From a4ace98a9733b3e83d971f4871c2908749c0e5c8 Mon Sep 17 00:00:00 2001 From: =?utf8?q?Marek=20Ol=C5=A1=C3=A1k?= Date: Mon, 19 Dec 2016 16:11:27 +0100 Subject: [PATCH] gallium: remove TGSI_OPCODE_ABS MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit It's redundant with the source modifier. Reviewed-by: Nicolai Hähnle --- src/gallium/auxiliary/gallivm/lp_bld_tgsi.c | 2 +- .../auxiliary/gallivm/lp_bld_tgsi_action.c | 16 +-------- .../auxiliary/gallivm/lp_bld_tgsi_aos.c | 2 +- src/gallium/auxiliary/nir/tgsi_to_nir.c | 1 - src/gallium/auxiliary/tgsi/tgsi_exec.c | 4 --- src/gallium/auxiliary/tgsi/tgsi_info.c | 2 +- src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h | 1 - src/gallium/auxiliary/tgsi/tgsi_util.c | 1 - .../drivers/freedreno/a2xx/fd2_compiler.c | 5 --- src/gallium/drivers/i915/i915_fpc_optimize.c | 1 - src/gallium/drivers/i915/i915_fpc_translate.c | 9 ----- src/gallium/drivers/ilo/shader/toy_tgsi.c | 4 --- .../nouveau/codegen/nv50_ir_from_tgsi.cpp | 3 -- .../drivers/nouveau/nv30/nvfx_fragprog.c | 3 -- .../drivers/nouveau/nv30/nvfx_vertprog.c | 3 -- src/gallium/drivers/r300/r300_tgsi_to_rc.c | 1 - src/gallium/drivers/r600/r600_shader.c | 9 ++--- .../drivers/radeonsi/si_shader_tgsi_alu.c | 2 -- src/gallium/drivers/svga/svga_tgsi_insn.c | 1 - src/gallium/drivers/svga/svga_tgsi_vgpu10.c | 24 ------------- src/gallium/include/pipe/p_shader_tokens.h | 2 +- src/mesa/state_tracker/st_glsl_to_tgsi.cpp | 35 +++++++++++++++---- src/mesa/state_tracker/st_mesa_to_tgsi.c | 6 ++-- 23 files changed, 41 insertions(+), 96 deletions(-) diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c index 68ac6953829..d368f38d098 100644 --- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c +++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c @@ -360,7 +360,7 @@ lp_build_emit_fetch( case TGSI_TYPE_DOUBLE: case TGSI_TYPE_UNTYPED: /* modifiers on movs assume data is float */ - res = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_ABS, res); + res = lp_build_abs(&bld_base->base, res); break; case TGSI_TYPE_UNSIGNED: case TGSI_TYPE_SIGNED: diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c index 9c6fc4b4ae6..7d939e8fe4d 100644 --- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c +++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c @@ -499,8 +499,7 @@ log_emit( LLVMValueRef abs_x, log_abs_x, flr_log_abs_x, ex2_flr_log_abs_x; /* abs( src0.x) */ - abs_x = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_ABS, - emit_data->args[0] /* src0.x */); + abs_x = lp_build_abs(&bld_base->base, emit_data->args[0] /* src0.x */); /* log( abs( src0.x ) ) */ log_abs_x = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_LG2, @@ -1412,18 +1411,6 @@ lp_set_default_actions(struct lp_build_tgsi_context * bld_base) * intrinsics. */ -/* TGSI_OPCODE_ABS (CPU Only)*/ - -static void -abs_emit_cpu( - const struct lp_build_tgsi_action * action, - struct lp_build_tgsi_context * bld_base, - struct lp_build_emit_data * emit_data) -{ - emit_data->output[emit_data->chan] = lp_build_abs(&bld_base->base, - emit_data->args[0]); -} - /* TGSI_OPCODE_ADD (CPU Only) */ static void add_emit_cpu( @@ -2588,7 +2575,6 @@ lp_set_default_actions_cpu( struct lp_build_tgsi_context * bld_base) { lp_set_default_actions(bld_base); - bld_base->op_actions[TGSI_OPCODE_ABS].emit = abs_emit_cpu; bld_base->op_actions[TGSI_OPCODE_ADD].emit = add_emit_cpu; bld_base->op_actions[TGSI_OPCODE_AND].emit = and_emit_cpu; bld_base->op_actions[TGSI_OPCODE_ARL].emit = arl_emit_cpu; diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_aos.