mesa/st: add tgsi-lowering code for depth-clamp

This is a TGSI pass that lowers depth-clamping into shader-operations,
by replacing the depth-value with 0 (a z-coordinate of zero will always
pass the OpenGL depth test conditions), and using a dedicated varying to
interpolate the real depth-value instead. Finally we replace the
depth-output in the fragment shader.

v1 implemented by Erik Faye-Lund <[email protected]>
v2: Add support for handling depth clip mode, and refactor code
v3: - Rename *_vs functions to *_last_vertex_stage (Erik)
    - Use 0.0 depth to avoid clipping (Erik)
v4: Fix inversion of bool value for clip control property

Signed-off-by: Gert Wollny <[email protected]>
Reviewed-by: Reviewed-by: Marek Olšák <[email protected]>

4 files changed, 451 insertions, 0 deletions

diff --git a/src/mesa/Makefile.sources b/src/mesa/Makefile.sources
index b0b43a78ce9..acb0c2b53cf 100644
--- a/src/mesa/Makefile.sources
+++ b/src/mesa/Makefile.sources
@@ -556,6 +556,8 @@ STATETRACKER_FILES = \
 	state_tracker/st_shader_cache.h \
 	state_tracker/st_texture.c \
 	state_tracker/st_texture.h \
+	state_tracker/st_tgsi_lower_depth_clamp.c \
+	state_tracker/st_tgsi_lower_depth_clamp.h \
 	state_tracker/st_tgsi_lower_yuv.c \
 	state_tracker/st_tgsi_lower_yuv.h \
 	state_tracker/st_util.h \
diff --git a/src/mesa/meson.build b/src/mesa/meson.build
index 4a89358a267..563bdd7c103 100644
--- a/src/mesa/meson.build
+++ b/src/mesa/meson.build
@@ -600,6 +600,8 @@ files_libmesa_gallium = files(
   'state_tracker/st_shader_cache.h',
   'state_tracker/st_texture.c',
   'state_tracker/st_texture.h',
+  'state_tracker/st_tgsi_lower_depth_clamp.c',
+  'state_tracker/st_tgsi_lower_depth_clamp.h',
   'state_tracker/st_tgsi_lower_yuv.c',
   'state_tracker/st_tgsi_lower_yuv.h',
   'state_tracker/st_util.h',
diff --git a/src/mesa/state_tracker/st_tgsi_lower_depth_clamp.c b/src/mesa/state_tracker/st_tgsi_lower_depth_clamp.c
new file mode 100644
index 00000000000..b89ee986d33
--- /dev/null
+++ b/src/mesa/state_tracker/st_tgsi_lower_depth_clamp.c
@@ -0,0 +1,408 @@
+/*
+ * Copyright © 2018 Collabora Ltd
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * 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 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 "st_tgsi_lower_depth_clamp.h"
+#include "tgsi/tgsi_transform.h"
+#include "tgsi/tgsi_scan.h"
+
+struct tgsi_depth_clamp_transform {
+   struct tgsi_transform_context base;
+
+   struct tgsi_shader_info info;
+
+   int depth_range_const;
+   int next_generic;
+   int imm;
+   int depth_var;
+   int pos_input;
+   int pos_output;
+   int pos_input_temp;
+   int pos_output_temp;
+   int depth_range_corrected;
+   bool depth_clip_minus_one_to_one;
+};
+
+static inline struct tgsi_depth_clamp_transform *
+tgsi_depth_clamp_transform(struct tgsi_transform_context *tctx)
+{
+   return (struct tgsi_depth_clamp_transform *)tctx;
+}
+
+static void
+transform_decl(struct tgsi_transform_context *tctx,
+               struct tgsi_full_declaration *decl)
+{
+   struct tgsi_depth_clamp_transform *ctx = tgsi_depth_clamp_transform(tctx);
+
+   /* find the next generic index usable for our inserted varying */
+   if (ctx->info.processor == PIPE_SHADER_FRAGMENT) {
+      if (decl->Declaration.File == TGSI_FILE_INPUT &&
+          decl->Semantic.Name == TGSI_SEMANTIC_GENERIC)
+         ctx->next_generic = MAX2(ctx->next_generic, decl->Semantic.Index + 1);
+   } else {
+      if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+          decl->Semantic.Name == TGSI_SEMANTIC_GENERIC)
+         ctx->next_generic = MAX2(ctx->next_generic, decl->Semantic.Index + 1);
+   }
+
