GS/HW: Purge Hitman Blood Money CRC

RTV(Render target view) is not bound so no need
to output anything, just clip/discard if needed.

Fixes dx11 api warning that pixel shader writes color
output but no rtv is bound.

.github/ISSUE_TEMPLATE/app_bug_report.yaml

@@ -70,6 +70,8 @@ body:
       - Windows 11
       - Windows 10 (64bit)
       - Linux (64bit) - Specify distro below
+      - macOS 26 (Tahoe)
+      - macOS 15 (Sequoia)
       - macOS 14 (Sonoma)
       - macOS 13 (Ventura)
       - macOS 12 (Monterey)

.github/ISSUE_TEMPLATE/emu_bug_report.yaml

@@ -87,6 +87,8 @@ body:
       - Windows 11
       - Windows 10 (64bit)
       - Linux (64bit) - Specify distro below
+      - macOS 26 (Tahoe)
+      - macOS 15 (Sequoia)
       - macOS 14 (Sonoma)
       - macOS 13 (Ventura)
       - macOS 12 (Monterey)
@@ -104,14 +106,14 @@ body:
     id: cpu
     attributes:
       label: CPU
-      placeholder: "Example: i5-7600"
+      placeholder: "Example: Intel i5 12400F"
     validations:
       required: true
   - type: input
     id: gpu
     attributes:
       label: GPU
-      placeholder: "Example: GTX 1070"
+      placeholder: "Example: Nvidia RTX 4060"
     validations:
       required: true
   - type: textarea

bin/docs/ThirdPartyLicenses.html

@@ -24,6 +24,7 @@
 <li><a href="#googletest">GoogleTest</a></li>
 <li><a href="#harfbuzz">HarfBuzz</a></li>
 <li><a href="#jsonformoderncpp">JSON for Modern C++</a></li>
+<li><a href="#kdbindings">KDBindings</a></li>
 <li><a href="#kddockwidgets">KDDockWidgets</a></li>
 <li><a href="#kissfft">kissfft</a></li>
 <li><a href="#libbacktrace">libbacktrace</a></li>
@@ -1910,6 +1911,7 @@ PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
 </pre>
 </div>
 
+<!-- Transitive dependency of KDDockWidgets. -->
 <div id="jsonformoderncpp">
 <h3>JSON for Modern C++ - <a href="https://json.nlohmann.me">https://json.nlohmann.me</a></h3>
 <pre>
@@ -1937,6 +1939,34 @@ SOFTWARE.
 </pre>
 </div>
 
+<!-- Transitive dependency of KDDockWidgets. -->
+<div id="kdbindings">
+<h3>KDBindings - <a href="https://github.com/KDAB/KDBindings">https://github.com/KDAB/KDBindings</a></h3>
+<pre>
+MIT License
+
+Copyright 2021 Klarälvdalens Datakonsult AB, a KDAB Group company &lt;info@kdab.com&gt;
+Copyright 2021 Jeremy Burns
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the &quot;Software&quot;), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
+the Software, and to permit persons to whom the Software is furnished to do so,
+subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED &quot;AS IS&quot;, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
+COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
+IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+</pre>
+</div>
+
 <div id="kddockwidgets">
 <h3>KDDockWidgets - <a href="https://github.com/KDAB/KDDockWidgets">https://github.com/KDAB/KDDockWidgets</a></h3>
 <pre>

bin/resources/shaders/dx11/convert.fx

@@ -110,48 +110,24 @@ uint ps_convert_rgba8_16bits(PS_INPUT input) : SV_Target0
 	return ((i.x & 0x00F8u) >> 3) | ((i.y & 0x00F8u) << 2) | ((i.z & 0x00f8u) << 7) | ((i.w & 0x80u) << 8);
 }
 
-PS_OUTPUT ps_datm1(PS_INPUT input)
+void ps_datm1(PS_INPUT input)
 {
-	PS_OUTPUT output;
-	
 	clip(sample_c(input.t).a - 127.5f / 255); // >= 0x80 pass
-	
-	output.c = 0;
-
-	return output;
 }
 
-PS_OUTPUT ps_datm0(PS_INPUT input)
+void ps_datm0(PS_INPUT input)
 {
-	PS_OUTPUT output;
-	
 	clip(127.5f / 255 - sample_c(input.t).a); // < 0x80 pass (== 0x80 should not pass)
-	
-	output.c = 0;
-
-	return output;
 }
 
