From 5fc0b580cecb1529659d5d3719412fb7cbffac0d Mon Sep 17 00:00:00 2001

From: Dave Airlie <airlied@redhat.com>

Date: Mon, 29 Jun 2020 13:26:56 +1000

Subject: [PATCH] glsl: fix constant packing for 64-bit big endian.

In a piglit run on s390 a lot of double tests fail, explicitly

packing/shifting things rather than using memcpy seems to help

---

 src/compiler/glsl/ir_constant_expression.cpp | 15 +++++++++++++++

 src/compiler/glsl/ir_expression_operation.py | 20 ++++++++++----------

 2 files changed, 25 insertions(+), 10 deletions(-)

diff --git a/src/compiler/glsl/ir_constant_expression.cpp b/src/compiler/glsl/ir_constant_expression.cpp

index 636196886b3..595cc821797 100644

--- a/src/compiler/glsl/ir_constant_expression.cpp

+++ b/src/compiler/glsl/ir_constant_expression.cpp

@@ -452,6 +452,21 @@ isub64_saturate(int64_t a, int64_t b)

    return a - b;

 }

+static uint64_t

+pack_2x32(uint32_t a, uint32_t b)

+{

+   uint64_t v = a;

+   v |= (uint64_t)b << 32;

+   return v;

+}

+

+static void

+unpack_2x32(uint64_t p, uint32_t *a, uint32_t *b)

+{

+   *a = p & 0xffffffff;

+   *b = (p >> 32);

+}

+

 /**

  * Get the constant that is ultimately referenced by an r-value, in a constant

  * expression evaluation context.

diff --git a/src/compiler/glsl/ir_expression_operation.py b/src/compiler/glsl/ir_expression_operation.py

index d2c4d41024f..1c4e6b358e1 100644

--- a/src/compiler/glsl/ir_expression_operation.py

+++ b/src/compiler/glsl/ir_expression_operation.py

@@ -560,14 +560,14 @@ ir_expression_operation = [

    operation("saturate", 1, printable_name="sat", source_types=(float_type,), c_expression="CLAMP({src0}, 0.0f, 1.0f)"),

    # Double packing, part of ARB_gpu_shader_fp64.

-   operation("pack_double_2x32", 1, printable_name="packDouble2x32", source_types=(uint_type,), dest_type=double_type, c_expression="memcpy(&data.d[0], &op[0]->value.u[0], sizeof(double))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("unpack_double_2x32", 1, printable_name="unpackDouble2x32", source_types=(double_type,), dest_type=uint_type, c_expression="memcpy(&data.u[0], &op[0]->value.d[0], sizeof(double))", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("pack_double_2x32", 1, printable_name="packDouble2x32", source_types=(uint_type,), dest_type=double_type, c_expression="data.u64[0] = pack_2x32(op[0]->value.u[0], op[0]->value.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("unpack_double_2x32", 1, printable_name="unpackDouble2x32", source_types=(double_type,), dest_type=uint_type, c_expression="unpack_2x32(op[0]->value.u64[0], &data.u[0], &data.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

    # Sampler/Image packing, part of ARB_bindless_texture.

-   operation("pack_sampler_2x32", 1, printable_name="packSampler2x32", source_types=(uint_type,), dest_type=uint64_type, c_expression="memcpy(&data.u64[0], &op[0]->value.u[0], sizeof(uint64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("pack_image_2x32", 1, printable_name="packImage2x32", source_types=(uint_type,), dest_type=uint64_type, c_expression="memcpy(&data.u64[0], &op[0]->value.u[0], sizeof(uint64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("unpack_sampler_2x32", 1, printable_name="unpackSampler2x32", source_types=(uint64_type,), dest_type=uint_type, c_expression="memcpy(&data.u[0], &op[0]->value.u64[0], sizeof(uint64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("unpack_image_2x32", 1, printable_name="unpackImage2x32", source_types=(uint64_type,), dest_type=uint_type, c_expression="memcpy(&data.u[0], &op[0]->value.u64[0], sizeof(uint64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("pack_sampler_2x32", 1, printable_name="packSampler2x32", source_types=(uint_type,), dest_type=uint64_type, c_expression="data.u64[0] = pack_2x32(op[0]->value.u[0], op[0]->value.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("pack_image_2x32", 1, printable_name="packImage2x32", source_types=(uint_type,), dest_type=uint64_type, c_expression="data.u64[0] = pack_2x32(op[0]->value.u[0], op[0]->value.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("unpack_sampler_2x32", 1, printable_name="unpackSampler2x32", source_types=(uint64_type,), dest_type=uint_type, c_expression="unpack_2x32(op[0]->value.u64[0], &data.u[0], &data.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("unpack_image_2x32", 1, printable_name="unpackImage2x32", source_types=(uint64_type,), dest_type=uint_type, c_expression="unpack_2x32(op[0]->value.u64[0], &data.u[0], &data.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

    operation("frexp_sig", 1),

    operation("frexp_exp", 1),

@@ -592,10 +592,10 @@ ir_expression_operation = [

    operation("ssbo_unsized_array_length", 1),

    # 64-bit integer packing ops.

-   operation("pack_int_2x32", 1, printable_name="packInt2x32", source_types=(int_type,), dest_type=int64_type, c_expression="memcpy(&data.i64[0], &op[0]->value.i[0], sizeof(int64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("pack_uint_2x32", 1, printable_name="packUint2x32", source_types=(uint_type,), dest_type=uint64_type, c_expression="memcpy(&data.u64[0], &op[0]->value.u[0], sizeof(uint64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("unpack_int_2x32", 1, printable_name="unpackInt2x32", source_types=(int64_type,), dest_type=int_type, c_expression="memcpy(&data.i[0], &op[0]->value.i64[0], sizeof(int64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

-   operation("unpack_uint_2x32", 1, printable_name="unpackUint2x32", source_types=(uint64_type,), dest_type=uint_type, c_expression="memcpy(&data.u[0], &op[0]->value.u64[0], sizeof(uint64_t))", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("pack_int_2x32", 1, printable_name="packInt2x32", source_types=(int_type,), dest_type=int64_type, c_expression="data.u64[0] = pack_2x32(op[0]->value.u[0], op[0]->value.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("pack_uint_2x32", 1, printable_name="packUint2x32", source_types=(uint_type,), dest_type=uint64_type, c_expression="data.u64[0] = pack_2x32(op[0]->value.u[0], op[0]->value.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("unpack_int_2x32", 1, printable_name="unpackInt2x32", source_types=(int64_type,), dest_type=int_type, c_expression="unpack_2x32(op[0]->value.u64[0], &data.u[0], &data.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

+   operation("unpack_uint_2x32", 1, printable_name="unpackUint2x32", source_types=(uint64_type,), dest_type=uint_type, c_expression="unpack_2x32(op[0]->value.u64[0], &data.u[0], &data.u[1])", flags=frozenset((horizontal_operation, non_assign_operation))),

    operation("add", 2, printable_name="+", source_types=numeric_types, c_expression="{src0} + {src1}", flags=vector_scalar_operation),

    operation("sub", 2, printable_name="-", source_types=numeric_types, c_expression="{src0} - {src1}", flags=vector_scalar_operation),

-- 

2.26.2