From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@freedesktop.org Subject: [Bug 82139] New: [r600g, bisected] multiple ubo piglit regressions Date: Mon, 04 Aug 2014 16:29:14 +0000 Message-ID: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============2073738124==" Return-path: Received: from culpepper.freedesktop.org (unknown [131.252.210.165]) by gabe.freedesktop.org (Postfix) with ESMTP id 846256E464 for ; Mon, 4 Aug 2014 09:29:14 -0700 (PDT) List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" To: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org --===============2073738124== Content-Type: multipart/alternative; boundary="1407169754.eb6E010.15676"; charset="us-ascii" --1407169754.eb6E010.15676 Date: Mon, 4 Aug 2014 16:29:14 +0000 MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" https://bugs.freedesktop.org/show_bug.cgi?id=82139 Priority: medium Bug ID: 82139 Assignee: dri-devel@lists.freedesktop.org Summary: [r600g, bisected] multiple ubo piglit regressions Severity: normal Classification: Unclassified OS: All Reporter: andreas.boll.dev@gmail.com Hardware: Other Status: NEW Version: git Component: Drivers/Gallium/r600 Product: Mesa I noticed multiple ubo piglit regressions on 10.3-devel / r600g / BARTS since this commit: commit 01bf8bb87565ed3677e43c6b6deeb90378d88647 Author: Brian Paul Date: Tue Jul 1 07:57:43 2014 -0600 st/mesa: don't use address register for constant-indexed ir_binop_ubo_load Before, we were always using the address register and indirect addressing to index into a UBO constant buffer. With this change we only do that when necessary. Using the piglit bin/arb_uniform_buffer_object-rendering test as an example: Shader code: uniform ub_rot {float rotation; }; ... m[1][1] = cos(rotation); Before: IMM[1] INT32 {0, 1, 0, 0} 1: UARL ADDR[0].x, IMM[1].xxxx 2: MOV TEMP[0].x, CONST[3][ADDR[0].x].xxxx 3: COS TEMP[1].x, TEMP[0].xxxx After: 0: COS TEMP[0].x, CONST[3][0].xxxx Reviewed-by: Roland Scheidegger diff --git a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp index d660a6b..69d67e1 100644 --- a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp +++ b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp @@ -1971,9 +1971,17 @@ glsl_to_tgsi_visitor::visit(ir_expression *ir) assert(ir->type->is_vector() || ir->type->is_scalar()); if (const_offset_ir) { - index_reg = st_src_reg_for_int(const_offset / 16); - } else { - emit(ir, TGSI_OPCODE_USHR, st_dst_reg(index_reg), op[1], st_src_reg_for_int(4)); + /* Constant index into constant buffer */ + cbuf.reladdr = NULL; + cbuf.index = const_offset / 16; + cbuf.has_index2 = true; + } + else { + /* Relative/variable index into constant buffer */ + emit(ir, TGSI_OPCODE_USHR, st_dst_reg(index_reg), op[1], + st_src_reg_for_int(4)); + cbuf.reladdr = ralloc(mem_ctx, st_src_reg); + memcpy(cbuf.reladdr, &index_reg, sizeof(index_reg)); } cbuf.swizzle = swizzle_for_size(ir->type->vector_elements); @@ -1982,9 +1990,6 @@ glsl_to_tgsi_visitor::visit(ir_expression *ir) const_offset % 16 / 4, const_offset % 16 / 4); - cbuf.reladdr = ralloc(mem_ctx, st_src_reg); - memcpy(cbuf.reladdr, &index_reg, sizeof(index_reg)); - if (ir->type->base_type == GLSL_TYPE_BOOL) { emit(ir, TGSI_OPCODE_USNE, result_dst, cbuf, st_src_reg_for_int(0)); } else { I guess it unhides a driver bug, since softpipe and llvmpipe are not affected. Regressed tests: spec/ARB_uniform_buffer_object/bufferstorage spec/ARB_uniform_buffer_object/maxblocks spec/ARB_uniform_buffer_object/rendering spec/glsl-1.40/uniform_buffer/fs-mat4 spec/glsl-1.40/uniform_buffer/fs-two-members spec/glsl-1.40/uniform_buffer/vs-mat4 spec/glsl-1.40/uniform_buffer/vs-two-members spec/glsl-1.50/uniform_buffer/gs-mat4 spec/glsl-1.50/uniform_buffer/gs-two-members Thanks, Andreas. -- You are receiving this mail because: You are the assignee for the bug. --1407169754.eb6E010.15676 Date: Mon, 4 Aug 2014 16:29:14 +0000 MIME-Version: 1.0 Content-Type: text/html; charset="UTF-8"
Priority medium
Bug ID 82139
Assignee dri-devel@lists.freedesktop.org
Summary [r600g, bisected] multiple ubo piglit regressions
Severity normal
Classification Unclassified
OS All
Reporter andreas.boll.dev@gmail.com
Hardware Other
Status NEW
Version git
Component Drivers/Gallium/r600
Product Mesa 