-PS_OUTPUT ps_datm1_rta_correction(PS_INPUT input)
+void ps_datm1_rta_correction(PS_INPUT input)
 {
-	PS_OUTPUT output;
-
 	clip(sample_c(input.t).a - 254.5f / 255); // >= 0x80 pass
-
-	output.c = 0;
-
-	return output;
 }
 
-PS_OUTPUT ps_datm0_rta_correction(PS_INPUT input)
+void ps_datm0_rta_correction(PS_INPUT input)
 {
-	PS_OUTPUT output;
-
 	clip(254.5f / 255 - sample_c(input.t).a); // < 0x80 pass (== 0x80 should not pass)
-
-	output.c = 0;
-
-	return output;
 }
 
 PS_OUTPUT ps_rta_correction(PS_INPUT input)

bin/resources/shaders/vulkan/convert.glsl

@@ -104,7 +104,6 @@ void ps_datm1()
 {
 	if(sample_c(v_tex).a < (127.5f / 255.0f)) // >= 0x80 pass
 		discard;
-
 }
 #endif
 

pcsx2-qt/Translations/pcsx2-qt_en.ts

@@ -10806,47 +10806,47 @@ Do you want to load this save and continue?</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="659"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="658"/>
         <source>CAS is not available, your graphics driver does not support the required functionality.</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="714"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="713"/>
         <source>with no compression</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="721"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="720"/>
         <source>with LZMA compression</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="728"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="727"/>
         <source>with Zstandard compression</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="734"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="733"/>
         <source>Saving {0} GS dump {1} to &apos;{2}&apos;</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="735"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="734"/>
         <source>single frame</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="735"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="734"/>
         <source>multi-frame</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="755"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="754"/>
         <source>Failed to render/download screenshot.</source>
         <translation type="unfinished"></translation>
     </message>
     <message>
-        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="766"/>
+        <location filename="../../pcsx2/GS/Renderers/Common/GSRenderer.cpp" line="765"/>
         <source>Saved GS dump to &apos;{}&apos;.</source>
         <translation type="unfinished"></translation>
     </message>

pcsx2/GS/GSState.cpp

@@ -4237,7 +4237,10 @@ void GSState::CalcAlphaMinMax(const int tex_alpha_min, const int tex_alpha_max)
 
 	if (IsCoverageAlpha())
 	{
-		min = 128;
+		// HW renderer doesn't currently support AA, so its min is 128.
+		// If we add AA support to the HW renderer, this will need to be changed.
+		// (Will probably only be supported with ROV/FBFetch so we would want to check for that.)
+		min = GSIsHardwareRenderer() ? 128 : 0;
 		max = 128;
 	}
 	else

pcsx2/GS/Renderers/HW/GSHwHack.cpp

@@ -822,23 +822,6 @@ bool GSHwHack::GSC_MetalGearSolid3(GSRendererHW& r, int& skip)
 	return true;
 }
 
-bool GSHwHack::GSC_HitmanBloodMoney(GSRendererHW& r, int& skip)
-{
-	// The game does a stupid thing where it backs up the last 2 pages of the framebuffer with shuffles, uploads a CT32 texture to it
-	// then copies the RGB back (keeping the new alpha only). It's pretty gross, I have no idea why they didn't just upload a new alpha.
-	// This is a real pain to emulate with the current state of things, so let's just clear the dirty area from the upload and pretend it wasn't there.
-	
-	// Catch the first draw of the copy back.
-	if (RFBP > 0 && RTPSM == PSMT8H && RFPSM == PSMCT32)
-	{
-		GSTextureCache::Target* target = g_texture_cache->FindOverlappingTarget(RFBP, RFBP + 1);
-		if (target)
-			target->m_dirty.clear();
-	}
-
-	return false;
-}
-
 bool GSHwHack::OI_PointListPalette(GSRendererHW& r, GSTexture* rt, GSTexture* ds, GSTextureCache::Source* t)
 {
 	const u32 n_vertices = r.m_vertex.next;
@@ -1328,7 +1311,6 @@ const GSHwHack::Entry<GSRendererHW::GSC_Ptr> GSHwHack::s_get_skip_count_function
 	CRC_F(GSC_NFSUndercover),
 	CRC_F(GSC_PolyphonyDigitalGames),
 	CRC_F(GSC_MetalGearSolid3),
-	CRC_F(GSC_HitmanBloodMoney),
 	CRC_F(GSC_Battlefield2),
 
