nir: Get rid of *_indirect variants of input/output load/store intrinsics

There is some special-casing needed in a competent back-end.  However, they
can do their special-casing easily enough based on whether or not the
offset is a constant.  In the mean time, having the *_indirect variants
adds special cases a number of places where they don't need to be and, in
general, only complicates things.  To complicate matters, NIR had no way to
convdert an indirect load/store to a direct one in the case that the
indirect was a constant so we would still not really get what the back-ends
wanted.  The best solution seems to be to get rid of the *_indirect
variants entirely.

This commit is a bunch of different changes squashed together:

 - nir: Get rid of *_indirect variants of input/output load/store intrinsics
 - nir/glsl: Stop handling UBO/SSBO load/stores differently depending on indirect
 - nir/lower_io: Get rid of load/store_foo_indirect
 - i965/fs: Get rid of load/store_foo_indirect
 - i965/vec4: Get rid of load/store_foo_indirect
 - tgsi_to_nir: Get rid of load/store_foo_indirect
 - ir3/nir: Use the new unified io intrinsics
 - vc4: Do all uniform loads with byte offsets
 - vc4/nir: Use the new unified io intrinsics
 - vc4: Fix load_user_clip_plane crash
 - vc4: add missing src for store outputs
 - vc4: Fix state uniforms
 - nir/lower_clip: Update to the new load/store intrinsics
 - nir/lower_two_sided_color: Update to the new load intrinsic

NIR and i965 changes are

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

NIR indirect declarations and vc4 changes are

Reviewed-by: Eric Anholt <eric@anholt.net>

ir3 changes are

Reviewed-by: Rob Clark <robdclark@gmail.com>

NIR changes are

Acked-by: Rob Clark <robdclark@gmail.com>

1 files changed, 31 insertions, 16 deletions

diff --git a/src/gallium/drivers/vc4/vc4_program.c b/src/gallium/drivers/vc4/vc4_program.c
index 31968bb..caad05c 100644
--- a/src/gallium/drivers/vc4/vc4_program.c
+++ b/src/gallium/drivers/vc4/vc4_program.c
@@ -115,8 +115,9 @@ nir_ssa_def *vc4_nir_get_state_uniform(struct nir_builder *b,
         nir_intrinsic_instr *intr =
                 nir_intrinsic_instr_create(b->shader,
                                            nir_intrinsic_load_uniform);
-        intr->const_index[0] = VC4_NIR_STATE_UNIFORM_OFFSET + contents;
+        intr->const_index[0] = (VC4_NIR_STATE_UNIFORM_OFFSET + contents) * 4;
         intr->num_components = 1;
+        intr->src[0] = nir_src_for_ssa(nir_imm_int(b, 0));
         nir_ssa_dest_init(&intr->instr, &intr->dest, 1, NULL);
         nir_builder_instr_insert(b, &intr->instr);
         return &intr->dest.ssa;
@@ -1516,6 +1517,8 @@ static void
 ntq_emit_intrinsic(struct vc4_compile *c, nir_intrinsic_instr *instr)
 {
         const nir_intrinsic_info *info = &nir_intrinsic_infos[instr->intrinsic];
+        nir_const_value *const_offset;
+        unsigned offset;
         struct qreg *dest = NULL;
 
         if (info->has_dest) {
@@ -1525,21 +1528,25 @@ ntq_emit_intrinsic(struct vc4_compile *c, nir_intrinsic_instr *instr)
         switch (instr->intrinsic) {
         case nir_intrinsic_load_uniform:
                 assert(instr->num_components == 1);
-                if (instr->const_index[0] < VC4_NIR_STATE_UNIFORM_OFFSET) {
-                        *dest = qir_uniform(c, QUNIFORM_UNIFORM,
-                                            instr->const_index[0]);
+                const_offset = nir_src_as_const_value(instr->src[0]);
+                if (const_offset) {
+                        offset = instr->const_index[0] + const_offset->u[0];
+                        assert(offset % 4 == 0);
+                        /* We need dwords */
+                        offset = offset / 4;
+                        if (offset < VC4_NIR_STATE_UNIFORM_OFFSET) {
+                                *dest = qir_uniform(c, QUNIFORM_UNIFORM,
+                                                    offset);
+                        } else {
+                                *dest = qir_uniform(c, offset -
+                                                    VC4_NIR_STATE_UNIFORM_OFFSET,
+                                                    0);
+                        }
                 } else {
-                        *dest = qir_uniform(c, instr->const_index[0] -
-                                            VC4_NIR_STATE_UNIFORM_OFFSET,
-                                            0);
+                        *dest = indirect_uniform_load(c, instr);
                 }
                 break;
 
-        case nir_intrinsic_load_uniform_indirect:
-                *dest = indirect_uniform_load(c, instr);
-
-                break;
-
         case nir_intrinsic_load_user_clip_plane:
                 *dest = qir_uniform(c, QUNIFORM_USER_CLIP_PLANE,
                                     instr->const_index[0]);
@@ -1551,7 +1558,10 @@ ntq_emit_intrinsic(struct vc4_compile *c, nir_intrinsic_instr *instr)
 
         case nir_intrinsic_load_input:
                 assert(instr->num_components == 1);
+                const_offset = nir_src_as_const_value(instr->src[0]);
+                assert(const_offset && "vc4 doesn't support indirect inputs");
                 if (instr->const_index[0] >= VC4_NIR_TLB_COLOR_READ_INPUT) {
+                        assert(const_offset->u[0] == 0);
                         /* Reads of the per-sample color need to be done in
                          * order.
                          */
@@ -1565,17 +1575,22 @@ ntq_emit_intrinsic(struct vc4_compile *c, nir_intrinsic_instr *instr)
                         }
                         *dest = c->color_reads[sample_index];
                 } else {
-                        *dest = c->inputs[instr->const_index[0]];
+                        offset = instr->const_index[0] + const_offset->u[0];
+                        *dest = c->inputs[offset];
                 }
                 break;
 
         case nir_intrinsic_store_output:
+                const_offset = nir_src_as_const_value(instr->src[1]);
+                assert(const_offset && "vc4 doesn't support indirect outputs");
+                offset = instr->const_index[0] + const_offset->u[0];
+
                 /* MSAA color outputs are the only case where we have an
                  * output that's not lowered to being a store of a single 32
                  * bit value.
                  */
                 if (c->stage == QSTAGE_FRAG && instr->num_components == 4) {
-                        assert(instr->const_index[0] == c->output_color_index);
+                        assert(offset == c->output_color_index);
                         for (int i = 0; i < 4; i++) {
                                 c->sample_colors[i] =
                                         qir_MOV(c, ntq_get_src(c, instr->src[0],
@@ -1583,9 +1598,9 @@ ntq_emit_intrinsic(struct vc4_compile *c, nir_intrinsic_instr *instr)
                         }
                 } else {
                         assert(instr->num_components == 1);
-                        c->outputs[instr->const_index[0]] =
+                        c->outputs[offset] =
                                 qir_MOV(c, ntq_get_src(c, instr->src[0], 0));
-                        c->num_outputs = MAX2(c->num_outputs, instr->const_index[0] + 1);
+                        c->num_outputs = MAX2(c->num_outputs, offset + 1);
                 }
                 break;