fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-24 21:50:12 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

i965/fs: Unpack count argument to 64-bit shift ops on Atom

Browse source 
64-bit operations on Atom parts have additional restrictions over their
big-core counterparts (validated by later patches).

Specifically, the restriction that "Source and Destination horizontal
stride must be aligned to the same qword" is violated by most shift
operations since NIR uses a 32-bit value as the shift count argument,
and this causes instructions like

   shl(8)          g19<1>Q         g5<4,4,1>Q      g23<4,4,1>UD

where src1 has a 32-bit stride, but the dest and src0 have a 64-bit
stride.

This caused ~4 pixels in the ARB_shader_ballot piglit test
fs-readInvocation-uint.shader_test to be incorrect. Unfortunately no
ARB_gpu_shader_int64 test hit this case because they operate on
uniforms, and their scalar regions are an exception to the restriction.

We work around this by effectively unpacking the shift count, so that we
can read it with a 64-bit stride in the shift instruction. Unfortunately
the unpack (a MOV with a dst stride of 2) is a partial write, and cannot
be copy-propagated or CSE'd.

Bugzilla: https://bugs.freedesktop.org/101984
This commit is contained in:
Matt Turner 2017-09-21 16:42:02 -07:00
parent 2082c32950
commit b541945c20
1 changed files with 28 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

34
src/intel/compiler/brw_fs_nir.cpp
View file

@ -1267,14 +1267,36 @@ fs_visitor::nir_emit_alu(const fs_builder &bld, nir_alu_instr *instr)
unreachable("not reached: should have been lowered");
case nir_op_ishl:
bld.SHL(result, op[0], op[1]);
break;
case nir_op_ishr:
bld.ASR(result, op[0], op[1]);
break;
case nir_op_ushr:
bld.SHR(result, op[0], op[1]);
case nir_op_ushr: {
fs_reg shift_count = op[1];
if (devinfo->is_cherryview || gen_device_info_is_9lp(devinfo)) {
if (op[1].file == VGRF &&
(result.type == BRW_REGISTER_TYPE_Q ||
result.type == BRW_REGISTER_TYPE_UQ)) {
shift_count = fs_reg(VGRF, alloc.allocate(dispatch_width / 4),
BRW_REGISTER_TYPE_UD);
shift_count.stride = 2;
bld.MOV(shift_count, op[1]);
}
}
switch (instr->op) {
case nir_op_ishl:
bld.SHL(result, op[0], shift_count);
break;
case nir_op_ishr:
bld.ASR(result, op[0], shift_count);
break;
case nir_op_ushr:
bld.SHR(result, op[0], shift_count);
break;
default:
unreachable("not reached");
}
break;
}
case nir_op_pack_half_2x16_split:
bld.emit(FS_OPCODE_PACK_HALF_2x16_SPLIT, result, op[0], op[1]);