Reviewed-by: Nicolai Hähnle <nicolai.haeh...@amd.com>
---
src/amd/common/ac_nir_to_llvm.c | 22 +++++++++++-----------
src/amd/common/ac_shader_abi.h | 2 ++
src/gallium/drivers/radeonsi/si_shader.c | 17 ++++++++---------
src/gallium/drivers/radeonsi/si_shader_internal.h | 2 --
4 files changed, 21 insertions(+), 22 deletions(-)
diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index 6060df75314..29c2bb26221 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -103,24 +103,22 @@ struct nir_to_llvm_context {
LLVMValueRef ls_out_layout;
LLVMValueRef es2gs_offset;
LLVMValueRef tcs_offchip_layout;
LLVMValueRef tcs_out_offsets;
LLVMValueRef tcs_out_layout;
LLVMValueRef tcs_in_layout;
LLVMValueRef oc_lds;
LLVMValueRef merged_wave_info;
LLVMValueRef tess_factor_offset;
- LLVMValueRef tcs_patch_id;
LLVMValueRef tcs_rel_ids;
LLVMValueRef tes_rel_patch_id;
- LLVMValueRef tes_patch_id;
LLVMValueRef tes_u;
LLVMValueRef tes_v;
LLVMValueRef gsvs_ring_stride;
LLVMValueRef gsvs_num_entries;
LLVMValueRef gs2vs_offset;
LLVMValueRef gs_wave_id;
LLVMValueRef gs_vtx_offset[6];
LLVMValueRef esgs_ring;
@@ -782,57 +780,57 @@ static void create_function(struct nir_to_llvm_context
*ctx,
radv_define_vs_user_sgprs_phase1(ctx, stage,
has_previous_stage, previous_stage, &args);
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->ls_out_layout); // ls out layout
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_offchip_layout); // tcs offchip layout
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_out_offsets); // tcs out offsets
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_out_layout); // tcs out layout
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_in_layout); // tcs in layout
if (ctx->shader_info->info.needs_multiview_view_index)
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->view_index);
- add_vgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_patch_id); // patch id
+ add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.tcs_patch_id); // patch id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_rel_ids); // rel ids;
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.vertex_id); // vertex id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->rel_auto_id); // rel auto id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->vs_prim_id); // vs prim id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.instance_id); // instance id
} else {
radv_define_common_user_sgprs_phase1(ctx, stage,
has_previous_stage, previous_stage, &user_sgpr_info, &args, &desc_sets);
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_offchip_layout); // tcs offchip layout
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_out_offsets); // tcs out offsets
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_out_layout); // tcs out layout
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_in_layout); // tcs in layout
if (ctx->shader_info->info.needs_multiview_view_index)
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->view_index);
add_sgpr_argument(&args, ctx->ac.i32, &ctx->oc_lds); //
param oc lds
add_sgpr_argument(&args, ctx->ac.i32,
&ctx->tess_factor_offset); // tess factor offset
- add_vgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_patch_id); // patch id
