Mesa (master): i965/fs: Support arbitrarily large sampler state indices on Haswell+.

[Kenneth Graunke]

Module: Mesa
Branch: master
Commit: 6943ac0bd9d0d07019cffbbcb5189aa8d67b729b
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=6943ac0bd9d0d07019cffbbcb5189aa8d67b729b

Author: Kenneth Graunke <kenneth at whitecape.org>
Date:   Sat Jan 18 13:29:39 2014 -0800

i965/fs: Support arbitrarily large sampler state indices on Haswell+.

The message descriptor's "Sampler Index" field is only 4 bits (on all
generations of hardware), so it can only represent indices 0 through 15.

Haswell introduced a new field in the message header - "Sampler State
Pointer".  Normally, this is copied straight from g0, but we can also
add a byte offset (as long as it's a multiple of 32).

This patch uses a "Sampler State Pointer" offset to select a group of
16 sampler states, and then uses the "Sampler Index" field to select
the state within that group.

Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
Reviewed-by: Matt Turner <mattst88 at gmail.com>
Reviewed-by: Chris Forbes <chrisf at ijw.co.nz>

---

 src/mesa/drivers/dri/i965/brw_fs_generator.cpp |   19 ++++++++++++++++++-
 src/mesa/drivers/dri/i965/brw_fs_visitor.cpp   |    5 ++++-
 2 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
index 38c7c5c..40d86eb 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
@@ -559,6 +559,23 @@ fs_generator::generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src
             brw_MOV(p, get_element_ud(header_reg, 2),
                        brw_imm_ud(inst->texture_offset));
          }
+
+         if (inst->sampler >= 16) {
+            /* The "Sampler Index" field can only store values between 0 and 15.
+             * However, we can add an offset to the "Sampler State Pointer"
+             * field, effectively selecting a different set of 16 samplers.
+             *
+             * The "Sampler State Pointer" needs to be aligned to a 32-byte
+             * offset, and each sampler state is only 16-bytes, so we can't
+             * exclusively use the offset - we have to use both.
+             */
+            assert(brw->is_haswell); /* field only exists on Haswell */
+            brw_ADD(p,
+                    get_element_ud(header_reg, 3),
+                    get_element_ud(brw_vec8_grf(0, 0), 3),
+                    brw_imm_ud(16 * (inst->sampler / 16) *
+                               sizeof(gen7_sampler_state)));
+         }
          brw_pop_insn_state(p);
       }
    }
@@ -573,7 +590,7 @@ fs_generator::generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src
 	      inst->base_mrf,
 	      src,
               surface_index,
-	      inst->sampler,
+	      inst->sampler % 16,
 	      msg_type,
 	      rlen,
 	      inst->mlen,
diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
index f49fcc7..28e4dd6 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
@@ -1250,13 +1250,16 @@ fs_visitor::emit_texture_gen7(ir_texture *ir, fs_reg dst, fs_reg coordinate,
    fs_reg payload = fs_reg(this, glsl_type::float_type);
    fs_reg next = payload;
 
-   if (ir->op == ir_tg4 || (ir->offset && ir->op != ir_txf)) {
+   if (ir->op == ir_tg4 || (ir->offset && ir->op != ir_txf) || sampler >= 16) {
       /* 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.
        *
        * * ir4_tg4 needs to place its channel select in the header,
        * for interaction with ARB_texture_swizzle
+       *
+       * The sampler index is only 4-bits, so for larger sampler numbers we
+       * need to offset the Sampler State Pointer in the header.
        */
       header_present = true;
       next.reg_offset++;