+   if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+       decl->Semantic.Name == TGSI_SEMANTIC_POSITION) {
+      assert(decl->Semantic.Index == 0);
+      ctx->pos_output = decl->Range.First;
+   } else if (decl->Declaration.File == TGSI_FILE_INPUT &&
+              decl->Semantic.Name == TGSI_SEMANTIC_POSITION) {
+      assert(decl->Semantic.Index == 0);
+      if (ctx->info.processor == PIPE_SHADER_FRAGMENT)
+         ctx->pos_input = decl->Range.First;
+   }
+
+   tctx->emit_declaration(tctx, decl);
+}
+
+static void
+prolog_common(struct tgsi_depth_clamp_transform *ctx)
+{
+   assert(ctx->depth_range_const >= 0);
+   if (ctx->info.const_file_max[0] < ctx->depth_range_const)
+      tgsi_transform_const_decl(&ctx->base, ctx->depth_range_const,
+                                ctx->depth_range_const);
+
+   /* declare a temp for the position-output */
+   ctx->pos_output_temp = ctx->info.file_max[TGSI_FILE_TEMPORARY] + 1;
+   tgsi_transform_temp_decl(&ctx->base, ctx->pos_output_temp);
+}
+
+static void
+prolog_last_vertex_stage(struct tgsi_transform_context *tctx)
+{
+   struct tgsi_depth_clamp_transform *ctx = tgsi_depth_clamp_transform(tctx);
+
+   prolog_common(ctx);
+
+   ctx->imm = ctx->info.immediate_count;
+   tgsi_transform_immediate_decl(tctx, 0.5, 0.0, 0.0, 0.0);
+
+   /* declare the output */
+   ctx->depth_var = ctx->info.num_outputs;
+   tgsi_transform_output_decl(tctx, ctx->depth_var,
+                              TGSI_SEMANTIC_GENERIC,
+                              ctx->next_generic,
+                              TGSI_INTERPOLATE_LINEAR);
+}
+
+static void
+epilog_last_vertex_stage(struct tgsi_transform_context *tctx)
+{
+   struct tgsi_depth_clamp_transform *ctx = tgsi_depth_clamp_transform(tctx);
+
+   int mad_dst_file = TGSI_FILE_TEMPORARY;
+   int mad_dst_index = ctx->pos_output_temp;
+
+   if (!ctx->depth_clip_minus_one_to_one) {
+      mad_dst_file = TGSI_FILE_OUTPUT;
+      mad_dst_index = ctx->depth_var;
+   }
+
+   /* move from temp-register to output */
+   tgsi_transform_op1_inst(tctx, TGSI_OPCODE_MOV,
+                           TGSI_FILE_OUTPUT, ctx->pos_output,
+                           TGSI_WRITEMASK_XYZW,
+                           TGSI_FILE_TEMPORARY, ctx->pos_output_temp);
+
+   /* Set gl_position.z to 0.0 to avoid clipping */
+   tgsi_transform_op1_swz_inst(tctx, TGSI_OPCODE_MOV,
+                               TGSI_FILE_OUTPUT, ctx->pos_output,
+                               TGSI_WRITEMASK_Z,
+                               TGSI_FILE_IMMEDIATE, ctx->imm,
+                               TGSI_SWIZZLE_Y);
+
+   /* Evaluate and pass true depth value in depthRange terms */
+   /* z = gl_Position.z / gl_Position.w */
+
+   struct tgsi_full_instruction inst;
+
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_DIV;
+   inst.Instruction.NumDstRegs = 1;
+   inst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+   inst.Dst[0].Register.Index = ctx->pos_output_temp;
+   inst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X;
+   inst.Instruction.NumSrcRegs = 2;
+   tgsi_transform_src_reg_xyzw(&inst.Src[0], TGSI_FILE_TEMPORARY, ctx->pos_output_temp);
+   tgsi_transform_src_reg_xyzw(&inst.Src[1], TGSI_FILE_TEMPORARY, ctx->pos_output_temp);
+   inst.Src[0].Register.SwizzleX =
+         inst.Src[0].Register.SwizzleY =
+         inst.Src[0].Register.SwizzleZ =
+         inst.Src[0].Register.SwizzleW = TGSI_SWIZZLE_Z;
+
+   inst.Src[1].Register.SwizzleX =
+         inst.Src[1].Register.SwizzleY =
+         inst.Src[1].Register.SwizzleZ =
+         inst.Src[1].Register.SwizzleW = TGSI_SWIZZLE_W;
+
+   tctx->emit_instruction(tctx, &inst);
+
+
+   /* OpenGL Core Profile 4.5 - 13.6.1
+    * The vertex's windows z coordinate zw is given by zw = s * z + b.
