From cc79945b215c3a78074905deeb27d0300034994a Mon Sep 17 00:00:00 2001 From: Samuel Pitoiset Date: Wed, 8 Apr 2020 08:39:08 +0200 Subject: [PATCH] aco: declare 8-bit/16-bit reduce operations The 8-bit float variants are only for consistency but are unused. Signed-off-by: Samuel Pitoiset Reviewed-by: Rhys Perry Part-of: --- src/amd/compiler/aco_ir.h | 26 +++++++++++++------------- src/amd/compiler/aco_print_ir.cpp | 26 ++++++++++++++++++++++++++ 2 files changed, 39 insertions(+), 13 deletions(-) diff --git a/src/amd/compiler/aco_ir.h b/src/amd/compiler/aco_ir.h index ba7f16e24fe..8fa6e48d452 100644 --- a/src/amd/compiler/aco_ir.h +++ b/src/amd/compiler/aco_ir.h @@ -1125,19 +1125,19 @@ struct Pseudo_barrier_instruction : public Instruction { static_assert(sizeof(Pseudo_barrier_instruction) == sizeof(Instruction) + 0, "Unexpected padding"); enum ReduceOp : uint16_t { - iadd32, iadd64, - imul32, imul64, - fadd32, fadd64, - fmul32, fmul64, - imin32, imin64, - imax32, imax64, - umin32, umin64, - umax32, umax64, - fmin32, fmin64, - fmax32, fmax64, - iand32, iand64, - ior32, ior64, - ixor32, ixor64, + iadd8, iadd16, iadd32, iadd64, + imul8, imul16, imul32, imul64, + fadd8, fadd16, fadd32, fadd64, + fmul8, fmul16, fmul32, fmul64, + imin8, imin16, imin32, imin64, + imax8, imax16, imax32, imax64, + umin8, umin16, umin32, umin64, + umax8, umax16, umax32, umax64, + fmin8, fmin16, fmin32, fmin64, + fmax8, fmax16, fmax32, fmax64, + iand8, iand16, iand32, iand64, + ior8, ior16, ior32, ior64, + ixor8, ixor16, ixor32, ixor64, gfx10_wave64_bpermute }; diff --git a/src/amd/compiler/aco_print_ir.cpp b/src/amd/compiler/aco_print_ir.cpp index 5bf91cf3a5c..e3c8cd81add 100644 --- a/src/amd/compiler/aco_print_ir.cpp +++ b/src/amd/compiler/aco_print_ir.cpp @@ -7,30 +7,56 @@ namespace aco { static const char *reduce_ops[] = { + [iadd8] = "iadd8", + [iadd16] = "iadd16", [iadd32] = "iadd32", [iadd64] = "iadd64", + [imul8] = "imul8", + [imul16] = "imul16", [imul32] = "imul32", [imul64] = "imul64", + [fadd8] = "fadd8", + [fadd16] = "fadd16", [fadd32] = "fadd32", [fadd64] = "fadd64", + [fmul8] = "fmul8", + [fmul16] = "fmul16", [fmul32] = "fmul32", [fmul64] = "fmul64", + [imin8] = "imin8", + [imin16] = "imin16", [imin32] = "imin32", [imin64] = "imin64", + [imax8] = "imax8", + [imax16] = "imax16", [imax32] = "imax32", [imax64] = "imax64", + [umin8] = "umin8", + [umin16] = "umin16", [umin32] = "umin32", [umin64] = "umin64", + [umax8] = "umax8", + [umax16] = "umax16", [umax32] = "umax32", [umax64] = "umax64", + [fmin8] = "fmin8", + [fmin16] = "fmin16", [fmin32] = "fmin32", [fmin64] = "fmin64", + [fmax8] = "fmax8", + [fmax16] = "fmax16", [fmax32] = "fmax32", [fmax64] = "fmax64", + [iand8] = "iand8", + [iand16] = "iand16", [iand32] = "iand32", [iand64] = "iand64", + [ior8] = "ior8", + [ior16] = "ior16", [ior32] = "ior32", [ior64] = "ior64", + [ixor8] = "ixor8", + [ixor16] = "ixor16", [ixor32] = "ixor32", [ixor64] = "ixor64", [gfx10_wave64_bpermute] = "gfx10_wave64_bpermute", -- 2.30.2