I noticed multiple ubo piglit regressions on 10.3-devel / r600g / BARTS
since this commit:

commit 01bf8bb87565ed3677e43c6b6deeb90378d88647
Author: Brian Paul <brianp@vmware.com>
Date:   Tue Jul 1 07:57:43 2014 -0600

    st/mesa: don't use address register for constant-indexed ir_binop_ubo_load

    Before, we were always using the address register and indirect addressing
    to index into a UBO constant buffer.  With this change we only do that
    when necessary.

    Using the piglit bin/arb_uniform_buffer_object-rendering test as an
    example:

    Shader code:
      uniform ub_rot {float rotation; };
      ...
      m[1][1] = cos(rotation);

    Before:
      IMM[1] INT32 {0, 1, 0, 0}
      1: UARL ADDR[0].x, IMM[1].xxxx
      2: MOV TEMP[0].x, CONST[3][ADDR[0].x].xxxx
      3: COS TEMP[1].x, TEMP[0].xxxx

    After:
      0: COS TEMP[0].x, CONST[3][0].xxxx

    Reviewed-by: Roland Scheidegger <sroland@vmware.com>

diff --git a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
index d660a6b..69d67e1 100644
--- a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
+++ b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
@@ -1971,9 +1971,17 @@ glsl_to_tgsi_visitor::visit(ir_expression *ir)
       assert(ir->type->is_vector() || ir->type->is_scalar());

       if (const_offset_ir) {
-         index_reg = st_src_reg_for_int(const_offset / 16);
-      } else {
-         emit(ir, TGSI_OPCODE_USHR, st_dst_reg(index_reg), op[1],
st_src_reg_for_int(4));
+         /* Constant index into constant buffer */
+         cbuf.reladdr = NULL;
+         cbuf.index = const_offset / 16;
+         cbuf.has_index2 = true;
+      }
+      else {
+         /* Relative/variable index into constant buffer */
+         emit(ir, TGSI_OPCODE_USHR, st_dst_reg(index_reg), op[1],
+              st_src_reg_for_int(4));
+         cbuf.reladdr = ralloc(mem_ctx, st_src_reg);
+         memcpy(cbuf.reladdr, &index_reg, sizeof(index_reg));
       }

       cbuf.swizzle = swizzle_for_size(ir->type->vector_elements);
@@ -1982,9 +1990,6 @@ glsl_to_tgsi_visitor::visit(ir_expression *ir)
                                     const_offset % 16 / 4,
                                     const_offset % 16 / 4);

-      cbuf.reladdr = ralloc(mem_ctx, st_src_reg);
-      memcpy(cbuf.reladdr, &index_reg, sizeof(index_reg));
-
       if (ir->type->base_type == GLSL_TYPE_BOOL) {
          emit(ir, TGSI_OPCODE_USNE, result_dst, cbuf, st_src_reg_for_int(0));
       } else {


I guess it unhides a driver bug, since softpipe and llvmpipe are not affected.

Regressed tests:

spec/ARB_uniform_buffer_object/bufferstorage
spec/ARB_uniform_buffer_object/maxblocks
spec/ARB_uniform_buffer_object/rendering

spec/glsl-1.40/uniform_buffer/fs-mat4
spec/glsl-1.40/uniform_buffer/fs-two-members
spec/glsl-1.40/uniform_buffer/vs-mat4
spec/glsl-1.40/uniform_buffer/vs-two-members

spec/glsl-1.50/uniform_buffer/gs-mat4
spec/glsl-1.50/uniform_buffer/gs-two-members


Thanks,
Andreas.

You are receiving this mail because:
  • You are the assignee for the bug.
--1407169754.eb6E010.15676-- --===============2073738124== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel --===============2073738124==--