 	// Channel Effect

pcsx2/GS/Renderers/HW/GSHwHack.h

@@ -26,7 +26,6 @@ public:
 	static bool GSC_Battlefield2(GSRendererHW& r, int& skip);
 	static bool GSC_PolyphonyDigitalGames(GSRendererHW& r, int& skip);
 	static bool GSC_MetalGearSolid3(GSRendererHW& r, int& skip);
-	static bool GSC_HitmanBloodMoney(GSRendererHW& r, int& skip);
 
 	static bool OI_PointListPalette(GSRendererHW& r, GSTexture* rt, GSTexture* ds, GSTextureCache::Source* t);
 	static bool OI_DBZBTGames(GSRendererHW& r, GSTexture* rt, GSTexture* ds, GSTextureCache::Source* t);

pcsx2/GS/Renderers/HW/GSRendererHW.cpp

@@ -3045,19 +3045,19 @@ void GSRendererHW::Draw()
 			if (scale_draw == 1)
 			{
 				target_scale = 1.0f;
-				m_downscale_source = src->m_from_target->GetScale() > 1.0f;
+				m_downscale_source = src->m_from_target ? src->m_from_target->GetScale() > 1.0f : false;
 			}
 			else
-				m_downscale_source = GSConfig.UserHacks_NativeScaling != GSNativeScaling::Aggressive ? false : src->m_from_target->GetScale() > 1.0f; // Bad for GTA + Full Spectrum Warrior, good for Sacred Blaze + Parappa.
+				m_downscale_source = (GSConfig.UserHacks_NativeScaling != GSNativeScaling::Aggressive || !src->m_from_target) ? false : src->m_from_target->GetScale() > 1.0f; // Bad for GTA + Full Spectrum Warrior, good for Sacred Blaze + Parappa.
 		}
 		else
 		{
 			// if it's directly copying keep the scale - Ratchet and clank hits this, stops edge garbage happening.
 			// Keep it to small targets of 256 or lower.
-			if (scale_draw == -1 && src && src->m_from_target && src->m_from_target->m_downscaled && ((static_cast<int>(m_cached_ctx.FRAME.FBW * 64) <= (PCRTCDisplays.GetResolution().x >> 1) &&
+			if (scale_draw == -1 && src && (!src->m_from_target || (src->m_from_target && src->m_from_target->m_downscaled)) && ((static_cast<int>(m_cached_ctx.FRAME.FBW * 64) <= (PCRTCDisplays.GetResolution().x >> 1) &&
 				(GSVector4i(m_vt.m_min.p).xyxy() == GSVector4i(m_vt.m_min.t).xyxy()).alltrue() && (GSVector4i(m_vt.m_max.p).xyxy() == GSVector4i(m_vt.m_max.t).xyxy()).alltrue()) || possible_shuffle))
 			{
-				target_scale = src->m_from_target->GetScale();
+				target_scale = src->m_from_target ? src->m_from_target->GetScale() : 1.0f;
 				scale_draw = 1;
 				scaled_copy = true;
 			}
@@ -3073,7 +3073,7 @@ void GSRendererHW::Draw()
 	// This upscaling hack is for games which construct P8 textures by drawing a bunch of small sprites in C32,
 	// then reinterpreting it as P8. We need to keep the off-screen intermediate textures at native resolution,
 	// but not propagate that through to the normal render targets. Test Case: Crash Wrath of Cortex.
-	if (no_ds && src && !m_channel_shuffle && src->m_from_target && (GSConfig.UserHacks_NativePaletteDraw || (src->m_from_target->m_downscaled && scale_draw <= 1)) &&
+	if (no_ds && src && !m_channel_shuffle && src->m_from_target && (GSConfig.UserHacks_NativePaletteDraw || (src->m_target_direct  && src->m_from_target->m_downscaled && scale_draw <= 1)) &&
 		src->m_scale == 1.0f && (src->m_TEX0.PSM == PSMT8 || src->m_TEX0.TBP0 == m_cached_ctx.FRAME.Block()))
 	{
 		GL_CACHE("HW: Using native resolution for target based on texture source");
@@ -3405,7 +3405,7 @@ void GSRendererHW::Draw()
 				return;
 			}
 