+ add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.tcs_patch_id); // patch id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_rel_ids); // rel ids;
}
break;
case MESA_SHADER_TESS_EVAL:
radv_define_common_user_sgprs_phase1(ctx, stage,
has_previous_stage, previous_stage, &user_sgpr_info, &args, &desc_sets);
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->tcs_offchip_layout); // tcs offchip layout
if (ctx->shader_info->info.needs_multiview_view_index ||
(!ctx->options->key.tes.as_es && ctx->options->key.has_multiview_view_index))
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->view_index);
if (ctx->options->key.tes.as_es) {
add_sgpr_argument(&args, ctx->ac.i32, &ctx->oc_lds); //
OC LDS
add_sgpr_argument(&args, ctx->ac.i32, NULL); //
add_sgpr_argument(&args, ctx->ac.i32,
&ctx->es2gs_offset); // es2gs offset
} else {
add_sgpr_argument(&args, ctx->ac.i32, NULL); //
add_sgpr_argument(&args, ctx->ac.i32, &ctx->oc_lds); //
OC LDS
}
add_vgpr_argument(&args, ctx->ac.f32, &ctx->tes_u); // tes_u
add_vgpr_argument(&args, ctx->ac.f32, &ctx->tes_v); // tes_v
add_vgpr_argument(&args, ctx->ac.i32, &ctx->tes_rel_patch_id);
// tes rel patch id
- add_vgpr_argument(&args, ctx->ac.i32, &ctx->tes_patch_id); //
tes patch id
+ add_vgpr_argument(&args, ctx->ac.i32, &ctx->abi.tes_patch_id);
// tes patch id
break;
case MESA_SHADER_GEOMETRY:
if (has_previous_stage) {
// First 6 system regs
add_sgpr_argument(&args, ctx->ac.i32,
&ctx->gs2vs_offset); // tess factor offset
add_sgpr_argument(&args, ctx->ac.i32,
&ctx->merged_wave_info); // merged wave info
add_sgpr_argument(&args, ctx->ac.i32, &ctx->oc_lds); //
param oc lds
add_sgpr_argument(&args, ctx->ac.i32, NULL); // scratch
offset
add_sgpr_argument(&args, ctx->ac.i32, NULL); // unknown
@@ -856,21 +854,21 @@ static void create_function(struct nir_to_llvm_context
*ctx,
if (previous_stage == MESA_SHADER_VERTEX) {
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.vertex_id); // vertex id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->rel_auto_id); // rel auto id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->vs_prim_id); // vs prim id
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.instance_id); // instance id
} else {
add_vgpr_argument(&args, ctx->ac.f32,
&ctx->tes_u); // tes_u
add_vgpr_argument(&args, ctx->ac.f32,
&ctx->tes_v); // tes_v
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->tes_rel_patch_id); // tes rel patch id
- add_vgpr_argument(&args, ctx->ac.i32,
&ctx->tes_patch_id); // tes patch id
+ add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.tes_patch_id); // tes patch id
}
} else {
radv_define_common_user_sgprs_phase1(ctx, stage,
has_previous_stage, previous_stage, &user_sgpr_info, &args, &desc_sets);
radv_define_vs_user_sgprs_phase1(ctx, stage,
has_previous_stage, previous_stage, &args);
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->gsvs_ring_stride); // gsvs stride
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->gsvs_num_entries); // gsvs num entires
if (ctx->shader_info->info.needs_multiview_view_index)
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->view_index);
add_sgpr_argument(&args, ctx->ac.i32,
&ctx->gs2vs_offset); // gs2vs offset
