Mesa (staging/20.1): glsl: fix constant packing for 64-bit big endian.

[GitLab Mirror]

Module: Mesa
Branch: staging/20.1
Commit: 047303d422f07243de70af90b139d16fa6ae542e
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=047303d422f07243de70af90b139d16fa6ae542e

Author: Dave Airlie <airlied at redhat.com>
Date:   Mon Jun 29 13:26:56 2020 +1000

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

Cc: <mesa-stable at lists.freedesktop.org>
Reviewed-by: Timothy Arceri <tarceri at itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5679>
(cherry picked from commit 928660527697bc37b475037f9f2579aa4d2205b4)

---

 .pick_status.json                            |  2 +-
 src/compiler/glsl/ir_constant_expression.cpp | 15 +++++++++++++++
 src/compiler/glsl/ir_expression_operation.py | 20 ++++++++++----------
 3 files changed, 26 insertions(+), 11 deletions(-)

diff --git a/.pick_status.json b/.pick_status.json
index 6c588a1b877..7182045f389 100644
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -121,7 +121,7 @@
         "description": "glsl: fix constant packing for 64-bit big endian.",
         "nominated": true,
         "nomination_type": 0,
-        "resolution": 0,
+        "resolution": 1,
         "master_sha": null,
         "because_sha": null
     },
diff --git a/src/compiler/glsl/ir_constant_expression.cpp b/src/compiler/glsl/ir_constant_expression.cpp
index 759924ed9e1..de7d4f6c425 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 8481e2ce640..160626e6ef6 100644
--- a/src/compiler/glsl/ir_expression_operation.py
+++ b/src/compiler/glsl/ir_expression_operation.py
@@ -555,14 +555,14 @@ 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),
@@ -587,10 +587,10 @@ 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),