-			rt = g_texture_cache->CreateTarget(FRAME_TEX0, t_size, GetValidSize(src, possible_shuffle), (GSConfig.UserHacks_NativeScaling != GSNativeScaling::Off && scale_draw < 0 && is_possible_mem_clear != ClearType::NormalClear) ? src->m_from_target->GetScale() : target_scale, 
+			rt = g_texture_cache->CreateTarget(FRAME_TEX0, t_size, GetValidSize(src, possible_shuffle), (GSConfig.UserHacks_NativeScaling != GSNativeScaling::Off && scale_draw < 0 && is_possible_mem_clear != ClearType::NormalClear) ? ((src && src->m_from_target) ? src->m_from_target->GetScale() : (ds ? ds->m_scale : 1.0f)) : target_scale, 
 												GSTextureCache::RenderTarget, true, fm, false, force_preload, preserve_rt_color || possible_shuffle, lookup_rect, src);
 
 			if (!rt) [[unlikely]]
@@ -4039,7 +4039,8 @@ void GSRendererHW::Draw()
 		bool valid_width_change = false;
 		if (rt && ((!is_possible_mem_clear || blending_cd) || rt->m_TEX0.PSM != FRAME_TEX0.PSM) && !m_in_target_draw)
 		{
-			valid_width_change = rt->m_TEX0.TBW != FRAME_TEX0.TBW;
+			const u32 frame_mask = (m_cached_ctx.FRAME.FBMSK & frame_psm.fmsk);
+			valid_width_change = rt->m_TEX0.TBW != FRAME_TEX0.TBW && (frame_mask != (frame_psm.fmsk & 0x00FFFFFF) || rt->m_valid_rgb == false);
 			if (valid_width_change && !m_cached_ctx.ZBUF.ZMSK && (m_cached_ctx.FRAME.FBMSK & 0xFF000000))
 			{
 				// Alpha could be a font, and since the width is changing it's no longer valid.
@@ -4178,6 +4179,10 @@ void GSRendererHW::Draw()
 
 		const int new_w = std::min(2048, std::max(new_size.x, std::max(rt ? rt->m_unscaled_size.x : 0, ds ? ds_size.x : 0)));
 		const int new_h = std::min(2048, std::max(new_size.y, std::max(rt ? rt->m_unscaled_size.y : 0, ds ? ds_size.y : 0)));
+
+		const bool full_cover_clear = is_possible_mem_clear && GSLocalMemory::IsPageAligned(m_cached_ctx.FRAME.PSM, m_r) && m_r.x == 0 && m_r.y == 0 && !preserve_rt_rgb &&
+									  !IsPageCopy() && m_r.width() == (m_cached_ctx.FRAME.FBW * 64);
+
 		if (rt)
 		{
 			const u32 old_end_block = rt->m_end_block;
@@ -4224,9 +4229,9 @@ void GSRendererHW::Draw()
 					g_texture_cache->AddDirtyRectTarget(rt, GSVector4i(rt->m_valid.x, rt->m_valid.w, rt->m_valid.z, new_h), rt->m_TEX0.PSM, rt->m_TEX0.TBW, mask, false);
 					g_texture_cache->GetTargetSize(rt->m_TEX0.TBP0, rt->m_TEX0.TBW, rt->m_TEX0.PSM, 0, new_h);
 				}
-
-				rt->ResizeValidity(rt->m_valid.rintersect(rt->GetUnscaledRect()));
-				rt->ResizeDrawn(rt->m_drawn_since_read.rintersect(rt->GetUnscaledRect()));
+				const bool rt_cover = full_cover_clear && (m_r.height() + frame_psm.pgs.y) >= rt->m_valid.height();
+				rt->ResizeValidity(rt_cover ? update_rect : rt->m_valid.rintersect(rt->GetUnscaledRect()));
+				rt->ResizeDrawn(rt_cover ? update_rect : rt->m_drawn_since_read.rintersect(rt->GetUnscaledRect()));
 			}
 
 			const bool rt_update = can_update_size || (m_texture_shuffle && (src && rt && src->m_from_target != rt));
@@ -4234,7 +4239,7 @@ void GSRendererHW::Draw()
 			// If it's updating from a texture shuffle, limit the size to the source size.
 			if (rt_update && !can_update_size)
 			{
-				if(src->m_from_target)
+				if (src->m_from_target)
 					update_rect = update_rect.rintersect(src->m_from_target->m_valid);
 