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_aos.c index 610283d7912..a5e439feb09 100644 --- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_aos.c +++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_aos.c @@ -521,7 +521,7 @@ lp_emit_instruction_aos( case TGSI_OPCODE_RSQ: /* TGSI_OPCODE_RECIPSQRT */ src0 = lp_build_emit_fetch(&bld->bld_base, inst, 0, LP_CHAN_ALL); - tmp0 = lp_build_emit_llvm_unary(&bld->bld_base, TGSI_OPCODE_ABS, src0); + tmp0 = lp_build_abs(&bld->bld_base.base, src0); dst0 = lp_build_rsqrt(&bld->bld_base.base, tmp0); break; diff --git a/src/gallium/auxiliary/nir/tgsi_to_nir.c b/src/gallium/auxiliary/nir/tgsi_to_nir.c index c7afe960f94..af4a6e034b6 100644 --- a/src/gallium/auxiliary/nir/tgsi_to_nir.c +++ b/src/gallium/auxiliary/nir/tgsi_to_nir.c @@ -1547,7 +1547,6 @@ static const nir_op op_trans[TGSI_OPCODE_LAST] = { [TGSI_OPCODE_LG2] = nir_op_flog2, [TGSI_OPCODE_POW] = nir_op_fpow, [TGSI_OPCODE_XPD] = 0, - [TGSI_OPCODE_ABS] = nir_op_fabs, [TGSI_OPCODE_DPH] = 0, [TGSI_OPCODE_COS] = nir_op_fcos, [TGSI_OPCODE_DDX] = nir_op_fddx, diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 7b5c56d9d02..2f89de663a4 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -5256,10 +5256,6 @@ exec_instruction( exec_xpd(mach, inst); break; - case TGSI_OPCODE_ABS: - exec_vector_unary(mach, inst, micro_abs, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); - break; - case TGSI_OPCODE_DPH: exec_dph(mach, inst); break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c index 37549aae78f..9b2431fe6b1 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.c +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -70,7 +70,7 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = { 1, 2, 0, 0, 0, 0, 0, REPL, "POW", TGSI_OPCODE_POW }, { 1, 2, 0, 0, 0, 0, 0, COMP, "XPD", TGSI_OPCODE_XPD }, { 1, 1, 0, 0, 0, 0, 0, COMP, "U2I64", TGSI_OPCODE_U2I64 }, - { 1, 1, 0, 0, 0, 0, 0, COMP, "ABS", TGSI_OPCODE_ABS }, + { 1, 1, 0, 0, 0, 0, 0, COMP, "", 33 }, /* removed */ { 1, 1, 0, 0, 0, 0, 0, COMP, "I2I64", TGSI_OPCODE_I2I64 }, { 1, 2, 0, 0, 0, 0, 0, REPL, "DPH", TGSI_OPCODE_DPH }, { 1, 1, 0, 0, 0, 0, 0, REPL, "COS", TGSI_OPCODE_COS }, diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h index d8752ce47d4..d78dd6615b2 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h +++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h @@ -85,7 +85,6 @@ OP11(EX2) OP11(LG2) OP12(POW) OP12(XPD) -OP11(ABS) OP12(DPH) OP11(COS) OP11(DDX) diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c index fbe29626a7f..4f02829e14d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_util.c +++ b/src/gallium/auxiliary/tgsi/tgsi_util.c @@ -201,7 +201,6 @@ tgsi_util_get_inst_usage_mask(const