On Wed, Dec 19, 2012 at 02:07:22PM -0500, Jerome Glisse wrote:
> On Wed, Dec 19, 2012 at 12:33 PM, Tom Stellard <t...@stellard.net> wrote:
> > On Sun, Dec 16, 2012 at 08:33:23PM +1000, Dave Airlie wrote:
> >> From: Dave Airlie <airl...@redhat.com>
> >>
> >> This adds TBO support to r600g, and with GLSL 1.40 enabled,
> >> we now get 3.1 core profiles advertised for r600g.
> >>
> >> This code is evergreen only so far, but I don't think there is
> >> much to make it work on r600/700/cayman other than testing.
> >>
> >> a) buffer txq is broken like cube map txq, this sucks, fix it the
> >> exact same way.
> >>
> >> b) buffer fetches are done with a vertex clause,
> >>
> >> c) vertex swizzling offsets are different than texture swizzles,
> >> but we still need to use the combiner, so make it configurable.
> >>
> >> d) add implementation of UCMP.
> >>
> >> TODO: r600/700/cayman testin
> >> Signed-off-by: Dave Airlie <airl...@redhat.com>
> >> ---
> >> src/gallium/drivers/r600/evergreen_state.c | 55 ++++++++++++++++++++
> >> src/gallium/drivers/r600/r600_asm.c | 2 +-
> >> src/gallium/drivers/r600/r600_asm.h | 2 +
> >> src/gallium/drivers/r600/r600_pipe.c | 4 +-
> >> src/gallium/drivers/r600/r600_pipe.h | 10 +++-
> >> src/gallium/drivers/r600/r600_shader.c | 75
> >> ++++++++++++++++++++++++++++
> >> src/gallium/drivers/r600/r600_shader.h | 1 +
> >> src/gallium/drivers/r600/r600_state_common.c | 58 +++++++++++++++++----
> >> src/gallium/drivers/r600/r600_texture.c | 16 ++++--
> >> 9 files changed, 204 insertions(+), 19 deletions(-)
> >>
> >
> > [snip]
> >
> >> diff --git a/src/gallium/drivers/r600/r600_shader.c
> >> b/src/gallium/drivers/r600/r600_shader.c
> >> index feb7001..60667e7 100644
> >> --- a/src/gallium/drivers/r600/r600_shader.c
> >> +++ b/src/gallium/drivers/r600/r600_shader.c
> >> @@ -3819,6 +3819,71 @@ static inline unsigned tgsi_tex_get_src_gpr(struct
> >> r600_shader_ctx *ctx,
> >> return ctx->file_offset[inst->Src[index].Register.File] +
> >> inst->Src[index].Register.Index;
> >> }
> >>
> >> +static int do_vtx_fetch_inst(struct r600_shader_ctx *ctx, boolean
> >> src_requires_loading)
> >> +{
> >> + struct r600_bytecode_vtx vtx;
> >> + struct r600_bytecode_alu alu;
> >> + struct tgsi_full_instruction *inst =
> >> &ctx->parse.FullToken.FullInstruction;
> >> + int src_gpr, r, i;
> >> +
> >> + src_gpr = tgsi_tex_get_src_gpr(ctx, 0);
> >> + if (src_requires_loading) {
> >> + for (i = 0; i < 4; i++) {
> >> + memset(&alu, 0, sizeof(struct r600_bytecode_alu));
> >> + alu.inst =
> >> CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
> >> + r600_bytecode_src(&alu.src[0], &ctx->src[0], i);
> >> + alu.dst.sel = ctx->temp_reg;
> >> + alu.dst.chan = i;
> >> + if (i == 3)
> >> + alu.last = 1;
> >> + alu.dst.write = 1;
> >> + r = r600_bytecode_add_alu(ctx->bc, &alu);
> >> + if (r)
> >> + return r;
> >> + }
> >> + src_gpr = ctx->temp_reg;
> >> + }
> >> +
> >> + memset(&vtx, 0, sizeof(vtx));
> >> + vtx.inst = 0;
> >> + vtx.buffer_id = tgsi_tex_get_src_gpr(ctx, 1) +
> >> R600_MAX_CONST_BUFFERS;;
> >> + vtx.fetch_type = 2; /* VTX_FETCH_NO_INDEX_OFFSET */
> >> + vtx.src_gpr = src_gpr;
> >> + vtx.mega_fetch_count = 16;
> >> + vtx.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] +
> >> inst->Dst[0].Register.Index;
> >> + vtx.dst_sel_x = (inst->Dst[0].Register.WriteMask & 1) ? 0 : 7;
> >> /* SEL_X */
> >> + vtx.dst_sel_y = (inst->Dst[0].Register.WriteMask & 2) ? 1 : 7;
> >> /* SEL_Y */
> >> + vtx.dst_sel_z = (inst->Dst[0].Register.WriteMask & 4) ? 2 : 7;
> >> /* SEL_Z */
> >> + vtx.dst_sel_w = (inst->Dst[0].Register.WriteMask & 8) ? 3 : 7;
> >> /* SEL_W */
> >> + vtx.use_const_fields = 1;
> >> + vtx.srf_mode_all = 1; /* SRF_MODE_NO_ZERO */
> >> +
> >
> > According to the docs, srf_mode_all will be ignored if use_const_fields
> > is set. However, based on my tests while running compute shaders, other
> > fields like data_format, which are supposed to be ignored weren't being
> > ignored unless the were set to zero. So, I think it would be safer
> > here to set srf_mode_all to zero and make sure that bit gets set on
> > the resource.