 				update_rect = update_rect.rintersect(GSVector4i::loadh(GSVector2i(new_w, new_h)));
@@ -4304,16 +4309,18 @@ void GSRendererHW::Draw()
 						DevCon.Warning("HW: Temporary depth buffer creation failed.");
 				}
 			}
+			const bool z_masked = m_cached_ctx.ZBUF.ZMSK;
+
 			if (!m_texture_shuffle && !m_channel_shuffle)
 			{
-				ds->ResizeValidity(ds->GetUnscaledRect());
-				ds->ResizeDrawn(ds->GetUnscaledRect());
+				const bool z_cover = full_cover_clear && (m_r.height() + GSLocalMemory::m_psm[m_cached_ctx.ZBUF.PSM].pgs.y) >= ds->m_valid.height();
+				ds->ResizeValidity(z_cover ? m_r : ds->GetUnscaledRect());
+				ds->ResizeDrawn(z_cover ? m_r : ds->GetUnscaledRect());
 			}
 
 			// Limit to 2x the vertical height of the resolution (for double buffering)
 			// Dark cloud writes to 424 when the buffer is only 416 high, but masks the Z.
 			// Updating the valid causes the Z to overlap the framebuffer, which is obviously incorrect.
-			const bool z_masked = m_cached_ctx.ZBUF.ZMSK;
 			const bool z_update = can_update_size && !z_masked;
 
 			if (rt && m_using_temp_z)
@@ -8974,6 +8981,9 @@ bool GSRendererHW::TextureCoversWithoutGapsNotEqual()
 
 int GSRendererHW::IsScalingDraw(GSTextureCache::Source* src, bool no_gaps)
 {
+	if (GSConfig.UserHacks_NativeScaling == GSNativeScaling::Off)
+		return 0;
+
 	const GSVector2i draw_size = GSVector2i(m_vt.m_max.p.x - m_vt.m_min.p.x, m_vt.m_max.p.y - m_vt.m_min.p.y);
 	GSVector2i tex_size = GSVector2i(m_vt.m_max.t.x - m_vt.m_min.t.x, m_vt.m_max.t.y - m_vt.m_min.t.y);
 
@@ -8984,10 +8994,13 @@ int GSRendererHW::IsScalingDraw(GSTextureCache::Source* src, bool no_gaps)
 
 	// Try to catch cases of stupid draws like Manhunt and Syphon Filter where they sample a single pixel.
 	// Also make sure it's grabbing most of the texture.
-	if (tex_size.x == 0 || tex_size.y == 0 || draw_size.x == 0 || draw_size.y == 0 || !is_target_src)
+	if (tex_size.x == 0 || tex_size.y == 0 || draw_size.x == 0 || draw_size.y == 0)
 		return 0;
 
-	if (is_target_src && src->m_from_target->m_downscaled && std::abs(draw_size.x - tex_size.x) <= 1 && std::abs(draw_size.y - tex_size.y) <= 1)
+	const bool no_resize = (std::abs(draw_size.x - tex_size.x) <= 1 && std::abs(draw_size.y - tex_size.y) <= 1);
+	const bool can_maintain = no_resize || (!is_target_src && m_index.tail == 2);
+
+	if (!src || ((!is_target_src || src->m_from_target->m_downscaled) && can_maintain))
 		return -1;
 
 	const GSDrawingContext& next_ctx = m_env.CTXT[m_env.PRIM.CTXT];
@@ -9000,8 +9013,9 @@ int GSRendererHW::IsScalingDraw(GSTextureCache::Source* src, bool no_gaps)
 	const bool is_downscale = m_cached_ctx.TEX0.TBW >= m_cached_ctx.FRAME.FBW && draw_size.x <= (tex_size.x * 0.75f) && draw_size.y <= (tex_size.y * 0.75f);
 	// Check we're getting most of the texture and not just stenciling a part of it.
 	// Only allow non-bilineared downscales if it's most of the target (misdetections of shadows in Naruto, Transformers etc), otherwise it's fine.
-	const GSVector2i tex_size_half = GSVector2i((src->GetRegion().HasX() ? src->GetRegionSize().x : src->m_from_target->m_valid.width()) / 2, (src->GetRegion().HasY() ? src->GetRegionSize().y : src->m_from_target->m_valid.height()) / 2);
-	const bool possible_downscale = m_context->TEX1.MMAG == 1 || src->m_from_target->m_downscaled || tex_size.x >= tex_size_half.x || tex_size.y >= tex_size_half.y;
+	const GSVector4i src_valid = src->m_from_target ? src->m_from_target->m_valid : src->m_valid_rect;
+	const GSVector2i tex_size_half = GSVector2i((src->GetRegion().HasX() ? src->GetRegionSize().x : src_valid.width()) / 2, (src->GetRegion().HasY() ? src->GetRegionSize().y : src_valid.height()) / 2);
+	const bool possible_downscale = m_context->TEX1.MMIN == 1 || !src->m_from_target || src->m_from_target->m_downscaled || tex_size.x >= tex_size_half.x || tex_size.y >= tex_size_half.y;
 