struct tgsi_full_instruction *inst, case TGSI_OPCODE_FLR: case TGSI_OPCODE_ROUND: case TGSI_OPCODE_POW: - case TGSI_OPCODE_ABS: case TGSI_OPCODE_COS: case TGSI_OPCODE_SIN: case TGSI_OPCODE_DDX: diff --git a/src/gallium/drivers/freedreno/a2xx/fd2_compiler.c b/src/gallium/drivers/freedreno/a2xx/fd2_compiler.c index 39418fca41f..2ffd8cd5a77 100644 --- a/src/gallium/drivers/freedreno/a2xx/fd2_compiler.c +++ b/src/gallium/drivers/freedreno/a2xx/fd2_compiler.c @@ -1080,11 +1080,6 @@ translate_instruction(struct fd2_compile_context *ctx, case TGSI_OPCODE_POW: translate_pow(ctx, inst); break; - case TGSI_OPCODE_ABS: - instr = ir2_instr_create_alu(cf, MAXv, ~0); - add_regs_vector_1(ctx, inst, instr); - instr->regs[1]->flags |= IR2_REG_NEGATE; /* src0 */ - break; case TGSI_OPCODE_COS: case TGSI_OPCODE_SIN: translate_trig(ctx, inst, opc); diff --git a/src/gallium/drivers/i915/i915_fpc_optimize.c b/src/gallium/drivers/i915/i915_fpc_optimize.c index a2b6d272939..7c3b9a9782f 100644 --- a/src/gallium/drivers/i915/i915_fpc_optimize.c +++ b/src/gallium/drivers/i915/i915_fpc_optimize.c @@ -76,7 +76,6 @@ static const struct { unsigned num_dst; unsigned num_src; } op_table [TGSI_OPCODE_LAST] = { - [ TGSI_OPCODE_ABS ] = { false, false, 0, 1, 1 }, [ TGSI_OPCODE_ADD ] = { false, true, TGSI_SWIZZLE_ZERO, 1, 2 }, [ TGSI_OPCODE_CEIL ] = { false, false, 0, 1, 1 }, [ TGSI_OPCODE_CMP ] = { false, false, 0, 1, 2 }, diff --git a/src/gallium/drivers/i915/i915_fpc_translate.c b/src/gallium/drivers/i915/i915_fpc_translate.c index 72b90929401..80caf31f383 100644 --- a/src/gallium/drivers/i915/i915_fpc_translate.c +++ b/src/gallium/drivers/i915/i915_fpc_translate.c @@ -500,15 +500,6 @@ i915_translate_instruction(struct i915_fp_compile *p, uint tmp = 0; switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ABS: - src0 = src_vector(p, &inst->Src[0], fs); - i915_emit_arith(p, - A0_MAX, - get_result_vector(p, &inst->Dst[0]), - get_result_flags(inst), 0, - src0, negate(src0, 1, 1, 1, 1), 0); - break; - case TGSI_OPCODE_ADD: emit_simple_arith(p, inst, A0_ADD, 2, fs); break; diff --git a/src/gallium/drivers/ilo/shader/toy_tgsi.c b/src/gallium/drivers/ilo/shader/toy_tgsi.c index e5fbb6e592a..a88f189ae88 100644 --- a/src/gallium/drivers/ilo/shader/toy_tgsi.c +++ b/src/gallium/drivers/ilo/shader/toy_tgsi.c @@ -58,7 +58,6 @@ static const struct { [TGSI_OPCODE_EX2] = { TOY_OPCODE_EXP, 1, 1 }, [TGSI_OPCODE_LG2] = { TOY_OPCODE_LOG, 1, 1 }, [TGSI_OPCODE_POW] = { TOY_OPCODE_POW, 1, 2 }, - [TGSI_OPCODE_ABS] = { GEN6_OPCODE_MOV, 1, 1 }, [TGSI_OPCODE_DPH] = { GEN6_OPCODE_DPH, 1, 2 }, [TGSI_OPCODE_COS] = { TOY_OPCODE_COS, 1, 1 }, [TGSI_OPCODE_KILL] = { TOY_OPCODE_KIL, 0, 0 }, @@ -148,7 +147,6 @@ aos_simple(struct toy_compiler *tc, case TGSI_OPCODE_SUB: src[1] = tsrc_negate(src[1]); break; - case TGSI_OPCODE_ABS: case TGSI_OPCODE_IABS: src[0] = tsrc_absolute(src[0]); break; @@ -790,7 +788,6 @@ static const toy_tgsi_translate aos_translate_table[TGSI_OPCODE_LAST] = { [TGSI_OPCODE_LG2] = aos_simple, [TGSI_OPCODE_POW] = aos_simple, [TGSI_OPCODE_XPD] = aos_XPD, - [TGSI_OPCODE_ABS] = aos_simple, [TGSI_OPCODE_DPH] = aos_simple, [TGSI_OPCODE_COS] = aos_simple, [TGSI_OPCODE_DDX] = aos_unsupported, @@ -1333,7 +1330,6 @@ static const toy_tgsi_translate soa_translate_table[TGSI_OPCODE_LAST] = { [TGSI_OPCODE_LG2] = soa_scalar_replicate, [TGSI_OPCODE_POW] = soa_scalar_replicate, [TGSI_OPCODE_XPD] = soa_XPD, - [TGSI_OPCODE_ABS] = soa_per_channel, [TGSI_OPCODE_DPH] = soa_dot_product, [TGSI_OPCODE_COS] = soa_scalar_replicate, [TGSI_OPCODE_DDX] = soa_partial_derivative, diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp index 91cef81aa0d..b9190980455 100644 --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp @@ -731,8 +731,6 @@ static nv50_ir::operation translateOpcode(uint opcode) NV50_IR_OPCODE_CASE(LG2, LG2); NV50_IR_OPCODE_CASE(POW, POW); - NV50_IR_OPCODE_CASE(ABS, ABS); - NV50_IR_OPCODE_CASE(COS, COS); NV50_IR_OPCODE_CASE(DDX, DFDX); NV50_IR_OPCODE_CASE(DDX_FINE, DFDX); @@ -3011,7 +3009,6 @@ Converter::handleInstruction(const struct tgsi_full_instruction *insn) } break; case TGSI_OPCODE_MOV: - case TGSI_OPCODE_ABS: case TGSI_OPCODE_CEIL: case TGSI_OPCODE_FLR: case TGSI_OPCODE_TRUNC: diff --git a/src/gallium/drivers/nouveau/nv30/nvfx_fragprog.c b/src/gallium/drivers/nouveau/nv30/nvfx_fragprog.c index dab42e17cfe..d031c6846ea 100644 --- a/src/gallium/drivers/nouveau/nv30/nvfx_fragprog.c +++ b/src/gallium/drivers/nouveau/nv30/nvfx_fragprog.c @@ -533,9 +533,6 @@ nvfx_fragprog_parse_instruction(struct nvfx_fpc *fpc, sat = finst->Instruction.Saturate; switch (finst->Instruction.Opcode) { - case TGSI_OPCODE_ABS: - nvfx_fp_emit(fpc, arith(sat, MOV, dst, mask, abs(src[0]), none, none)); - break; case TGSI_OPCODE_ADD: nvfx_fp_emit(fpc, arith(sat, ADD, dst, mask, src[0], src[1], none)); break; diff --git a/src/gallium/drivers/nouveau/nv30/nvfx_vertprog.c b/src/gallium/drivers/nouveau/nv30/nvfx_vertprog.c index b39c4b74eca..a802c43e22b 100644 --- a/src/gallium/drivers/nouveau/nv30/nvfx_vertprog.c +++ b/src/gallium/drivers/nouveau/nv30/nvfx_vertprog.c @@ -549,9 +549,6 @@ nvfx_vertprog_parse_instruction(struct nvfx_vpc *vpc, } switch (finst->Instruction.Opcode) { - case TGSI_OPCODE_ABS: - nvfx_vp_emit(vpc, arith(sat, VEC, MOV, dst, mask, abs(src[0]), none, none)); - break; case TGSI_OPCODE_ADD: nvfx_vp_emit(vpc, arith(sat, VEC, ADD, dst, mask, src[0], none, src[1])); break; diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c index 23ed2cf2532..9d1e59f18b6 100644 --- a/src/gallium/drivers/r300/r300_tgsi_to_rc.c +++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c @@ -61,7 +61,6 @@ static unsigned translate_opcode(unsigned opcode) case TGSI_OPCODE_LG2: return RC_OPCODE_LG2; case TGSI_OPCODE_POW: return RC_OPCODE_POW; case TGSI_OPCODE_XPD: return RC_OPCODE_XPD; - case TGSI_OPCODE_ABS: return RC_OPCODE_ABS; case TGSI_OPCODE_DPH: return RC_OPCODE_DPH; case TGSI_OPCODE_COS: return RC_OPCODE_COS; case TGSI_OPCODE_DDX: return RC_OPCODE_DDX; diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c index b5e7b7d7223..c2996aa3085 100644 --- a/src/gallium/drivers/r600/r600_shader.c +++ b/src/gallium/drivers/r600/r600_shader.c @@ -3936,9 +3936,6 @@ static int tgsi_op2_s(struct r600_shader_ctx *ctx, int swap, int trans_only) case