> >
> >
> >> + if ((r = r600_bytecode_add_vtx(ctx->bc, &vtx)))
> >> + return r;
> >> + return 0;
> >> +}
> >> +
> >
> > Otherwise, this code for vtx fetch looks good to me. One problem I ran into
> > with vtx fetch instructions while working on compute shaders was that
> > the GPU will hang if you write to vtx.src_gpr in the
> > instruction group following the vtx fetch. Here is a simple example:
> >
> > %T2_X<def> = MOV %ZERO
> > %T3_X<def> = VTX_READ_eg %T2_X<kill>, 24
> > %T2_X<def> = MOV %ZERO
> >
> > I'm not sure if this happens on all GPU variants, but I was able to
> > consistently reproduce this on my SUMO. You may want to keep an eye
> > out for this in case you run into any unexplainable hangs.
> >
>
> The vtx fetch group had the barrier flag set ?
>
Yes, I've attached the full dump if you are interested.
-Tom
target datalayout =
"e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f80:32:32-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n8:16:32:64"
target triple = "r600--"
define void @loop_hang(i32 addrspace(1)* nocapture %out, i32 %iterations)
nounwind {
entry:
%0 = call i32 @llvm.r600.read.local.size.x() nounwind
%1 = call i32 @llvm.r600.read.tgid.x() nounwind
%2 = call i32 @llvm.r600.read.tidig.x() nounwind
%mul.i = mul i32 %0, %1
%add.i = add i32 %mul.i, %2
%cmp = icmp eq i32 %iterations, 0
br i1 %cmp, label %if.else, label %if.then
if.then: ; preds = %entry
%arrayidx = getelementptr inbounds i32 addrspace(1)* %out, i32 %add.i
store i32 %iterations, i32 addrspace(1)* %arrayidx, align 4
br label %if.end
if.else: ; preds = %entry
%add = add i32 %add.i, 1
%arrayidx1 = getelementptr inbounds i32 addrspace(1)* %out, i32 %add
store i32 0, i32 addrspace(1)* %arrayidx1, align 4
br label %if.end
if.end: ; preds = %if.else, %if.then
ret void
}
declare i32 @llvm.r600.read.tgid.x() nounwind readnone
declare i32 @llvm.r600.read.local.size.x() nounwind readnone
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
!opencl.kernels = !{!0}
!0 = metadata !{void (i32 addrspace(1)*, i32)* @loop_hang}
# Machine code for function loop_hang: Post SSA, not tracking liveness
Function Live Ins: %T0_X in %vreg3, %T1_X in %vreg4
BB#0: derived from LLVM BB %entry
Live Ins: %T0_X %T1_X
%T2_X<def> = MOV %ZERO
%T3_X<def> = VTX_READ_PARAM_i32_eg %T2_X<kill>, 24
%T2_X<def> = MOV %ZERO
%T0_Y<def> = MULLO_INT_eg %T3_X<kill>, %T1_X<kill>
%T0_Y<def> = ADD_INT %T0_Y<kill>, %T0_X<kill>
%T0_X<def> = VTX_READ_PARAM_i32_eg %T2_X, 36
%T1_X<def> = VTX_READ_PARAM_i32_eg %T2_X<kill>, 40
%T0_Z<def> = MOV_IMM_I32 %ALU_LITERAL_X, 0
%T0_Z<def> = SETE_INT %T1_X, %T0_Z<kill>
IF_LOGICALNZ_i32 %T0_Z
%T0_Z<def> = MOV_IMM_I32 %ALU_LITERAL_X, 2
%T0_Y<def> = LSHL_eg %T0_Y<kill>, %T0_Z<kill>
%T0_X<def> = ADD_INT %T0_Y<kill>, %T0_X<kill>
%T0_Y<def> = MOV_IMM_I32 %ALU_LITERAL_X, 4
%T0_X<def> = ADD_INT %T0_X<kill>, %T0_Y<kill>
%T1_X<def> = MOV_IMM_I32 %ALU_LITERAL_X, 0
%T0_Y<def> = MOV %ALU_LITERAL_X, 2
%T0_X<def> = LSHR_eg %T0_X<kill>, %T0_Y<kill>
RAT_WRITE_CACHELESS_eg %T1_X<kill>, %T0_X<kill>
ELSE
%T0_Z<def> = MOV_IMM_I32 %ALU_LITERAL_X, 2
