[Mesa-dev] [PATCH V2 4/4] i965: Add lowering pass for splitting textureGatherOffsets

[Kenneth Graunke]

On 10/14/2013 01:33 AM, Chris Forbes wrote:
> Rewrites textureGatherOffsets(s, p, offsets) into
> 
>    gvec4(
>       textureGatherOffset(s, p, offsets[0]).w,
>       textureGatherOffset(s, p, offsets[1]).w,
>       textureGatherOffset(s, p, offsets[2]).w,
>       textureGatherOffset(s, p, offsets[3]).w
>       )
> 
> Signed-off-by: Chris Forbes <chrisf at ijw.co.nz>
> ---
>  src/mesa/drivers/dri/i965/Makefile.sources         |  1 +
>  src/mesa/drivers/dri/i965/brw_context.h            |  1 +
>  .../drivers/dri/i965/brw_lower_offset_array.cpp    | 93 ++++++++++++++++++++++
>  src/mesa/drivers/dri/i965/brw_shader.cpp           |  1 +
>  4 files changed, 96 insertions(+)
>  create mode 100644 src/mesa/drivers/dri/i965/brw_lower_offset_array.cpp
> 
> diff --git a/src/mesa/drivers/dri/i965/Makefile.sources b/src/mesa/drivers/dri/i965/Makefile.sources
> index b8e83ef..816ecc1 100644
> --- a/src/mesa/drivers/dri/i965/Makefile.sources
> +++ b/src/mesa/drivers/dri/i965/Makefile.sources
> @@ -68,6 +68,7 @@ i965_FILES = \
>  	brw_gs_surface_state.c \
>  	brw_interpolation_map.c \
>  	brw_lower_texture_gradients.cpp \
> +	brw_lower_offset_array.cpp \
>  	brw_misc_state.c \
>  	brw_object_purgeable.c \
>  	brw_performance_monitor.c \
> diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
> index 5725ef6..84ff552 100644
> --- a/src/mesa/drivers/dri/i965/brw_context.h
> +++ b/src/mesa/drivers/dri/i965/brw_context.h
> @@ -1731,6 +1731,7 @@ brw_program_reloc(struct brw_context *brw, uint32_t state_offset,
>  bool brw_do_cubemap_normalize(struct exec_list *instructions);
>  bool brw_lower_texture_gradients(struct brw_context *brw,
>                                   struct exec_list *instructions);
> +bool brw_do_lower_offset_arrays(struct exec_list *instructions);
>  
>  struct opcode_desc {
>      char    *name;
> diff --git a/src/mesa/drivers/dri/i965/brw_lower_offset_array.cpp b/src/mesa/drivers/dri/i965/brw_lower_offset_array.cpp
> new file mode 100644
> index 0000000..c512406
> --- /dev/null
> +++ b/src/mesa/drivers/dri/i965/brw_lower_offset_array.cpp
> @@ -0,0 +1,93 @@
> +/*
> + * Copyright © 2013 Intel Corporation
> + *
> + * 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.
> + */
> +
> +/**
> + * \file brw_lower_offset_array.cpp
> + *
> + * IR lower pass to decompose ir_texture ir_tg4 with an array of offsets
> + * into four ir_tg4s with a single ivec2 offset, select the .w component of each,
> + * and return those four values packed into a gvec4.
> + *
> + * \author Chris Forbes <chrisf at ijw.co.nz>
> + */
> +
> +#include "glsl/glsl_types.h"
> +#include "glsl/ir.h"
> +#include "program/prog_instruction.h" /* For WRITEMASK_* */
> +
> +class brw_lower_offset_array_visitor : public ir_hierarchical_visitor {
> +public:
> +   brw_lower_offset_array_visitor()
> +   {
> +      progress = false;
> +   }
> +
> +   ir_visitor_status visit_leave(ir_texture *ir);
> +
> +   bool progress;
> +};
> +
> +ir_visitor_status
> +brw_lower_offset_array_visitor::visit_leave(ir_texture *ir)
> +{
> +   if (ir->op != ir_tg4 || !ir->offset || !ir->offset->type->is_array())
> +      return visit_continue;
> +
> +   void *mem_ctx = ralloc_parent(ir);
> +
> +   ir_variable *var = new (mem_ctx) ir_variable(ir->type, "result", ir_var_auto);
> +   base_ir->insert_before(var);
> +
> +   for (int i = 0; i < 4; i++) {
> +      ir_texture *tex = ir->clone(mem_ctx, NULL);
> +      tex->offset = new (mem_ctx) ir_dereference_array(tex->offset,
> +            new (mem_ctx) ir_constant(i));
> +
> +      ir_assignment *assign = new (mem_ctx) ir_assignment(
> +            new (mem_ctx) ir_dereference_variable(var),
> +            new (mem_ctx) ir_swizzle(tex, 3, 0, 0, 0, 1));     /* .w */
> +
> +      assign->write_mask = 1 << i;

You might consider using ir_builder...it's much easier to work with.
For example, the above four lines become:

assign(var, swizzle_w(tex), 1 << i);

That's probably all it actually buys you at this point, but it might be
worth it just for that.

Either way, this series is:
Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

> +
> +      base_ir->insert_before(assign);
> +   }
> +
> +   base_ir->replace_with(new (mem_ctx) ir_dereference_variable(var));
> +
> +   progress = true;
> +   return visit_continue;
> +}
> +
> +extern "C" {
> +
> +bool
> +brw_do_lower_offset_arrays(exec_list *instructions)
> +{
> +   brw_lower_offset_array_visitor v;
> +
> +   visit_list_elements(&v, instructions);
> +
> +   return v.progress;
> +}
> +
> +}
> diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
> index 5da1b0f..5090d22 100644
> --- a/src/mesa/drivers/dri/i965/brw_shader.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
> @@ -168,6 +168,7 @@ brw_link_shader(struct gl_context *ctx, struct gl_shader_program *shProg)
>        do_vec_index_to_cond_assign(shader->ir);
>        lower_vector_insert(shader->ir, true);
>        brw_do_cubemap_normalize(shader->ir);
> +      brw_do_lower_offset_arrays(shader->ir);
>        lower_noise(shader->ir);
>        lower_quadop_vector(shader->ir, false);
>  
>