TGSI_OPCODE_SUB: r600_bytecode_src_toggle_neg(&alu.src[1]); break; - case TGSI_OPCODE_ABS: - r600_bytecode_src_set_abs(&alu.src[0]); - break; default: break; } @@ -9025,7 +9022,7 @@ static const struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] [TGSI_OPCODE_POW] = { ALU_OP0_NOP, tgsi_pow}, [TGSI_OPCODE_XPD] = { ALU_OP0_NOP, tgsi_xpd}, [32] = { ALU_OP0_NOP, tgsi_unsupported}, - [TGSI_OPCODE_ABS] = { ALU_OP1_MOV, tgsi_op2}, + [33] = { ALU_OP0_NOP, tgsi_unsupported}, [34] = { ALU_OP0_NOP, tgsi_unsupported}, [TGSI_OPCODE_DPH] = { ALU_OP2_DOT4, tgsi_dp}, [TGSI_OPCODE_COS] = { ALU_OP1_COS, tgsi_trig}, @@ -9224,7 +9221,7 @@ static const struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction[] = [TGSI_OPCODE_POW] = { ALU_OP0_NOP, tgsi_pow}, [TGSI_OPCODE_XPD] = { ALU_OP0_NOP, tgsi_xpd}, [32] = { ALU_OP0_NOP, tgsi_unsupported}, - [TGSI_OPCODE_ABS] = { ALU_OP1_MOV, tgsi_op2}, + [33] = { ALU_OP0_NOP, tgsi_unsupported}, [34] = { ALU_OP0_NOP, tgsi_unsupported}, [TGSI_OPCODE_DPH] = { ALU_OP2_DOT4, tgsi_dp}, [TGSI_OPCODE_COS] = { ALU_OP1_COS, tgsi_trig}, @@ -9447,7 +9444,7 @@ static const struct r600_shader_tgsi_instruction cm_shader_tgsi_instruction[] = [TGSI_OPCODE_POW] = { ALU_OP0_NOP, cayman_pow}, [TGSI_OPCODE_XPD] = { ALU_OP0_NOP, tgsi_xpd}, [32] = { ALU_OP0_NOP, tgsi_unsupported}, - [TGSI_OPCODE_ABS] = { ALU_OP1_MOV, tgsi_op2}, + [33] = { ALU_OP0_NOP, tgsi_unsupported}, [34] = { ALU_OP0_NOP, tgsi_unsupported}, [TGSI_OPCODE_DPH] = { ALU_OP2_DOT4, tgsi_dp}, [TGSI_OPCODE_COS] = { ALU_OP1_COS, cayman_trig}, diff --git a/src/gallium/drivers/radeonsi/si_shader_tgsi_alu.c b/src/gallium/drivers/radeonsi/si_shader_tgsi_alu.c index 1966752cc0a..0a49bc2b2c2 100644 --- a/src/gallium/drivers/radeonsi/si_shader_tgsi_alu.c +++ b/src/gallium/drivers/radeonsi/si_shader_tgsi_alu.c @@ -767,8 +767,6 @@ void si_shader_context_init_alu(struct lp_build_tgsi_context *bld_base) { lp_set_default_actions(bld_base); - bld_base->op_actions[TGSI_OPCODE_ABS].emit = build_tgsi_intrinsic_nomem; - bld_base->op_actions[TGSI_OPCODE_ABS].intr_name = "llvm.fabs.f32"; bld_base->op_actions[TGSI_OPCODE_AND].emit = emit_and; bld_base->op_actions[TGSI_OPCODE_ARL].emit = emit_arl; bld_base->op_actions[TGSI_OPCODE_BFI].emit = emit_bfi; diff --git a/src/gallium/drivers/svga/svga_tgsi_insn.c b/src/gallium/drivers/svga/svga_tgsi_insn.c index 0e200136b6c..47a0afc8ad0 100644 --- a/src/gallium/drivers/svga/svga_tgsi_insn.c +++ b/src/gallium/drivers/svga/svga_tgsi_insn.c @@ -43,7 +43,6 @@ static unsigned translate_opcode(uint opcode) { switch (opcode) { - case TGSI_OPCODE_ABS: return SVGA3DOP_ABS; case TGSI_OPCODE_ADD: return SVGA3DOP_ADD; case TGSI_OPCODE_DP2A: return SVGA3DOP_DP2ADD; case TGSI_OPCODE_DP3: return SVGA3DOP_DP3; diff --git a/src/gallium/drivers/svga/svga_tgsi_vgpu10.c b/src/gallium/drivers/svga/svga_tgsi_vgpu10.c index 981251af628..e7cfb404a2e 100644 --- a/src/gallium/drivers/svga/svga_tgsi_vgpu10.c +++ b/src/gallium/drivers/svga/svga_tgsi_vgpu10.c @@ -3455,28 +3455,6 @@ emit_puint_to_sscaled(struct svga_shader_emitter_v10 *emit, } -/** - * Emit code for TGSI_OPCODE_ABS instruction. - */ -static boolean -emit_abs(struct svga_shader_emitter_v10 *emit, - const struct tgsi_full_instruction *inst) -{ - /* dst = ABS(s0): - * dst = abs(s0) - * Translates into: - * MOV dst, abs(s0) - */ - struct tgsi_full_src_register abs_src0 = absolute_src(&inst->Src[0]); - - /* MOV dst, abs(s0) */ - emit_instruction_op1(emit, VGPU10_OPCODE_MOV, &inst->Dst[0], - &abs_src0, inst->Instruction.Saturate); - - return TRUE; -} - - /** * Emit code for TGSI_OPCODE_ARL or TGSI_OPCODE_UARL instruction. */ @@ -5756,8 +5734,6 @@ emit_vgpu10_instruction(struct svga_shader_emitter_v10 *emit, return emit_vertex(emit, inst); case TGSI_OPCODE_ENDPRIM: return emit_endprim(emit, inst); - case TGSI_OPCODE_ABS: - return emit_abs(emit, inst); case TGSI_OPCODE_IABS: return emit_iabs(emit, inst); case TGSI_OPCODE_ARL: diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index ee59df0c393..3538090081e 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -361,7 +361,7 @@ struct tgsi_property_data { #define TGSI_OPCODE_POW 30 #define TGSI_OPCODE_XPD 31 #define TGSI_OPCODE_U2I64 32 -#define TGSI_OPCODE_ABS 33 +/* gap */ #define TGSI_OPCODE_I2I64 34 #define TGSI_OPCODE_DPH 35 #define TGSI_OPCODE_COS 36 diff --git a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp index b348783ebc2..3daf60a31e6 100644 --- a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp +++ b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp @@ -98,6 +98,7 @@ public: this->index = index; this->swizzle = swizzle_for_type(type, component); this->negate = 0; + this->abs = 0; this->index2D = 0; this->type = type ? type->base_type : GLSL_TYPE_ERROR; this->reladdr = NULL; @@ -117,6 +118,7 @@ public: this->index2D = 0; this->swizzle = SWIZZLE_XYZW; this->negate = 0; + this->abs = 0; this->reladdr = NULL; this->reladdr2 = NULL; this->has_index2 = false; @@ -134,6 +136,7 @@ public: this->index2D = index2D; this->swizzle = SWIZZLE_XYZW; this->negate = 0; + this->abs = 0; this->reladdr = NULL; this->reladdr2 = NULL; this->has_index2 = false; @@ -150,6 +153,7 @@ public: this->index2D = 0; this->swizzle = 0; this->negate = 0; + this->abs = 0; this->reladdr = NULL; this->reladdr2 = NULL; this->has_index2 = false; @@ -164,6 +168,7 @@ public: int16_t index2D; uint16_t swizzle; /**< SWIZZLE_XYZWONEZERO swizzles from Mesa. */ int negate:4; /**< NEGATE_XYZW mask from mesa */ + unsigned abs:1; enum glsl_base_type type:4; /** GLSL_TYPE_* from GLSL IR (enum glsl_base_type) */ unsigned has_index2:1; gl_register_file file:5; /**< PROGRAM_* from Mesa */ @@ -178,6 +183,14 @@ public: /** Register index should be offset by the integer in this reg. */ st_src_reg *reladdr; st_src_reg *reladdr2; + + st_src_reg get_abs() + { + st_src_reg reg = *this; + reg.negate = 0; + reg.abs = 1; + return reg; + } }; class st_dst_reg { @@ -245,6 +258,7 @@ st_src_reg::st_src_reg(st_dst_reg reg) this->index = reg.index; this->swizzle = SWIZZLE_XYZW; this->negate = 0; + this->abs = 0; this->reladdr = reg.reladdr; this->index2D = reg.index2D; this->reladdr2 = reg.reladdr2; @@ -954,7 +968,6 @@ glsl_to_tgsi_visitor::get_opcode(unsigned