%T0_Y<def> = LSHL_eg %T0_Y<kill>, %T0_Z<kill>
%T0_X<def> = ADD_INT %T0_X<kill>, %T0_Y<kill>
%T0_Y<def> = MOV %ALU_LITERAL_X, 2
%T0_X<def> = LSHR_eg %T0_X<kill>, %T0_Y<kill>
RAT_WRITE_CACHELESS_eg %T1_X<kill>, %T0_X<kill>
ENDIF
RETURN
# End machine code for function loop_hang.
bytecode 90 dw -- 4 gprs ---------------------
E
0000 0000000D ALU ADDR:26 KCACHE_MODE0:0 KCACHE_BANK0:0 KCACHE_BANK1:0
0001 A0000000 ALU INST:0x8 KCACHE_MODE1:0 KCACHE_ADDR0:0 KCACHE_ADDR1:0 COUNT:1
0026 800000F8 SRC0(SEL:248 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0027 00400C90 * INST:0x19 DST(SEL:2 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0002 0000000E TEX/VTX ADDR:28
0003 80400000 TEX/VTX INST:0x1 COUNT:1
0028 10020040 INST:0 FETCH_TYPE:2 BUFFER_ID:0
0029 135FF003 SRC(GPR:2 SEL_X:0) MEGA_FETCH_COUNT:4 DST(GPR:3 SEL_X:0 SEL_Y:7
SEL_Z:7 SEL_W:7) USE_CONST_FIELDS:0 FORMAT(DATA:13 NUM:1 COMP:0 MODE:0)
0030 00080018 ENDIAN:0 OFFSET:24
0031 00000000
0004 00000010 ALU ADDR:32 KCACHE_MODE0:0 KCACHE_BANK0:0 KCACHE_BANK1:0
0005 A0080000 ALU INST:0x8 KCACHE_MODE1:0 KCACHE_ADDR0:0 KCACHE_ADDR1:0 COUNT:3
0032 000000F8 SRC0(SEL:248 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:0)
0033 00400C90 INST:0x19 DST(SEL:2 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0034 80002003 SRC0(SEL:3 REL:0 CHAN:0 NEG:0) SRC1(SEL:1 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0035 20004790 * INST:0x8f DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0036 800000FF SRC0(SEL:255 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0037 20001A10 * INST:0x34 DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0006 00000014 TEX/VTX ADDR:40
0007 80400000 TEX/VTX INST:0x1 COUNT:1
0040 10020040 INST:0 FETCH_TYPE:2 BUFFER_ID:0
0041 135FF000 SRC(GPR:2 SEL_X:0) MEGA_FETCH_COUNT:4 DST(GPR:0 SEL_X:0 SEL_Y:7
SEL_Z:7 SEL_W:7) USE_CONST_FIELDS:0 FORMAT(DATA:13 NUM:1 COMP:0 MODE:0)
0042 00080024 ENDIAN:0 OFFSET:36
0043 00000000
0008 00000016 TEX/VTX ADDR:44
0009 80400000 TEX/VTX INST:0x1 COUNT:1
0044 10020040 INST:0 FETCH_TYPE:2 BUFFER_ID:0
0045 135FF001 SRC(GPR:2 SEL_X:0) MEGA_FETCH_COUNT:4 DST(GPR:1 SEL_X:0 SEL_Y:7
SEL_Z:7 SEL_W:7) USE_CONST_FIELDS:0 FORMAT(DATA:13 NUM:1 COMP:0 MODE:0)
0046 00080028 ENDIAN:0 OFFSET:40
0047 00000000
0010 00000018 ALU ADDR:48 KCACHE_MODE0:0 KCACHE_BANK0:0 KCACHE_BANK1:0
0011 A4080000 ALU INST:0x9 KCACHE_MODE1:0 KCACHE_ADDR0:0 KCACHE_ADDR1:0 COUNT:3
0048 800000F8 SRC0(SEL:248 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0049 40000C90 * INST:0x19 DST(SEL:0 CHAN:2 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0050 811FC001 SRC0(SEL:1 REL:0 CHAN:0 NEG:0) SRC1(SEL:254 REL:0 CHAN:2 NEG:0
IM:0) LAST:1)
0051 40001D10 * INST:0x3a DST(SEL:0 CHAN:2 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0052 801F08FE SRC0(SEL:254 REL:0 CHAN:2 NEG:0) SRC1(SEL:248 REL:0 CHAN:0
NEG:0 IM:0) LAST:1)
0053 0000228C * INST:0x45 DST(SEL:0 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:0 OMOD:0 EXECUTE_MASK:1 UPDATE_PRED:1
0012 00000009 CF ADDR:18
0013 82800000 CF INST:0xa COND:0 POP_COUNT:0
