diff options
author | Kenneth Graunke <[email protected]> | 2014-09-01 01:58:06 -0700 |
---|---|---|
committer | Kenneth Graunke <[email protected]> | 2014-10-15 17:04:45 -0700 |
commit | 1f76fcf231cf7c8c19586e93ba8e4d800ad3be72 (patch) | |
tree | 1ee4b8f3affecb2950abcd83d94a020c191accaa /src/mesa/drivers | |
parent | cee2027574579d63cdf9693156542e7862b27131 (diff) |
i965/fs: Move texel offset handling to visit(ir_texture *).
This moves the handling of non-constant texel offset subexpression trees
to the place where we visit other such subtrees. It also removes some
uses of ir->offset in emit_texture_gen7, which will be useful when we
write the backend for our new upcoming IR.
Based on a patch by Connor Abbott.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
Diffstat (limited to 'src/mesa/drivers')
-rw-r--r-- | src/mesa/drivers/dri/i965/brw_fs.h | 3 | ||||
-rw-r--r-- | src/mesa/drivers/dri/i965/brw_fs_fp.cpp | 3 | ||||
-rw-r--r-- | src/mesa/drivers/dri/i965/brw_fs_visitor.cpp | 34 |
3 files changed, 29 insertions, 11 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h index 943e6387b8e..bba8d08ca89 100644 --- a/src/mesa/drivers/dri/i965/brw_fs.h +++ b/src/mesa/drivers/dri/i965/brw_fs.h @@ -478,7 +478,8 @@ public: fs_reg sample_index, uint32_t sampler); fs_inst *emit_texture_gen7(ir_texture *ir, fs_reg dst, fs_reg coordinate, fs_reg shadow_comp, fs_reg lod, fs_reg lod2, - fs_reg sample_index, fs_reg mcs, fs_reg sampler); + fs_reg sample_index, fs_reg mcs, fs_reg sampler, + fs_reg offset_value); fs_reg emit_mcs_fetch(ir_texture *ir, fs_reg coordinate, fs_reg sampler); void emit_gen6_gather_wa(uint8_t wa, fs_reg dst); fs_reg fix_math_operand(fs_reg src); diff --git a/src/mesa/drivers/dri/i965/brw_fs_fp.cpp b/src/mesa/drivers/dri/i965/brw_fs_fp.cpp index 9f0c0c7ac48..fb24c5897c2 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_fp.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_fp.cpp @@ -404,6 +404,7 @@ fs_visitor::emit_fragment_program_code() fs_reg coordinate = src[0]; fs_reg shadow_c; fs_reg sample_index; + fs_reg texel_offset; /* No offsets; leave as BAD_FILE. */ switch (fpi->Opcode) { case OPCODE_TEX: @@ -487,7 +488,7 @@ fs_visitor::emit_fragment_program_code() fs_inst *inst; if (brw->gen >= 7) { - inst = emit_texture_gen7(ir, dst, coordinate, shadow_c, lod, dpdy, sample_index, fs_reg(0u), fs_reg(fpi->TexSrcUnit)); + inst = emit_texture_gen7(ir, dst, coordinate, shadow_c, lod, dpdy, sample_index, fs_reg(0u), fs_reg(fpi->TexSrcUnit), texel_offset); } else if (brw->gen >= 5) { inst = emit_texture_gen5(ir, dst, coordinate, shadow_c, lod, dpdy, sample_index, fpi->TexSrcUnit); } else { diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp index 163d6423d9e..e4b87bbd5da 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp @@ -1481,7 +1481,8 @@ is_high_sampler(struct brw_context *brw, fs_reg sampler) fs_inst * fs_visitor::emit_texture_gen7(ir_texture *ir, fs_reg dst, fs_reg coordinate, fs_reg shadow_c, fs_reg lod, fs_reg lod2, - fs_reg sample_index, fs_reg mcs, fs_reg sampler) + fs_reg sample_index, fs_reg mcs, fs_reg sampler, + fs_reg offset_value) { int reg_width = dispatch_width / 8; bool header_present = false; @@ -1492,7 +1493,8 @@ fs_visitor::emit_texture_gen7(ir_texture *ir, fs_reg dst, fs_reg coordinate, } int length = 0; - if (ir->op == ir_tg4 || ir->offset || is_high_sampler(brw, sampler)) { + if (ir->op == ir_tg4 || offset_value.file != BAD_FILE || + is_high_sampler(brw, sampler)) { /* For general texture offsets (no txf workaround), we need a header to * put them in. Note that for SIMD16 we're making space for two actual * hardware registers here, so the emit will have to fix up for this. @@ -1513,7 +1515,8 @@ fs_visitor::emit_texture_gen7(ir_texture *ir, fs_reg dst, fs_reg coordinate, length++; } - bool has_nonconstant_offset = ir->offset && !ir->offset->as_constant(); + bool has_nonconstant_offset = + offset_value.file != BAD_FILE && offset_value.file != IMM; bool coordinate_done = false; /* Set up the LOD info */ @@ -1607,9 +1610,6 @@ fs_visitor::emit_texture_gen7(ir_texture *ir, fs_reg dst, fs_reg coordinate, no16("Gen7 does not support gather4_po_c in SIMD16 mode."); /* More crazy intermixing */ - ir->offset->accept(this); - fs_reg offset_value = this->result; - for (int i = 0; i < 2; i++) { /* u, v */ emit(MOV(sources[length], coordinate)); coordinate = offset(coordinate, 1); @@ -1935,6 +1935,21 @@ fs_visitor::visit(ir_texture *ir) shadow_comparitor = this->result; } + fs_reg offset_value; + if (ir->offset) { + ir_constant *const_offset = ir->offset->as_constant(); + if (const_offset) { + /* Store the header bitfield in an IMM register. This allows us to + * use offset_value.file to distinguish between no offset, a constant + * offset, and a non-constant offset. + */ + offset_value = fs_reg(brw_texture_offset(ctx, const_offset)); + } else { + ir->offset->accept(this); + offset_value = this->result; + } + } + fs_reg lod, lod2, sample_index, mcs; switch (ir->op) { case ir_tex: @@ -1979,7 +1994,8 @@ fs_visitor::visit(ir_texture *ir) if (brw->gen >= 7) { inst = emit_texture_gen7(ir, dst, coordinate, shadow_comparitor, - lod, lod2, sample_index, mcs, sampler_reg); + lod, lod2, sample_index, mcs, sampler_reg, + offset_value); } else if (brw->gen >= 5) { inst = emit_texture_gen5(ir, dst, coordinate, shadow_comparitor, lod, lod2, sample_index, sampler); @@ -1988,8 +2004,8 @@ fs_visitor::visit(ir_texture *ir) lod, lod2, sampler); } - if (ir->offset != NULL) - inst->texture_offset = brw_texture_offset(ctx, ir->offset->as_constant()); + if (offset_value.file == IMM) + inst->texture_offset = offset_value.fixed_hw_reg.dw1.ud; if (ir->op == ir_tg4) inst->texture_offset |= gather_channel(ir, sampler) << 16; // M0.2:16-17 |