+    *
+    * *  With clip control depth mode ZERO_TO_ONE
+    *      s = f - n, b = n, and hence
+    *
+    *     zw_0_1 = z * gl_DepthRange.diff + gl_DepthRange.near
+    */
+   tgsi_transform_op3_swz_inst(tctx, TGSI_OPCODE_MAD,
+                               mad_dst_file, mad_dst_index,
+                               TGSI_WRITEMASK_X,
+                               TGSI_FILE_TEMPORARY, ctx->pos_output_temp,
+                               TGSI_SWIZZLE_X,
+                               false,
+                               TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                               TGSI_SWIZZLE_Z,
+                               TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                               TGSI_SWIZZLE_X);
+
+   /* If clip control depth mode is NEGATIVE_ONE_TO_ONE, then
+   *     s = 0.5 * (f - n), b = 0.5 * (n + f), and hence
+   *
+   *     zw_m1_1 = 0.5 * (zw_01 + gl_DepthRange.far)
+   */
+   if (ctx->depth_clip_minus_one_to_one) {
+       /* z += gl_DepthRange.far */
+      tgsi_transform_op2_swz_inst(tctx, TGSI_OPCODE_ADD,
+                                  TGSI_FILE_TEMPORARY, ctx->pos_output_temp,
+                                  TGSI_WRITEMASK_X,
+                                  TGSI_FILE_TEMPORARY, ctx->pos_output_temp,
+                                  TGSI_SWIZZLE_X,
+                                  TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                                  TGSI_SWIZZLE_Y, false);
+      /* z *=  0.5 */
+      tgsi_transform_op2_swz_inst(tctx, TGSI_OPCODE_MUL,
+                                  TGSI_FILE_OUTPUT, ctx->depth_var,
+                                  TGSI_WRITEMASK_X,
+                                  TGSI_FILE_TEMPORARY, ctx->pos_output_temp,
+                                  TGSI_SWIZZLE_X,
+                                  TGSI_FILE_IMMEDIATE, ctx->imm,
+                                  TGSI_SWIZZLE_X, false);
+   }
+}
+
+
+static void
+prolog_fs(struct tgsi_transform_context *tctx)
+{
+   struct tgsi_depth_clamp_transform *ctx = tgsi_depth_clamp_transform(tctx);
+
+   prolog_common(ctx);
+
+   ctx->depth_range_corrected = ctx->info.file_max[TGSI_FILE_TEMPORARY] + 2;
+   tgsi_transform_temp_decl(tctx, ctx->depth_range_corrected);
+
+   /* declare the input */
+   ctx->depth_var = ctx->info.num_inputs;
+   tgsi_transform_input_decl(tctx, ctx->depth_var,
+                             TGSI_SEMANTIC_GENERIC,
+                             ctx->next_generic,
+                             TGSI_INTERPOLATE_LINEAR);
+
+   /* declare the output */
+   if (ctx->pos_output < 0) {
+      ctx->pos_output = ctx->info.num_outputs;
+      tgsi_transform_output_decl(tctx, ctx->pos_output,
+                                 TGSI_SEMANTIC_POSITION,
+                                 0,
+                                 TGSI_INTERPOLATE_LINEAR);
+   }
+
+   if (ctx->info.reads_z) {
+      ctx->pos_input_temp = ctx->info.file_max[TGSI_FILE_TEMPORARY] + 3;
+      tgsi_transform_temp_decl(tctx, ctx->pos_input_temp);
+
+      assert(ctx->pos_input_temp >= 0);
+      /* copy normal position */
+      tgsi_transform_op1_inst(tctx, TGSI_OPCODE_MOV,
+                              TGSI_FILE_TEMPORARY, ctx->pos_input_temp,
+                              TGSI_WRITEMASK_XYZW,
+                              TGSI_FILE_INPUT, ctx->pos_input);
+      /* replace z-component with varying */
+      tgsi_transform_op1_swz_inst(tctx, TGSI_OPCODE_MOV,
+                                  TGSI_FILE_TEMPORARY, ctx->pos_input_temp,
+                                  TGSI_WRITEMASK_Z,