op, case2iu(ISHR, USHR); case3fid(SSG, ISSG, DSSG); - case3fid(ABS, IABS, DABS); case2iu(IBFE, UBFE); case2iu(IMSB, UMSB); @@ -1228,6 +1241,7 @@ glsl_to_tgsi_visitor::get_temp(const glsl_type *type) src.type = native_integers ? type->base_type : GLSL_TYPE_FLOAT; src.reladdr = NULL; src.negate = 0; + src.abs = 0; if (!options->EmitNoIndirectTemp && type_has_array_or_matrix(type)) { if (next_array >= max_num_arrays) { @@ -1593,7 +1607,12 @@ glsl_to_tgsi_visitor::visit_expression(ir_expression* ir, st_src_reg *op) emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0]); break; case ir_unop_abs: - emit_asm(ir, TGSI_OPCODE_ABS, result_dst, op[0]); + if (result_dst.type == GLSL_TYPE_FLOAT) + emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0].get_abs()); + else if (result_dst.type == GLSL_TYPE_DOUBLE) + emit_asm(ir, TGSI_OPCODE_DABS, result_dst, op[0]); + else + emit_asm(ir, TGSI_OPCODE_IABS, result_dst, op[0]); break; case ir_unop_sign: emit_asm(ir, TGSI_OPCODE_SSG, result_dst, op[0]); @@ -1919,8 +1938,7 @@ glsl_to_tgsi_visitor::visit_expression(ir_expression* ir, st_src_reg *op) * we want, I choose to use ABS to match DX9 and pre-GLSL RSQ * behavior. */ - emit_scalar(ir, TGSI_OPCODE_ABS, result_dst, op[0]); - emit_scalar(ir, TGSI_OPCODE_RSQ, result_dst, result_src); + emit_scalar(ir, TGSI_OPCODE_RSQ, result_dst, op[0].get_abs()); emit_scalar(ir, TGSI_OPCODE_RCP, result_dst, result_src); } break; @@ -1974,7 +1992,7 @@ glsl_to_tgsi_visitor::visit_expression(ir_expression* ir, st_src_reg *op) case ir_unop_bitcast_f2i: case ir_unop_bitcast_f2u: /* Make sure we don't propagate the negate modifier to integer opcodes. */ - if (op[0].negate) + if (op[0].negate || op[0].abs) emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0]); else result_src = op[0]; @@ -2076,6 +2094,7 @@ glsl_to_tgsi_visitor::visit_expression(ir_expression* ir, st_src_reg *op) cbuf.index = 0; cbuf.reladdr = NULL; cbuf.negate = 0; + cbuf.abs = 0; assert(ir->type->is_vector() || ir->type->is_scalar()); @@ -4843,7 +4862,8 @@ glsl_to_tgsi_visitor::copy_propagate(void) inst->src[0].file != PROGRAM_ARRAY && !inst->src[0].reladdr && !inst->src[0].reladdr2 && - !inst->src[0].negate) { + !inst->src[0].negate && + !inst->src[0].abs) { for (int i = 0; i < 4; i++) { if (inst->dst[0].writemask & (1 << i)) { acp[4 * inst->dst[0].index + i] = inst; @@ -5516,6 +5536,9 @@ translate_src(struct st_translate *t, const st_src_reg *src_reg) GET_SWZ(src_reg->swizzle, 2) & 0x3, GET_SWZ(src_reg->swizzle, 3) & 0x3); + if (src_reg->abs) + src = ureg_abs(src); + if ((src_reg->negate & 0xf) == NEGATE_XYZW) src = ureg_negate(src); diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 0a225ae0809..1768356413c 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -427,8 +427,6 @@ translate_opcode( unsigned op ) switch( op ) { case OPCODE_ARL: return TGSI_OPCODE_ARL; - case OPCODE_ABS: - return TGSI_OPCODE_ABS; case OPCODE_ADD: return TGSI_OPCODE_ADD; case OPCODE_CMP: @@ -564,6 +562,10 @@ compile_instruction( ureg_RSQ( ureg, dst[0], ureg_abs(src[0]) ); break; + case OPCODE_ABS: + ureg_MOV(ureg, dst[0], ureg_abs(src[0])); + break; + default: ureg_insn( ureg, translate_opcode( inst->Opcode ), -- 2.30.2