From 74cff779eb5217fe2b791f2a23405096901f45fd Mon Sep 17 00:00:00 2001 From: Matt Turner Date: Thu, 7 Jan 2016 16:16:35 -0800 Subject: [PATCH] nir: Change bfm's semantics to match Intel/AMD/SM5. Intel/AMD's hardware instructions do not handle arguments of 32. Constant evaluation should not produce a result different from the hardware instruction. The s/1ull/1u/ change is intentional: previously we wanted defined behavior for the "1 << 32" case, but we're making this case undefined so we can make it 1u and save ourselves a 64-bit operation. Reviewed-by: Ian Romanick --- src/glsl/nir/nir_opcodes.py | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/src/glsl/nir/nir_opcodes.py b/src/glsl/nir/nir_opcodes.py index d7ba0b62375..1c65def691a 100644 --- a/src/glsl/nir/nir_opcodes.py +++ b/src/glsl/nir/nir_opcodes.py @@ -511,12 +511,15 @@ binop("fpow", tfloat, "", "powf(src0, src1)") binop_horiz("pack_half_2x16_split", 1, tuint, 1, tfloat, 1, tfloat, "pack_half_1x16(src0.x) | (pack_half_1x16(src1.x) << 16)") +# bfm implements the behavior of the first operation of the SM5 "bfi" assembly +# and that of the "bfi1" i965 instruction. That is, it has undefined behavior +# if either of its arguments are 32. binop_convert("bfm", tuint, tint, "", """ int bits = src0, offset = src1; -if (offset < 0 || bits < 0 || offset + bits > 32) - dst = 0; /* undefined per the spec */ +if (offset < 0 || bits < 0 || offset > 31 || bits > 31 || offset + bits > 32) + dst = 0; /* undefined */ else - dst = ((1ull << bits) - 1) << offset; + dst = ((1u << bits) - 1) << offset; """) opcode("ldexp", 0, tfloat, [0, 0], [tfloat, tint], "", """ -- 2.30.2