---
src/amd/common/ac_llvm_build.c | 1 +
src/amd/common/ac_llvm_build.h | 1 +
src/amd/common/ac_nir_to_llvm.c | 20 +++++++++-----------
3 files changed, 11 insertions(+), 11 deletions(-)
diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c
index ea238fa006..1a0d44bcdd 100644
--- a/src/amd/common/ac_llvm_build.c
+++ b/src/amd/common/ac_llvm_build.c
@@ -59,20 +59,21 @@ ac_llvm_context_init(struct ac_llvm_context *ctx,
LLVMContextRef context,
ctx->voidt = LLVMVoidTypeInContext(ctx->context);
ctx->i1 = LLVMInt1TypeInContext(ctx->context);
ctx->i8 = LLVMInt8TypeInContext(ctx->context);
ctx->i16 = LLVMIntTypeInContext(ctx->context, 16);
ctx->i32 = LLVMIntTypeInContext(ctx->context, 32);
ctx->i64 = LLVMIntTypeInContext(ctx->context, 64);
ctx->f16 = LLVMHalfTypeInContext(ctx->context);
ctx->f32 = LLVMFloatTypeInContext(ctx->context);
ctx->f64 = LLVMDoubleTypeInContext(ctx->context);
+ ctx->v2i32 = LLVMVectorType(ctx->i32, 2);
ctx->v4i32 = LLVMVectorType(ctx->i32, 4);
ctx->v4f32 = LLVMVectorType(ctx->f32, 4);
ctx->v8i32 = LLVMVectorType(ctx->i32, 8);
ctx->i32_0 = LLVMConstInt(ctx->i32, 0, false);
ctx->i32_1 = LLVMConstInt(ctx->i32, 1, false);
ctx->f32_0 = LLVMConstReal(ctx->f32, 0.0);
ctx->f32_1 = LLVMConstReal(ctx->f32, 1.0);
ctx->i1false = LLVMConstInt(ctx->i1, 0, false);
diff --git a/src/amd/common/ac_llvm_build.h b/src/amd/common/ac_llvm_build.h
index f790619827..7fc336c3f9 100644
--- a/src/amd/common/ac_llvm_build.h
+++ b/src/amd/common/ac_llvm_build.h
@@ -45,20 +45,21 @@ struct ac_llvm_context {
LLVMTypeRef voidt;
LLVMTypeRef i1;
LLVMTypeRef i8;
LLVMTypeRef i16;
LLVMTypeRef i32;
LLVMTypeRef i64;
LLVMTypeRef f16;
LLVMTypeRef f32;
LLVMTypeRef f64;
+ LLVMTypeRef v2i32;
LLVMTypeRef v4i32;
LLVMTypeRef v4f32;
LLVMTypeRef v8i32;
LLVMValueRef i32_0;
LLVMValueRef i32_1;
LLVMValueRef f32_0;
LLVMValueRef f32_1;
LLVMValueRef i1true;
LLVMValueRef i1false;
diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index 2437ea05c1..0bcb0b0525 100644
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -127,21 +127,20 @@ struct nir_to_llvm_context {
LLVMValueRef esgs_ring;
LLVMValueRef gsvs_ring;
LLVMValueRef hs_ring_tess_offchip;
LLVMValueRef hs_ring_tess_factor;
LLVMValueRef prim_mask;
LLVMValueRef sample_pos_offset;
LLVMValueRef persp_sample, persp_center, persp_centroid;
LLVMValueRef linear_sample, linear_center, linear_centroid;
- LLVMTypeRef v2i32;
LLVMTypeRef v3i32;
LLVMTypeRef v4i32;
LLVMTypeRef v8i32;
LLVMTypeRef f64;
LLVMTypeRef f32;
LLVMTypeRef f16;
LLVMTypeRef v2f32;
LLVMTypeRef v4f32;
unsigned uniform_md_kind;
@@ -870,27 +869,27 @@ static void create_function(struct nir_to_llvm_context
*ctx,
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->gs_vtx_offset[4]);
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->gs_vtx_offset[5]);
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->gs_invocation_id);
}
break;
case MESA_SHADER_FRAGMENT:
radv_define_common_user_sgprs_phase1(ctx, stage, has_previous_stage,
previous_stage, &user_sgpr_info, &args, &desc_sets);
if (ctx->shader_info->info.ps.needs_sample_positions)
add_user_sgpr_argument(&args, ctx->ac.i32,
&ctx->sample_pos_offset); /* sample position offset */
add_sgpr_argument(&args, ctx->ac.i32, &ctx->prim_mask); /* prim
mask */
- add_vgpr_argument(&args, ctx->v2i32, &ctx->persp_sample); /*
persp sample */
- add_vgpr_argument(&args, ctx->v2i32, &ctx->persp_center); /*
persp center */
- add_vgpr_argument(&args, ctx->v2i32, &ctx->persp_centroid); /*
persp centroid */
