diff options
Diffstat (limited to 'src/gallium/drivers/radeonsi/si_shader.c')
| -rw-r--r-- | src/gallium/drivers/radeonsi/si_shader.c | 52 |
1 files changed, 47 insertions, 5 deletions
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c index bc0f2171a52..27b3c72d1c9 100644 --- a/src/gallium/drivers/radeonsi/si_shader.c +++ b/src/gallium/drivers/radeonsi/si_shader.c @@ -2854,6 +2854,35 @@ static void build_txq_intrinsic(const struct lp_build_tgsi_action * action, } } +/* + * SI implements derivatives using the local data store (LDS) + * All writes to the LDS happen in all executing threads at + * the same time. TID is the Thread ID for the current + * thread and is a value between 0 and 63, representing + * the thread's position in the wavefront. + * + * For the pixel shader threads are grouped into quads of four pixels. + * The TIDs of the pixels of a quad are: + * + * +------+------+ + * |4n + 0|4n + 1| + * +------+------+ + * |4n + 2|4n + 3| + * +------+------+ + * + * So, masking the TID with 0xfffffffc yields the TID of the top left pixel + * of the quad, masking with 0xfffffffd yields the TID of the top pixel of + * the current pixel's column, and masking with 0xfffffffe yields the TID + * of the left pixel of the current pixel's row. + * + * Adding 1 yields the TID of the pixel to the right of the left pixel, and + * adding 2 yields the TID of the pixel below the top pixel. + */ +/* masks for thread ID. */ +#define TID_MASK_TOP_LEFT 0xfffffffc +#define TID_MASK_TOP 0xfffffffd +#define TID_MASK_LEFT 0xfffffffe + static void si_llvm_emit_ddxy( const struct lp_build_tgsi_action * action, struct lp_build_tgsi_context * bld_base, @@ -2870,6 +2899,8 @@ static void si_llvm_emit_ddxy( LLVMTypeRef i32; unsigned swizzle[4]; unsigned c; + int idx; + unsigned mask; i32 = LLVMInt32TypeInContext(gallivm->context); @@ -2879,15 +2910,22 @@ static void si_llvm_emit_ddxy( store_ptr = LLVMBuildGEP(gallivm->builder, si_shader_ctx->lds, indices, 2, ""); + if (opcode == TGSI_OPCODE_DDX_FINE) + mask = TID_MASK_LEFT; + else if (opcode == TGSI_OPCODE_DDY_FINE) + mask = TID_MASK_TOP; + else + mask = TID_MASK_TOP_LEFT; + indices[1] = LLVMBuildAnd(gallivm->builder, indices[1], - lp_build_const_int32(gallivm, 0xfffffffc), ""); + lp_build_const_int32(gallivm, mask), ""); load_ptr0 = LLVMBuildGEP(gallivm->builder, si_shader_ctx->lds, indices, 2, ""); + /* for DDX we want to next X pixel, DDY next Y pixel. */ + idx = (opcode == TGSI_OPCODE_DDX || opcode == TGSI_OPCODE_DDX_FINE) ? 1 : 2; indices[1] = LLVMBuildAdd(gallivm->builder, indices[1], - lp_build_const_int32(gallivm, - opcode == TGSI_OPCODE_DDX ? 1 : 2), - ""); + lp_build_const_int32(gallivm, idx), ""); load_ptr1 = LLVMBuildGEP(gallivm->builder, si_shader_ctx->lds, indices, 2, ""); @@ -3229,7 +3267,9 @@ static void create_function(struct si_shader_context *si_shader_ctx) if (bld_base->info && (bld_base->info->opcode_count[TGSI_OPCODE_DDX] > 0 || - bld_base->info->opcode_count[TGSI_OPCODE_DDY] > 0)) + bld_base->info->opcode_count[TGSI_OPCODE_DDY] > 0 || + bld_base->info->opcode_count[TGSI_OPCODE_DDX_FINE] > 0 || + bld_base->info->opcode_count[TGSI_OPCODE_DDY_FINE] > 0)) si_shader_ctx->lds = LLVMAddGlobalInAddressSpace(gallivm->module, LLVMArrayType(i32, 64), @@ -3722,6 +3762,8 @@ int si_shader_create(struct si_screen *sscreen, LLVMTargetMachineRef tm, bld_base->op_actions[TGSI_OPCODE_DDX].emit = si_llvm_emit_ddxy; bld_base->op_actions[TGSI_OPCODE_DDY].emit = si_llvm_emit_ddxy; + bld_base->op_actions[TGSI_OPCODE_DDX_FINE].emit = si_llvm_emit_ddxy; + bld_base->op_actions[TGSI_OPCODE_DDY_FINE].emit = si_llvm_emit_ddxy; bld_base->op_actions[TGSI_OPCODE_EMIT].emit = si_llvm_emit_vertex; bld_base->op_actions[TGSI_OPCODE_ENDPRIM].emit = si_llvm_emit_primitive; |