https://bugs.freedesktop.org/show_bug.cgi?id=92945
Bug ID: 92945
Summary: [llvmpipe] [softpipe] piglit fs-ldexp-dvec4 regression
Product: Mesa
Version: git
Hardware: x86-64 (AMD64)
OS: Linux (All)
Status: NEW
Keywords: bisected, regression
Severity: normal
Priority: medium
Component: Mesa core
Assignee: mesa-dev@lists.freedesktop.org
Reporter: v...@freedesktop.org
QA Contact: mesa-dev@lists.freedesktop.org
CC: jljus...@gmail.com, k...@bitplanet.net
mesa: 3694d58e6c4a39bd84e8aef0d8e67c3ae9447f33 (master 11.1.0-devel)
$ ./bin/shader_runner
tests/spec/arb_gpu_shader_fp64/execution/built-in-functions/fs-ldexp-dvec4.shader_t$
ir_dereference_variable @ 0xfe9b90 specifies undeclared variable `color' @
0xfee5f0
Aborted (core dumped)
(gdb) bt
#0 0x00007f721d432267 in __GI_raise (sig=sig@entry=6) at
../sysdeps/unix/sysv/linux/raise.c:55
#1 0x00007f721d433eca in __GI_abort () at abort.c:89
#2 0x00007f721baefee1 in (anonymous namespace)::ir_validate::visit
(this=0x7ffe25e692b0, ir=0xfe9b90) at$
#3 0x00007f721baebc7e in ir_dereference_variable::accept (this=0xfe9b90,
v=0x7ffe25e692b0) at src/glsl/i$
#4 0x00007f721baebeaa in ir_assignment::accept (this=0xfea070,
v=0x7ffe25e692b0) at src/glsl/ir_hv_accep$
#5 0x00007f721baeb40a in visit_list_elements (v=0x7ffe25e692b0, l=0xf90548,
statement_list=true) at src/$
#6 0x00007f721baeb60f in ir_function_signature::accept (this=0xf90500,
v=0x7ffe25e692b0) at src/glsl/ir_$
#7 0x00007f721baeb40a in visit_list_elements (v=0x7ffe25e692b0, l=0xfee3e8,
statement_list=false) at src$
#8 0x00007f721baeb6a3 in ir_function::accept (this=0xfee3c0, v=0x7ffe25e692b0)
at src/glsl/ir_hv_accept.$
#9 0x00007f721baeb40a in visit_list_elements (v=0x7ffe25e692b0, l=0xfe7d90,
statement_list=true) at src/$
#10 0x00007f721baeb2dc in ir_hierarchical_visitor::run (this=0x7ffe25e692b0,
instructions=0xfe7d90)
at src/glsl/ir_hierarchical_visitor.cpp:364
#11 0x00007f721baf3885 in validate_ir_tree (instructions=0xfe7d90) at
src/glsl/ir_validate.cpp:939
#12 0x00007f721bafd2e0 in link_shaders (ctx=0x7f721df9c010, prog=0xf87fc0) at
src/glsl/linker.cpp:4467
#13 0x00007f721b93405f in _mesa_glsl_link_shader (ctx=0x7f721df9c010,
prog=0xf87fc0) at src/mesa/program/$
#14 0x00007f721b887345 in link_program (ctx=0x7f721df9c010, program=3) at
src/mesa/main/shaderapi.c:1047
#15 0x00007f721b8884fb in _mesa_LinkProgram (programObj=3) at
src/mesa/main/shaderapi.c:1517
#16 0x00007f721da964cc in stub_glLinkProgram (program=3) at
piglit/tests/util/piglit-dispatch-gen.c:32599
#17 0x0000000000406f81 in link_and_use_shaders () at
piglit/tests/shaders/shader_runner.c:965
#18 0x000000000040d471 in piglit_init (argc=2, argv=0x7ffe25e69748) at
piglit/tests/shaders/shader_runner$
#19 0x00007f721db04f84 in run_test (gl_fw=0x7f721dda00e0 <glut_fw>, argc=2,
argv=0x7ffe25e69748)
at piglit/tests/util/piglit-framework-gl/piglit_glut_framework.c:285
#20 0x00007f721daeb014 in piglit_gl_test_run (argc=2, argv=0x7ffe25e69748,
config=0x7ffe25e69600)
at piglit/tests/util/piglit-framework-gl.c:199
#21 0x0000000000405630 in main (argc=2, argv=0x7ffe25e69748) at
piglit/tests/shaders/shader_runner.c:54
(gdb) frame 2
#2 0x00007f721baefee1 in (anonymous namespace)::ir_validate::visit
(this=0x7ffe25e692b0, ir=0xfe9b90) at$
102 abort();
(gdb) l
97
98 if (_mesa_set_search(ir_set, ir->var) == NULL) {
99 printf("ir_dereference_variable @ %p specifies undeclared
variable "
100 "`%s' @ %p\n",
101 (void *) ir, ir->var->name, (void *) ir->var);
102 abort();
103 }
104
105 this->validate_ir(ir, this->data_enter);
106
96b22fb080894ba1840af2372f28a46cc0f40c76 is the first bad commit
commit 96b22fb080894ba1840af2372f28a46cc0f40c76
Author: Kristian Høgsberg Kristensen <k...@bitplanet.net>
Date: Wed Nov 4 14:58:54 2015 -0800
glsl: Use array deref for access to vector components
We've assumed that we could lower per-component vector access from
vec[i] = scalar
to
vec = ir_triop_vector_insert(vec, scalar, i)
but with SSBOs (and compute shader SLM and tesselation outputs) this is
no longer valid. If a vector is "externally visible", multiple threads
can write independent components simultaneously. With lowering to
ir_triop_vector_insert, each thread read the entire vector, changes one
component, then writes out the entire vector. This is racy.
Instead of generating a ir_binop_vector_extract when we see v[i], we
generate ir_dereference_array. We then add a lowering pass to lower the
ir_dereference_array to ir_binop_vector_extract for rvalues and for to
vector_insert for lvalues in a separate lowering pass.
The resulting IR is the same as before, but we now have a window between
ast->ir conversion and the lowering pass where v[i] appears in the IR as
an array deref. This lets us run lowering passes that lower the vector
access to I/O (eg for SSBO load/store) before we lower the per-component
access to full vector writes.
Reviewed-by: Jordan Justen <jordan.l.jus...@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <k...@bitplanet.net>
:040000 040000 e2df93f05e94cac81ae3222837a463d7bcb41cc8
e4f382406684faa0fcafea9e9c293ef0ccffdd96 M src
bisect run success
--
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
