From: Matt Turner Date: Thu, 16 Jan 2020 19:17:14 +0000 (-0800) Subject: intel/compiler: Move Gen4/5 rounding to visitor X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=88a0523bd2d3c635c41f0d0f6844bd8cf43933cc;p=mesa.git intel/compiler: Move Gen4/5 rounding to visitor Gen4/5's rounding instructions operate differently than later Gens'. They all return the floor of the input and the "Round-increment" conditional modifier answers whether the result should be incremented by 1.0 to get the appropriate result for the operation (and thus its behavior is determined by the round opcode; e.g., RNDZ vs RNDE). Since this requires a second instruciton (a predicated ADD) that consumes the result of the round instruction, the round instruction cannot write its result directly to the (write-only) message registers. By emitting the ADD in the generator, the backend thinks it's safe to store the round's result directly to the message register file. To avoid this, we move the emission of the ADD instruction to the NIR translator so that the backend has the information it needs. I suspect this also fixes code generated for RNDZ.SAT but since Gen4/5 don't support GLSL 1.30 which adds the trunc() function, I couldn't write a piglit test to confirm. My thinking is that if x=-0.5: sat(trunc(-0.5)) = 0.0 But on Gen4/5 where sat(trunc(x)) is implemented as rndz.r.f0 result, x // result = floor(x) // set f0 if increment needed (+f0) add result, result, 1.0 // fixup so result = trunc(x) then putting saturate on both instructions will give the wrong result. floor(-0.5) = -1.0 sat(floor(-0.5)) = 0.0 // +1 increment would be needed since floor(-0.5) != trunc(-0.5) sat(sat(floor(-0.5)) + 1.0) = 1.0 Fixes: 6f394343b1f ("nir/algebraic: i2f(f2i()) -> trunc()") Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2355 Reviewed-by: Ian Romanick Tested-by: Marge Bot Part-of: --- diff --git a/src/intel/compiler/brw_eu.h b/src/intel/compiler/brw_eu.h index ea4ee7748f6..591d01386fe 100644 --- a/src/intel/compiler/brw_eu.h +++ b/src/intel/compiler/brw_eu.h @@ -199,9 +199,6 @@ brw_inst *brw_##OP(struct brw_codegen *p, \ struct brw_reg src1, \ struct brw_reg src2); -#define ROUND(OP) \ -void brw_##OP(struct brw_codegen *p, struct brw_reg dest, struct brw_reg src0); - ALU1(MOV) ALU2(SEL) ALU1(NOT) @@ -222,6 +219,8 @@ ALU2(AVG) ALU2(MUL) ALU1(FRC) ALU1(RNDD) +ALU1(RNDE) +ALU1(RNDZ) ALU2(MAC) ALU2(MACH) ALU1(LZD) @@ -244,13 +243,9 @@ ALU2(ADDC) ALU2(SUBB) ALU2(MAC) -ROUND(RNDZ) -ROUND(RNDE) - #undef ALU1 #undef ALU2 #undef ALU3 -#undef ROUND /* Helpers for SEND instruction: diff --git a/src/intel/compiler/brw_eu_emit.c b/src/intel/compiler/brw_eu_emit.c index 44577de62e8..427b01e789b 100644 --- a/src/intel/compiler/brw_eu_emit.c +++ b/src/intel/compiler/brw_eu_emit.c @@ -1020,33 +1020,6 @@ brw_inst *brw_##OP(struct brw_codegen *p, \ return brw_alu3(p, BRW_OPCODE_##OP, dest, src0, src1, src2); \ } -/* Rounding operations (other than RNDD) require two instructions - the first - * stores a rounded value (possibly the wrong way) in the dest register, but - * also sets a per-channel "increment bit" in