUT_BATCH(0); 61 } 62 /* Pointer to the constant buffer. Covered by the set of state flags 63 * from gen6_prepare_wm_contants 64 */ 65 if (brw->gen >= 9 && active) { 66 OUT_BATCH(0); 67 OUT_BATCH(0); 68 OUT_BATCH(0); 69 OUT_BATCH(0); 70 /* XXX: When using buffers other than 0, you need to specify the 71 * graphics virtual address regardless of INSPM/debug bits 72 */ 73 OUT_RELOC64(brw->batch.bo, I915_GEM_DOMAIN_RENDER, 0, 74 stage_state->push_const_offset); 75 OUT_BATCH(0); 76 OUT_BATCH(0); 77 } else if (brw->gen >= 8) { 78 OUT_BATCH(active ? (stage_state->push_const_offset | mocs) : 0); 79 OUT_BATCH(0); 80 OUT_BATCH(0); 81 OUT_BATCH(0); 82 OUT_BATCH(0); 83 OUT_BATCH(0); 84 OUT_BATCH(0); 85 OUT_BATCH(0); 86 } else { 87 OUT_BATCH(active ? (stage_state->push_const_offset | mocs) : 0); 88 OUT_BATCH(0); 89 OUT_BATCH(0); 90 OUT_BATCH(0); 91 } 92 93 ADVANCE_BATCH(); 94 95 /* On SKL+ the new constants don't take effect until the next corresponding 96 * 3DSTATE_BINDING_TABLE_POINTER_* command is parsed so we need to ensure 97 * that is sent 98 */ 99 if (brw->gen >= 9) 100 brw->ctx.NewDriverState |= BRW_NEW_SURFACES; 101 } 102 103 /** 104 * Creates a streamed BO containing the push constants for the VS or GS on 105 * gen6+. 106 * 107 * Push constants are constant values (such as GLSL uniforms) that are 108 * pre-loaded into a shader stage's register space at thread spawn time. 109 * 110 * Not all GLSL uniforms will be uploaded as push constants: The hardware has 111 * a limitation of 32 or 64 EU registers (256 or 512 floats) per stage to be 112 * uploaded as push constants, while GL 4.4 requires at least 1024 components 113 * to be usable for the VS. Plus, currently we always use pull constants 114 * instead of push constants when doing variable-index array access. 115 * 116 * See brw_curbe.c for the equivalent gen4/5 code. 117 */ 118 void 119 gen6_upload_push_constants(struct brw_context *brw, 120 const struct gl_program *prog, 121 const struct brw_stage_prog_data *prog_data, 122 struct brw_stage_state *stage_state, 123 enum aub_state_struct_type type) 124 { 125 struct gl_context *ctx = &brw->ctx; 126 127 if (prog_data->nr_params == 0) { 128 stage_state->push_const_size = 0; 129 } else { 130 /* Updates the ParamaterValues[i] pointers for all parameters of the 131 * basic type of PROGRAM_STATE_VAR. 132 */ 133 /* XXX: Should this happen somewhere before to get our state flag set? */ 134 if (prog) 135 _mesa_load_state_parameters(ctx, prog->Parameters); 136 137 gl_constant_value *param; 138 int i; 139 140 param = brw_state_batch(brw, type, 141 prog_data->nr_params * sizeof(gl_constant_value), 142 32, &stage_state->push_const_offset); 143 144 STATIC_ASSERT(sizeof(gl_constant_value) == sizeof(float)); 145 146 /* _NEW_PROGRAM_CONSTANTS 147 * 148 * Also _NEW_TRANSFORM -- we may reference clip planes other than as a 149 * side effect of dereferencing uniforms, so _NEW_PROGRAM_CONSTANTS 150 * wouldn't be set for them. 151 */ 152 for (i = 0; i < prog_data->nr_params; i++) { 153 param[i] = *prog_data->param[i]; 154 } 155 156 if (0) { 157 fprintf(stderr, "%s constants:\n", 158 _mesa_shader_stage_to_string(stage_state->stage)); 159 for (i = 0; i < prog_data->nr_params; i++) { 160 if ((i & 7) == 0) 161 fprintf(stderr, "g%d: ", 162 prog_data->dispatch_grf_start_reg + i / 8); 163 fprintf(stderr, "%8f ", param[i].f); 164 if ((i & 7) == 7) 165 fprintf(stderr, "\n"); 166 } 167 if ((i & 7) != 0) 168 fprintf(stderr, "\n"); 169 fprintf(stderr, "\n"); 170 } 171 172 stage_state->push_const_size = ALIGN(prog_data->nr_params, 8) / 8; 173 /* We can only push 32 registers of constants at a time. */ 174 175 /* From the SNB PRM (vol2, part 1, section 3.2.1.4: 3DSTATE_CONSTANT_VS: 176 * 177 * "The sum of all four read length fields (each incremented to 178 * represent the actual read length) must be less than or equal to 179 * 32" 180 * 181 * From the IVB PRM (vol2, part 1, section 3.2.1.3: 3DSTATE_CONSTANT_VS: 182 * 183 * "The sum of all four read length fields must be less than or 184 * equal to the size of 64" 185 * 186 * The other shader stages all match the VS's limits. 187 */ 188 assert(stage_state->push_const_size <= 32); 189 } 190 } 191