 	if (is_downscale && (draw_size.x >= PCRTCDisplays.GetResolution().x || !possible_downscale))
 		return 0;
@@ -9019,6 +9033,39 @@ int GSRendererHW::IsScalingDraw(GSTextureCache::Source* src, bool no_gaps)
 		return is_upscale ? 2 : 1;
 	}
 
+	// Last ditched check if it's doing a lot of small draws exactly the same which could be recursive lighting bloom.
+	if (m_vt.m_primclass == GS_SPRITE_CLASS && m_index.tail > 2 && !no_gaps_or_single_sprite && m_context->TEX1.MMAG == 1 && !m_context->ALPHA.IsOpaque())
+	{
+		GSVertex* v = &m_vertex.buff[0];
+		float tw = 1 << src->m_TEX0.TW;
+		float th = 1 << src->m_TEX0.TH;
+
+		const int first_u = (PRIM->FST) ? (v[1].U - v[0].U) >> 4 : std::floor(static_cast<int>(tw * v[1].ST.S) - static_cast<int>(tw * v[0].ST.S));
+		const int first_v = (PRIM->FST) ? (v[1].V - v[0].V) >> 4 : std::floor(static_cast<int>(th * v[1].ST.T) - static_cast<int>(th * v[0].ST.T));
+		const int first_x = (v[1].XYZ.X - v[0].XYZ.X) >> 4;
+		const int first_y = (v[1].XYZ.Y - v[0].XYZ.Y) >> 4;
+
+		if (first_x > first_u && first_y > first_v && !no_resize && std::abs(draw_size.x - first_x) <= 4 && std::abs(draw_size.y - first_y) <= 4)
+		{
+			for (u32 i = 2; i < m_index.tail; i += 2)
+			{
+				const int next_u = (PRIM->FST) ? (v[i + 1].U - v[i].U) >> 4 : std::floor(static_cast<int>(tw * v[i + 1].ST.S) - static_cast<int>(tw * v[i].ST.S));
+				const int next_v = (PRIM->FST) ? (v[i + 1].V - v[i].V) >> 4 : std::floor(static_cast<int>(th * v[i + 1].ST.T) - static_cast<int>(th * v[i].ST.T));
+				const int next_x = (v[i + 1].XYZ.X - v[i].XYZ.X) >> 4;
+				const int next_y = (v[i + 1].XYZ.Y - v[i].XYZ.Y) >> 4;
+
+				if (std::abs(draw_size.x - next_x) > 4 || std::abs(draw_size.y - next_y) > 4)
+					break;
+
+				if (next_u != first_u || next_v != first_v || next_x != first_x || next_y != first_y)
+					break;
+
+				if (i + 2 >= m_index.tail)
+					return 2;
+			}
+		}
+	}
+
 	return 0;
 }
 