+ add_vgpr_argument(&args, ctx->ac.v2i32, &ctx->persp_sample); /*
persp sample */
+ add_vgpr_argument(&args, ctx->ac.v2i32, &ctx->persp_center); /*
persp center */
+ add_vgpr_argument(&args, ctx->ac.v2i32, &ctx->persp_centroid);
/* persp centroid */
add_vgpr_argument(&args, ctx->v3i32, NULL); /* persp pull model
*/
- add_vgpr_argument(&args, ctx->v2i32, &ctx->linear_sample); /*
linear sample */
- add_vgpr_argument(&args, ctx->v2i32, &ctx->linear_center); /*
linear center */
- add_vgpr_argument(&args, ctx->v2i32, &ctx->linear_centroid); /*
linear centroid */
+ add_vgpr_argument(&args, ctx->ac.v2i32, &ctx->linear_sample);
/* linear sample */
+ add_vgpr_argument(&args, ctx->ac.v2i32, &ctx->linear_center);
/* linear center */
+ add_vgpr_argument(&args, ctx->ac.v2i32, &ctx->linear_centroid);
/* linear centroid */
add_vgpr_argument(&args, ctx->f32, NULL); /* line stipple tex
*/
add_vgpr_argument(&args, ctx->f32, &ctx->abi.frag_pos[0]); /*
pos x float */
add_vgpr_argument(&args, ctx->f32, &ctx->abi.frag_pos[1]); /*
pos y float */
add_vgpr_argument(&args, ctx->f32, &ctx->abi.frag_pos[2]); /*
pos z float */
add_vgpr_argument(&args, ctx->f32, &ctx->abi.frag_pos[3]); /*
pos w float */
add_vgpr_argument(&args, ctx->ac.i32, &ctx->abi.front_face);
/* front face */
add_vgpr_argument(&args, ctx->ac.i32, &ctx->abi.ancillary); /*
ancillary */
add_vgpr_argument(&args, ctx->ac.i32,
&ctx->abi.sample_coverage); /* sample coverage */
add_vgpr_argument(&args, ctx->ac.i32, NULL); /* fixed pt */
break;
@@ -985,21 +984,20 @@ static void create_function(struct nir_to_llvm_context
*ctx,
break;
default:
unreachable("Shader stage not implemented");
}
ctx->shader_info->num_user_sgprs = user_sgpr_idx;
}
static void setup_types(struct nir_to_llvm_context *ctx)
{
- ctx->v2i32 = LLVMVectorType(ctx->ac.i32, 2);
ctx->v3i32 = LLVMVectorType(ctx->ac.i32, 3);
ctx->v4i32 = LLVMVectorType(ctx->ac.i32, 4);
ctx->v8i32 = LLVMVectorType(ctx->ac.i32, 8);
ctx->f32 = LLVMFloatTypeInContext(ctx->context);
ctx->f16 = LLVMHalfTypeInContext(ctx->context);
ctx->f64 = LLVMDoubleTypeInContext(ctx->context);
ctx->v2f32 = LLVMVectorType(ctx->f32, 2);
ctx->v4f32 = LLVMVectorType(ctx->f32, 4);
ctx->uniform_md_kind =
@@ -1901,39 +1899,39 @@ static void visit_alu(struct ac_nir_context *ctx, const
nir_alu_instr *instr)
case nir_op_fddx_fine:
case nir_op_fddy_fine:
case nir_op_fddx_coarse:
case nir_op_fddy_coarse:
result = emit_ddxy(ctx, instr->op, src[0]);
break;
case nir_op_unpack_64_2x32_split_x: {
assert(instr->src[0].src.ssa->num_components == 1);
LLVMValueRef tmp = LLVMBuildBitCast(ctx->ac.builder, src[0],
- LLVMVectorType(ctx->ac.i32,
2),
+ ctx->ac.v2i32,
"");
result = LLVMBuildExtractElement(ctx->ac.builder, tmp,
ctx->ac.i32_0, "");
break;
}
case nir_op_unpack_64_2x32_split_y: {
assert(instr->src[0].src.ssa->num_components == 1);
LLVMValueRef tmp = LLVMBuildBitCast(ctx->ac.builder, src[0],
- LLVMVectorType(ctx->ac.i32,
2),
+ ctx->ac.v2i32,
"");
result = LLVMBuildExtractElement(ctx->ac.builder, tmp,
ctx->ac.i32_1, "");
break;
}
case nir_op_pack_64_2x32_split: {
- LLVMValueRef tmp = LLVMGetUndef(LLVMVectorType(ctx->ac.i32, 2));
+ LLVMValueRef tmp = LLVMGetUndef(ctx->ac.v2i32);
tmp = LLVMBuildInsertElement(ctx->ac.builder, tmp,
src[0], ctx->ac.i32_0, "");
tmp = LLVMBuildInsertElement(ctx->ac.builder, tmp,
src[1], ctx->ac.i32_1, "");
result = LLVMBuildBitCast(ctx->ac.builder, tmp, ctx->ac.i64,
"");
break;
}
default:
fprintf(stderr, "Unknown NIR alu instr: ");