add_sgpr_argument(&args, ctx->ac.i32,
&ctx->gs_wave_id); // wave id
@@ -4089,25 +4087,27 @@ static void visit_intrinsic(struct ac_nir_context *ctx,
result = unpack_param(&ctx->ac, ctx->nctx->tcs_rel_ids,
8, 5);
else
result = ctx->abi->gs_invocation_id;
break;
case nir_intrinsic_load_primitive_id:
if (ctx->stage == MESA_SHADER_GEOMETRY) {
if (ctx->nctx)
ctx->nctx->shader_info->gs.uses_prim_id = true;
result = ctx->abi->gs_prim_id;
} else if (ctx->stage == MESA_SHADER_TESS_CTRL) {
- ctx->nctx->shader_info->tcs.uses_prim_id = true;
- result = ctx->nctx->tcs_patch_id;
+ if (ctx->nctx)
+ ctx->nctx->shader_info->tcs.uses_prim_id = true;
+ result = ctx->abi->tcs_patch_id;
} else if (ctx->stage == MESA_SHADER_TESS_EVAL) {
- ctx->nctx->shader_info->tcs.uses_prim_id = true;
- result = ctx->nctx->tes_patch_id;
+ if (ctx->nctx)
+ ctx->nctx->shader_info->tcs.uses_prim_id = true;
+ result = ctx->abi->tes_patch_id;
} else
fprintf(stderr, "Unknown primitive id intrinsic: %d",
ctx->stage);
break;
case nir_intrinsic_load_sample_id:
result = unpack_param(&ctx->ac, ctx->abi->ancillary, 8, 4);
break;
case nir_intrinsic_load_sample_pos:
result = load_sample_pos(ctx);
break;
case nir_intrinsic_load_sample_mask_in:
@@ -6447,21 +6447,21 @@ ac_nir_get_max_workgroup_size(enum chip_class
chip_class,
static void ac_nir_fixup_ls_hs_input_vgprs(struct nir_to_llvm_context *ctx)
{
LLVMValueRef count = ac_build_bfe(&ctx->ac, ctx->merged_wave_info,
LLVMConstInt(ctx->ac.i32, 8, false),
LLVMConstInt(ctx->ac.i32, 8, false),
false);
LLVMValueRef hs_empty = LLVMBuildICmp(ctx->ac.builder, LLVMIntEQ, count,
LLVMConstInt(ctx->ac.i32, 0,
false), "");
ctx->abi.instance_id = LLVMBuildSelect(ctx->ac.builder, hs_empty,
ctx->rel_auto_id, ctx->abi.instance_id, "");
ctx->vs_prim_id = LLVMBuildSelect(ctx->ac.builder, hs_empty,
ctx->abi.vertex_id, ctx->vs_prim_id, "");
ctx->rel_auto_id = LLVMBuildSelect(ctx->ac.builder, hs_empty,
ctx->tcs_rel_ids, ctx->rel_auto_id, "");
- ctx->abi.vertex_id = LLVMBuildSelect(ctx->ac.builder, hs_empty,
ctx->tcs_patch_id, ctx->abi.vertex_id, "");
+ ctx->abi.vertex_id = LLVMBuildSelect(ctx->ac.builder, hs_empty,
ctx->abi.tcs_patch_id, ctx->abi.vertex_id, "");
}
static void prepare_gs_input_vgprs(struct nir_to_llvm_context *ctx)
{
for(int i = 5; i >= 0; --i) {
ctx->gs_vtx_offset[i] = ac_build_bfe(&ctx->ac,
ctx->gs_vtx_offset[i & ~1],
LLVMConstInt(ctx->ac.i32,
(i & 1) * 16, false),
LLVMConstInt(ctx->ac.i32,
16, false), false);
}
diff --git a/src/amd/common/ac_shader_abi.h b/src/amd/common/ac_shader_abi.h
index fd2ec06fb14..6f526d9f254 100644
--- a/src/amd/common/ac_shader_abi.h
+++ b/src/amd/common/ac_shader_abi.h
@@ -35,20 +35,22 @@ enum ac_descriptor_type {
/* Document the shader ABI during compilation. This is what allows radeonsi and
* radv to share a compiler backend.
*/
struct ac_shader_abi {
LLVMValueRef base_vertex;
LLVMValueRef start_instance;
LLVMValueRef draw_id;
LLVMValueRef vertex_id;
LLVMValueRef instance_id;
+ LLVMValueRef tcs_patch_id;
+ LLVMValueRef tes_patch_id;
LLVMValueRef gs_prim_id;
LLVMValueRef gs_invocation_id;
LLVMValueRef frag_pos[4];
LLVMValueRef front_face;
LLVMValueRef ancillary;
LLVMValueRef sample_coverage;
/* For VS and PS: pre-loaded shader inputs.
