Work around dxc matrix layout behavior (#694)

The Slang compiler allows the default matrix layout convention (row-major vs. column-major) to be specified via the command line or API.
When generating output HLSL, Slang emits a `#pragma pack_matrix` directive for the chosen default convention, so that a user can generate plain HLSL output and still have it encode their desired defaults.

The problem that has arisen is that many released versions of dxc (including those in the most recent Windows SDK at this time) *ignore* the `#pragma pack_matrix` directive (the feature has since been added to top-of-tree dxc).

The main fix here is to instead pass the `-Zpr` option in to dxc when invoking it if the row-major (non-default) convention is requested.
This will solve the problem for clients that use Slang to generate DXIL, but not for clients who use Slang to generate plain HLSL that they then pass into dxc (those clients are assumed to be able to work around the problem for themselves).

In order to test the change, I added a test that fills a constant buffer with sequential integers, and then reads out the rows/columns of an `int3x4` matrix with both row- and column-major layout, as well as an integer placed *after* the matrix, so we can see the offset it was given.

The `render-test` application did not yet support generating code via dxc/DXIL, so I added an option for that.
This ends up assuming that anybody who is running the D3D12 tests will also have a version of dxc available.

2 files changed, 78 insertions, 0 deletions

diff --git a/tests/compute/matrix-layout.hlsl b/tests/compute/matrix-layout.hlsl
new file mode 100644
index 000000000..68366fc5b
--- /dev/null
+++ b/tests/compute/matrix-layout.hlsl
@@ -0,0 +1,66 @@
+// dxc-matrix-layout.hlsl
+
+// This test tries to ensure that a `row_major` layout default specified on
+// the command line makes it through to affect code generation on all targets.
+// The test was created because it was found that released versions of dxc
+// were ignoring the `#pragma pack_matrix` directive.
+
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -xslang -matrix-layout-row-major
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -xslang -matrix-layout-row-major
+
+// Not testing on Vulkan because of lack of support
+// for integer matrices in GLSL. Slang needs to
+// supporting lowering of such matrix before we
+// can run this test.
+//
+//NO_TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -xslang -matrix-layout-row-major
+
+struct S
+{
+    int3x4 a;
+    int    b;
+};
+
+//TEST_INPUT:cbuffer(data=[1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24]):dxbinding(0),glbinding(0)
+cbuffer C0
+{
+    S s;
+};
+
+//TEST_INPUT:cbuffer(data=[1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24]):dxbinding(1),glbinding(1)
+cbuffer C1
+{
+    column_major
+    int3x4 cc;
+    int    dd;
+};
+
+int test(int val)
+{
+    int N = 256;
+
+    // Note: using `val %3` here instead of `val %4` in order
+    // to work around a code generation issue in dxc.
+
+    int a = s.a[val / 4][val % 3];
+    int b = s.b;
+
+    int c = cc[val / 4][val % 4];
+    int d = dd;
+
+    return ((a*N + b) * N + c) * N + d;
+}
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0 0 0 0 0 0 0 0 0], stride=4):dxbinding(0),glbinding(2),out
+RWStructuredBuffer<int> buffer;
+
+[numthreads(12, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    uint tid = dispatchThreadID.x;
+
+    int val = tid;
+    val = test(val);
+
+    buffer[tid] = val;
+}
diff --git a/tests/compute/matrix-layout.hlsl.expected.txt b/tests/compute/matrix-layout.hlsl.expected.txt
new file mode 100644
index 000000000..cb8e2cae7
--- /dev/null
+++ b/tests/compute/matrix-layout.hlsl.expected.txt
@@ -0,0 +1,12 @@
+10D0111
+20D0511
+30D0911
+10D0D11
+60D0211
+70D0611
+50D0A11
+60D0E11
+B0D0311
+90D0711
+A0D0B11
+B0D0F11