From 32308fe9f1c6c9ef6716be4d9812ad41794f48b4 Mon Sep 17 00:00:00 2001 From: Connor Abbott Date: Fri, 27 Oct 2023 14:37:53 +0200 Subject: [PATCH] ir3/nir: Fix imadsh_mix16 definition The constant-folding definition and comments say that it takes the high 16 bits of the first source and low 16 bits of the second source, but actually it's the opposite. The algebraic optimization, which actually happens and needs to be correct, was correct but the comment above it was wrong. Note that in the way we use it when lowering multiplications, the ordering doesn't matter. Part-of: --- src/compiler/nir/nir_opcodes.py | 4 ++-- src/compiler/nir/nir_opt_algebraic.py | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/src/compiler/nir/nir_opcodes.py b/src/compiler/nir/nir_opcodes.py index 7bd7a776b11..b04c2f363ef 100644 --- a/src/compiler/nir/nir_opcodes.py +++ b/src/compiler/nir/nir_opcodes.py @@ -1248,11 +1248,11 @@ dst.p = src15.x; binop("amul", tint, _2src_commutative + associative, "src0 * src1") # ir3-specific instruction that maps directly to mul-add shift high mix, -# (IMADSH_MIX16 i.e. ah * bl << 16 + c). It is used for lowering integer +# (IMADSH_MIX16 i.e. al * bh << 16 + c). It is used for lowering integer # multiplication (imul) on Freedreno backend.. opcode("imadsh_mix16", 0, tint32, [0, 0, 0], [tint32, tint32, tint32], False, "", """ -dst = ((((src0 & 0xffff0000) >> 16) * (src1 & 0x0000ffff)) << 16) + src2; +dst = ((((src0 & 0x0000ffff) << 16) * (src1 & 0xffff0000)) >> 16) + src2; """) # ir3-specific instruction that maps directly to ir3 mad.s24. diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py index 20e3836ff58..6446c4915b7 100644 --- a/src/compiler/nir/nir_opt_algebraic.py +++ b/src/compiler/nir/nir_opt_algebraic.py @@ -2751,7 +2751,7 @@ for op in ['fddx', 'fddx_fine', 'fddx_coarse', optimizations += [ # 'al * bl': If either 'al' or 'bl' is zero, return zero. (('umul_low', '#a(is_lower_half_zero)', 'b'), (0)), - # '(ah * bl) << 16 + c': If either 'ah' or 'bl' is zero, return 'c'. + # '(al * bh) << 16 + c': If either 'al' or 'bh' is zero, return 'c'. (('imadsh_mix16', '#a@32(is_lower_half_zero)', 'b@32', 'c@32'), ('c')), (('imadsh_mix16', 'a@32', '#b@32(is_upper_half_zero)', 'c@32'), ('c')), ]