+                                  TGSI_FILE_INPUT, ctx->depth_var,
+                                  TGSI_SWIZZLE_X);
+   }
+}
+
+static void
+epilog_fs(struct tgsi_transform_context *tctx)
+{
+   struct tgsi_depth_clamp_transform *ctx = tgsi_depth_clamp_transform(tctx);
+
+   unsigned src0_file = TGSI_FILE_INPUT;
+   unsigned src0_index = ctx->depth_var;
+   unsigned src0_swizzle = TGSI_SWIZZLE_X;
+
+   if (ctx->info.writes_z) {
+      src0_file = TGSI_FILE_TEMPORARY;
+      src0_index = ctx->pos_output_temp;
+      src0_swizzle = TGSI_SWIZZLE_Z;
+   }
+
+   /* it is possible to have gl_DepthRange.near > gl_DepthRange.far, so first
+    * we have to sort the two */
+   tgsi_transform_op2_swz_inst(tctx, TGSI_OPCODE_MIN,
+                               TGSI_FILE_TEMPORARY, ctx->depth_range_corrected,
+                               TGSI_WRITEMASK_X,
+                               TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                               TGSI_SWIZZLE_X,
+                               TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                               TGSI_SWIZZLE_Y,
+                               false);
+
+   tgsi_transform_op2_swz_inst(tctx, TGSI_OPCODE_MAX,
+                               TGSI_FILE_TEMPORARY, ctx->depth_range_corrected,
+                               TGSI_WRITEMASK_Y,
+                               TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                               TGSI_SWIZZLE_X,
+                               TGSI_FILE_CONSTANT, ctx->depth_range_const,
+                               TGSI_SWIZZLE_Y,
+                               false);
+
+   /* gl_FragDepth = max(gl_FragDepth, min(gl_DepthRange.near, gl_DepthRange.far)) */
+   tgsi_transform_op2_swz_inst(tctx, TGSI_OPCODE_MAX,
+                               TGSI_FILE_TEMPORARY, ctx->pos_output_temp,
+                               TGSI_WRITEMASK_X,
+                               src0_file, src0_index, src0_swizzle,
+                               TGSI_FILE_TEMPORARY, ctx->depth_range_corrected,
+                               TGSI_SWIZZLE_X, false);
+
+   /* gl_FragDepth = min(gl_FragDepth, max(gl_DepthRange.near, gl_DepthRange.far)) */
+   tgsi_transform_op2_swz_inst(tctx, TGSI_OPCODE_MIN,
+                               TGSI_FILE_OUTPUT, ctx->pos_output,
+                               TGSI_WRITEMASK_Z,
+                               TGSI_FILE_TEMPORARY, ctx->pos_output_temp,
+                               TGSI_SWIZZLE_X,
+                               TGSI_FILE_TEMPORARY, ctx->depth_range_corrected,
+                               TGSI_SWIZZLE_Y, false);
+}
+
+static void
+transform_instr(struct tgsi_transform_context *tctx,
+                struct tgsi_full_instruction *inst)
+{
+   struct tgsi_depth_clamp_transform *ctx = tgsi_depth_clamp_transform(tctx);
+
+   if (ctx->pos_output >= 0) {
+      /* replace writes to gl_Position / gl_FragDepth with a temp-variable
+       */
+      for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) {
+         if (inst->Dst[i].Register.File == TGSI_FILE_OUTPUT &&
+             inst->Dst[i].Register.Index == ctx->pos_output) {
+            inst->Dst[i].Register.File = TGSI_FILE_TEMPORARY;
+            inst->Dst[i].Register.Index = ctx->pos_output_temp;
+         }
+      }
+   }
+
+   if (ctx->info.reads_z) {
+      /* replace reads from gl_FragCoord with temp-variable
+       */
+      assert(ctx->pos_input_temp >= 0);
+      for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) {
+         if (inst->Src[i].Register.File == TGSI_FILE_INPUT &&
