[Piglit] [PATCH 10/16] msaa/formats: Adjust tolerance appropriately for floating-point formats.

[Paul Berry]

When setting up a floating-point format, the GL implementation
responds to queries GL_ATTACHMENT_{RED,GREEN,BLUE,ALPHA}_SIZE with the
total number of bits in the floating-point value, regardless of
whether those bits are used for sign, exponent, or mantissa.  However,
for the purpose of setting a reasonable tolerance for pass/fail
comparisons, we really should only account for mantissa bits.

This patch modifies the MSAA "formats" test to detect when the buffer
format is GL_FLOAT, and if so, convert the channel size into the
number of mantissa bits, so that we don't test floating-point formats
to an unreasonably tight tolerance.
---
 tests/spec/ext_framebuffer_multisample/formats.cpp |   34 ++++++++++++++++++--
 1 files changed, 31 insertions(+), 3 deletions(-)

diff --git a/tests/spec/ext_framebuffer_multisample/formats.cpp b/tests/spec/ext_framebuffer_multisample/formats.cpp
index a55f053..b8e8352 100644
--- a/tests/spec/ext_framebuffer_multisample/formats.cpp
+++ b/tests/spec/ext_framebuffer_multisample/formats.cpp
@@ -168,6 +168,27 @@ PatternRenderer::try_setup(GLenum internalformat)
 
 
 /**
+ * Return the number of mantissa bits available in an n-bit floating
+ * point format.
+ */
+int
+get_mantissa_bits(int n)
+{
+	switch (n) {
+	case 32: return 23;
+	case 16: return 10;
+	case 11: return 6;
+	case 10: return 5;
+	case 0: return 0; /* Unused channel */
+	default:
+		printf("Unrecognized floating point format (%d bits)\n", n);
+		piglit_report_result(PIGLIT_FAIL);
+		return 0;
+	}
+}
+
+
+/**
  * Set the piglit tolerance appropriately based on the number of bits
  * in each channel.
  */
@@ -176,20 +197,27 @@ void PatternRenderer::set_piglit_tolerance()
 	int tolerance_bits[4];
 
 	for (int i = 0; i < 4; ++i) {
-		if (color_bits[i] == 0) {
+		int bits = color_bits[i];
+		if (component_type == GL_FLOAT) {
+			/* We only want to count mantissa bits for the
+			 * purpose of setting the test tolerance.
+			 */
+			bits = get_mantissa_bits(bits);
+		}
+		if (bits == 0) {
 			/* For channels that have 0 bits, test to 8
 			 * bits precision so we can verify that the
 			 * blit puts in the appropriate value.
 			 */
 			tolerance_bits[i] = 8;
-		} else if (color_bits[i] > 8) {
+		} else if (bits > 8) {
 			/* For channels that have >8 bits, test to 8
 			 * bits precision because we only use an 8-bit
 			 * reference image.
 			 */
 			tolerance_bits[i] = 8;
 		} else {
-			tolerance_bits[i] = color_bits[i];
+			tolerance_bits[i] = bits;
 		}
 	}
 
-- 
1.7.7.6