0014 0000001B ALU ADDR:54 KCACHE_MODE0:0 KCACHE_BANK0:0 KCACHE_BANK1:0
0015 A0280000 ALU INST:0x8 KCACHE_MODE1:0 KCACHE_ADDR0:0 KCACHE_ADDR1:0 COUNT:11
0054 800000FD SRC0(SEL:253 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0055 40000C90 * INST:0x19 DST(SEL:0 CHAN:2 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0056 00000002 0.000000 (2)
0058 811FC400 SRC0(SEL:0 REL:0 CHAN:1 NEG:0) SRC1(SEL:254 REL:0 CHAN:2 NEG:0
IM:0) LAST:1)
0059 20000B90 * INST:0x17 DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0060 000004FE SRC0(SEL:254 REL:0 CHAN:1 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:0)
0061 00001A10 INST:0x34 DST(SEL:0 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0062 800000FD SRC0(SEL:253 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0063 20000C90 * INST:0x19 DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0064 00000004 0.000000 (4)
0066 009FC0FE SRC0(SEL:254 REL:0 CHAN:0 NEG:0) SRC1(SEL:254 REL:0 CHAN:1
NEG:0 IM:0) LAST:0)
0067 00001A10 INST:0x34 DST(SEL:0 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0068 000000FD SRC0(SEL:253 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:0)
0069 20000C90 INST:0x19 DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0070 800000F8 SRC0(SEL:248 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0071 00200C90 * INST:0x19 DST(SEL:1 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0072 00000002 0.000000 (2)
0074 809FC0FE SRC0(SEL:254 REL:0 CHAN:0 NEG:0) SRC1(SEL:254 REL:0 CHAN:1
NEG:0 IM:0) LAST:1)
0075 00000B10 * INST:0x16 DST(SEL:0 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0016 0000A020 CF NATIVE
0017 95C01000 CF NATIVE
0018 0000000D CF ADDR:26
0019 83400001 CF INST:0xd COND:0 POP_COUNT:1
0020 00000026 ALU ADDR:76 KCACHE_MODE0:0 KCACHE_BANK0:0 KCACHE_BANK1:0
0021 A0180000 ALU INST:0x8 KCACHE_MODE1:0 KCACHE_ADDR0:0 KCACHE_ADDR1:0 COUNT:7
0076 800000FD SRC0(SEL:253 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0077 40000C90 * INST:0x19 DST(SEL:0 CHAN:2 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0078 00000002 0.000000 (2)
0080 811FC400 SRC0(SEL:0 REL:0 CHAN:1 NEG:0) SRC1(SEL:254 REL:0 CHAN:2 NEG:0
IM:0) LAST:1)
0081 20000B90 * INST:0x17 DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0082 009FC000 SRC0(SEL:0 REL:0 CHAN:0 NEG:0) SRC1(SEL:254 REL:0 CHAN:1 NEG:0
IM:0) LAST:0)
0083 00001A10 INST:0x34 DST(SEL:0 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0084 800000FD SRC0(SEL:253 REL:0 CHAN:0 NEG:0) SRC1(SEL:0 REL:0 CHAN:0 NEG:0
IM:0) LAST:1)
0085 20000C90 * INST:0x19 DST(SEL:0 CHAN:1 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0086 00000002 0.000000 (2)
0088 809FC0FE SRC0(SEL:254 REL:0 CHAN:0 NEG:0) SRC1(SEL:254 REL:0 CHAN:1
NEG:0 IM:0) LAST:1)
0089 00000B10 * INST:0x16 DST(SEL:0 CHAN:0 REL:0 CLAMP:0) BANK_SWIZZLE:0
SRC0_ABS:0 SRC1_ABS:0 WRITE_MASK:1 OMOD:0 EXECUTE_MASK:0 UPDATE_PRED:0
0022 0000A020 CF NATIVE
0023 95C01000 CF NATIVE
0024 0000000D CF ADDR:26
0025 83800001 CF INST:0xe COND:0 POP_COUNT:1
--------------------------------------
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