+             inst->Src[i].Register.Index == ctx->pos_input) {
+            inst->Src[i].Register.File = TGSI_FILE_TEMPORARY;
+            inst->Src[i].Register.Index = ctx->pos_input_temp;
+         }
+      }
+   }
+
+   /* In a GS each we have to add the z-write opilog for each emit
+    */
+   if (ctx->info.processor == PIPE_SHADER_GEOMETRY &&
+       inst->Instruction.Opcode == TGSI_OPCODE_EMIT)
+      epilog_last_vertex_stage(tctx);
+
+   tctx->emit_instruction(tctx, inst);
+}
+
+const struct tgsi_token *
+st_tgsi_lower_depth_clamp(const struct tgsi_token *tokens,
+                          int depth_range_const,
+                          bool clip_negative_one_to_one)
+{
+   struct tgsi_depth_clamp_transform ctx = {};
+   struct tgsi_token *newtoks;
+   int newlen;
+
+   tgsi_scan_shader(tokens, &ctx.info);
+
+   /* we only want to do this for the fragment shader, and the shader-stage
+    * right before it, but in the first pass there might be no "next" shader
+    */
+   if (ctx.info.processor != PIPE_SHADER_FRAGMENT &&
+       ctx.info.processor != PIPE_SHADER_GEOMETRY &&
+       ctx.info.processor != PIPE_SHADER_VERTEX &&
+       ctx.info.processor != PIPE_SHADER_TESS_EVAL &&
+       (ctx.info.properties[TGSI_PROPERTY_NEXT_SHADER] > PIPE_SHADER_VERTEX &&
+       (ctx.info.properties[TGSI_PROPERTY_NEXT_SHADER] != PIPE_SHADER_FRAGMENT)))  {
+      return tokens;
+   }
+
+   ctx.base.transform_declaration = transform_decl;
+   ctx.base.transform_instruction = transform_instr;
+
+   if (ctx.info.processor == PIPE_SHADER_FRAGMENT) {
+      ctx.base.prolog = prolog_fs;
+      ctx.base.epilog = epilog_fs;
+   } else {
+      ctx.base.prolog = prolog_last_vertex_stage;
+      ctx.base.epilog = epilog_last_vertex_stage;
+   }
+
+   ctx.pos_output = ctx.pos_input = -1;
+   ctx.depth_range_const = depth_range_const;
+   ctx.depth_clip_minus_one_to_one = clip_negative_one_to_one;
+
+   /* We add approximately 30 tokens per Z write, so add this per vertex in
+    * a GS and some additional tokes for VS and TES
+    */
+   newlen = tgsi_num_tokens(tokens) +
+            30 * ctx.info.properties[TGSI_PROPERTY_GS_MAX_OUTPUT_VERTICES] +
+            120;
+
+   newtoks = tgsi_alloc_tokens(newlen);
+   if (!newtoks)
+      return tokens;
+
+   tgsi_transform_shader(tokens, newtoks, newlen, &ctx.base);
+
+   return newtoks;
+}
+
+const struct tgsi_token *
+st_tgsi_lower_depth_clamp_fs(const struct tgsi_token *tokens,
+                             int depth_range_const)
+{
+   return st_tgsi_lower_depth_clamp(tokens, depth_range_const, false);
+}
\ No newline at end of file
diff --git a/src/mesa/state_tracker/st_tgsi_lower_depth_clamp.h b/src/mesa/state_tracker/st_tgsi_lower_depth_clamp.h
new file mode 100644
index 00000000000..acfb97569f6
--- /dev/null
+++ b/src/mesa/state_tracker/st_tgsi_lower_depth_clamp.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright © 2018 Collabora Ltd
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * 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 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.
+ */
+
+#ifndef ST_TGSI_LOWER_DEPTH_CLAMP_H
+#define ST_TGSI_LOWER_DEPTH_CLAMP_H
+
+#include <stdbool.h>
+struct tgsi_token;
+
+const struct tgsi_token *
+st_tgsi_lower_depth_clamp(const struct tgsi_token *tokens,
+                          int depth_range_const,
+                          bool clip_negative_one_to_one);
+
+const struct tgsi_token *
+st_tgsi_lower_depth_clamp_fs(const struct tgsi_token *tokens,
+                             int depth_range_const);
+
+#endif /* ST_TGSI_LOWER_DEPTH_CLAMP_H */