*
* Currently only used for NIR shaders; indexed by variables'
diff --git a/src/gallium/drivers/radeonsi/si_shader.c
b/src/gallium/drivers/radeonsi/si_shader.c
index a8f425b50ee..ed9cb8a2261 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -755,25 +755,23 @@ static LLVMValueRef get_primitive_id(struct
si_shader_context *ctx,
unsigned swizzle)
{
if (swizzle > 0)
return ctx->i32_0;
switch (ctx->type) {
case PIPE_SHADER_VERTEX:
return LLVMGetParam(ctx->main_fn,
ctx->param_vs_prim_id);
case PIPE_SHADER_TESS_CTRL:
- return LLVMGetParam(ctx->main_fn,
- ctx->param_tcs_patch_id);
+ return ctx->abi.tcs_patch_id;
case PIPE_SHADER_TESS_EVAL:
- return LLVMGetParam(ctx->main_fn,
- ctx->param_tes_patch_id);
+ return ctx->abi.tes_patch_id;
case PIPE_SHADER_GEOMETRY:
return ctx->abi.gs_prim_id;
default:
assert(0);
return ctx->i32_0;
}
}
/**
* Return the value of tgsi_ind_register for indexing.
@@ -3361,22 +3359,23 @@ static void si_set_ls_return_value_for_tcs(struct
si_shader_context *ctx)
ret = si_insert_input_ret(ctx, ret, ctx->param_tcs_factor_addr_base64k,
8 + GFX9_SGPR_TCS_FACTOR_ADDR_BASE64K);
unsigned desc_param = ctx->param_tcs_factor_addr_base64k + 2;
ret = si_insert_input_ptr_as_2xi32(ctx, ret, desc_param,
8 +
GFX9_SGPR_TCS_CONST_AND_SHADER_BUFFERS);
ret = si_insert_input_ptr_as_2xi32(ctx, ret, desc_param + 1,
8 +
GFX9_SGPR_TCS_SAMPLERS_AND_IMAGES);
unsigned vgpr = 8 + GFX9_TCS_NUM_USER_SGPR;
- ret = si_insert_input_ret_float(ctx, ret,
- ctx->param_tcs_patch_id, vgpr++);
+ ret = LLVMBuildInsertValue(ctx->ac.builder, ret,
+ ac_to_float(&ctx->ac, ctx->abi.tcs_patch_id),
+ vgpr++, "");
ret = si_insert_input_ret_float(ctx, ret,
ctx->param_tcs_rel_ids, vgpr++);
ctx->return_value = ret;
}
/* Pass GS inputs from ES to GS on GFX9. */
static void si_set_es_return_value_for_gs(struct si_shader_context *ctx)
{
LLVMValueRef ret = ctx->return_value;
@@ -4649,21 +4648,21 @@ static void declare_vs_input_vgprs(struct
si_shader_context *ctx,
*num_prolog_vgprs += shader->selector->info.num_inputs;
}
}
static void declare_tes_input_vgprs(struct si_shader_context *ctx,
struct si_function_info *fninfo)
{
ctx->param_tes_u = add_arg(fninfo, ARG_VGPR, ctx->f32);
ctx->param_tes_v = add_arg(fninfo, ARG_VGPR, ctx->f32);
ctx->param_tes_rel_patch_id = add_arg(fninfo, ARG_VGPR, ctx->i32);
- ctx->param_tes_patch_id = add_arg(fninfo, ARG_VGPR, ctx->i32);
+ add_arg_assign(fninfo, ARG_VGPR, ctx->i32, &ctx->abi.tes_patch_id);
}
enum {
/* Convenient merged shader definitions. */
SI_SHADER_MERGED_VERTEX_TESSCTRL = PIPE_SHADER_TYPES,
SI_SHADER_MERGED_VERTEX_OR_TESSEVAL_GEOMETRY,
};
static void create_function(struct si_shader_context *ctx)
{
@@ -4746,21 +4745,21 @@ static void create_function(struct si_shader_context
*ctx)
ctx->param_tcs_offchip_layout = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_tcs_out_lds_offsets = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_tcs_out_lds_layout = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_vs_state_bits = add_arg(&fninfo, ARG_SGPR, ctx->i32);
ctx->param_tcs_offchip_addr_base64k = add_arg(&fninfo,
ARG_SGPR, ctx->i32);
ctx->param_tcs_factor_addr_base64k = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_tcs_offchip_offset = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_tcs_factor_offset = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
/* VGPRs */
- ctx->param_tcs_patch_id = add_arg(&fninfo, ARG_VGPR, ctx->i32);
+ add_arg_assign(&fninfo, ARG_VGPR, ctx->i32,
&ctx->abi.tcs_patch_id);
ctx->param_tcs_rel_ids = add_arg(&fninfo, ARG_VGPR, ctx->i32);
/* param_tcs_offchip_offset and param_tcs_factor_offset are
* placed after the user SGPRs.
