<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_NEW " title="NEW - Geometry Shaders output wrong vertices on Sandy Bridge" href="https://bugs.freedesktop.org/show_bug.cgi?id=93418#c2">Comment # 2</a> on <a class="bz_bug_link bz_status_NEW " title="NEW - Geometry Shaders output wrong vertices on Sandy Bridge" href="https://bugs.freedesktop.org/show_bug.cgi?id=93418">bug 93418</a> from <a class="email" href="mailto:imirkin@alum.mit.edu" title="Ilia Mirkin <imirkin@alum.mit.edu>"> Ilia Mirkin</a> <pre>And with master I'm getting: brw_vec4.cpp:1797: void brw::vec4_visitor::convert_to_hw_regs(): Assertion `brw_is_single_value_swizzle(inst->src[i].swizzle)' failed. Program received signal SIGABRT, Aborted. 0x00007ffff607f167 in raise () from /lib64/libc.so.6 (gdb) bt #0 0x00007ffff607f167 in raise () from /lib64/libc.so.6 #1 0x00007ffff60804ca in abort () from /lib64/libc.so.6 #2 0x00007ffff6078296 in __assert_fail_base () from /lib64/libc.so.6 #3 0x00007ffff6078342 in __assert_fail () from /lib64/libc.so.6 #4 0x00007ffff286d630 in brw::vec4_visitor::convert_to_hw_regs (this=this@entry=0x4b59490) at brw_vec4.cpp:1797 #5 0x00007ffff286d991 in brw::vec4_visitor::run (this=this@entry=0x4b59490) at brw_vec4.cpp:1960 #6 0x00007ffff287b2d4 in brw::brw_compile_gs (compiler=0x9185d0, log_data=log_data@entry=0x7ffff7fd0040, mem_ctx=mem_ctx@entry=0x4bc6510, key=key@entry=0x7fffffffd910, prog_data=prog_data@entry=0x7fffffffd6e0, src_shader=<optimized out>, shader_prog=shader_prog@entry=0x4b4cac0, shader_time_index=shader_time_index@entry=-1, final_assembly_size=final_assembly_size@entry=0x7fffffffd6d4, error_str=error_str@entry=0x7fffffffd6d8) at brw_vec4_gs_visitor.cpp:914 #7 0x00007ffff27a308c in brw_codegen_gs_prog (brw=brw@entry=0x7ffff7fd0040, prog=prog@entry=0x4b4cac0, gp=gp@entry=0x4b1d940, key=key@entry=0x7fffffffd910) at brw_gs.c:160 #8 0x00007ffff27a3604 in brw_gs_precompile (ctx=0x7ffff7fd0040, shader_prog=0x4b4cac0, prog=0x4b1d940) at brw_gs.c:288 #9 0x00007ffff27a3dc9 in brw_shader_precompile (sh_prog=0x4b4cac0, ctx=0x7ffff7fd0040) at brw_link.cpp:53 #10 brw_link_shader (ctx=0x7ffff7fd0040, shProg=<optimized out>) at brw_link.cpp:282 #11 0x00007ffff266833e in _mesa_glsl_link_shader (ctx=0x7ffff7fd0040, prog=0x4b4cac0) at program/ir_to_mesa.cpp:2962 #12 0x00007ffff256f35a in link_program (ctx=0x7ffff7fd0040, program=<optimized out>) at main/shaderapi.c:1048 The shader in question is, I believe, the same one as the one used in draw call 813290. Here are some debug prints from INTEL_DEBUG=gs (I skipped the GLSL IR bits): NIR (SSA form) for geometry shader: shader: MESA_SHADER_GEOMETRY name: GLSL214 inputs: 0 outputs: 0 uniforms: 0 decl_var uniform INTERP_QUALIFIER_NONE ivec4 ctexoffset (2, 0) decl_var uniform INTERP_QUALIFIER_NONE vec4 clinept (1, 0) decl_var uniform INTERP_QUALIFIER_NONE vec4 cstereo (0, 0) decl_var shader_in INTERP_QUALIFIER_FLAT vec4[1] colors_0 (VARYING_SLOT_VAR0, 26) decl_var shader_in INTERP_QUALIFIER_FLAT vec4[1] pos (VARYING_SLOT_VAR1, 27) decl_var shader_out INTERP_QUALIFIER_NONE vec4 gl_Position (VARYING_SLOT_POS, 0) decl_var shader_out INTERP_QUALIFIER_NONE vec4 colors_0@0 (VARYING_SLOT_VAR0, 26) decl_function main returning void impl main { block block_0: /* preds: */ vec1 ssa_0 = load_const (0x00000000 /* 0.000000 */) vec2 ssa_1 = load_const (0xbf800000 /* -1.000000 */, 0x3f800000 /* 1.000000 */) vec2 ssa_2 = load_const (0x3f800000 /* 1.000000 */, 0xbf800000 /* -1.000000 */) vec1 ssa_3 = load_const (0x00000010 /* 0.000000 */) vec4 ssa_4 = intrinsic load_ubo (ssa_0, ssa_3) () () vec1 ssa_5 = frcp ssa_4 vec1 ssa_6 = fmul ssa_4.w, ssa_5 vec1 ssa_7 = frcp ssa_4.y vec1 ssa_8 = fmul -ssa_4.w, ssa_7 vec2 ssa_9 = vec2 ssa_6, ssa_8 vec4 ssa_10 = intrinsic load_per_vertex_input (ssa_0, ssa_0) () (27) /* pos */ vec2 ssa_11 = fmul ssa_9, ssa_10.ww vec4 ssa_12 = intrinsic load_per_vertex_input (ssa_0, ssa_0) () (26) /* colors_0 */ vec2 ssa_13 = fadd ssa_10, -ssa_11 vec4 ssa_14 = vec4 ssa_13, ssa_13.y, ssa_10.z, ssa_10.w vec2 ssa_15 = ffma ssa_2, ssa_11, ssa_10 vec4 ssa_16 = vec4 ssa_15, ssa_15.y, ssa_10.z, ssa_10.w vec2 ssa_17 = ffma ssa_1, ssa_11, ssa_10 vec4 ssa_18 = vec4 ssa_17, ssa_17.y, ssa_10.z, ssa_10.w vec2 ssa_19 = fadd ssa_10, ssa_11 vec4 ssa_20 = vec4 ssa_19, ssa_19.y, ssa_10.z, ssa_10.w intrinsic store_output (ssa_14, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_0) () (0) vec1 ssa_21 = load_const (0x00000001 /* 0.000000 */) intrinsic store_output (ssa_16, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_21) () (0) vec1 ssa_22 = load_const (0x00000002 /* 0.000000 */) intrinsic store_output (ssa_18, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_22) () (0) vec1 ssa_23 = load_const (0x00000003 /* 0.000000 */) intrinsic store_output (ssa_20, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_23) () (0) vec1 ssa_24 = load_const (0x00000004 /* 0.000000 */) intrinsic end_primitive_with_counter (ssa_24) () (0) intrinsic set_vertex_count (ssa_24) () () /* succs: block_1 */ block block_1: } NIR (final form) for geometry shader: shader: MESA_SHADER_GEOMETRY name: GLSL214 inputs: 0 outputs: 0 uniforms: 0 decl_var uniform INTERP_QUALIFIER_NONE ivec4 ctexoffset (2, 0) decl_var uniform INTERP_QUALIFIER_NONE vec4 clinept (1, 0) decl_var uniform INTERP_QUALIFIER_NONE vec4 cstereo (0, 0) decl_var shader_in INTERP_QUALIFIER_FLAT vec4[1] colors_0 (VARYING_SLOT_VAR0, 26) decl_var shader_in INTERP_QUALIFIER_FLAT vec4[1] pos (VARYING_SLOT_VAR1, 27) decl_var shader_out INTERP_QUALIFIER_NONE vec4 gl_Position (VARYING_SLOT_POS, 0) decl_var shader_out INTERP_QUALIFIER_NONE vec4 colors_0@0 (VARYING_SLOT_VAR0, 26) decl_function main returning void impl main { decl_reg vec2 r0 decl_reg vec4 r1 decl_reg vec4 r2 decl_reg vec4 r3 decl_reg vec4 r4 block block_0: /* preds: */ vec1 ssa_0 = load_const (0x00000000 /* 0.000000 */) vec2 ssa_1 = load_const (0xbf800000 /* -1.000000 */, 0x3f800000 /* 1.000000 */) vec2 ssa_2 = load_const (0x3f800000 /* 1.000000 */, 0xbf800000 /* -1.000000 */) vec1 ssa_3 = load_const (0x00000010 /* 0.000000 */) vec4 ssa_4 = intrinsic load_ubo (ssa_0, ssa_3) () () vec1 ssa_5 = frcp ssa_4 r0.x = fmul ssa_4.w, ssa_5 vec1 ssa_7 = frcp ssa_4.y r0.y = fmul -ssa_4.w, ssa_7.x vec4 ssa_10 = intrinsic load_per_vertex_input (ssa_0, ssa_0) () (27) /* pos */ vec2 ssa_11 = fmul r0, ssa_10.ww vec4 ssa_12 = intrinsic load_per_vertex_input (ssa_0, ssa_0) () (26) /* colors_0 */ r1.xy = fadd ssa_10, -ssa_11 r1.zw = imov ssa_10 r2.xy = ffma ssa_2, ssa_11, ssa_10 r2.zw = imov r1 r3.xy = ffma ssa_1, ssa_11, ssa_10 r3.zw = imov r2 r4.xy = fadd ssa_10, ssa_11 r4.zw = imov r3 intrinsic store_output (r1, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_0) () (0) vec1 ssa_21 = load_const (0x00000001 /* 0.000000 */) intrinsic store_output (r2, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_21) () (0) vec1 ssa_22 = load_const (0x00000002 /* 0.000000 */) intrinsic store_output (r3, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_22) () (0) vec1 ssa_23 = load_const (0x00000003 /* 0.000000 */) intrinsic store_output (r4, ssa_0) () (0, 15) /* gl_Position */ intrinsic store_output (ssa_12, ssa_0) () (26, 15) /* colors_0 */ intrinsic emit_vertex_with_counter (ssa_23) () (0) vec1 ssa_24 = load_const (0x00000004 /* 0.000000 */) intrinsic end_primitive_with_counter (ssa_24) () (0) intrinsic set_vertex_count (ssa_24) () () /* succs: block_1 */ block block_1: } GS Input VUE map (4 slots, non-SSO) [0] VARYING_SLOT_PSIZ [1] VARYING_SLOT_POS [2] VARYING_SLOT_VAR0 [3] VARYING_SLOT_VAR1 GS Output VUE map (3 slots, non-SSO) [0] VARYING_SLOT_PSIZ [1] VARYING_SLOT_POS [2] VARYING_SLOT_VAR0 GS estimated execution time: 138 cycles</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the QA Contact for the bug.</li> </ul> </body> </html>