KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

radeonsi: use DDX/DDY directly in si_llvm_emit_ddxy_interp 

We can finally do this, because the opcodes are scalar now.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
This commit is contained in:
Marek Olšák 2016-10-02 21:56:00 +02:00
parent b57aef8033
commit b2a694f079
1 changed files with 7 additions and 49 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
src/gallium/drivers/radeonsi/si_shader.c
View File

@ -5074,56 +5074,14 @@ static LLVMValueRef si_llvm_emit_ddxy_interp(
{
struct si_shader_context *ctx = si_shader_context(bld_base);
struct gallivm_state *gallivm = bld_base->base.gallivm;
LLVMValueRef store_ptr, load_ptr_x, load_ptr_y, load_ptr_ddx, load_ptr_ddy, temp, temp2;
LLVMValueRef tl, tr, bl, result[4], thread_id;
unsigned c;
LLVMValueRef result[4], a;
unsigned i;
thread_id = get_thread_id(ctx);
store_ptr = build_gep0(ctx, ctx->lds, thread_id);
temp = LLVMBuildAnd(gallivm->builder, thread_id,
lp_build_const_int32(gallivm, TID_MASK_LEFT), "");
temp2 = LLVMBuildAnd(gallivm->builder, thread_id,
lp_build_const_int32(gallivm, TID_MASK_TOP), "");
load_ptr_x = build_gep0(ctx, ctx->lds, temp);
load_ptr_y = build_gep0(ctx, ctx->lds, temp2);
load_ptr_ddx = build_gep0(ctx, ctx->lds,
LLVMBuildAdd(gallivm->builder, temp,
lp_build_const_int32(gallivm, 1), ""));
load_ptr_ddy = build_gep0(ctx, ctx->lds,
LLVMBuildAdd(gallivm->builder, temp2,
lp_build_const_int32(gallivm, 2), ""));
for (c = 0; c < 2; ++c) {
LLVMValueRef store_val;
LLVMValueRef c_ll = lp_build_const_int32(gallivm, c);
store_val = LLVMBuildExtractElement(gallivm->builder,
interp_ij, c_ll, "");
LLVMBuildStore(gallivm->builder,
store_val,
store_ptr);
tl = LLVMBuildLoad(gallivm->builder, load_ptr_x, "");
tl = LLVMBuildBitCast(gallivm->builder, tl, ctx->f32, "");
tr = LLVMBuildLoad(gallivm->builder, load_ptr_ddx, "");
tr = LLVMBuildBitCast(gallivm->builder, tr, ctx->f32, "");
result[c] = LLVMBuildFSub(gallivm->builder, tr, tl, "");
tl = LLVMBuildLoad(gallivm->builder, load_ptr_y, "");
tl = LLVMBuildBitCast(gallivm->builder, tl, ctx->f32, "");
bl = LLVMBuildLoad(gallivm->builder, load_ptr_ddy, "");
bl = LLVMBuildBitCast(gallivm->builder, bl, ctx->f32, "");
result[c + 2] = LLVMBuildFSub(gallivm->builder, bl, tl, "");
for (i = 0; i < 2; i++) {
a = LLVMBuildExtractElement(gallivm->builder, interp_ij,
LLVMConstInt(ctx->i32, i, 0), "");
result[i] = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_DDX, a);
result[2+i] = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_DDY, a);
}
return lp_build_gather_values(gallivm, result, 4);