*/
for (i = 0; i < GFX6_TCS_NUM_USER_SGPR + 2; i++)
returns[num_returns++] = ctx->i32; /* SGPRs */
for (i = 0; i < 11; i++)
returns[num_returns++] = ctx->f32; /* VGPRs */
break;
@@ -4785,21 +4784,21 @@ static void create_function(struct si_shader_context
*ctx)
ctx->param_tcs_out_lds_offsets = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_tcs_out_lds_layout = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
ctx->param_tcs_offchip_addr_base64k = add_arg(&fninfo,
ARG_SGPR, ctx->i32);
ctx->param_tcs_factor_addr_base64k = add_arg(&fninfo, ARG_SGPR,
ctx->i32);
add_arg(&fninfo, ARG_SGPR, ctx->i32); /* unused */
declare_per_stage_desc_pointers(ctx, &fninfo,
ctx->type ==
PIPE_SHADER_TESS_CTRL);
/* VGPRs (first TCS, then VS) */
- ctx->param_tcs_patch_id = add_arg(&fninfo, ARG_VGPR, ctx->i32);
+ add_arg_assign(&fninfo, ARG_VGPR, ctx->i32,
&ctx->abi.tcs_patch_id);
ctx->param_tcs_rel_ids = add_arg(&fninfo, ARG_VGPR, ctx->i32);
if (ctx->type == PIPE_SHADER_VERTEX) {
declare_vs_input_vgprs(ctx, &fninfo,
&num_prolog_vgprs);
/* LS return values are inputs to the TCS main shader
part. */
for (i = 0; i < 8 + GFX9_TCS_NUM_USER_SGPR; i++)
returns[num_returns++] = ctx->i32; /* SGPRs */
for (i = 0; i < 2; i++)
diff --git a/src/gallium/drivers/radeonsi/si_shader_internal.h
b/src/gallium/drivers/radeonsi/si_shader_internal.h
index 3fbfea7752b..39a074cb833 100644
--- a/src/gallium/drivers/radeonsi/si_shader_internal.h
+++ b/src/gallium/drivers/radeonsi/si_shader_internal.h
@@ -162,28 +162,26 @@ struct si_shader_context {
/* Layout of TCS outputs / TES inputs:
* [0:12] = stride between output patches in DW, num_outputs *
num_vertices * 4
* max = 32*32*4 + 32*4
* [26:31] = gl_PatchVerticesIn, max = 32
*/
int param_tcs_out_lds_layout;
int param_tcs_offchip_addr_base64k;
int param_tcs_factor_addr_base64k;
int param_tcs_offchip_offset;
int param_tcs_factor_offset;
- int param_tcs_patch_id;
int param_tcs_rel_ids;
/* API TES */
int param_tes_u;
int param_tes_v;
int param_tes_rel_patch_id;
- int param_tes_patch_id;
/* HW ES */
int param_es2gs_offset;
/* API GS */
int param_gs2vs_offset;
int param_gs_wave_id; /* GFX6 */
LLVMValueRef gs_vtx_offset[6]; /* in dwords (GFX6) */
int param_gs_vtx01_offset; /* in dwords (GFX9) */
int param_gs_vtx23_offset; /* in dwords (GFX9) */
int param_gs_vtx45_offset; /* in dwords (GFX9) */
/* CS */
--
2.14.3
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