pcsx2/GS/Renderers/HW/GSTextureCache.cpp

@@ -424,7 +424,7 @@ GSVector4i GSTextureCache::TranslateAlignedRectByPage(u32 tbp, u32 tebp, u32 tbw
 			}
 		}
 	}
-	else // Widths match
+	else if (!block_offset) // Widths match
 	{
 		const int horizontal_dst_page_offset = page_offset % clamped_tbw;
 		const int vertical_dst_page_offset = page_offset / clamped_tbw;
@@ -1626,14 +1626,12 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const bool is_color, const
 					{
 						// It is a complex to convert the code in shader. As a reference, let's do it on the CPU,
 						// it will be slow but can work even with upscaling, also fine tune it so it's not enabled when not needed.
-						if (psm == PSMT4 || (GSConfig.UserHacks_CPUFBConversion && psm == PSMT8 && (!possible_shuffle || GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp != 32)) ||
-							(psm == PSMT8H && GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp == 16))
+						if (psm == PSMT4 || (psm == PSMT8H && GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp == 16))
 						{
-							// Forces 4-bit and 8-bit frame buffer conversion to be done on the CPU instead of the GPU, but performance will be slower.
-							// There is no dedicated shader to handle 4-bit conversion (Beyond Good and Evil and Stuntman).
+							// Enable readbacks on PSMT4 as we don't have a dedicated shader (Beyond Good and Evil and Stuntman).
+							// Enable readbacks on PSMT8H 16bit as we don't have a dedicated shader (History Channel - Battle for the Pacific, Sea World - Shamu's Big Adventure). 
 							// Note: Stuntman no longer hits the PSMT4 code path.
-							// Direct3D10/11 and OpenGL support 8-bit fb conversion but don't render some corner cases properly (Harry Potter games).
-							// The hack can fix glitches in some games.
+							// Note2: Harry Potter is now properly handled with shader conversion and no need to enable frame buffer conversion.
 							if (!t->m_drawn_since_read.rempty())
 							{
 								t->UnscaleRTAlpha();
@@ -2043,11 +2041,29 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const bool is_color, const
 			{
 				rect.y /= 2;
 				rect.w /= 2;
+
+				if (region.HasY())
+				{
+					const u32 min_y = region.GetMinY() / 2;
+					const u32 max_y = region.GetMaxY() / 2;
+
+					region.ClearY();
+					region.SetY(min_y, max_y);
+				}
 			}
 			else
 			{
 				rect.x /= 2;
 				rect.z /= 2;
+
+				if (region.HasX())
+				{
+					const u32 min_x = region.GetMinX() / 2;
+					const u32 max_x = region.GetMaxX() / 2;
+
+					region.ClearX();
+					region.SetX(min_x, max_x);
+				}
 			}
 			if (TEX0.TBP0 == frame.Block())
 			{

pcsx2/GS/Renderers/HW/GSTextureCache.h

@@ -45,6 +45,9 @@ public:
 		bool HasY() const { return static_cast<u32>(bits >> 32) != 0; }
 		bool HasEither() const { return (bits != 0); }
 
+		void ClearX() { bits &= ~0xFFFFFFFFULL; }
+		void ClearY() { bits &= 0xFFFFFFFFULL; }
+
 		void SetX(s32 min, s32 max) { bits |= (static_cast<u64>(static_cast<u16>(min)) | (static_cast<u64>(static_cast<u16>(max) << 16))); }
 		void SetY(s32 min, s32 max) { bits |= ((static_cast<u64>(static_cast<u16>(min)) << 32) | (static_cast<u64>(static_cast<u16>(max)) << 48)); }
 

pcsx2/ImGui/ImGuiOverlays.cpp

@@ -416,10 +416,10 @@ __ri void ImGuiManager::DrawSettingsOverlay(float scale, float margin, float spa
 
 		APPEND("B={} PL={} ", static_cast<unsigned>(GSConfig.AccurateBlendingUnit), static_cast<unsigned>(GSConfig.TexturePreloading));
 		if (GSConfig.GPUPaletteConversion)
-			APPEND("PT ");
+			APPEND("PLTX ");
 
 		if (GSConfig.HWDownloadMode != GSHardwareDownloadMode::Enabled)
-			APPEND("DL={} ", static_cast<unsigned>(GSConfig.HWDownloadMode));
+			APPEND("HWDM={} ", static_cast<unsigned>(GSConfig.HWDownloadMode));
 
 		if (GSConfig.HWMipmap)
 			APPEND("MM ");
@@ -466,7 +466,7 @@ __ri void ImGuiManager::DrawSettingsOverlay(float scale, float margin, float spa
 		if (GSConfig.UserHacks_CPUFBConversion)
 			APPEND("FBC ");
 		if (GSConfig.UserHacks_ReadTCOnClose)
-			APPEND("FTC ");
+			APPEND("RTOC ");
 		if (GSConfig.UserHacks_DisableDepthSupport)
 			APPEND("DDC ");
 		if (GSConfig.UserHacks_DisablePartialInvalidation)

pcsx2/ShaderCacheVersion.h

@@ -3,4 +3,4 @@
 
 /// Version number for GS and other shaders. Increment whenever any of the contents of the
 /// shaders change, to invalidate the cache.
-static constexpr u32 SHADER_CACHE_VERSION = 66;
+static constexpr u32 SHADER_CACHE_VERSION = 67;