the flag register. A predicated - * add of 1.0 fixes dest to contain the desired result. - * - * Sandybridge and later appear to round correctly without an ADD. - */ -#define ROUND(OP) \ -void brw_##OP(struct brw_codegen *p, \ - struct brw_reg dest, \ - struct brw_reg src) \ -{ \ - const struct gen_device_info *devinfo = p->devinfo; \ - brw_inst *rnd, *add; \ - rnd = next_insn(p, BRW_OPCODE_##OP); \ - brw_set_dest(p, rnd, dest); \ - brw_set_src0(p, rnd, src); \ - \ - if (devinfo->gen < 6) { \ - /* turn on round-increments */ \ - brw_inst_set_cond_modifier(devinfo, rnd, BRW_CONDITIONAL_R); \ - add = brw_ADD(p, dest, dest, brw_imm_f(1.0f)); \ - brw_inst_set_pred_control(devinfo, add, BRW_PREDICATE_NORMAL); \ - } \ -} - - ALU2(SEL) ALU1(NOT) ALU2(AND) @@ -1061,6 +1034,8 @@ ALU2(ROR) ALU3(CSEL) ALU1(FRC) ALU1(RNDD) +ALU1(RNDE) +ALU1(RNDZ) ALU2(MAC) ALU2(MACH) ALU1(LZD) @@ -1080,9 +1055,6 @@ ALU1(CBIT) ALU2(ADDC) ALU2(SUBB) -ROUND(RNDZ) -ROUND(RNDE) - brw_inst * brw_MOV(struct brw_codegen *p, struct brw_reg dest, struct brw_reg src0) { diff --git a/src/intel/compiler/brw_fs_nir.cpp b/src/intel/compiler/brw_fs_nir.cpp index a861a1d938d..4cfac360874 100644 --- a/src/intel/compiler/brw_fs_nir.cpp +++ b/src/intel/compiler/brw_fs_nir.cpp @@ -1593,6 +1593,12 @@ fs_visitor::nir_emit_alu(const fs_builder &bld, nir_alu_instr *instr, case nir_op_ftrunc: inst = bld.RNDZ(result, op[0]); + if (devinfo->gen < 6) { + set_condmod(BRW_CONDITIONAL_R, inst); + set_predicate(BRW_PREDICATE_NORMAL, + bld.ADD(result, result, brw_imm_f(1.0f))); + inst = bld.MOV(result, result); /* for potential saturation */ + } inst->saturate = instr->dest.saturate; break; @@ -1615,6 +1621,12 @@ fs_visitor::nir_emit_alu(const fs_builder &bld, nir_alu_instr *instr, break; case nir_op_fround_even: inst = bld.RNDE(result, op[0]); + if (devinfo->gen < 6) { + set_condmod(BRW_CONDITIONAL_R, inst); + set_predicate(BRW_PREDICATE_NORMAL, + bld.ADD(result, result, brw_imm_f(1.0f))); + inst = bld.MOV(result, result); /* for potential saturation */ + } inst->saturate = instr->dest.saturate; break; diff --git a/src/intel/compiler/brw_vec4_nir.cpp b/src/intel/compiler/brw_vec4_nir.cpp index 03fda5e9aaa..d32f65a9dbe 100644 --- a/src/intel/compiler/brw_vec4_nir.cpp +++ b/src/intel/compiler/brw_vec4_nir.cpp @@ -1379,6 +1379,12 @@ vec4_visitor::nir_emit_alu(nir_alu_instr *instr) case nir_op_ftrunc: inst = emit(RNDZ(dst, op[0])); + if (devinfo->gen < 6) { + inst->conditional_mod = BRW_CONDITIONAL_R; + inst = emit(ADD(dst, src_reg(dst), brw_imm_f(1.0f))); + inst->predicate = BRW_PREDICATE_NORMAL; + inst = emit(MOV(dst, src_reg(dst))); /* for potential saturation */ + } inst->saturate = instr->dest.saturate; break; @@ -1409,6 +1415,12 @@ vec4_visitor::nir_emit_alu(nir_alu_instr *instr) case nir_op_fround_even: inst = emit(RNDE(dst, op[0])); + if (devinfo->gen < 6) { + inst->conditional_mod = BRW_CONDITIONAL_R; + inst = emit(ADD(dst, src_reg(dst), brw_imm_f(1.0f))); + inst->predicate = BRW_PREDICATE_NORMAL; + inst = emit(MOV(dst, src_reg(dst))); /* for potential saturation */ + } inst->saturate = instr->dest.saturate; break;